1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static DEFINE_MUTEX(ext4_li_mtx);
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static void ext4_update_super(struct super_block *sb);
69 static int ext4_commit_super(struct super_block *sb);
70 static int ext4_mark_recovery_complete(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_clear_journal_err(struct super_block *sb,
73 struct ext4_super_block *es);
74 static int ext4_sync_fs(struct super_block *sb, int wait);
75 static int ext4_remount(struct super_block *sb, int *flags, char *data);
76 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
77 static int ext4_unfreeze(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
87 static struct inode *ext4_get_journal_inode(struct super_block *sb,
88 unsigned int journal_inum);
93 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
94 * i_mmap_rwsem (inode->i_mmap_rwsem)!
97 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
98 * page lock -> i_data_sem (rw)
100 * buffered write path:
101 * sb_start_write -> i_mutex -> mmap_lock
102 * sb_start_write -> i_mutex -> transaction start -> page lock ->
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
107 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
111 * sb_start_write -> i_mutex -> mmap_lock
112 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
115 * transaction start -> page lock(s) -> i_data_sem (rw)
118 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
119 static struct file_system_type ext2_fs_type = {
120 .owner = THIS_MODULE,
123 .kill_sb = kill_block_super,
124 .fs_flags = FS_REQUIRES_DEV,
126 MODULE_ALIAS_FS("ext2");
127 MODULE_ALIAS("ext2");
128 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
130 #define IS_EXT2_SB(sb) (0)
134 static struct file_system_type ext3_fs_type = {
135 .owner = THIS_MODULE,
138 .kill_sb = kill_block_super,
139 .fs_flags = FS_REQUIRES_DEV,
141 MODULE_ALIAS_FS("ext3");
142 MODULE_ALIAS("ext3");
143 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
146 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
150 * buffer's verified bit is no longer valid after reading from
151 * disk again due to write out error, clear it to make sure we
152 * recheck the buffer contents.
154 clear_buffer_verified(bh);
156 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
158 submit_bh(REQ_OP_READ, op_flags, bh);
161 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
164 BUG_ON(!buffer_locked(bh));
166 if (ext4_buffer_uptodate(bh)) {
170 __ext4_read_bh(bh, op_flags, end_io);
173 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
175 BUG_ON(!buffer_locked(bh));
177 if (ext4_buffer_uptodate(bh)) {
182 __ext4_read_bh(bh, op_flags, end_io);
185 if (buffer_uptodate(bh))
190 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
192 if (trylock_buffer(bh)) {
194 return ext4_read_bh(bh, op_flags, NULL);
195 ext4_read_bh_nowait(bh, op_flags, NULL);
200 if (buffer_uptodate(bh))
208 * This works like __bread_gfp() except it uses ERR_PTR for error
209 * returns. Currently with sb_bread it's impossible to distinguish
210 * between ENOMEM and EIO situations (since both result in a NULL
213 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
214 sector_t block, int op_flags,
217 struct buffer_head *bh;
220 bh = sb_getblk_gfp(sb, block, gfp);
222 return ERR_PTR(-ENOMEM);
223 if (ext4_buffer_uptodate(bh))
226 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
234 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
237 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
240 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
243 return __ext4_sb_bread_gfp(sb, block, 0, 0);
246 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
248 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
251 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
256 static int ext4_verify_csum_type(struct super_block *sb,
257 struct ext4_super_block *es)
259 if (!ext4_has_feature_metadata_csum(sb))
262 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
265 static __le32 ext4_superblock_csum(struct super_block *sb,
266 struct ext4_super_block *es)
268 struct ext4_sb_info *sbi = EXT4_SB(sb);
269 int offset = offsetof(struct ext4_super_block, s_checksum);
272 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
274 return cpu_to_le32(csum);
277 static int ext4_superblock_csum_verify(struct super_block *sb,
278 struct ext4_super_block *es)
280 if (!ext4_has_metadata_csum(sb))
283 return es->s_checksum == ext4_superblock_csum(sb, es);
286 void ext4_superblock_csum_set(struct super_block *sb)
288 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
290 if (!ext4_has_metadata_csum(sb))
293 es->s_checksum = ext4_superblock_csum(sb, es);
296 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
297 struct ext4_group_desc *bg)
299 return le32_to_cpu(bg->bg_block_bitmap_lo) |
300 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
301 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
304 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
305 struct ext4_group_desc *bg)
307 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
308 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
309 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
312 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
313 struct ext4_group_desc *bg)
315 return le32_to_cpu(bg->bg_inode_table_lo) |
316 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
317 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
320 __u32 ext4_free_group_clusters(struct super_block *sb,
321 struct ext4_group_desc *bg)
323 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
324 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
325 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
328 __u32 ext4_free_inodes_count(struct super_block *sb,
329 struct ext4_group_desc *bg)
331 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
332 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
333 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
336 __u32 ext4_used_dirs_count(struct super_block *sb,
337 struct ext4_group_desc *bg)
339 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
340 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
341 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
344 __u32 ext4_itable_unused_count(struct super_block *sb,
345 struct ext4_group_desc *bg)
347 return le16_to_cpu(bg->bg_itable_unused_lo) |
348 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
349 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
352 void ext4_block_bitmap_set(struct super_block *sb,
353 struct ext4_group_desc *bg, ext4_fsblk_t blk)
355 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
356 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
357 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
360 void ext4_inode_bitmap_set(struct super_block *sb,
361 struct ext4_group_desc *bg, ext4_fsblk_t blk)
363 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
364 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
365 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
368 void ext4_inode_table_set(struct super_block *sb,
369 struct ext4_group_desc *bg, ext4_fsblk_t blk)
371 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
372 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
373 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
376 void ext4_free_group_clusters_set(struct super_block *sb,
377 struct ext4_group_desc *bg, __u32 count)
379 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
380 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
381 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
384 void ext4_free_inodes_set(struct super_block *sb,
385 struct ext4_group_desc *bg, __u32 count)
387 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
388 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
389 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
392 void ext4_used_dirs_set(struct super_block *sb,
393 struct ext4_group_desc *bg, __u32 count)
395 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
396 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
397 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
400 void ext4_itable_unused_set(struct super_block *sb,
401 struct ext4_group_desc *bg, __u32 count)
403 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
404 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
405 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
408 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
410 now = clamp_val(now, 0, (1ull << 40) - 1);
412 *lo = cpu_to_le32(lower_32_bits(now));
413 *hi = upper_32_bits(now);
416 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
418 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
420 #define ext4_update_tstamp(es, tstamp) \
421 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
422 ktime_get_real_seconds())
423 #define ext4_get_tstamp(es, tstamp) \
424 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
427 * The del_gendisk() function uninitializes the disk-specific data
428 * structures, including the bdi structure, without telling anyone
429 * else. Once this happens, any attempt to call mark_buffer_dirty()
430 * (for example, by ext4_commit_super), will cause a kernel OOPS.
431 * This is a kludge to prevent these oops until we can put in a proper
432 * hook in del_gendisk() to inform the VFS and file system layers.
434 static int block_device_ejected(struct super_block *sb)
436 struct inode *bd_inode = sb->s_bdev->bd_inode;
437 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
439 return bdi->dev == NULL;
442 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
444 struct super_block *sb = journal->j_private;
445 struct ext4_sb_info *sbi = EXT4_SB(sb);
446 int error = is_journal_aborted(journal);
447 struct ext4_journal_cb_entry *jce;
449 BUG_ON(txn->t_state == T_FINISHED);
451 ext4_process_freed_data(sb, txn->t_tid);
453 spin_lock(&sbi->s_md_lock);
454 while (!list_empty(&txn->t_private_list)) {
455 jce = list_entry(txn->t_private_list.next,
456 struct ext4_journal_cb_entry, jce_list);
457 list_del_init(&jce->jce_list);
458 spin_unlock(&sbi->s_md_lock);
459 jce->jce_func(sb, jce, error);
460 spin_lock(&sbi->s_md_lock);
462 spin_unlock(&sbi->s_md_lock);
466 * This writepage callback for write_cache_pages()
467 * takes care of a few cases after page cleaning.
469 * write_cache_pages() already checks for dirty pages
470 * and calls clear_page_dirty_for_io(), which we want,
471 * to write protect the pages.
473 * However, we may have to redirty a page (see below.)
475 static int ext4_journalled_writepage_callback(struct page *page,
476 struct writeback_control *wbc,
479 transaction_t *transaction = (transaction_t *) data;
480 struct buffer_head *bh, *head;
481 struct journal_head *jh;
483 bh = head = page_buffers(page);
486 * We have to redirty a page in these cases:
487 * 1) If buffer is dirty, it means the page was dirty because it
488 * contains a buffer that needs checkpointing. So the dirty bit
489 * needs to be preserved so that checkpointing writes the buffer
491 * 2) If buffer is not part of the committing transaction
492 * (we may have just accidentally come across this buffer because
493 * inode range tracking is not exact) or if the currently running
494 * transaction already contains this buffer as well, dirty bit
495 * needs to be preserved so that the buffer gets writeprotected
496 * properly on running transaction's commit.
499 if (buffer_dirty(bh) ||
500 (jh && (jh->b_transaction != transaction ||
501 jh->b_next_transaction))) {
502 redirty_page_for_writepage(wbc, page);
505 } while ((bh = bh->b_this_page) != head);
508 return AOP_WRITEPAGE_ACTIVATE;
511 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
513 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
514 struct writeback_control wbc = {
515 .sync_mode = WB_SYNC_ALL,
516 .nr_to_write = LONG_MAX,
517 .range_start = jinode->i_dirty_start,
518 .range_end = jinode->i_dirty_end,
521 return write_cache_pages(mapping, &wbc,
522 ext4_journalled_writepage_callback,
523 jinode->i_transaction);
526 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
530 if (ext4_should_journal_data(jinode->i_vfs_inode))
531 ret = ext4_journalled_submit_inode_data_buffers(jinode);
533 ret = jbd2_journal_submit_inode_data_buffers(jinode);
538 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
542 if (!ext4_should_journal_data(jinode->i_vfs_inode))
543 ret = jbd2_journal_finish_inode_data_buffers(jinode);
548 static bool system_going_down(void)
550 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
551 || system_state == SYSTEM_RESTART;
554 struct ext4_err_translation {
559 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
561 static struct ext4_err_translation err_translation[] = {
562 EXT4_ERR_TRANSLATE(EIO),
563 EXT4_ERR_TRANSLATE(ENOMEM),
564 EXT4_ERR_TRANSLATE(EFSBADCRC),
565 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
566 EXT4_ERR_TRANSLATE(ENOSPC),
567 EXT4_ERR_TRANSLATE(ENOKEY),
568 EXT4_ERR_TRANSLATE(EROFS),
569 EXT4_ERR_TRANSLATE(EFBIG),
570 EXT4_ERR_TRANSLATE(EEXIST),
571 EXT4_ERR_TRANSLATE(ERANGE),
572 EXT4_ERR_TRANSLATE(EOVERFLOW),
573 EXT4_ERR_TRANSLATE(EBUSY),
574 EXT4_ERR_TRANSLATE(ENOTDIR),
575 EXT4_ERR_TRANSLATE(ENOTEMPTY),
576 EXT4_ERR_TRANSLATE(ESHUTDOWN),
577 EXT4_ERR_TRANSLATE(EFAULT),
580 static int ext4_errno_to_code(int errno)
584 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
585 if (err_translation[i].errno == errno)
586 return err_translation[i].code;
587 return EXT4_ERR_UNKNOWN;
590 static void save_error_info(struct super_block *sb, int error,
591 __u32 ino, __u64 block,
592 const char *func, unsigned int line)
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
596 /* We default to EFSCORRUPTED error... */
598 error = EFSCORRUPTED;
600 spin_lock(&sbi->s_error_lock);
601 sbi->s_add_error_count++;
602 sbi->s_last_error_code = error;
603 sbi->s_last_error_line = line;
604 sbi->s_last_error_ino = ino;
605 sbi->s_last_error_block = block;
606 sbi->s_last_error_func = func;
607 sbi->s_last_error_time = ktime_get_real_seconds();
608 if (!sbi->s_first_error_time) {
609 sbi->s_first_error_code = error;
610 sbi->s_first_error_line = line;
611 sbi->s_first_error_ino = ino;
612 sbi->s_first_error_block = block;
613 sbi->s_first_error_func = func;
614 sbi->s_first_error_time = sbi->s_last_error_time;
616 spin_unlock(&sbi->s_error_lock);
619 /* Deal with the reporting of failure conditions on a filesystem such as
620 * inconsistencies detected or read IO failures.
622 * On ext2, we can store the error state of the filesystem in the
623 * superblock. That is not possible on ext4, because we may have other
624 * write ordering constraints on the superblock which prevent us from
625 * writing it out straight away; and given that the journal is about to
626 * be aborted, we can't rely on the current, or future, transactions to
627 * write out the superblock safely.
629 * We'll just use the jbd2_journal_abort() error code to record an error in
630 * the journal instead. On recovery, the journal will complain about
631 * that error until we've noted it down and cleared it.
633 * If force_ro is set, we unconditionally force the filesystem into an
634 * ABORT|READONLY state, unless the error response on the fs has been set to
635 * panic in which case we take the easy way out and panic immediately. This is
636 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
637 * at a critical moment in log management.
639 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
640 __u32 ino, __u64 block,
641 const char *func, unsigned int line)
643 journal_t *journal = EXT4_SB(sb)->s_journal;
644 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
646 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
647 if (test_opt(sb, WARN_ON_ERROR))
650 if (!continue_fs && !sb_rdonly(sb)) {
651 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
653 jbd2_journal_abort(journal, -EIO);
656 if (!bdev_read_only(sb->s_bdev)) {
657 save_error_info(sb, error, ino, block, func, line);
659 * In case the fs should keep running, we need to writeout
660 * superblock through the journal. Due to lock ordering
661 * constraints, it may not be safe to do it right here so we
662 * defer superblock flushing to a workqueue.
665 schedule_work(&EXT4_SB(sb)->s_error_work);
667 ext4_commit_super(sb);
671 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
672 * could panic during 'reboot -f' as the underlying device got already
675 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
676 panic("EXT4-fs (device %s): panic forced after error\n",
680 if (sb_rdonly(sb) || continue_fs)
683 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
685 * Make sure updated value of ->s_mount_flags will be visible before
689 sb->s_flags |= SB_RDONLY;
692 static void flush_stashed_error_work(struct work_struct *work)
694 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
696 journal_t *journal = sbi->s_journal;
700 * If the journal is still running, we have to write out superblock
701 * through the journal to avoid collisions of other journalled sb
704 * We use directly jbd2 functions here to avoid recursing back into
705 * ext4 error handling code during handling of previous errors.
707 if (!sb_rdonly(sbi->s_sb) && journal) {
708 handle = jbd2_journal_start(journal, 1);
711 if (jbd2_journal_get_write_access(handle, sbi->s_sbh)) {
712 jbd2_journal_stop(handle);
715 ext4_update_super(sbi->s_sb);
716 if (jbd2_journal_dirty_metadata(handle, sbi->s_sbh)) {
717 jbd2_journal_stop(handle);
720 jbd2_journal_stop(handle);
721 ext4_notify_error_sysfs(sbi);
726 * Write through journal failed. Write sb directly to get error info
727 * out and hope for the best.
729 ext4_commit_super(sbi->s_sb);
730 ext4_notify_error_sysfs(sbi);
733 #define ext4_error_ratelimit(sb) \
734 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
737 void __ext4_error(struct super_block *sb, const char *function,
738 unsigned int line, bool force_ro, int error, __u64 block,
739 const char *fmt, ...)
741 struct va_format vaf;
744 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
747 trace_ext4_error(sb, function, line);
748 if (ext4_error_ratelimit(sb)) {
753 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
754 sb->s_id, function, line, current->comm, &vaf);
757 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
760 void __ext4_error_inode(struct inode *inode, const char *function,
761 unsigned int line, ext4_fsblk_t block, int error,
762 const char *fmt, ...)
765 struct va_format vaf;
767 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
770 trace_ext4_error(inode->i_sb, function, line);
771 if (ext4_error_ratelimit(inode->i_sb)) {
776 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
777 "inode #%lu: block %llu: comm %s: %pV\n",
778 inode->i_sb->s_id, function, line, inode->i_ino,
779 block, current->comm, &vaf);
781 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
782 "inode #%lu: comm %s: %pV\n",
783 inode->i_sb->s_id, function, line, inode->i_ino,
784 current->comm, &vaf);
787 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
791 void __ext4_error_file(struct file *file, const char *function,
792 unsigned int line, ext4_fsblk_t block,
793 const char *fmt, ...)
796 struct va_format vaf;
797 struct inode *inode = file_inode(file);
798 char pathname[80], *path;
800 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
803 trace_ext4_error(inode->i_sb, function, line);
804 if (ext4_error_ratelimit(inode->i_sb)) {
805 path = file_path(file, pathname, sizeof(pathname));
813 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
814 "block %llu: comm %s: path %s: %pV\n",
815 inode->i_sb->s_id, function, line, inode->i_ino,
816 block, current->comm, path, &vaf);
819 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
820 "comm %s: path %s: %pV\n",
821 inode->i_sb->s_id, function, line, inode->i_ino,
822 current->comm, path, &vaf);
825 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
829 const char *ext4_decode_error(struct super_block *sb, int errno,
836 errstr = "Corrupt filesystem";
839 errstr = "Filesystem failed CRC";
842 errstr = "IO failure";
845 errstr = "Out of memory";
848 if (!sb || (EXT4_SB(sb)->s_journal &&
849 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
850 errstr = "Journal has aborted";
852 errstr = "Readonly filesystem";
855 /* If the caller passed in an extra buffer for unknown
856 * errors, textualise them now. Else we just return
859 /* Check for truncated error codes... */
860 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
869 /* __ext4_std_error decodes expected errors from journaling functions
870 * automatically and invokes the appropriate error response. */
872 void __ext4_std_error(struct super_block *sb, const char *function,
873 unsigned int line, int errno)
878 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
881 /* Special case: if the error is EROFS, and we're not already
882 * inside a transaction, then there's really no point in logging
884 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
887 if (ext4_error_ratelimit(sb)) {
888 errstr = ext4_decode_error(sb, errno, nbuf);
889 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
890 sb->s_id, function, line, errstr);
893 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
896 void __ext4_msg(struct super_block *sb,
897 const char *prefix, const char *fmt, ...)
899 struct va_format vaf;
902 atomic_inc(&EXT4_SB(sb)->s_msg_count);
903 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
909 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
913 static int ext4_warning_ratelimit(struct super_block *sb)
915 atomic_inc(&EXT4_SB(sb)->s_warning_count);
916 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
920 void __ext4_warning(struct super_block *sb, const char *function,
921 unsigned int line, const char *fmt, ...)
923 struct va_format vaf;
926 if (!ext4_warning_ratelimit(sb))
932 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
933 sb->s_id, function, line, &vaf);
937 void __ext4_warning_inode(const struct inode *inode, const char *function,
938 unsigned int line, const char *fmt, ...)
940 struct va_format vaf;
943 if (!ext4_warning_ratelimit(inode->i_sb))
949 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
950 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
951 function, line, inode->i_ino, current->comm, &vaf);
955 void __ext4_grp_locked_error(const char *function, unsigned int line,
956 struct super_block *sb, ext4_group_t grp,
957 unsigned long ino, ext4_fsblk_t block,
958 const char *fmt, ...)
962 struct va_format vaf;
965 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
968 trace_ext4_error(sb, function, line);
969 if (ext4_error_ratelimit(sb)) {
973 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
974 sb->s_id, function, line, grp);
976 printk(KERN_CONT "inode %lu: ", ino);
978 printk(KERN_CONT "block %llu:",
979 (unsigned long long) block);
980 printk(KERN_CONT "%pV\n", &vaf);
984 if (test_opt(sb, ERRORS_CONT)) {
985 if (test_opt(sb, WARN_ON_ERROR))
987 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
988 if (!bdev_read_only(sb->s_bdev)) {
989 save_error_info(sb, EFSCORRUPTED, ino, block, function,
991 schedule_work(&EXT4_SB(sb)->s_error_work);
995 ext4_unlock_group(sb, grp);
996 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
998 * We only get here in the ERRORS_RO case; relocking the group
999 * may be dangerous, but nothing bad will happen since the
1000 * filesystem will have already been marked read/only and the
1001 * journal has been aborted. We return 1 as a hint to callers
1002 * who might what to use the return value from
1003 * ext4_grp_locked_error() to distinguish between the
1004 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1005 * aggressively from the ext4 function in question, with a
1006 * more appropriate error code.
1008 ext4_lock_group(sb, grp);
1012 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1016 struct ext4_sb_info *sbi = EXT4_SB(sb);
1017 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1018 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1021 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1022 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1025 percpu_counter_sub(&sbi->s_freeclusters_counter,
1029 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1030 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1035 count = ext4_free_inodes_count(sb, gdp);
1036 percpu_counter_sub(&sbi->s_freeinodes_counter,
1042 void ext4_update_dynamic_rev(struct super_block *sb)
1044 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1046 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1050 "updating to rev %d because of new feature flag, "
1051 "running e2fsck is recommended",
1054 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1055 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1056 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1057 /* leave es->s_feature_*compat flags alone */
1058 /* es->s_uuid will be set by e2fsck if empty */
1061 * The rest of the superblock fields should be zero, and if not it
1062 * means they are likely already in use, so leave them alone. We
1063 * can leave it up to e2fsck to clean up any inconsistencies there.
1068 * Open the external journal device
1070 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1072 struct block_device *bdev;
1074 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1080 ext4_msg(sb, KERN_ERR,
1081 "failed to open journal device unknown-block(%u,%u) %ld",
1082 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1087 * Release the journal device
1089 static void ext4_blkdev_put(struct block_device *bdev)
1091 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1094 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1096 struct block_device *bdev;
1097 bdev = sbi->s_journal_bdev;
1099 ext4_blkdev_put(bdev);
1100 sbi->s_journal_bdev = NULL;
1104 static inline struct inode *orphan_list_entry(struct list_head *l)
1106 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1109 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1111 struct list_head *l;
1113 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1114 le32_to_cpu(sbi->s_es->s_last_orphan));
1116 printk(KERN_ERR "sb_info orphan list:\n");
1117 list_for_each(l, &sbi->s_orphan) {
1118 struct inode *inode = orphan_list_entry(l);
1120 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1121 inode->i_sb->s_id, inode->i_ino, inode,
1122 inode->i_mode, inode->i_nlink,
1123 NEXT_ORPHAN(inode));
1128 static int ext4_quota_off(struct super_block *sb, int type);
1130 static inline void ext4_quota_off_umount(struct super_block *sb)
1134 /* Use our quota_off function to clear inode flags etc. */
1135 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1136 ext4_quota_off(sb, type);
1140 * This is a helper function which is used in the mount/remount
1141 * codepaths (which holds s_umount) to fetch the quota file name.
1143 static inline char *get_qf_name(struct super_block *sb,
1144 struct ext4_sb_info *sbi,
1147 return rcu_dereference_protected(sbi->s_qf_names[type],
1148 lockdep_is_held(&sb->s_umount));
1151 static inline void ext4_quota_off_umount(struct super_block *sb)
1156 static void ext4_put_super(struct super_block *sb)
1158 struct ext4_sb_info *sbi = EXT4_SB(sb);
1159 struct ext4_super_block *es = sbi->s_es;
1160 struct buffer_head **group_desc;
1161 struct flex_groups **flex_groups;
1165 ext4_unregister_li_request(sb);
1166 ext4_quota_off_umount(sb);
1168 flush_work(&sbi->s_error_work);
1169 destroy_workqueue(sbi->rsv_conversion_wq);
1172 * Unregister sysfs before destroying jbd2 journal.
1173 * Since we could still access attr_journal_task attribute via sysfs
1174 * path which could have sbi->s_journal->j_task as NULL
1176 ext4_unregister_sysfs(sb);
1178 if (sbi->s_journal) {
1179 jbd2_journal_unregister_shrinker(sbi->s_journal);
1180 aborted = is_journal_aborted(sbi->s_journal);
1181 err = jbd2_journal_destroy(sbi->s_journal);
1182 sbi->s_journal = NULL;
1183 if ((err < 0) && !aborted) {
1184 ext4_abort(sb, -err, "Couldn't clean up the journal");
1188 ext4_es_unregister_shrinker(sbi);
1189 del_timer_sync(&sbi->s_err_report);
1190 ext4_release_system_zone(sb);
1191 ext4_mb_release(sb);
1192 ext4_ext_release(sb);
1194 if (!sb_rdonly(sb) && !aborted) {
1195 ext4_clear_feature_journal_needs_recovery(sb);
1196 es->s_state = cpu_to_le16(sbi->s_mount_state);
1199 ext4_commit_super(sb);
1202 group_desc = rcu_dereference(sbi->s_group_desc);
1203 for (i = 0; i < sbi->s_gdb_count; i++)
1204 brelse(group_desc[i]);
1206 flex_groups = rcu_dereference(sbi->s_flex_groups);
1208 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1209 kvfree(flex_groups[i]);
1210 kvfree(flex_groups);
1213 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1214 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1215 percpu_counter_destroy(&sbi->s_dirs_counter);
1216 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1217 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1218 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1220 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1221 kfree(get_qf_name(sb, sbi, i));
1224 /* Debugging code just in case the in-memory inode orphan list
1225 * isn't empty. The on-disk one can be non-empty if we've
1226 * detected an error and taken the fs readonly, but the
1227 * in-memory list had better be clean by this point. */
1228 if (!list_empty(&sbi->s_orphan))
1229 dump_orphan_list(sb, sbi);
1230 ASSERT(list_empty(&sbi->s_orphan));
1232 sync_blockdev(sb->s_bdev);
1233 invalidate_bdev(sb->s_bdev);
1234 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1236 * Invalidate the journal device's buffers. We don't want them
1237 * floating about in memory - the physical journal device may
1238 * hotswapped, and it breaks the `ro-after' testing code.
1240 sync_blockdev(sbi->s_journal_bdev);
1241 invalidate_bdev(sbi->s_journal_bdev);
1242 ext4_blkdev_remove(sbi);
1245 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1246 sbi->s_ea_inode_cache = NULL;
1248 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1249 sbi->s_ea_block_cache = NULL;
1251 ext4_stop_mmpd(sbi);
1254 sb->s_fs_info = NULL;
1256 * Now that we are completely done shutting down the
1257 * superblock, we need to actually destroy the kobject.
1259 kobject_put(&sbi->s_kobj);
1260 wait_for_completion(&sbi->s_kobj_unregister);
1261 if (sbi->s_chksum_driver)
1262 crypto_free_shash(sbi->s_chksum_driver);
1263 kfree(sbi->s_blockgroup_lock);
1264 fs_put_dax(sbi->s_daxdev);
1265 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1266 #ifdef CONFIG_UNICODE
1267 utf8_unload(sb->s_encoding);
1272 static struct kmem_cache *ext4_inode_cachep;
1275 * Called inside transaction, so use GFP_NOFS
1277 static struct inode *ext4_alloc_inode(struct super_block *sb)
1279 struct ext4_inode_info *ei;
1281 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1285 inode_set_iversion(&ei->vfs_inode, 1);
1286 spin_lock_init(&ei->i_raw_lock);
1287 INIT_LIST_HEAD(&ei->i_prealloc_list);
1288 atomic_set(&ei->i_prealloc_active, 0);
1289 spin_lock_init(&ei->i_prealloc_lock);
1290 ext4_es_init_tree(&ei->i_es_tree);
1291 rwlock_init(&ei->i_es_lock);
1292 INIT_LIST_HEAD(&ei->i_es_list);
1293 ei->i_es_all_nr = 0;
1294 ei->i_es_shk_nr = 0;
1295 ei->i_es_shrink_lblk = 0;
1296 ei->i_reserved_data_blocks = 0;
1297 spin_lock_init(&(ei->i_block_reservation_lock));
1298 ext4_init_pending_tree(&ei->i_pending_tree);
1300 ei->i_reserved_quota = 0;
1301 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1304 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1305 spin_lock_init(&ei->i_completed_io_lock);
1307 ei->i_datasync_tid = 0;
1308 atomic_set(&ei->i_unwritten, 0);
1309 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1310 ext4_fc_init_inode(&ei->vfs_inode);
1311 mutex_init(&ei->i_fc_lock);
1312 return &ei->vfs_inode;
1315 static int ext4_drop_inode(struct inode *inode)
1317 int drop = generic_drop_inode(inode);
1320 drop = fscrypt_drop_inode(inode);
1322 trace_ext4_drop_inode(inode, drop);
1326 static void ext4_free_in_core_inode(struct inode *inode)
1328 fscrypt_free_inode(inode);
1329 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1330 pr_warn("%s: inode %ld still in fc list",
1331 __func__, inode->i_ino);
1333 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1336 static void ext4_destroy_inode(struct inode *inode)
1338 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1339 ext4_msg(inode->i_sb, KERN_ERR,
1340 "Inode %lu (%p): orphan list check failed!",
1341 inode->i_ino, EXT4_I(inode));
1342 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1343 EXT4_I(inode), sizeof(struct ext4_inode_info),
1349 static void init_once(void *foo)
1351 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1353 INIT_LIST_HEAD(&ei->i_orphan);
1354 init_rwsem(&ei->xattr_sem);
1355 init_rwsem(&ei->i_data_sem);
1356 init_rwsem(&ei->i_mmap_sem);
1357 inode_init_once(&ei->vfs_inode);
1358 ext4_fc_init_inode(&ei->vfs_inode);
1361 static int __init init_inodecache(void)
1363 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1364 sizeof(struct ext4_inode_info), 0,
1365 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1367 offsetof(struct ext4_inode_info, i_data),
1368 sizeof_field(struct ext4_inode_info, i_data),
1370 if (ext4_inode_cachep == NULL)
1375 static void destroy_inodecache(void)
1378 * Make sure all delayed rcu free inodes are flushed before we
1382 kmem_cache_destroy(ext4_inode_cachep);
1385 void ext4_clear_inode(struct inode *inode)
1388 invalidate_inode_buffers(inode);
1390 ext4_discard_preallocations(inode, 0);
1391 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1393 if (EXT4_I(inode)->jinode) {
1394 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1395 EXT4_I(inode)->jinode);
1396 jbd2_free_inode(EXT4_I(inode)->jinode);
1397 EXT4_I(inode)->jinode = NULL;
1399 fscrypt_put_encryption_info(inode);
1400 fsverity_cleanup_inode(inode);
1403 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1404 u64 ino, u32 generation)
1406 struct inode *inode;
1409 * Currently we don't know the generation for parent directory, so
1410 * a generation of 0 means "accept any"
1412 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1414 return ERR_CAST(inode);
1415 if (generation && inode->i_generation != generation) {
1417 return ERR_PTR(-ESTALE);
1423 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1424 int fh_len, int fh_type)
1426 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1427 ext4_nfs_get_inode);
1430 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1431 int fh_len, int fh_type)
1433 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1434 ext4_nfs_get_inode);
1437 static int ext4_nfs_commit_metadata(struct inode *inode)
1439 struct writeback_control wbc = {
1440 .sync_mode = WB_SYNC_ALL
1443 trace_ext4_nfs_commit_metadata(inode);
1444 return ext4_write_inode(inode, &wbc);
1447 #ifdef CONFIG_FS_ENCRYPTION
1448 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1450 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1451 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1454 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1457 handle_t *handle = fs_data;
1458 int res, res2, credits, retries = 0;
1461 * Encrypting the root directory is not allowed because e2fsck expects
1462 * lost+found to exist and be unencrypted, and encrypting the root
1463 * directory would imply encrypting the lost+found directory as well as
1464 * the filename "lost+found" itself.
1466 if (inode->i_ino == EXT4_ROOT_INO)
1469 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1472 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1475 res = ext4_convert_inline_data(inode);
1480 * If a journal handle was specified, then the encryption context is
1481 * being set on a new inode via inheritance and is part of a larger
1482 * transaction to create the inode. Otherwise the encryption context is
1483 * being set on an existing inode in its own transaction. Only in the
1484 * latter case should the "retry on ENOSPC" logic be used.
1488 res = ext4_xattr_set_handle(handle, inode,
1489 EXT4_XATTR_INDEX_ENCRYPTION,
1490 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1493 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1494 ext4_clear_inode_state(inode,
1495 EXT4_STATE_MAY_INLINE_DATA);
1497 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1498 * S_DAX may be disabled
1500 ext4_set_inode_flags(inode, false);
1505 res = dquot_initialize(inode);
1509 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1514 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1516 return PTR_ERR(handle);
1518 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1519 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1522 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1524 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1525 * S_DAX may be disabled
1527 ext4_set_inode_flags(inode, false);
1528 res = ext4_mark_inode_dirty(handle, inode);
1530 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1532 res2 = ext4_journal_stop(handle);
1534 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1541 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1543 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1546 static bool ext4_has_stable_inodes(struct super_block *sb)
1548 return ext4_has_feature_stable_inodes(sb);
1551 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1552 int *ino_bits_ret, int *lblk_bits_ret)
1554 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1555 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1558 static const struct fscrypt_operations ext4_cryptops = {
1559 .key_prefix = "ext4:",
1560 .get_context = ext4_get_context,
1561 .set_context = ext4_set_context,
1562 .get_dummy_policy = ext4_get_dummy_policy,
1563 .empty_dir = ext4_empty_dir,
1564 .max_namelen = EXT4_NAME_LEN,
1565 .has_stable_inodes = ext4_has_stable_inodes,
1566 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1571 static const char * const quotatypes[] = INITQFNAMES;
1572 #define QTYPE2NAME(t) (quotatypes[t])
1574 static int ext4_write_dquot(struct dquot *dquot);
1575 static int ext4_acquire_dquot(struct dquot *dquot);
1576 static int ext4_release_dquot(struct dquot *dquot);
1577 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1578 static int ext4_write_info(struct super_block *sb, int type);
1579 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1580 const struct path *path);
1581 static int ext4_quota_on_mount(struct super_block *sb, int type);
1582 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1583 size_t len, loff_t off);
1584 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1585 const char *data, size_t len, loff_t off);
1586 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1587 unsigned int flags);
1588 static int ext4_enable_quotas(struct super_block *sb);
1590 static struct dquot **ext4_get_dquots(struct inode *inode)
1592 return EXT4_I(inode)->i_dquot;
1595 static const struct dquot_operations ext4_quota_operations = {
1596 .get_reserved_space = ext4_get_reserved_space,
1597 .write_dquot = ext4_write_dquot,
1598 .acquire_dquot = ext4_acquire_dquot,
1599 .release_dquot = ext4_release_dquot,
1600 .mark_dirty = ext4_mark_dquot_dirty,
1601 .write_info = ext4_write_info,
1602 .alloc_dquot = dquot_alloc,
1603 .destroy_dquot = dquot_destroy,
1604 .get_projid = ext4_get_projid,
1605 .get_inode_usage = ext4_get_inode_usage,
1606 .get_next_id = dquot_get_next_id,
1609 static const struct quotactl_ops ext4_qctl_operations = {
1610 .quota_on = ext4_quota_on,
1611 .quota_off = ext4_quota_off,
1612 .quota_sync = dquot_quota_sync,
1613 .get_state = dquot_get_state,
1614 .set_info = dquot_set_dqinfo,
1615 .get_dqblk = dquot_get_dqblk,
1616 .set_dqblk = dquot_set_dqblk,
1617 .get_nextdqblk = dquot_get_next_dqblk,
1621 static const struct super_operations ext4_sops = {
1622 .alloc_inode = ext4_alloc_inode,
1623 .free_inode = ext4_free_in_core_inode,
1624 .destroy_inode = ext4_destroy_inode,
1625 .write_inode = ext4_write_inode,
1626 .dirty_inode = ext4_dirty_inode,
1627 .drop_inode = ext4_drop_inode,
1628 .evict_inode = ext4_evict_inode,
1629 .put_super = ext4_put_super,
1630 .sync_fs = ext4_sync_fs,
1631 .freeze_fs = ext4_freeze,
1632 .unfreeze_fs = ext4_unfreeze,
1633 .statfs = ext4_statfs,
1634 .remount_fs = ext4_remount,
1635 .show_options = ext4_show_options,
1637 .quota_read = ext4_quota_read,
1638 .quota_write = ext4_quota_write,
1639 .get_dquots = ext4_get_dquots,
1643 static const struct export_operations ext4_export_ops = {
1644 .fh_to_dentry = ext4_fh_to_dentry,
1645 .fh_to_parent = ext4_fh_to_parent,
1646 .get_parent = ext4_get_parent,
1647 .commit_metadata = ext4_nfs_commit_metadata,
1651 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1652 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1653 Opt_nouid32, Opt_debug, Opt_removed,
1654 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1655 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1656 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1657 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1658 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1659 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1661 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1662 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1663 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1664 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1665 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1666 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1667 Opt_nowarn_on_error, Opt_mblk_io_submit,
1668 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1669 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1670 Opt_inode_readahead_blks, Opt_journal_ioprio,
1671 Opt_dioread_nolock, Opt_dioread_lock,
1672 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1673 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1674 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1675 #ifdef CONFIG_EXT4_DEBUG
1676 Opt_fc_debug_max_replay, Opt_fc_debug_force
1680 static const match_table_t tokens = {
1681 {Opt_bsd_df, "bsddf"},
1682 {Opt_minix_df, "minixdf"},
1683 {Opt_grpid, "grpid"},
1684 {Opt_grpid, "bsdgroups"},
1685 {Opt_nogrpid, "nogrpid"},
1686 {Opt_nogrpid, "sysvgroups"},
1687 {Opt_resgid, "resgid=%u"},
1688 {Opt_resuid, "resuid=%u"},
1690 {Opt_err_cont, "errors=continue"},
1691 {Opt_err_panic, "errors=panic"},
1692 {Opt_err_ro, "errors=remount-ro"},
1693 {Opt_nouid32, "nouid32"},
1694 {Opt_debug, "debug"},
1695 {Opt_removed, "oldalloc"},
1696 {Opt_removed, "orlov"},
1697 {Opt_user_xattr, "user_xattr"},
1698 {Opt_nouser_xattr, "nouser_xattr"},
1700 {Opt_noacl, "noacl"},
1701 {Opt_noload, "norecovery"},
1702 {Opt_noload, "noload"},
1703 {Opt_removed, "nobh"},
1704 {Opt_removed, "bh"},
1705 {Opt_commit, "commit=%u"},
1706 {Opt_min_batch_time, "min_batch_time=%u"},
1707 {Opt_max_batch_time, "max_batch_time=%u"},
1708 {Opt_journal_dev, "journal_dev=%u"},
1709 {Opt_journal_path, "journal_path=%s"},
1710 {Opt_journal_checksum, "journal_checksum"},
1711 {Opt_nojournal_checksum, "nojournal_checksum"},
1712 {Opt_journal_async_commit, "journal_async_commit"},
1713 {Opt_abort, "abort"},
1714 {Opt_data_journal, "data=journal"},
1715 {Opt_data_ordered, "data=ordered"},
1716 {Opt_data_writeback, "data=writeback"},
1717 {Opt_data_err_abort, "data_err=abort"},
1718 {Opt_data_err_ignore, "data_err=ignore"},
1719 {Opt_offusrjquota, "usrjquota="},
1720 {Opt_usrjquota, "usrjquota=%s"},
1721 {Opt_offgrpjquota, "grpjquota="},
1722 {Opt_grpjquota, "grpjquota=%s"},
1723 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1724 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1725 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1726 {Opt_grpquota, "grpquota"},
1727 {Opt_noquota, "noquota"},
1728 {Opt_quota, "quota"},
1729 {Opt_usrquota, "usrquota"},
1730 {Opt_prjquota, "prjquota"},
1731 {Opt_barrier, "barrier=%u"},
1732 {Opt_barrier, "barrier"},
1733 {Opt_nobarrier, "nobarrier"},
1734 {Opt_i_version, "i_version"},
1736 {Opt_dax_always, "dax=always"},
1737 {Opt_dax_inode, "dax=inode"},
1738 {Opt_dax_never, "dax=never"},
1739 {Opt_stripe, "stripe=%u"},
1740 {Opt_delalloc, "delalloc"},
1741 {Opt_warn_on_error, "warn_on_error"},
1742 {Opt_nowarn_on_error, "nowarn_on_error"},
1743 {Opt_lazytime, "lazytime"},
1744 {Opt_nolazytime, "nolazytime"},
1745 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1746 {Opt_nodelalloc, "nodelalloc"},
1747 {Opt_removed, "mblk_io_submit"},
1748 {Opt_removed, "nomblk_io_submit"},
1749 {Opt_block_validity, "block_validity"},
1750 {Opt_noblock_validity, "noblock_validity"},
1751 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1752 {Opt_journal_ioprio, "journal_ioprio=%u"},
1753 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1754 {Opt_auto_da_alloc, "auto_da_alloc"},
1755 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1756 {Opt_dioread_nolock, "dioread_nolock"},
1757 {Opt_dioread_lock, "nodioread_nolock"},
1758 {Opt_dioread_lock, "dioread_lock"},
1759 {Opt_discard, "discard"},
1760 {Opt_nodiscard, "nodiscard"},
1761 {Opt_init_itable, "init_itable=%u"},
1762 {Opt_init_itable, "init_itable"},
1763 {Opt_noinit_itable, "noinit_itable"},
1764 #ifdef CONFIG_EXT4_DEBUG
1765 {Opt_fc_debug_force, "fc_debug_force"},
1766 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1768 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1769 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1770 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1771 {Opt_inlinecrypt, "inlinecrypt"},
1772 {Opt_nombcache, "nombcache"},
1773 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1774 {Opt_removed, "prefetch_block_bitmaps"},
1775 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1776 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1777 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1778 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1779 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1780 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1781 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1785 static ext4_fsblk_t get_sb_block(void **data)
1787 ext4_fsblk_t sb_block;
1788 char *options = (char *) *data;
1790 if (!options || strncmp(options, "sb=", 3) != 0)
1791 return 1; /* Default location */
1794 /* TODO: use simple_strtoll with >32bit ext4 */
1795 sb_block = simple_strtoul(options, &options, 0);
1796 if (*options && *options != ',') {
1797 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1801 if (*options == ',')
1803 *data = (void *) options;
1808 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1809 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1811 static const char deprecated_msg[] =
1812 "Mount option \"%s\" will be removed by %s\n"
1813 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1816 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1818 struct ext4_sb_info *sbi = EXT4_SB(sb);
1819 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1822 if (sb_any_quota_loaded(sb) && !old_qname) {
1823 ext4_msg(sb, KERN_ERR,
1824 "Cannot change journaled "
1825 "quota options when quota turned on");
1828 if (ext4_has_feature_quota(sb)) {
1829 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1830 "ignored when QUOTA feature is enabled");
1833 qname = match_strdup(args);
1835 ext4_msg(sb, KERN_ERR,
1836 "Not enough memory for storing quotafile name");
1840 if (strcmp(old_qname, qname) == 0)
1843 ext4_msg(sb, KERN_ERR,
1844 "%s quota file already specified",
1848 if (strchr(qname, '/')) {
1849 ext4_msg(sb, KERN_ERR,
1850 "quotafile must be on filesystem root");
1853 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1861 static int clear_qf_name(struct super_block *sb, int qtype)
1864 struct ext4_sb_info *sbi = EXT4_SB(sb);
1865 char *old_qname = get_qf_name(sb, sbi, qtype);
1867 if (sb_any_quota_loaded(sb) && old_qname) {
1868 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1869 " when quota turned on");
1872 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1879 #define MOPT_SET 0x0001
1880 #define MOPT_CLEAR 0x0002
1881 #define MOPT_NOSUPPORT 0x0004
1882 #define MOPT_EXPLICIT 0x0008
1883 #define MOPT_CLEAR_ERR 0x0010
1884 #define MOPT_GTE0 0x0020
1887 #define MOPT_QFMT 0x0040
1889 #define MOPT_Q MOPT_NOSUPPORT
1890 #define MOPT_QFMT MOPT_NOSUPPORT
1892 #define MOPT_DATAJ 0x0080
1893 #define MOPT_NO_EXT2 0x0100
1894 #define MOPT_NO_EXT3 0x0200
1895 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1896 #define MOPT_STRING 0x0400
1897 #define MOPT_SKIP 0x0800
1898 #define MOPT_2 0x1000
1900 static const struct mount_opts {
1904 } ext4_mount_opts[] = {
1905 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1906 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1907 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1908 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1909 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1910 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1911 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1912 MOPT_EXT4_ONLY | MOPT_SET},
1913 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1914 MOPT_EXT4_ONLY | MOPT_CLEAR},
1915 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1916 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1917 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1918 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1919 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1920 MOPT_EXT4_ONLY | MOPT_CLEAR},
1921 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1922 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1923 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1924 MOPT_EXT4_ONLY | MOPT_CLEAR},
1925 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1926 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1927 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1928 EXT4_MOUNT_JOURNAL_CHECKSUM),
1929 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1930 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1931 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1932 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1933 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1934 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1936 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1938 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1939 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1940 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1941 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1942 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1943 {Opt_commit, 0, MOPT_GTE0},
1944 {Opt_max_batch_time, 0, MOPT_GTE0},
1945 {Opt_min_batch_time, 0, MOPT_GTE0},
1946 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1947 {Opt_init_itable, 0, MOPT_GTE0},
1948 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1949 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1950 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1951 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1952 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1953 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1954 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1955 {Opt_stripe, 0, MOPT_GTE0},
1956 {Opt_resuid, 0, MOPT_GTE0},
1957 {Opt_resgid, 0, MOPT_GTE0},
1958 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1959 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1960 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1961 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1962 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1963 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1964 MOPT_NO_EXT2 | MOPT_DATAJ},
1965 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1966 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1967 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1968 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1969 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1971 {Opt_acl, 0, MOPT_NOSUPPORT},
1972 {Opt_noacl, 0, MOPT_NOSUPPORT},
1974 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1975 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1976 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1977 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1978 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1980 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1982 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1984 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1985 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1986 MOPT_CLEAR | MOPT_Q},
1987 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1988 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1989 {Opt_offusrjquota, 0, MOPT_Q},
1990 {Opt_offgrpjquota, 0, MOPT_Q},
1991 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1992 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1993 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1994 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1995 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1996 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1997 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1999 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2000 #ifdef CONFIG_EXT4_DEBUG
2001 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2002 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2003 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2008 #ifdef CONFIG_UNICODE
2009 static const struct ext4_sb_encodings {
2013 } ext4_sb_encoding_map[] = {
2014 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2017 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2018 const struct ext4_sb_encodings **encoding,
2021 __u16 magic = le16_to_cpu(es->s_encoding);
2024 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2025 if (magic == ext4_sb_encoding_map[i].magic)
2028 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2031 *encoding = &ext4_sb_encoding_map[i];
2032 *flags = le16_to_cpu(es->s_encoding_flags);
2038 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2040 const substring_t *arg,
2043 #ifdef CONFIG_FS_ENCRYPTION
2044 struct ext4_sb_info *sbi = EXT4_SB(sb);
2048 * This mount option is just for testing, and it's not worthwhile to
2049 * implement the extra complexity (e.g. RCU protection) that would be
2050 * needed to allow it to be set or changed during remount. We do allow
2051 * it to be specified during remount, but only if there is no change.
2053 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2054 ext4_msg(sb, KERN_WARNING,
2055 "Can't set test_dummy_encryption on remount");
2058 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2059 &sbi->s_dummy_enc_policy);
2062 ext4_msg(sb, KERN_WARNING,
2063 "Can't change test_dummy_encryption on remount");
2064 else if (err == -EINVAL)
2065 ext4_msg(sb, KERN_WARNING,
2066 "Value of option \"%s\" is unrecognized", opt);
2068 ext4_msg(sb, KERN_WARNING,
2069 "Error processing option \"%s\" [%d]",
2073 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2075 ext4_msg(sb, KERN_WARNING,
2076 "Test dummy encryption mount option ignored");
2081 struct ext4_parsed_options {
2082 unsigned long journal_devnum;
2083 unsigned int journal_ioprio;
2084 int mb_optimize_scan;
2087 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2088 substring_t *args, struct ext4_parsed_options *parsed_opts,
2091 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 const struct mount_opts *m;
2098 if (token == Opt_usrjquota)
2099 return set_qf_name(sb, USRQUOTA, &args[0]);
2100 else if (token == Opt_grpjquota)
2101 return set_qf_name(sb, GRPQUOTA, &args[0]);
2102 else if (token == Opt_offusrjquota)
2103 return clear_qf_name(sb, USRQUOTA);
2104 else if (token == Opt_offgrpjquota)
2105 return clear_qf_name(sb, GRPQUOTA);
2109 case Opt_nouser_xattr:
2110 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2113 return 1; /* handled by get_sb_block() */
2115 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2118 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2121 sb->s_flags |= SB_I_VERSION;
2124 sb->s_flags |= SB_LAZYTIME;
2126 case Opt_nolazytime:
2127 sb->s_flags &= ~SB_LAZYTIME;
2129 case Opt_inlinecrypt:
2130 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2131 sb->s_flags |= SB_INLINECRYPT;
2133 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2138 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2139 if (token == m->token)
2142 if (m->token == Opt_err) {
2143 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2144 "or missing value", opt);
2148 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2149 ext4_msg(sb, KERN_ERR,
2150 "Mount option \"%s\" incompatible with ext2", opt);
2153 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2154 ext4_msg(sb, KERN_ERR,
2155 "Mount option \"%s\" incompatible with ext3", opt);
2159 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2161 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2163 if (m->flags & MOPT_EXPLICIT) {
2164 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2165 set_opt2(sb, EXPLICIT_DELALLOC);
2166 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2167 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2171 if (m->flags & MOPT_CLEAR_ERR)
2172 clear_opt(sb, ERRORS_MASK);
2173 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2174 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2175 "options when quota turned on");
2179 if (m->flags & MOPT_NOSUPPORT) {
2180 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2181 } else if (token == Opt_commit) {
2183 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2184 else if (arg > INT_MAX / HZ) {
2185 ext4_msg(sb, KERN_ERR,
2186 "Invalid commit interval %d, "
2187 "must be smaller than %d",
2191 sbi->s_commit_interval = HZ * arg;
2192 } else if (token == Opt_debug_want_extra_isize) {
2195 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2196 ext4_msg(sb, KERN_ERR,
2197 "Invalid want_extra_isize %d", arg);
2200 sbi->s_want_extra_isize = arg;
2201 } else if (token == Opt_max_batch_time) {
2202 sbi->s_max_batch_time = arg;
2203 } else if (token == Opt_min_batch_time) {
2204 sbi->s_min_batch_time = arg;
2205 } else if (token == Opt_inode_readahead_blks) {
2206 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2207 ext4_msg(sb, KERN_ERR,
2208 "EXT4-fs: inode_readahead_blks must be "
2209 "0 or a power of 2 smaller than 2^31");
2212 sbi->s_inode_readahead_blks = arg;
2213 } else if (token == Opt_init_itable) {
2214 set_opt(sb, INIT_INODE_TABLE);
2216 arg = EXT4_DEF_LI_WAIT_MULT;
2217 sbi->s_li_wait_mult = arg;
2218 } else if (token == Opt_max_dir_size_kb) {
2219 sbi->s_max_dir_size_kb = arg;
2220 #ifdef CONFIG_EXT4_DEBUG
2221 } else if (token == Opt_fc_debug_max_replay) {
2222 sbi->s_fc_debug_max_replay = arg;
2224 } else if (token == Opt_stripe) {
2225 sbi->s_stripe = arg;
2226 } else if (token == Opt_resuid) {
2227 uid = make_kuid(current_user_ns(), arg);
2228 if (!uid_valid(uid)) {
2229 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2232 sbi->s_resuid = uid;
2233 } else if (token == Opt_resgid) {
2234 gid = make_kgid(current_user_ns(), arg);
2235 if (!gid_valid(gid)) {
2236 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2239 sbi->s_resgid = gid;
2240 } else if (token == Opt_journal_dev) {
2242 ext4_msg(sb, KERN_ERR,
2243 "Cannot specify journal on remount");
2246 parsed_opts->journal_devnum = arg;
2247 } else if (token == Opt_journal_path) {
2249 struct inode *journal_inode;
2254 ext4_msg(sb, KERN_ERR,
2255 "Cannot specify journal on remount");
2258 journal_path = match_strdup(&args[0]);
2259 if (!journal_path) {
2260 ext4_msg(sb, KERN_ERR, "error: could not dup "
2261 "journal device string");
2265 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2267 ext4_msg(sb, KERN_ERR, "error: could not find "
2268 "journal device path: error %d", error);
2269 kfree(journal_path);
2273 journal_inode = d_inode(path.dentry);
2274 if (!S_ISBLK(journal_inode->i_mode)) {
2275 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2276 "is not a block device", journal_path);
2278 kfree(journal_path);
2282 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2284 kfree(journal_path);
2285 } else if (token == Opt_journal_ioprio) {
2287 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2291 parsed_opts->journal_ioprio =
2292 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2293 } else if (token == Opt_test_dummy_encryption) {
2294 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2296 } else if (m->flags & MOPT_DATAJ) {
2298 if (!sbi->s_journal)
2299 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2300 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2301 ext4_msg(sb, KERN_ERR,
2302 "Cannot change data mode on remount");
2306 clear_opt(sb, DATA_FLAGS);
2307 sbi->s_mount_opt |= m->mount_opt;
2310 } else if (m->flags & MOPT_QFMT) {
2311 if (sb_any_quota_loaded(sb) &&
2312 sbi->s_jquota_fmt != m->mount_opt) {
2313 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2314 "quota options when quota turned on");
2317 if (ext4_has_feature_quota(sb)) {
2318 ext4_msg(sb, KERN_INFO,
2319 "Quota format mount options ignored "
2320 "when QUOTA feature is enabled");
2323 sbi->s_jquota_fmt = m->mount_opt;
2325 } else if (token == Opt_dax || token == Opt_dax_always ||
2326 token == Opt_dax_inode || token == Opt_dax_never) {
2327 #ifdef CONFIG_FS_DAX
2330 case Opt_dax_always:
2332 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2333 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2334 fail_dax_change_remount:
2335 ext4_msg(sb, KERN_ERR, "can't change "
2336 "dax mount option while remounting");
2340 (test_opt(sb, DATA_FLAGS) ==
2341 EXT4_MOUNT_JOURNAL_DATA)) {
2342 ext4_msg(sb, KERN_ERR, "can't mount with "
2343 "both data=journal and dax");
2346 ext4_msg(sb, KERN_WARNING,
2347 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2348 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2349 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2353 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2354 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2355 goto fail_dax_change_remount;
2356 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2357 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2361 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2362 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2363 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2364 goto fail_dax_change_remount;
2365 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2366 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2367 /* Strictly for printing options */
2368 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2372 ext4_msg(sb, KERN_INFO, "dax option not supported");
2373 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2374 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2377 } else if (token == Opt_data_err_abort) {
2378 sbi->s_mount_opt |= m->mount_opt;
2379 } else if (token == Opt_data_err_ignore) {
2380 sbi->s_mount_opt &= ~m->mount_opt;
2381 } else if (token == Opt_mb_optimize_scan) {
2382 if (arg != 0 && arg != 1) {
2383 ext4_msg(sb, KERN_WARNING,
2384 "mb_optimize_scan should be set to 0 or 1.");
2387 parsed_opts->mb_optimize_scan = arg;
2391 if (m->flags & MOPT_CLEAR)
2393 else if (unlikely(!(m->flags & MOPT_SET))) {
2394 ext4_msg(sb, KERN_WARNING,
2395 "buggy handling of option %s", opt);
2399 if (m->flags & MOPT_2) {
2401 sbi->s_mount_opt2 |= m->mount_opt;
2403 sbi->s_mount_opt2 &= ~m->mount_opt;
2406 sbi->s_mount_opt |= m->mount_opt;
2408 sbi->s_mount_opt &= ~m->mount_opt;
2414 static int parse_options(char *options, struct super_block *sb,
2415 struct ext4_parsed_options *ret_opts,
2418 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2419 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2420 substring_t args[MAX_OPT_ARGS];
2426 while ((p = strsep(&options, ",")) != NULL) {
2430 * Initialize args struct so we know whether arg was
2431 * found; some options take optional arguments.
2433 args[0].to = args[0].from = NULL;
2434 token = match_token(p, tokens, args);
2435 if (handle_mount_opt(sb, p, token, args, ret_opts,
2441 * We do the test below only for project quotas. 'usrquota' and
2442 * 'grpquota' mount options are allowed even without quota feature
2443 * to support legacy quotas in quota files.
2445 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2446 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2447 "Cannot enable project quota enforcement.");
2450 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2451 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2452 if (usr_qf_name || grp_qf_name) {
2453 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2454 clear_opt(sb, USRQUOTA);
2456 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2457 clear_opt(sb, GRPQUOTA);
2459 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2460 ext4_msg(sb, KERN_ERR, "old and new quota "
2465 if (!sbi->s_jquota_fmt) {
2466 ext4_msg(sb, KERN_ERR, "journaled quota format "
2472 if (test_opt(sb, DIOREAD_NOLOCK)) {
2474 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2475 if (blocksize < PAGE_SIZE)
2476 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2477 "experimental mount option 'dioread_nolock' "
2478 "for blocksize < PAGE_SIZE");
2483 static inline void ext4_show_quota_options(struct seq_file *seq,
2484 struct super_block *sb)
2486 #if defined(CONFIG_QUOTA)
2487 struct ext4_sb_info *sbi = EXT4_SB(sb);
2488 char *usr_qf_name, *grp_qf_name;
2490 if (sbi->s_jquota_fmt) {
2493 switch (sbi->s_jquota_fmt) {
2504 seq_printf(seq, ",jqfmt=%s", fmtname);
2508 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2509 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2511 seq_show_option(seq, "usrjquota", usr_qf_name);
2513 seq_show_option(seq, "grpjquota", grp_qf_name);
2518 static const char *token2str(int token)
2520 const struct match_token *t;
2522 for (t = tokens; t->token != Opt_err; t++)
2523 if (t->token == token && !strchr(t->pattern, '='))
2530 * - it's set to a non-default value OR
2531 * - if the per-sb default is different from the global default
2533 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2536 struct ext4_sb_info *sbi = EXT4_SB(sb);
2537 struct ext4_super_block *es = sbi->s_es;
2538 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2539 const struct mount_opts *m;
2540 char sep = nodefs ? '\n' : ',';
2542 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2543 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2545 if (sbi->s_sb_block != 1)
2546 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2548 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2549 int want_set = m->flags & MOPT_SET;
2550 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2551 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2553 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2554 continue; /* skip if same as the default */
2556 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2557 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2558 continue; /* select Opt_noFoo vs Opt_Foo */
2559 SEQ_OPTS_PRINT("%s", token2str(m->token));
2562 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2563 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2564 SEQ_OPTS_PRINT("resuid=%u",
2565 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2566 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2567 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2568 SEQ_OPTS_PRINT("resgid=%u",
2569 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2570 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2571 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2572 SEQ_OPTS_PUTS("errors=remount-ro");
2573 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2574 SEQ_OPTS_PUTS("errors=continue");
2575 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2576 SEQ_OPTS_PUTS("errors=panic");
2577 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2578 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2579 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2580 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2581 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2582 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2583 if (sb->s_flags & SB_I_VERSION)
2584 SEQ_OPTS_PUTS("i_version");
2585 if (nodefs || sbi->s_stripe)
2586 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2587 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2588 (sbi->s_mount_opt ^ def_mount_opt)) {
2589 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2590 SEQ_OPTS_PUTS("data=journal");
2591 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2592 SEQ_OPTS_PUTS("data=ordered");
2593 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2594 SEQ_OPTS_PUTS("data=writeback");
2597 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2598 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2599 sbi->s_inode_readahead_blks);
2601 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2602 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2603 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2604 if (nodefs || sbi->s_max_dir_size_kb)
2605 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2606 if (test_opt(sb, DATA_ERR_ABORT))
2607 SEQ_OPTS_PUTS("data_err=abort");
2609 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2611 if (sb->s_flags & SB_INLINECRYPT)
2612 SEQ_OPTS_PUTS("inlinecrypt");
2614 if (test_opt(sb, DAX_ALWAYS)) {
2616 SEQ_OPTS_PUTS("dax");
2618 SEQ_OPTS_PUTS("dax=always");
2619 } else if (test_opt2(sb, DAX_NEVER)) {
2620 SEQ_OPTS_PUTS("dax=never");
2621 } else if (test_opt2(sb, DAX_INODE)) {
2622 SEQ_OPTS_PUTS("dax=inode");
2624 ext4_show_quota_options(seq, sb);
2628 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2630 return _ext4_show_options(seq, root->d_sb, 0);
2633 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2635 struct super_block *sb = seq->private;
2638 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2639 rc = _ext4_show_options(seq, sb, 1);
2640 seq_puts(seq, "\n");
2644 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2647 struct ext4_sb_info *sbi = EXT4_SB(sb);
2650 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2651 ext4_msg(sb, KERN_ERR, "revision level too high, "
2652 "forcing read-only mode");
2658 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2659 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2660 "running e2fsck is recommended");
2661 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2662 ext4_msg(sb, KERN_WARNING,
2663 "warning: mounting fs with errors, "
2664 "running e2fsck is recommended");
2665 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2666 le16_to_cpu(es->s_mnt_count) >=
2667 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2668 ext4_msg(sb, KERN_WARNING,
2669 "warning: maximal mount count reached, "
2670 "running e2fsck is recommended");
2671 else if (le32_to_cpu(es->s_checkinterval) &&
2672 (ext4_get_tstamp(es, s_lastcheck) +
2673 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2674 ext4_msg(sb, KERN_WARNING,
2675 "warning: checktime reached, "
2676 "running e2fsck is recommended");
2677 if (!sbi->s_journal)
2678 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2679 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2680 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2681 le16_add_cpu(&es->s_mnt_count, 1);
2682 ext4_update_tstamp(es, s_mtime);
2684 ext4_set_feature_journal_needs_recovery(sb);
2686 err = ext4_commit_super(sb);
2688 if (test_opt(sb, DEBUG))
2689 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2690 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2692 sbi->s_groups_count,
2693 EXT4_BLOCKS_PER_GROUP(sb),
2694 EXT4_INODES_PER_GROUP(sb),
2695 sbi->s_mount_opt, sbi->s_mount_opt2);
2697 cleancache_init_fs(sb);
2701 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2703 struct ext4_sb_info *sbi = EXT4_SB(sb);
2704 struct flex_groups **old_groups, **new_groups;
2707 if (!sbi->s_log_groups_per_flex)
2710 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2711 if (size <= sbi->s_flex_groups_allocated)
2714 new_groups = kvzalloc(roundup_pow_of_two(size *
2715 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2717 ext4_msg(sb, KERN_ERR,
2718 "not enough memory for %d flex group pointers", size);
2721 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2722 new_groups[i] = kvzalloc(roundup_pow_of_two(
2723 sizeof(struct flex_groups)),
2725 if (!new_groups[i]) {
2726 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2727 kvfree(new_groups[j]);
2729 ext4_msg(sb, KERN_ERR,
2730 "not enough memory for %d flex groups", size);
2735 old_groups = rcu_dereference(sbi->s_flex_groups);
2737 memcpy(new_groups, old_groups,
2738 (sbi->s_flex_groups_allocated *
2739 sizeof(struct flex_groups *)));
2741 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2742 sbi->s_flex_groups_allocated = size;
2744 ext4_kvfree_array_rcu(old_groups);
2748 static int ext4_fill_flex_info(struct super_block *sb)
2750 struct ext4_sb_info *sbi = EXT4_SB(sb);
2751 struct ext4_group_desc *gdp = NULL;
2752 struct flex_groups *fg;
2753 ext4_group_t flex_group;
2756 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2757 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2758 sbi->s_log_groups_per_flex = 0;
2762 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2766 for (i = 0; i < sbi->s_groups_count; i++) {
2767 gdp = ext4_get_group_desc(sb, i, NULL);
2769 flex_group = ext4_flex_group(sbi, i);
2770 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2771 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2772 atomic64_add(ext4_free_group_clusters(sb, gdp),
2773 &fg->free_clusters);
2774 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2782 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2783 struct ext4_group_desc *gdp)
2785 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2787 __le32 le_group = cpu_to_le32(block_group);
2788 struct ext4_sb_info *sbi = EXT4_SB(sb);
2790 if (ext4_has_metadata_csum(sbi->s_sb)) {
2791 /* Use new metadata_csum algorithm */
2793 __u16 dummy_csum = 0;
2795 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2797 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2798 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2799 sizeof(dummy_csum));
2800 offset += sizeof(dummy_csum);
2801 if (offset < sbi->s_desc_size)
2802 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2803 sbi->s_desc_size - offset);
2805 crc = csum32 & 0xFFFF;
2809 /* old crc16 code */
2810 if (!ext4_has_feature_gdt_csum(sb))
2813 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2814 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2815 crc = crc16(crc, (__u8 *)gdp, offset);
2816 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2817 /* for checksum of struct ext4_group_desc do the rest...*/
2818 if (ext4_has_feature_64bit(sb) &&
2819 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2820 crc = crc16(crc, (__u8 *)gdp + offset,
2821 le16_to_cpu(sbi->s_es->s_desc_size) -
2825 return cpu_to_le16(crc);
2828 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2829 struct ext4_group_desc *gdp)
2831 if (ext4_has_group_desc_csum(sb) &&
2832 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2838 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2839 struct ext4_group_desc *gdp)
2841 if (!ext4_has_group_desc_csum(sb))
2843 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2846 /* Called at mount-time, super-block is locked */
2847 static int ext4_check_descriptors(struct super_block *sb,
2848 ext4_fsblk_t sb_block,
2849 ext4_group_t *first_not_zeroed)
2851 struct ext4_sb_info *sbi = EXT4_SB(sb);
2852 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2853 ext4_fsblk_t last_block;
2854 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2855 ext4_fsblk_t block_bitmap;
2856 ext4_fsblk_t inode_bitmap;
2857 ext4_fsblk_t inode_table;
2858 int flexbg_flag = 0;
2859 ext4_group_t i, grp = sbi->s_groups_count;
2861 if (ext4_has_feature_flex_bg(sb))
2864 ext4_debug("Checking group descriptors");
2866 for (i = 0; i < sbi->s_groups_count; i++) {
2867 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2869 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2870 last_block = ext4_blocks_count(sbi->s_es) - 1;
2872 last_block = first_block +
2873 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2875 if ((grp == sbi->s_groups_count) &&
2876 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2879 block_bitmap = ext4_block_bitmap(sb, gdp);
2880 if (block_bitmap == sb_block) {
2881 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2882 "Block bitmap for group %u overlaps "
2887 if (block_bitmap >= sb_block + 1 &&
2888 block_bitmap <= last_bg_block) {
2889 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2890 "Block bitmap for group %u overlaps "
2891 "block group descriptors", i);
2895 if (block_bitmap < first_block || block_bitmap > last_block) {
2896 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2897 "Block bitmap for group %u not in group "
2898 "(block %llu)!", i, block_bitmap);
2901 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2902 if (inode_bitmap == sb_block) {
2903 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2904 "Inode bitmap for group %u overlaps "
2909 if (inode_bitmap >= sb_block + 1 &&
2910 inode_bitmap <= last_bg_block) {
2911 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2912 "Inode bitmap for group %u overlaps "
2913 "block group descriptors", i);
2917 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2918 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2919 "Inode bitmap for group %u not in group "
2920 "(block %llu)!", i, inode_bitmap);
2923 inode_table = ext4_inode_table(sb, gdp);
2924 if (inode_table == sb_block) {
2925 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2926 "Inode table for group %u overlaps "
2931 if (inode_table >= sb_block + 1 &&
2932 inode_table <= last_bg_block) {
2933 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2934 "Inode table for group %u overlaps "
2935 "block group descriptors", i);
2939 if (inode_table < first_block ||
2940 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2941 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2942 "Inode table for group %u not in group "
2943 "(block %llu)!", i, inode_table);
2946 ext4_lock_group(sb, i);
2947 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2948 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2949 "Checksum for group %u failed (%u!=%u)",
2950 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2951 gdp)), le16_to_cpu(gdp->bg_checksum));
2952 if (!sb_rdonly(sb)) {
2953 ext4_unlock_group(sb, i);
2957 ext4_unlock_group(sb, i);
2959 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2961 if (NULL != first_not_zeroed)
2962 *first_not_zeroed = grp;
2966 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2967 * the superblock) which were deleted from all directories, but held open by
2968 * a process at the time of a crash. We walk the list and try to delete these
2969 * inodes at recovery time (only with a read-write filesystem).
2971 * In order to keep the orphan inode chain consistent during traversal (in
2972 * case of crash during recovery), we link each inode into the superblock
2973 * orphan list_head and handle it the same way as an inode deletion during
2974 * normal operation (which journals the operations for us).
2976 * We only do an iget() and an iput() on each inode, which is very safe if we
2977 * accidentally point at an in-use or already deleted inode. The worst that
2978 * can happen in this case is that we get a "bit already cleared" message from
2979 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2980 * e2fsck was run on this filesystem, and it must have already done the orphan
2981 * inode cleanup for us, so we can safely abort without any further action.
2983 static void ext4_orphan_cleanup(struct super_block *sb,
2984 struct ext4_super_block *es)
2986 unsigned int s_flags = sb->s_flags;
2987 int ret, nr_orphans = 0, nr_truncates = 0;
2989 int quota_update = 0;
2992 if (!es->s_last_orphan) {
2993 jbd_debug(4, "no orphan inodes to clean up\n");
2997 if (bdev_read_only(sb->s_bdev)) {
2998 ext4_msg(sb, KERN_ERR, "write access "
2999 "unavailable, skipping orphan cleanup");
3003 /* Check if feature set would not allow a r/w mount */
3004 if (!ext4_feature_set_ok(sb, 0)) {
3005 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3006 "unknown ROCOMPAT features");
3010 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3011 /* don't clear list on RO mount w/ errors */
3012 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3013 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3014 "clearing orphan list.\n");
3015 es->s_last_orphan = 0;
3017 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3021 if (s_flags & SB_RDONLY) {
3022 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3023 sb->s_flags &= ~SB_RDONLY;
3027 * Turn on quotas which were not enabled for read-only mounts if
3028 * filesystem has quota feature, so that they are updated correctly.
3030 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3031 int ret = ext4_enable_quotas(sb);
3036 ext4_msg(sb, KERN_ERR,
3037 "Cannot turn on quotas: error %d", ret);
3040 /* Turn on journaled quotas used for old sytle */
3041 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3042 if (EXT4_SB(sb)->s_qf_names[i]) {
3043 int ret = ext4_quota_on_mount(sb, i);
3048 ext4_msg(sb, KERN_ERR,
3049 "Cannot turn on journaled "
3050 "quota: type %d: error %d", i, ret);
3055 while (es->s_last_orphan) {
3056 struct inode *inode;
3059 * We may have encountered an error during cleanup; if
3060 * so, skip the rest.
3062 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3063 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3064 es->s_last_orphan = 0;
3068 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3069 if (IS_ERR(inode)) {
3070 es->s_last_orphan = 0;
3074 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3075 dquot_initialize(inode);
3076 if (inode->i_nlink) {
3077 if (test_opt(sb, DEBUG))
3078 ext4_msg(sb, KERN_DEBUG,
3079 "%s: truncating inode %lu to %lld bytes",
3080 __func__, inode->i_ino, inode->i_size);
3081 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3082 inode->i_ino, inode->i_size);
3084 truncate_inode_pages(inode->i_mapping, inode->i_size);
3085 ret = ext4_truncate(inode);
3088 * We need to clean up the in-core orphan list
3089 * manually if ext4_truncate() failed to get a
3090 * transaction handle.
3092 ext4_orphan_del(NULL, inode);
3093 ext4_std_error(inode->i_sb, ret);
3095 inode_unlock(inode);
3098 if (test_opt(sb, DEBUG))
3099 ext4_msg(sb, KERN_DEBUG,
3100 "%s: deleting unreferenced inode %lu",
3101 __func__, inode->i_ino);
3102 jbd_debug(2, "deleting unreferenced inode %lu\n",
3106 iput(inode); /* The delete magic happens here! */
3109 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3112 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3113 PLURAL(nr_orphans));
3115 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3116 PLURAL(nr_truncates));
3118 /* Turn off quotas if they were enabled for orphan cleanup */
3120 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3121 if (sb_dqopt(sb)->files[i])
3122 dquot_quota_off(sb, i);
3126 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3130 * Maximal extent format file size.
3131 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3132 * extent format containers, within a sector_t, and within i_blocks
3133 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3134 * so that won't be a limiting factor.
3136 * However there is other limiting factor. We do store extents in the form
3137 * of starting block and length, hence the resulting length of the extent
3138 * covering maximum file size must fit into on-disk format containers as
3139 * well. Given that length is always by 1 unit bigger than max unit (because
3140 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3142 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3144 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3147 loff_t upper_limit = MAX_LFS_FILESIZE;
3149 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3151 if (!has_huge_files) {
3152 upper_limit = (1LL << 32) - 1;
3154 /* total blocks in file system block size */
3155 upper_limit >>= (blkbits - 9);
3156 upper_limit <<= blkbits;
3160 * 32-bit extent-start container, ee_block. We lower the maxbytes
3161 * by one fs block, so ee_len can cover the extent of maximum file
3164 res = (1LL << 32) - 1;
3167 /* Sanity check against vm- & vfs- imposed limits */
3168 if (res > upper_limit)
3175 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3176 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3177 * We need to be 1 filesystem block less than the 2^48 sector limit.
3179 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3181 loff_t res = EXT4_NDIR_BLOCKS;
3184 /* This is calculated to be the largest file size for a dense, block
3185 * mapped file such that the file's total number of 512-byte sectors,
3186 * including data and all indirect blocks, does not exceed (2^48 - 1).
3188 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3189 * number of 512-byte sectors of the file.
3192 if (!has_huge_files) {
3194 * !has_huge_files or implies that the inode i_block field
3195 * represents total file blocks in 2^32 512-byte sectors ==
3196 * size of vfs inode i_blocks * 8
3198 upper_limit = (1LL << 32) - 1;
3200 /* total blocks in file system block size */
3201 upper_limit >>= (bits - 9);
3205 * We use 48 bit ext4_inode i_blocks
3206 * With EXT4_HUGE_FILE_FL set the i_blocks
3207 * represent total number of blocks in
3208 * file system block size
3210 upper_limit = (1LL << 48) - 1;
3214 /* indirect blocks */
3216 /* double indirect blocks */
3217 meta_blocks += 1 + (1LL << (bits-2));
3218 /* tripple indirect blocks */
3219 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3221 upper_limit -= meta_blocks;
3222 upper_limit <<= bits;
3224 res += 1LL << (bits-2);
3225 res += 1LL << (2*(bits-2));
3226 res += 1LL << (3*(bits-2));
3228 if (res > upper_limit)
3231 if (res > MAX_LFS_FILESIZE)
3232 res = MAX_LFS_FILESIZE;
3237 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3238 ext4_fsblk_t logical_sb_block, int nr)
3240 struct ext4_sb_info *sbi = EXT4_SB(sb);
3241 ext4_group_t bg, first_meta_bg;
3244 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3246 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3247 return logical_sb_block + nr + 1;
3248 bg = sbi->s_desc_per_block * nr;
3249 if (ext4_bg_has_super(sb, bg))
3253 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3254 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3255 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3258 if (sb->s_blocksize == 1024 && nr == 0 &&
3259 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3262 return (has_super + ext4_group_first_block_no(sb, bg));
3266 * ext4_get_stripe_size: Get the stripe size.
3267 * @sbi: In memory super block info
3269 * If we have specified it via mount option, then
3270 * use the mount option value. If the value specified at mount time is
3271 * greater than the blocks per group use the super block value.
3272 * If the super block value is greater than blocks per group return 0.
3273 * Allocator needs it be less than blocks per group.
3276 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3278 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3279 unsigned long stripe_width =
3280 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3283 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3284 ret = sbi->s_stripe;
3285 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3287 else if (stride && stride <= sbi->s_blocks_per_group)
3293 * If the stripe width is 1, this makes no sense and
3294 * we set it to 0 to turn off stripe handling code.
3303 * Check whether this filesystem can be mounted based on
3304 * the features present and the RDONLY/RDWR mount requested.
3305 * Returns 1 if this filesystem can be mounted as requested,
3306 * 0 if it cannot be.
3308 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3310 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3311 ext4_msg(sb, KERN_ERR,
3312 "Couldn't mount because of "
3313 "unsupported optional features (%x)",
3314 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3315 ~EXT4_FEATURE_INCOMPAT_SUPP));
3319 #ifndef CONFIG_UNICODE
3320 if (ext4_has_feature_casefold(sb)) {
3321 ext4_msg(sb, KERN_ERR,
3322 "Filesystem with casefold feature cannot be "
3323 "mounted without CONFIG_UNICODE");
3331 if (ext4_has_feature_readonly(sb)) {
3332 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3333 sb->s_flags |= SB_RDONLY;
3337 /* Check that feature set is OK for a read-write mount */
3338 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3339 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3340 "unsupported optional features (%x)",
3341 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3342 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3345 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3346 ext4_msg(sb, KERN_ERR,
3347 "Can't support bigalloc feature without "
3348 "extents feature\n");
3352 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3353 if (!readonly && (ext4_has_feature_quota(sb) ||
3354 ext4_has_feature_project(sb))) {
3355 ext4_msg(sb, KERN_ERR,
3356 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3359 #endif /* CONFIG_QUOTA */
3364 * This function is called once a day if we have errors logged
3365 * on the file system
3367 static void print_daily_error_info(struct timer_list *t)
3369 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3370 struct super_block *sb = sbi->s_sb;
3371 struct ext4_super_block *es = sbi->s_es;
3373 if (es->s_error_count)
3374 /* fsck newer than v1.41.13 is needed to clean this condition. */
3375 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3376 le32_to_cpu(es->s_error_count));
3377 if (es->s_first_error_time) {
3378 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3380 ext4_get_tstamp(es, s_first_error_time),
3381 (int) sizeof(es->s_first_error_func),
3382 es->s_first_error_func,
3383 le32_to_cpu(es->s_first_error_line));
3384 if (es->s_first_error_ino)
3385 printk(KERN_CONT ": inode %u",
3386 le32_to_cpu(es->s_first_error_ino));
3387 if (es->s_first_error_block)
3388 printk(KERN_CONT ": block %llu", (unsigned long long)
3389 le64_to_cpu(es->s_first_error_block));
3390 printk(KERN_CONT "\n");
3392 if (es->s_last_error_time) {
3393 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3395 ext4_get_tstamp(es, s_last_error_time),
3396 (int) sizeof(es->s_last_error_func),
3397 es->s_last_error_func,
3398 le32_to_cpu(es->s_last_error_line));
3399 if (es->s_last_error_ino)
3400 printk(KERN_CONT ": inode %u",
3401 le32_to_cpu(es->s_last_error_ino));
3402 if (es->s_last_error_block)
3403 printk(KERN_CONT ": block %llu", (unsigned long long)
3404 le64_to_cpu(es->s_last_error_block));
3405 printk(KERN_CONT "\n");
3407 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3410 /* Find next suitable group and run ext4_init_inode_table */
3411 static int ext4_run_li_request(struct ext4_li_request *elr)
3413 struct ext4_group_desc *gdp = NULL;
3414 struct super_block *sb = elr->lr_super;
3415 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3416 ext4_group_t group = elr->lr_next_group;
3417 unsigned long timeout = 0;
3418 unsigned int prefetch_ios = 0;
3421 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3422 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3423 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3425 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3427 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3429 if (group >= elr->lr_next_group) {
3431 if (elr->lr_first_not_zeroed != ngroups &&
3432 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3433 elr->lr_next_group = elr->lr_first_not_zeroed;
3434 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3441 for (; group < ngroups; group++) {
3442 gdp = ext4_get_group_desc(sb, group, NULL);
3448 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3452 if (group >= ngroups)
3457 ret = ext4_init_inode_table(sb, group,
3458 elr->lr_timeout ? 0 : 1);
3459 trace_ext4_lazy_itable_init(sb, group);
3460 if (elr->lr_timeout == 0) {
3461 timeout = (jiffies - timeout) *
3462 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3463 elr->lr_timeout = timeout;
3465 elr->lr_next_sched = jiffies + elr->lr_timeout;
3466 elr->lr_next_group = group + 1;
3472 * Remove lr_request from the list_request and free the
3473 * request structure. Should be called with li_list_mtx held
3475 static void ext4_remove_li_request(struct ext4_li_request *elr)
3480 list_del(&elr->lr_request);
3481 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3485 static void ext4_unregister_li_request(struct super_block *sb)
3487 mutex_lock(&ext4_li_mtx);
3488 if (!ext4_li_info) {
3489 mutex_unlock(&ext4_li_mtx);
3493 mutex_lock(&ext4_li_info->li_list_mtx);
3494 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3495 mutex_unlock(&ext4_li_info->li_list_mtx);
3496 mutex_unlock(&ext4_li_mtx);
3499 static struct task_struct *ext4_lazyinit_task;
3502 * This is the function where ext4lazyinit thread lives. It walks
3503 * through the request list searching for next scheduled filesystem.
3504 * When such a fs is found, run the lazy initialization request
3505 * (ext4_rn_li_request) and keep track of the time spend in this
3506 * function. Based on that time we compute next schedule time of
3507 * the request. When walking through the list is complete, compute
3508 * next waking time and put itself into sleep.
3510 static int ext4_lazyinit_thread(void *arg)
3512 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3513 struct list_head *pos, *n;
3514 struct ext4_li_request *elr;
3515 unsigned long next_wakeup, cur;
3517 BUG_ON(NULL == eli);
3521 next_wakeup = MAX_JIFFY_OFFSET;
3523 mutex_lock(&eli->li_list_mtx);
3524 if (list_empty(&eli->li_request_list)) {
3525 mutex_unlock(&eli->li_list_mtx);
3528 list_for_each_safe(pos, n, &eli->li_request_list) {
3531 elr = list_entry(pos, struct ext4_li_request,
3534 if (time_before(jiffies, elr->lr_next_sched)) {
3535 if (time_before(elr->lr_next_sched, next_wakeup))
3536 next_wakeup = elr->lr_next_sched;
3539 if (down_read_trylock(&elr->lr_super->s_umount)) {
3540 if (sb_start_write_trylock(elr->lr_super)) {
3543 * We hold sb->s_umount, sb can not
3544 * be removed from the list, it is
3545 * now safe to drop li_list_mtx
3547 mutex_unlock(&eli->li_list_mtx);
3548 err = ext4_run_li_request(elr);
3549 sb_end_write(elr->lr_super);
3550 mutex_lock(&eli->li_list_mtx);
3553 up_read((&elr->lr_super->s_umount));
3555 /* error, remove the lazy_init job */
3557 ext4_remove_li_request(elr);
3561 elr->lr_next_sched = jiffies +
3563 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3565 if (time_before(elr->lr_next_sched, next_wakeup))
3566 next_wakeup = elr->lr_next_sched;
3568 mutex_unlock(&eli->li_list_mtx);
3573 if ((time_after_eq(cur, next_wakeup)) ||
3574 (MAX_JIFFY_OFFSET == next_wakeup)) {
3579 schedule_timeout_interruptible(next_wakeup - cur);
3581 if (kthread_should_stop()) {
3582 ext4_clear_request_list();
3589 * It looks like the request list is empty, but we need
3590 * to check it under the li_list_mtx lock, to prevent any
3591 * additions into it, and of course we should lock ext4_li_mtx
3592 * to atomically free the list and ext4_li_info, because at
3593 * this point another ext4 filesystem could be registering
3596 mutex_lock(&ext4_li_mtx);
3597 mutex_lock(&eli->li_list_mtx);
3598 if (!list_empty(&eli->li_request_list)) {
3599 mutex_unlock(&eli->li_list_mtx);
3600 mutex_unlock(&ext4_li_mtx);
3603 mutex_unlock(&eli->li_list_mtx);
3604 kfree(ext4_li_info);
3605 ext4_li_info = NULL;
3606 mutex_unlock(&ext4_li_mtx);
3611 static void ext4_clear_request_list(void)
3613 struct list_head *pos, *n;
3614 struct ext4_li_request *elr;
3616 mutex_lock(&ext4_li_info->li_list_mtx);
3617 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3618 elr = list_entry(pos, struct ext4_li_request,
3620 ext4_remove_li_request(elr);
3622 mutex_unlock(&ext4_li_info->li_list_mtx);
3625 static int ext4_run_lazyinit_thread(void)
3627 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3628 ext4_li_info, "ext4lazyinit");
3629 if (IS_ERR(ext4_lazyinit_task)) {
3630 int err = PTR_ERR(ext4_lazyinit_task);
3631 ext4_clear_request_list();
3632 kfree(ext4_li_info);
3633 ext4_li_info = NULL;
3634 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3635 "initialization thread\n",
3639 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3644 * Check whether it make sense to run itable init. thread or not.
3645 * If there is at least one uninitialized inode table, return
3646 * corresponding group number, else the loop goes through all
3647 * groups and return total number of groups.
3649 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3651 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3652 struct ext4_group_desc *gdp = NULL;
3654 if (!ext4_has_group_desc_csum(sb))
3657 for (group = 0; group < ngroups; group++) {
3658 gdp = ext4_get_group_desc(sb, group, NULL);
3662 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3669 static int ext4_li_info_new(void)
3671 struct ext4_lazy_init *eli = NULL;
3673 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3677 INIT_LIST_HEAD(&eli->li_request_list);
3678 mutex_init(&eli->li_list_mtx);
3680 eli->li_state |= EXT4_LAZYINIT_QUIT;
3687 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3690 struct ext4_li_request *elr;
3692 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3697 elr->lr_first_not_zeroed = start;
3698 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3699 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3700 elr->lr_next_group = start;
3702 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3706 * Randomize first schedule time of the request to
3707 * spread the inode table initialization requests
3710 elr->lr_next_sched = jiffies + (prandom_u32() %
3711 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3715 int ext4_register_li_request(struct super_block *sb,
3716 ext4_group_t first_not_zeroed)
3718 struct ext4_sb_info *sbi = EXT4_SB(sb);
3719 struct ext4_li_request *elr = NULL;
3720 ext4_group_t ngroups = sbi->s_groups_count;
3723 mutex_lock(&ext4_li_mtx);
3724 if (sbi->s_li_request != NULL) {
3726 * Reset timeout so it can be computed again, because
3727 * s_li_wait_mult might have changed.
3729 sbi->s_li_request->lr_timeout = 0;
3733 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3734 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3735 !test_opt(sb, INIT_INODE_TABLE)))
3738 elr = ext4_li_request_new(sb, first_not_zeroed);
3744 if (NULL == ext4_li_info) {
3745 ret = ext4_li_info_new();
3750 mutex_lock(&ext4_li_info->li_list_mtx);
3751 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3752 mutex_unlock(&ext4_li_info->li_list_mtx);
3754 sbi->s_li_request = elr;
3756 * set elr to NULL here since it has been inserted to
3757 * the request_list and the removal and free of it is
3758 * handled by ext4_clear_request_list from now on.
3762 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3763 ret = ext4_run_lazyinit_thread();
3768 mutex_unlock(&ext4_li_mtx);
3775 * We do not need to lock anything since this is called on
3778 static void ext4_destroy_lazyinit_thread(void)
3781 * If thread exited earlier
3782 * there's nothing to be done.
3784 if (!ext4_li_info || !ext4_lazyinit_task)
3787 kthread_stop(ext4_lazyinit_task);
3790 static int set_journal_csum_feature_set(struct super_block *sb)
3793 int compat, incompat;
3794 struct ext4_sb_info *sbi = EXT4_SB(sb);
3796 if (ext4_has_metadata_csum(sb)) {
3797 /* journal checksum v3 */
3799 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3801 /* journal checksum v1 */
3802 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3806 jbd2_journal_clear_features(sbi->s_journal,
3807 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3808 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3809 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3810 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3811 ret = jbd2_journal_set_features(sbi->s_journal,
3813 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3815 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3816 ret = jbd2_journal_set_features(sbi->s_journal,
3819 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3820 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3822 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3823 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3830 * Note: calculating the overhead so we can be compatible with
3831 * historical BSD practice is quite difficult in the face of
3832 * clusters/bigalloc. This is because multiple metadata blocks from
3833 * different block group can end up in the same allocation cluster.
3834 * Calculating the exact overhead in the face of clustered allocation
3835 * requires either O(all block bitmaps) in memory or O(number of block
3836 * groups**2) in time. We will still calculate the superblock for
3837 * older file systems --- and if we come across with a bigalloc file
3838 * system with zero in s_overhead_clusters the estimate will be close to
3839 * correct especially for very large cluster sizes --- but for newer
3840 * file systems, it's better to calculate this figure once at mkfs
3841 * time, and store it in the superblock. If the superblock value is
3842 * present (even for non-bigalloc file systems), we will use it.
3844 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3847 struct ext4_sb_info *sbi = EXT4_SB(sb);
3848 struct ext4_group_desc *gdp;
3849 ext4_fsblk_t first_block, last_block, b;
3850 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3851 int s, j, count = 0;
3853 if (!ext4_has_feature_bigalloc(sb))
3854 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3855 sbi->s_itb_per_group + 2);
3857 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3858 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3859 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3860 for (i = 0; i < ngroups; i++) {
3861 gdp = ext4_get_group_desc(sb, i, NULL);
3862 b = ext4_block_bitmap(sb, gdp);
3863 if (b >= first_block && b <= last_block) {
3864 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3867 b = ext4_inode_bitmap(sb, gdp);
3868 if (b >= first_block && b <= last_block) {
3869 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3872 b = ext4_inode_table(sb, gdp);
3873 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3874 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3875 int c = EXT4_B2C(sbi, b - first_block);
3876 ext4_set_bit(c, buf);
3882 if (ext4_bg_has_super(sb, grp)) {
3883 ext4_set_bit(s++, buf);
3886 j = ext4_bg_num_gdb(sb, grp);
3887 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3888 ext4_error(sb, "Invalid number of block group "
3889 "descriptor blocks: %d", j);
3890 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3894 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3898 return EXT4_CLUSTERS_PER_GROUP(sb) -
3899 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3903 * Compute the overhead and stash it in sbi->s_overhead
3905 int ext4_calculate_overhead(struct super_block *sb)
3907 struct ext4_sb_info *sbi = EXT4_SB(sb);
3908 struct ext4_super_block *es = sbi->s_es;
3909 struct inode *j_inode;
3910 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3911 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3912 ext4_fsblk_t overhead = 0;
3913 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3919 * Compute the overhead (FS structures). This is constant
3920 * for a given filesystem unless the number of block groups
3921 * changes so we cache the previous value until it does.
3925 * All of the blocks before first_data_block are overhead
3927 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3930 * Add the overhead found in each block group
3932 for (i = 0; i < ngroups; i++) {
3935 blks = count_overhead(sb, i, buf);
3938 memset(buf, 0, PAGE_SIZE);
3943 * Add the internal journal blocks whether the journal has been
3946 if (sbi->s_journal && !sbi->s_journal_bdev)
3947 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3948 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3949 /* j_inum for internal journal is non-zero */
3950 j_inode = ext4_get_journal_inode(sb, j_inum);
3952 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3953 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3956 ext4_msg(sb, KERN_ERR, "can't get journal size");
3959 sbi->s_overhead = overhead;
3961 free_page((unsigned long) buf);
3965 static void ext4_set_resv_clusters(struct super_block *sb)
3967 ext4_fsblk_t resv_clusters;
3968 struct ext4_sb_info *sbi = EXT4_SB(sb);
3971 * There's no need to reserve anything when we aren't using extents.
3972 * The space estimates are exact, there are no unwritten extents,
3973 * hole punching doesn't need new metadata... This is needed especially
3974 * to keep ext2/3 backward compatibility.
3976 if (!ext4_has_feature_extents(sb))
3979 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3980 * This should cover the situations where we can not afford to run
3981 * out of space like for example punch hole, or converting
3982 * unwritten extents in delalloc path. In most cases such
3983 * allocation would require 1, or 2 blocks, higher numbers are
3986 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3987 sbi->s_cluster_bits);
3989 do_div(resv_clusters, 50);
3990 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3992 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3995 static const char *ext4_quota_mode(struct super_block *sb)
3998 if (!ext4_quota_capable(sb))
4001 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4002 return "journalled";
4010 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
4012 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
4013 char *orig_data = kstrdup(data, GFP_KERNEL);
4014 struct buffer_head *bh, **group_desc;
4015 struct ext4_super_block *es = NULL;
4016 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4017 struct flex_groups **flex_groups;
4019 ext4_fsblk_t sb_block = get_sb_block(&data);
4020 ext4_fsblk_t logical_sb_block;
4021 unsigned long offset = 0;
4022 unsigned long def_mount_opts;
4026 int blocksize, clustersize;
4027 unsigned int db_count;
4029 int needs_recovery, has_huge_files;
4032 ext4_group_t first_not_zeroed;
4033 struct ext4_parsed_options parsed_opts;
4035 /* Set defaults for the variables that will be set during parsing */
4036 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4037 parsed_opts.journal_devnum = 0;
4038 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
4040 if ((data && !orig_data) || !sbi)
4043 sbi->s_daxdev = dax_dev;
4044 sbi->s_blockgroup_lock =
4045 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4046 if (!sbi->s_blockgroup_lock)
4049 sb->s_fs_info = sbi;
4051 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4052 sbi->s_sb_block = sb_block;
4053 sbi->s_sectors_written_start =
4054 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4056 /* Cleanup superblock name */
4057 strreplace(sb->s_id, '/', '!');
4059 /* -EINVAL is default */
4061 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4063 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4068 * The ext4 superblock will not be buffer aligned for other than 1kB
4069 * block sizes. We need to calculate the offset from buffer start.
4071 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4072 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4073 offset = do_div(logical_sb_block, blocksize);
4075 logical_sb_block = sb_block;
4078 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4080 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4085 * Note: s_es must be initialized as soon as possible because
4086 * some ext4 macro-instructions depend on its value
4088 es = (struct ext4_super_block *) (bh->b_data + offset);
4090 sb->s_magic = le16_to_cpu(es->s_magic);
4091 if (sb->s_magic != EXT4_SUPER_MAGIC)
4093 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4095 /* Warn if metadata_csum and gdt_csum are both set. */
4096 if (ext4_has_feature_metadata_csum(sb) &&
4097 ext4_has_feature_gdt_csum(sb))
4098 ext4_warning(sb, "metadata_csum and uninit_bg are "
4099 "redundant flags; please run fsck.");
4101 /* Check for a known checksum algorithm */
4102 if (!ext4_verify_csum_type(sb, es)) {
4103 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4104 "unknown checksum algorithm.");
4109 /* Load the checksum driver */
4110 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4111 if (IS_ERR(sbi->s_chksum_driver)) {
4112 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4113 ret = PTR_ERR(sbi->s_chksum_driver);
4114 sbi->s_chksum_driver = NULL;
4118 /* Check superblock checksum */
4119 if (!ext4_superblock_csum_verify(sb, es)) {
4120 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4121 "invalid superblock checksum. Run e2fsck?");
4127 /* Precompute checksum seed for all metadata */
4128 if (ext4_has_feature_csum_seed(sb))
4129 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4130 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4131 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4132 sizeof(es->s_uuid));
4134 /* Set defaults before we parse the mount options */
4135 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4136 set_opt(sb, INIT_INODE_TABLE);
4137 if (def_mount_opts & EXT4_DEFM_DEBUG)
4139 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4141 if (def_mount_opts & EXT4_DEFM_UID16)
4142 set_opt(sb, NO_UID32);
4143 /* xattr user namespace & acls are now defaulted on */
4144 set_opt(sb, XATTR_USER);
4145 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4146 set_opt(sb, POSIX_ACL);
4148 if (ext4_has_feature_fast_commit(sb))
4149 set_opt2(sb, JOURNAL_FAST_COMMIT);
4150 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4151 if (ext4_has_metadata_csum(sb))
4152 set_opt(sb, JOURNAL_CHECKSUM);
4154 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4155 set_opt(sb, JOURNAL_DATA);
4156 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4157 set_opt(sb, ORDERED_DATA);
4158 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4159 set_opt(sb, WRITEBACK_DATA);
4161 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4162 set_opt(sb, ERRORS_PANIC);
4163 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4164 set_opt(sb, ERRORS_CONT);
4166 set_opt(sb, ERRORS_RO);
4167 /* block_validity enabled by default; disable with noblock_validity */
4168 set_opt(sb, BLOCK_VALIDITY);
4169 if (def_mount_opts & EXT4_DEFM_DISCARD)
4170 set_opt(sb, DISCARD);
4172 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4173 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4174 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4175 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4176 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4178 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4179 set_opt(sb, BARRIER);
4182 * enable delayed allocation by default
4183 * Use -o nodelalloc to turn it off
4185 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4186 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4187 set_opt(sb, DELALLOC);
4190 * set default s_li_wait_mult for lazyinit, for the case there is
4191 * no mount option specified.
4193 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4195 if (le32_to_cpu(es->s_log_block_size) >
4196 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4197 ext4_msg(sb, KERN_ERR,
4198 "Invalid log block size: %u",
4199 le32_to_cpu(es->s_log_block_size));
4202 if (le32_to_cpu(es->s_log_cluster_size) >
4203 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4204 ext4_msg(sb, KERN_ERR,
4205 "Invalid log cluster size: %u",
4206 le32_to_cpu(es->s_log_cluster_size));
4210 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4212 if (blocksize == PAGE_SIZE)
4213 set_opt(sb, DIOREAD_NOLOCK);
4215 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4216 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4217 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4219 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4220 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4221 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4222 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4226 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4227 (!is_power_of_2(sbi->s_inode_size)) ||
4228 (sbi->s_inode_size > blocksize)) {
4229 ext4_msg(sb, KERN_ERR,
4230 "unsupported inode size: %d",
4232 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4236 * i_atime_extra is the last extra field available for
4237 * [acm]times in struct ext4_inode. Checking for that
4238 * field should suffice to ensure we have extra space
4241 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4242 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4243 sb->s_time_gran = 1;
4244 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4246 sb->s_time_gran = NSEC_PER_SEC;
4247 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4249 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4251 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4252 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4253 EXT4_GOOD_OLD_INODE_SIZE;
4254 if (ext4_has_feature_extra_isize(sb)) {
4255 unsigned v, max = (sbi->s_inode_size -
4256 EXT4_GOOD_OLD_INODE_SIZE);
4258 v = le16_to_cpu(es->s_want_extra_isize);
4260 ext4_msg(sb, KERN_ERR,
4261 "bad s_want_extra_isize: %d", v);
4264 if (sbi->s_want_extra_isize < v)
4265 sbi->s_want_extra_isize = v;
4267 v = le16_to_cpu(es->s_min_extra_isize);
4269 ext4_msg(sb, KERN_ERR,
4270 "bad s_min_extra_isize: %d", v);
4273 if (sbi->s_want_extra_isize < v)
4274 sbi->s_want_extra_isize = v;
4278 if (sbi->s_es->s_mount_opts[0]) {
4279 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4280 sizeof(sbi->s_es->s_mount_opts),
4284 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4285 ext4_msg(sb, KERN_WARNING,
4286 "failed to parse options in superblock: %s",
4289 kfree(s_mount_opts);
4291 sbi->s_def_mount_opt = sbi->s_mount_opt;
4292 if (!parse_options((char *) data, sb, &parsed_opts, 0))
4295 #ifdef CONFIG_UNICODE
4296 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4297 const struct ext4_sb_encodings *encoding_info;
4298 struct unicode_map *encoding;
4299 __u16 encoding_flags;
4301 if (ext4_sb_read_encoding(es, &encoding_info,
4303 ext4_msg(sb, KERN_ERR,
4304 "Encoding requested by superblock is unknown");
4308 encoding = utf8_load(encoding_info->version);
4309 if (IS_ERR(encoding)) {
4310 ext4_msg(sb, KERN_ERR,
4311 "can't mount with superblock charset: %s-%s "
4312 "not supported by the kernel. flags: 0x%x.",
4313 encoding_info->name, encoding_info->version,
4317 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4318 "%s-%s with flags 0x%hx", encoding_info->name,
4319 encoding_info->version?:"\b", encoding_flags);
4321 sb->s_encoding = encoding;
4322 sb->s_encoding_flags = encoding_flags;
4326 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4327 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4328 /* can't mount with both data=journal and dioread_nolock. */
4329 clear_opt(sb, DIOREAD_NOLOCK);
4330 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4331 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4332 ext4_msg(sb, KERN_ERR, "can't mount with "
4333 "both data=journal and delalloc");
4336 if (test_opt(sb, DAX_ALWAYS)) {
4337 ext4_msg(sb, KERN_ERR, "can't mount with "
4338 "both data=journal and dax");
4341 if (ext4_has_feature_encrypt(sb)) {
4342 ext4_msg(sb, KERN_WARNING,
4343 "encrypted files will use data=ordered "
4344 "instead of data journaling mode");
4346 if (test_opt(sb, DELALLOC))
4347 clear_opt(sb, DELALLOC);
4349 sb->s_iflags |= SB_I_CGROUPWB;
4352 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4353 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4355 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4356 (ext4_has_compat_features(sb) ||
4357 ext4_has_ro_compat_features(sb) ||
4358 ext4_has_incompat_features(sb)))
4359 ext4_msg(sb, KERN_WARNING,
4360 "feature flags set on rev 0 fs, "
4361 "running e2fsck is recommended");
4363 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4364 set_opt2(sb, HURD_COMPAT);
4365 if (ext4_has_feature_64bit(sb)) {
4366 ext4_msg(sb, KERN_ERR,
4367 "The Hurd can't support 64-bit file systems");
4372 * ea_inode feature uses l_i_version field which is not
4373 * available in HURD_COMPAT mode.
4375 if (ext4_has_feature_ea_inode(sb)) {
4376 ext4_msg(sb, KERN_ERR,
4377 "ea_inode feature is not supported for Hurd");
4382 if (IS_EXT2_SB(sb)) {
4383 if (ext2_feature_set_ok(sb))
4384 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4385 "using the ext4 subsystem");
4388 * If we're probing be silent, if this looks like
4389 * it's actually an ext[34] filesystem.
4391 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4393 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4394 "to feature incompatibilities");
4399 if (IS_EXT3_SB(sb)) {
4400 if (ext3_feature_set_ok(sb))
4401 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4402 "using the ext4 subsystem");
4405 * If we're probing be silent, if this looks like
4406 * it's actually an ext4 filesystem.
4408 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4410 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4411 "to feature incompatibilities");
4417 * Check feature flags regardless of the revision level, since we
4418 * previously didn't change the revision level when setting the flags,
4419 * so there is a chance incompat flags are set on a rev 0 filesystem.
4421 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4424 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4425 ext4_msg(sb, KERN_ERR,
4426 "Number of reserved GDT blocks insanely large: %d",
4427 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4431 if (bdev_dax_supported(sb->s_bdev, blocksize))
4432 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4434 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4435 if (ext4_has_feature_inline_data(sb)) {
4436 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4437 " that may contain inline data");
4440 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4441 ext4_msg(sb, KERN_ERR,
4442 "DAX unsupported by block device.");
4447 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4448 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4449 es->s_encryption_level);
4453 if (sb->s_blocksize != blocksize) {
4455 * bh must be released before kill_bdev(), otherwise
4456 * it won't be freed and its page also. kill_bdev()
4457 * is called by sb_set_blocksize().
4460 /* Validate the filesystem blocksize */
4461 if (!sb_set_blocksize(sb, blocksize)) {
4462 ext4_msg(sb, KERN_ERR, "bad block size %d",
4468 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4469 offset = do_div(logical_sb_block, blocksize);
4470 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4472 ext4_msg(sb, KERN_ERR,
4473 "Can't read superblock on 2nd try");
4478 es = (struct ext4_super_block *)(bh->b_data + offset);
4480 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4481 ext4_msg(sb, KERN_ERR,
4482 "Magic mismatch, very weird!");
4487 has_huge_files = ext4_has_feature_huge_file(sb);
4488 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4490 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4492 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4493 if (ext4_has_feature_64bit(sb)) {
4494 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4495 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4496 !is_power_of_2(sbi->s_desc_size)) {
4497 ext4_msg(sb, KERN_ERR,
4498 "unsupported descriptor size %lu",
4503 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4505 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4506 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4508 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4509 if (sbi->s_inodes_per_block == 0)
4511 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4512 sbi->s_inodes_per_group > blocksize * 8) {
4513 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4514 sbi->s_inodes_per_group);
4517 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4518 sbi->s_inodes_per_block;
4519 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4521 sbi->s_mount_state = le16_to_cpu(es->s_state);
4522 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4523 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4525 for (i = 0; i < 4; i++)
4526 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4527 sbi->s_def_hash_version = es->s_def_hash_version;
4528 if (ext4_has_feature_dir_index(sb)) {
4529 i = le32_to_cpu(es->s_flags);
4530 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4531 sbi->s_hash_unsigned = 3;
4532 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4533 #ifdef __CHAR_UNSIGNED__
4536 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4537 sbi->s_hash_unsigned = 3;
4541 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4546 /* Handle clustersize */
4547 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4548 if (ext4_has_feature_bigalloc(sb)) {
4549 if (clustersize < blocksize) {
4550 ext4_msg(sb, KERN_ERR,
4551 "cluster size (%d) smaller than "
4552 "block size (%d)", clustersize, blocksize);
4555 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4556 le32_to_cpu(es->s_log_block_size);
4557 sbi->s_clusters_per_group =
4558 le32_to_cpu(es->s_clusters_per_group);
4559 if (sbi->s_clusters_per_group > blocksize * 8) {
4560 ext4_msg(sb, KERN_ERR,
4561 "#clusters per group too big: %lu",
4562 sbi->s_clusters_per_group);
4565 if (sbi->s_blocks_per_group !=
4566 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4567 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4568 "clusters per group (%lu) inconsistent",
4569 sbi->s_blocks_per_group,
4570 sbi->s_clusters_per_group);
4574 if (clustersize != blocksize) {
4575 ext4_msg(sb, KERN_ERR,
4576 "fragment/cluster size (%d) != "
4577 "block size (%d)", clustersize, blocksize);
4580 if (sbi->s_blocks_per_group > blocksize * 8) {
4581 ext4_msg(sb, KERN_ERR,
4582 "#blocks per group too big: %lu",
4583 sbi->s_blocks_per_group);
4586 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4587 sbi->s_cluster_bits = 0;
4589 sbi->s_cluster_ratio = clustersize / blocksize;
4591 /* Do we have standard group size of clustersize * 8 blocks ? */
4592 if (sbi->s_blocks_per_group == clustersize << 3)
4593 set_opt2(sb, STD_GROUP_SIZE);
4596 * Test whether we have more sectors than will fit in sector_t,
4597 * and whether the max offset is addressable by the page cache.
4599 err = generic_check_addressable(sb->s_blocksize_bits,
4600 ext4_blocks_count(es));
4602 ext4_msg(sb, KERN_ERR, "filesystem"
4603 " too large to mount safely on this system");
4607 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4610 /* check blocks count against device size */
4611 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4612 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4613 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4614 "exceeds size of device (%llu blocks)",
4615 ext4_blocks_count(es), blocks_count);
4620 * It makes no sense for the first data block to be beyond the end
4621 * of the filesystem.
4623 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4624 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4625 "block %u is beyond end of filesystem (%llu)",
4626 le32_to_cpu(es->s_first_data_block),
4627 ext4_blocks_count(es));
4630 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4631 (sbi->s_cluster_ratio == 1)) {
4632 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4633 "block is 0 with a 1k block and cluster size");
4637 blocks_count = (ext4_blocks_count(es) -
4638 le32_to_cpu(es->s_first_data_block) +
4639 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4640 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4641 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4642 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4643 "(block count %llu, first data block %u, "
4644 "blocks per group %lu)", blocks_count,
4645 ext4_blocks_count(es),
4646 le32_to_cpu(es->s_first_data_block),
4647 EXT4_BLOCKS_PER_GROUP(sb));
4650 sbi->s_groups_count = blocks_count;
4651 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4652 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4653 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4654 le32_to_cpu(es->s_inodes_count)) {
4655 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4656 le32_to_cpu(es->s_inodes_count),
4657 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4661 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4662 EXT4_DESC_PER_BLOCK(sb);
4663 if (ext4_has_feature_meta_bg(sb)) {
4664 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4665 ext4_msg(sb, KERN_WARNING,
4666 "first meta block group too large: %u "
4667 "(group descriptor block count %u)",
4668 le32_to_cpu(es->s_first_meta_bg), db_count);
4672 rcu_assign_pointer(sbi->s_group_desc,
4673 kvmalloc_array(db_count,
4674 sizeof(struct buffer_head *),
4676 if (sbi->s_group_desc == NULL) {
4677 ext4_msg(sb, KERN_ERR, "not enough memory");
4682 bgl_lock_init(sbi->s_blockgroup_lock);
4684 /* Pre-read the descriptors into the buffer cache */
4685 for (i = 0; i < db_count; i++) {
4686 block = descriptor_loc(sb, logical_sb_block, i);
4687 ext4_sb_breadahead_unmovable(sb, block);
4690 for (i = 0; i < db_count; i++) {
4691 struct buffer_head *bh;
4693 block = descriptor_loc(sb, logical_sb_block, i);
4694 bh = ext4_sb_bread_unmovable(sb, block);
4696 ext4_msg(sb, KERN_ERR,
4697 "can't read group descriptor %d", i);
4703 rcu_dereference(sbi->s_group_desc)[i] = bh;
4706 sbi->s_gdb_count = db_count;
4707 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4708 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4709 ret = -EFSCORRUPTED;
4713 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4714 spin_lock_init(&sbi->s_error_lock);
4715 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4717 /* Register extent status tree shrinker */
4718 if (ext4_es_register_shrinker(sbi))
4721 sbi->s_stripe = ext4_get_stripe_size(sbi);
4722 sbi->s_extent_max_zeroout_kb = 32;
4725 * set up enough so that it can read an inode
4727 sb->s_op = &ext4_sops;
4728 sb->s_export_op = &ext4_export_ops;
4729 sb->s_xattr = ext4_xattr_handlers;
4730 #ifdef CONFIG_FS_ENCRYPTION
4731 sb->s_cop = &ext4_cryptops;
4733 #ifdef CONFIG_FS_VERITY
4734 sb->s_vop = &ext4_verityops;
4737 sb->dq_op = &ext4_quota_operations;
4738 if (ext4_has_feature_quota(sb))
4739 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4741 sb->s_qcop = &ext4_qctl_operations;
4742 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4744 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4746 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4747 mutex_init(&sbi->s_orphan_lock);
4749 /* Initialize fast commit stuff */
4750 atomic_set(&sbi->s_fc_subtid, 0);
4751 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4752 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4753 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4754 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4755 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4756 sbi->s_fc_bytes = 0;
4757 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4758 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4759 spin_lock_init(&sbi->s_fc_lock);
4760 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4761 sbi->s_fc_replay_state.fc_regions = NULL;
4762 sbi->s_fc_replay_state.fc_regions_size = 0;
4763 sbi->s_fc_replay_state.fc_regions_used = 0;
4764 sbi->s_fc_replay_state.fc_regions_valid = 0;
4765 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4766 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4767 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4771 needs_recovery = (es->s_last_orphan != 0 ||
4772 ext4_has_feature_journal_needs_recovery(sb));
4774 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4775 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4776 goto failed_mount3a;
4779 * The first inode we look at is the journal inode. Don't try
4780 * root first: it may be modified in the journal!
4782 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4783 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4785 goto failed_mount3a;
4786 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4787 ext4_has_feature_journal_needs_recovery(sb)) {
4788 ext4_msg(sb, KERN_ERR, "required journal recovery "
4789 "suppressed and not mounted read-only");
4790 goto failed_mount_wq;
4792 /* Nojournal mode, all journal mount options are illegal */
4793 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4794 ext4_msg(sb, KERN_ERR, "can't mount with "
4795 "journal_checksum, fs mounted w/o journal");
4796 goto failed_mount_wq;
4798 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4799 ext4_msg(sb, KERN_ERR, "can't mount with "
4800 "journal_async_commit, fs mounted w/o journal");
4801 goto failed_mount_wq;
4803 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4804 ext4_msg(sb, KERN_ERR, "can't mount with "
4805 "commit=%lu, fs mounted w/o journal",
4806 sbi->s_commit_interval / HZ);
4807 goto failed_mount_wq;
4809 if (EXT4_MOUNT_DATA_FLAGS &
4810 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4811 ext4_msg(sb, KERN_ERR, "can't mount with "
4812 "data=, fs mounted w/o journal");
4813 goto failed_mount_wq;
4815 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4816 clear_opt(sb, JOURNAL_CHECKSUM);
4817 clear_opt(sb, DATA_FLAGS);
4818 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4819 sbi->s_journal = NULL;
4824 if (ext4_has_feature_64bit(sb) &&
4825 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4826 JBD2_FEATURE_INCOMPAT_64BIT)) {
4827 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4828 goto failed_mount_wq;
4831 if (!set_journal_csum_feature_set(sb)) {
4832 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4834 goto failed_mount_wq;
4837 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4838 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4839 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4840 ext4_msg(sb, KERN_ERR,
4841 "Failed to set fast commit journal feature");
4842 goto failed_mount_wq;
4845 /* We have now updated the journal if required, so we can
4846 * validate the data journaling mode. */
4847 switch (test_opt(sb, DATA_FLAGS)) {
4849 /* No mode set, assume a default based on the journal
4850 * capabilities: ORDERED_DATA if the journal can
4851 * cope, else JOURNAL_DATA
4853 if (jbd2_journal_check_available_features
4854 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4855 set_opt(sb, ORDERED_DATA);
4856 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4858 set_opt(sb, JOURNAL_DATA);
4859 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4863 case EXT4_MOUNT_ORDERED_DATA:
4864 case EXT4_MOUNT_WRITEBACK_DATA:
4865 if (!jbd2_journal_check_available_features
4866 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4867 ext4_msg(sb, KERN_ERR, "Journal does not support "
4868 "requested data journaling mode");
4869 goto failed_mount_wq;
4876 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4877 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4878 ext4_msg(sb, KERN_ERR, "can't mount with "
4879 "journal_async_commit in data=ordered mode");
4880 goto failed_mount_wq;
4883 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4885 sbi->s_journal->j_submit_inode_data_buffers =
4886 ext4_journal_submit_inode_data_buffers;
4887 sbi->s_journal->j_finish_inode_data_buffers =
4888 ext4_journal_finish_inode_data_buffers;
4891 if (!test_opt(sb, NO_MBCACHE)) {
4892 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4893 if (!sbi->s_ea_block_cache) {
4894 ext4_msg(sb, KERN_ERR,
4895 "Failed to create ea_block_cache");
4896 goto failed_mount_wq;
4899 if (ext4_has_feature_ea_inode(sb)) {
4900 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4901 if (!sbi->s_ea_inode_cache) {
4902 ext4_msg(sb, KERN_ERR,
4903 "Failed to create ea_inode_cache");
4904 goto failed_mount_wq;
4909 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4910 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4911 goto failed_mount_wq;
4914 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4915 !ext4_has_feature_encrypt(sb)) {
4916 ext4_set_feature_encrypt(sb);
4917 ext4_commit_super(sb);
4921 * Get the # of file system overhead blocks from the
4922 * superblock if present.
4924 if (es->s_overhead_clusters)
4925 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4927 err = ext4_calculate_overhead(sb);
4929 goto failed_mount_wq;
4933 * The maximum number of concurrent works can be high and
4934 * concurrency isn't really necessary. Limit it to 1.
4936 EXT4_SB(sb)->rsv_conversion_wq =
4937 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4938 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4939 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4945 * The jbd2_journal_load will have done any necessary log recovery,
4946 * so we can safely mount the rest of the filesystem now.
4949 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4951 ext4_msg(sb, KERN_ERR, "get root inode failed");
4952 ret = PTR_ERR(root);
4956 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4957 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4962 sb->s_root = d_make_root(root);
4964 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4969 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4970 if (ret == -EROFS) {
4971 sb->s_flags |= SB_RDONLY;
4974 goto failed_mount4a;
4976 ext4_set_resv_clusters(sb);
4978 if (test_opt(sb, BLOCK_VALIDITY)) {
4979 err = ext4_setup_system_zone(sb);
4981 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4983 goto failed_mount4a;
4986 ext4_fc_replay_cleanup(sb);
4991 * Enable optimize_scan if number of groups is > threshold. This can be
4992 * turned off by passing "mb_optimize_scan=0". This can also be
4993 * turned on forcefully by passing "mb_optimize_scan=1".
4995 if (parsed_opts.mb_optimize_scan == 1)
4996 set_opt2(sb, MB_OPTIMIZE_SCAN);
4997 else if (parsed_opts.mb_optimize_scan == 0)
4998 clear_opt2(sb, MB_OPTIMIZE_SCAN);
4999 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5000 set_opt2(sb, MB_OPTIMIZE_SCAN);
5002 err = ext4_mb_init(sb);
5004 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5010 * We can only set up the journal commit callback once
5011 * mballoc is initialized
5014 sbi->s_journal->j_commit_callback =
5015 ext4_journal_commit_callback;
5017 block = ext4_count_free_clusters(sb);
5018 ext4_free_blocks_count_set(sbi->s_es,
5019 EXT4_C2B(sbi, block));
5020 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5023 unsigned long freei = ext4_count_free_inodes(sb);
5024 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5025 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5029 err = percpu_counter_init(&sbi->s_dirs_counter,
5030 ext4_count_dirs(sb), GFP_KERNEL);
5032 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5035 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5038 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5041 ext4_msg(sb, KERN_ERR, "insufficient memory");
5045 if (ext4_has_feature_flex_bg(sb))
5046 if (!ext4_fill_flex_info(sb)) {
5047 ext4_msg(sb, KERN_ERR,
5048 "unable to initialize "
5049 "flex_bg meta info!");
5054 err = ext4_register_li_request(sb, first_not_zeroed);
5058 err = ext4_register_sysfs(sb);
5063 /* Enable quota usage during mount. */
5064 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5065 err = ext4_enable_quotas(sb);
5069 #endif /* CONFIG_QUOTA */
5072 * Save the original bdev mapping's wb_err value which could be
5073 * used to detect the metadata async write error.
5075 spin_lock_init(&sbi->s_bdev_wb_lock);
5076 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5077 &sbi->s_bdev_wb_err);
5078 sb->s_bdev->bd_super = sb;
5079 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5080 ext4_orphan_cleanup(sb, es);
5081 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5082 if (needs_recovery) {
5083 ext4_msg(sb, KERN_INFO, "recovery complete");
5084 err = ext4_mark_recovery_complete(sb, es);
5088 if (EXT4_SB(sb)->s_journal) {
5089 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5090 descr = " journalled data mode";
5091 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5092 descr = " ordered data mode";
5094 descr = " writeback data mode";
5096 descr = "out journal";
5098 if (test_opt(sb, DISCARD)) {
5099 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5100 if (!blk_queue_discard(q))
5101 ext4_msg(sb, KERN_WARNING,
5102 "mounting with \"discard\" option, but "
5103 "the device does not support discard");
5106 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5107 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5108 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5109 (int) sizeof(sbi->s_es->s_mount_opts),
5110 sbi->s_es->s_mount_opts,
5111 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5112 ext4_quota_mode(sb));
5114 if (es->s_error_count)
5115 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5117 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5118 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5119 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5120 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5121 atomic_set(&sbi->s_warning_count, 0);
5122 atomic_set(&sbi->s_msg_count, 0);
5129 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5133 ext4_unregister_sysfs(sb);
5134 kobject_put(&sbi->s_kobj);
5136 ext4_unregister_li_request(sb);
5138 ext4_mb_release(sb);
5140 flex_groups = rcu_dereference(sbi->s_flex_groups);
5142 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5143 kvfree(flex_groups[i]);
5144 kvfree(flex_groups);
5147 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5148 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5149 percpu_counter_destroy(&sbi->s_dirs_counter);
5150 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5151 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5152 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5154 ext4_ext_release(sb);
5155 ext4_release_system_zone(sb);
5160 ext4_msg(sb, KERN_ERR, "mount failed");
5161 if (EXT4_SB(sb)->rsv_conversion_wq)
5162 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5164 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5165 sbi->s_ea_inode_cache = NULL;
5167 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5168 sbi->s_ea_block_cache = NULL;
5170 if (sbi->s_journal) {
5171 jbd2_journal_unregister_shrinker(sbi->s_journal);
5172 jbd2_journal_destroy(sbi->s_journal);
5173 sbi->s_journal = NULL;
5176 ext4_es_unregister_shrinker(sbi);
5178 flush_work(&sbi->s_error_work);
5179 del_timer_sync(&sbi->s_err_report);
5180 ext4_stop_mmpd(sbi);
5183 group_desc = rcu_dereference(sbi->s_group_desc);
5184 for (i = 0; i < db_count; i++)
5185 brelse(group_desc[i]);
5189 if (sbi->s_chksum_driver)
5190 crypto_free_shash(sbi->s_chksum_driver);
5192 #ifdef CONFIG_UNICODE
5193 utf8_unload(sb->s_encoding);
5197 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5198 kfree(get_qf_name(sb, sbi, i));
5200 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5201 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5203 ext4_blkdev_remove(sbi);
5205 sb->s_fs_info = NULL;
5206 kfree(sbi->s_blockgroup_lock);
5210 fs_put_dax(dax_dev);
5211 return err ? err : ret;
5215 * Setup any per-fs journal parameters now. We'll do this both on
5216 * initial mount, once the journal has been initialised but before we've
5217 * done any recovery; and again on any subsequent remount.
5219 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5221 struct ext4_sb_info *sbi = EXT4_SB(sb);
5223 journal->j_commit_interval = sbi->s_commit_interval;
5224 journal->j_min_batch_time = sbi->s_min_batch_time;
5225 journal->j_max_batch_time = sbi->s_max_batch_time;
5226 ext4_fc_init(sb, journal);
5228 write_lock(&journal->j_state_lock);
5229 if (test_opt(sb, BARRIER))
5230 journal->j_flags |= JBD2_BARRIER;
5232 journal->j_flags &= ~JBD2_BARRIER;
5233 if (test_opt(sb, DATA_ERR_ABORT))
5234 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5236 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5237 write_unlock(&journal->j_state_lock);
5240 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5241 unsigned int journal_inum)
5243 struct inode *journal_inode;
5246 * Test for the existence of a valid inode on disk. Bad things
5247 * happen if we iget() an unused inode, as the subsequent iput()
5248 * will try to delete it.
5250 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5251 if (IS_ERR(journal_inode)) {
5252 ext4_msg(sb, KERN_ERR, "no journal found");
5255 if (!journal_inode->i_nlink) {
5256 make_bad_inode(journal_inode);
5257 iput(journal_inode);
5258 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5262 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5263 journal_inode, journal_inode->i_size);
5264 if (!S_ISREG(journal_inode->i_mode)) {
5265 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5266 iput(journal_inode);
5269 return journal_inode;
5272 static journal_t *ext4_get_journal(struct super_block *sb,
5273 unsigned int journal_inum)
5275 struct inode *journal_inode;
5278 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5281 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5285 journal = jbd2_journal_init_inode(journal_inode);
5287 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5288 iput(journal_inode);
5291 journal->j_private = sb;
5292 ext4_init_journal_params(sb, journal);
5296 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5299 struct buffer_head *bh;
5303 int hblock, blocksize;
5304 ext4_fsblk_t sb_block;
5305 unsigned long offset;
5306 struct ext4_super_block *es;
5307 struct block_device *bdev;
5309 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5312 bdev = ext4_blkdev_get(j_dev, sb);
5316 blocksize = sb->s_blocksize;
5317 hblock = bdev_logical_block_size(bdev);
5318 if (blocksize < hblock) {
5319 ext4_msg(sb, KERN_ERR,
5320 "blocksize too small for journal device");
5324 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5325 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5326 set_blocksize(bdev, blocksize);
5327 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5328 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5329 "external journal");
5333 es = (struct ext4_super_block *) (bh->b_data + offset);
5334 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5335 !(le32_to_cpu(es->s_feature_incompat) &
5336 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5337 ext4_msg(sb, KERN_ERR, "external journal has "
5343 if ((le32_to_cpu(es->s_feature_ro_compat) &
5344 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5345 es->s_checksum != ext4_superblock_csum(sb, es)) {
5346 ext4_msg(sb, KERN_ERR, "external journal has "
5347 "corrupt superblock");
5352 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5353 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5358 len = ext4_blocks_count(es);
5359 start = sb_block + 1;
5360 brelse(bh); /* we're done with the superblock */
5362 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5363 start, len, blocksize);
5365 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5368 journal->j_private = sb;
5369 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5370 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5373 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5374 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5375 "user (unsupported) - %d",
5376 be32_to_cpu(journal->j_superblock->s_nr_users));
5379 EXT4_SB(sb)->s_journal_bdev = bdev;
5380 ext4_init_journal_params(sb, journal);
5384 jbd2_journal_destroy(journal);
5386 ext4_blkdev_put(bdev);
5390 static int ext4_load_journal(struct super_block *sb,
5391 struct ext4_super_block *es,
5392 unsigned long journal_devnum)
5395 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5398 int really_read_only;
5401 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5402 return -EFSCORRUPTED;
5404 if (journal_devnum &&
5405 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5406 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5407 "numbers have changed");
5408 journal_dev = new_decode_dev(journal_devnum);
5410 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5412 if (journal_inum && journal_dev) {
5413 ext4_msg(sb, KERN_ERR,
5414 "filesystem has both journal inode and journal device!");
5419 journal = ext4_get_journal(sb, journal_inum);
5423 journal = ext4_get_dev_journal(sb, journal_dev);
5428 journal_dev_ro = bdev_read_only(journal->j_dev);
5429 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5431 if (journal_dev_ro && !sb_rdonly(sb)) {
5432 ext4_msg(sb, KERN_ERR,
5433 "journal device read-only, try mounting with '-o ro'");
5439 * Are we loading a blank journal or performing recovery after a
5440 * crash? For recovery, we need to check in advance whether we
5441 * can get read-write access to the device.
5443 if (ext4_has_feature_journal_needs_recovery(sb)) {
5444 if (sb_rdonly(sb)) {
5445 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5446 "required on readonly filesystem");
5447 if (really_read_only) {
5448 ext4_msg(sb, KERN_ERR, "write access "
5449 "unavailable, cannot proceed "
5450 "(try mounting with noload)");
5454 ext4_msg(sb, KERN_INFO, "write access will "
5455 "be enabled during recovery");
5459 if (!(journal->j_flags & JBD2_BARRIER))
5460 ext4_msg(sb, KERN_INFO, "barriers disabled");
5462 if (!ext4_has_feature_journal_needs_recovery(sb))
5463 err = jbd2_journal_wipe(journal, !really_read_only);
5465 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5467 memcpy(save, ((char *) es) +
5468 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5469 err = jbd2_journal_load(journal);
5471 memcpy(((char *) es) + EXT4_S_ERR_START,
5472 save, EXT4_S_ERR_LEN);
5477 ext4_msg(sb, KERN_ERR, "error loading journal");
5481 EXT4_SB(sb)->s_journal = journal;
5482 err = ext4_clear_journal_err(sb, es);
5484 EXT4_SB(sb)->s_journal = NULL;
5485 jbd2_journal_destroy(journal);
5489 if (!really_read_only && journal_devnum &&
5490 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5491 es->s_journal_dev = cpu_to_le32(journal_devnum);
5493 /* Make sure we flush the recovery flag to disk. */
5494 ext4_commit_super(sb);
5497 err = jbd2_journal_register_shrinker(journal);
5499 EXT4_SB(sb)->s_journal = NULL;
5506 jbd2_journal_destroy(journal);
5510 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5511 static void ext4_update_super(struct super_block *sb)
5513 struct ext4_sb_info *sbi = EXT4_SB(sb);
5514 struct ext4_super_block *es = sbi->s_es;
5515 struct buffer_head *sbh = sbi->s_sbh;
5519 * If the file system is mounted read-only, don't update the
5520 * superblock write time. This avoids updating the superblock
5521 * write time when we are mounting the root file system
5522 * read/only but we need to replay the journal; at that point,
5523 * for people who are east of GMT and who make their clock
5524 * tick in localtime for Windows bug-for-bug compatibility,
5525 * the clock is set in the future, and this will cause e2fsck
5526 * to complain and force a full file system check.
5528 if (!(sb->s_flags & SB_RDONLY))
5529 ext4_update_tstamp(es, s_wtime);
5530 es->s_kbytes_written =
5531 cpu_to_le64(sbi->s_kbytes_written +
5532 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5533 sbi->s_sectors_written_start) >> 1));
5534 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5535 ext4_free_blocks_count_set(es,
5536 EXT4_C2B(sbi, percpu_counter_sum_positive(
5537 &sbi->s_freeclusters_counter)));
5538 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5539 es->s_free_inodes_count =
5540 cpu_to_le32(percpu_counter_sum_positive(
5541 &sbi->s_freeinodes_counter));
5542 /* Copy error information to the on-disk superblock */
5543 spin_lock(&sbi->s_error_lock);
5544 if (sbi->s_add_error_count > 0) {
5545 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5546 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5547 __ext4_update_tstamp(&es->s_first_error_time,
5548 &es->s_first_error_time_hi,
5549 sbi->s_first_error_time);
5550 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5551 sizeof(es->s_first_error_func));
5552 es->s_first_error_line =
5553 cpu_to_le32(sbi->s_first_error_line);
5554 es->s_first_error_ino =
5555 cpu_to_le32(sbi->s_first_error_ino);
5556 es->s_first_error_block =
5557 cpu_to_le64(sbi->s_first_error_block);
5558 es->s_first_error_errcode =
5559 ext4_errno_to_code(sbi->s_first_error_code);
5561 __ext4_update_tstamp(&es->s_last_error_time,
5562 &es->s_last_error_time_hi,
5563 sbi->s_last_error_time);
5564 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5565 sizeof(es->s_last_error_func));
5566 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5567 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5568 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5569 es->s_last_error_errcode =
5570 ext4_errno_to_code(sbi->s_last_error_code);
5572 * Start the daily error reporting function if it hasn't been
5575 if (!es->s_error_count)
5576 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5577 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5578 sbi->s_add_error_count = 0;
5580 spin_unlock(&sbi->s_error_lock);
5582 ext4_superblock_csum_set(sb);
5586 static int ext4_commit_super(struct super_block *sb)
5588 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5593 if (block_device_ejected(sb))
5596 ext4_update_super(sb);
5598 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5600 * Oh, dear. A previous attempt to write the
5601 * superblock failed. This could happen because the
5602 * USB device was yanked out. Or it could happen to
5603 * be a transient write error and maybe the block will
5604 * be remapped. Nothing we can do but to retry the
5605 * write and hope for the best.
5607 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5608 "superblock detected");
5609 clear_buffer_write_io_error(sbh);
5610 set_buffer_uptodate(sbh);
5612 BUFFER_TRACE(sbh, "marking dirty");
5613 mark_buffer_dirty(sbh);
5614 error = __sync_dirty_buffer(sbh,
5615 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5616 if (buffer_write_io_error(sbh)) {
5617 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5619 clear_buffer_write_io_error(sbh);
5620 set_buffer_uptodate(sbh);
5626 * Have we just finished recovery? If so, and if we are mounting (or
5627 * remounting) the filesystem readonly, then we will end up with a
5628 * consistent fs on disk. Record that fact.
5630 static int ext4_mark_recovery_complete(struct super_block *sb,
5631 struct ext4_super_block *es)
5634 journal_t *journal = EXT4_SB(sb)->s_journal;
5636 if (!ext4_has_feature_journal(sb)) {
5637 if (journal != NULL) {
5638 ext4_error(sb, "Journal got removed while the fs was "
5640 return -EFSCORRUPTED;
5644 jbd2_journal_lock_updates(journal);
5645 err = jbd2_journal_flush(journal, 0);
5649 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5650 ext4_clear_feature_journal_needs_recovery(sb);
5651 ext4_commit_super(sb);
5654 jbd2_journal_unlock_updates(journal);
5659 * If we are mounting (or read-write remounting) a filesystem whose journal
5660 * has recorded an error from a previous lifetime, move that error to the
5661 * main filesystem now.
5663 static int ext4_clear_journal_err(struct super_block *sb,
5664 struct ext4_super_block *es)
5670 if (!ext4_has_feature_journal(sb)) {
5671 ext4_error(sb, "Journal got removed while the fs was mounted!");
5672 return -EFSCORRUPTED;
5675 journal = EXT4_SB(sb)->s_journal;
5678 * Now check for any error status which may have been recorded in the
5679 * journal by a prior ext4_error() or ext4_abort()
5682 j_errno = jbd2_journal_errno(journal);
5686 errstr = ext4_decode_error(sb, j_errno, nbuf);
5687 ext4_warning(sb, "Filesystem error recorded "
5688 "from previous mount: %s", errstr);
5689 ext4_warning(sb, "Marking fs in need of filesystem check.");
5691 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5692 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5693 ext4_commit_super(sb);
5695 jbd2_journal_clear_err(journal);
5696 jbd2_journal_update_sb_errno(journal);
5702 * Force the running and committing transactions to commit,
5703 * and wait on the commit.
5705 int ext4_force_commit(struct super_block *sb)
5712 journal = EXT4_SB(sb)->s_journal;
5713 return ext4_journal_force_commit(journal);
5716 static int ext4_sync_fs(struct super_block *sb, int wait)
5720 bool needs_barrier = false;
5721 struct ext4_sb_info *sbi = EXT4_SB(sb);
5723 if (unlikely(ext4_forced_shutdown(sbi)))
5726 trace_ext4_sync_fs(sb, wait);
5727 flush_workqueue(sbi->rsv_conversion_wq);
5729 * Writeback quota in non-journalled quota case - journalled quota has
5732 dquot_writeback_dquots(sb, -1);
5734 * Data writeback is possible w/o journal transaction, so barrier must
5735 * being sent at the end of the function. But we can skip it if
5736 * transaction_commit will do it for us.
5738 if (sbi->s_journal) {
5739 target = jbd2_get_latest_transaction(sbi->s_journal);
5740 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5741 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5742 needs_barrier = true;
5744 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5746 ret = jbd2_log_wait_commit(sbi->s_journal,
5749 } else if (wait && test_opt(sb, BARRIER))
5750 needs_barrier = true;
5751 if (needs_barrier) {
5753 err = blkdev_issue_flush(sb->s_bdev);
5762 * LVM calls this function before a (read-only) snapshot is created. This
5763 * gives us a chance to flush the journal completely and mark the fs clean.
5765 * Note that only this function cannot bring a filesystem to be in a clean
5766 * state independently. It relies on upper layer to stop all data & metadata
5769 static int ext4_freeze(struct super_block *sb)
5777 journal = EXT4_SB(sb)->s_journal;
5780 /* Now we set up the journal barrier. */
5781 jbd2_journal_lock_updates(journal);
5784 * Don't clear the needs_recovery flag if we failed to
5785 * flush the journal.
5787 error = jbd2_journal_flush(journal, 0);
5791 /* Journal blocked and flushed, clear needs_recovery flag. */
5792 ext4_clear_feature_journal_needs_recovery(sb);
5795 error = ext4_commit_super(sb);
5798 /* we rely on upper layer to stop further updates */
5799 jbd2_journal_unlock_updates(journal);
5804 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5805 * flag here, even though the filesystem is not technically dirty yet.
5807 static int ext4_unfreeze(struct super_block *sb)
5809 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5812 if (EXT4_SB(sb)->s_journal) {
5813 /* Reset the needs_recovery flag before the fs is unlocked. */
5814 ext4_set_feature_journal_needs_recovery(sb);
5817 ext4_commit_super(sb);
5822 * Structure to save mount options for ext4_remount's benefit
5824 struct ext4_mount_options {
5825 unsigned long s_mount_opt;
5826 unsigned long s_mount_opt2;
5829 unsigned long s_commit_interval;
5830 u32 s_min_batch_time, s_max_batch_time;
5833 char *s_qf_names[EXT4_MAXQUOTAS];
5837 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5839 struct ext4_super_block *es;
5840 struct ext4_sb_info *sbi = EXT4_SB(sb);
5841 unsigned long old_sb_flags, vfs_flags;
5842 struct ext4_mount_options old_opts;
5843 int enable_quota = 0;
5848 char *to_free[EXT4_MAXQUOTAS];
5850 char *orig_data = kstrdup(data, GFP_KERNEL);
5851 struct ext4_parsed_options parsed_opts;
5853 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5854 parsed_opts.journal_devnum = 0;
5856 if (data && !orig_data)
5859 /* Store the original options */
5860 old_sb_flags = sb->s_flags;
5861 old_opts.s_mount_opt = sbi->s_mount_opt;
5862 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5863 old_opts.s_resuid = sbi->s_resuid;
5864 old_opts.s_resgid = sbi->s_resgid;
5865 old_opts.s_commit_interval = sbi->s_commit_interval;
5866 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5867 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5869 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5870 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5871 if (sbi->s_qf_names[i]) {
5872 char *qf_name = get_qf_name(sb, sbi, i);
5874 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5875 if (!old_opts.s_qf_names[i]) {
5876 for (j = 0; j < i; j++)
5877 kfree(old_opts.s_qf_names[j]);
5882 old_opts.s_qf_names[i] = NULL;
5884 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5885 parsed_opts.journal_ioprio =
5886 sbi->s_journal->j_task->io_context->ioprio;
5889 * Some options can be enabled by ext4 and/or by VFS mount flag
5890 * either way we need to make sure it matches in both *flags and
5891 * s_flags. Copy those selected flags from *flags to s_flags
5893 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5894 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5896 if (!parse_options(data, sb, &parsed_opts, 1)) {
5901 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5902 test_opt(sb, JOURNAL_CHECKSUM)) {
5903 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5904 "during remount not supported; ignoring");
5905 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5908 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5909 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5910 ext4_msg(sb, KERN_ERR, "can't mount with "
5911 "both data=journal and delalloc");
5915 if (test_opt(sb, DIOREAD_NOLOCK)) {
5916 ext4_msg(sb, KERN_ERR, "can't mount with "
5917 "both data=journal and dioread_nolock");
5921 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5922 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5923 ext4_msg(sb, KERN_ERR, "can't mount with "
5924 "journal_async_commit in data=ordered mode");
5930 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5931 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5936 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5937 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5939 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5940 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5944 if (sbi->s_journal) {
5945 ext4_init_journal_params(sb, sbi->s_journal);
5946 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5949 /* Flush outstanding errors before changing fs state */
5950 flush_work(&sbi->s_error_work);
5952 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5953 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5958 if (*flags & SB_RDONLY) {
5959 err = sync_filesystem(sb);
5962 err = dquot_suspend(sb, -1);
5967 * First of all, the unconditional stuff we have to do
5968 * to disable replay of the journal when we next remount
5970 sb->s_flags |= SB_RDONLY;
5973 * OK, test if we are remounting a valid rw partition
5974 * readonly, and if so set the rdonly flag and then
5975 * mark the partition as valid again.
5977 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5978 (sbi->s_mount_state & EXT4_VALID_FS))
5979 es->s_state = cpu_to_le16(sbi->s_mount_state);
5981 if (sbi->s_journal) {
5983 * We let remount-ro finish even if marking fs
5984 * as clean failed...
5986 ext4_mark_recovery_complete(sb, es);
5988 ext4_stop_mmpd(sbi);
5990 /* Make sure we can mount this feature set readwrite */
5991 if (ext4_has_feature_readonly(sb) ||
5992 !ext4_feature_set_ok(sb, 0)) {
5997 * Make sure the group descriptor checksums
5998 * are sane. If they aren't, refuse to remount r/w.
6000 for (g = 0; g < sbi->s_groups_count; g++) {
6001 struct ext4_group_desc *gdp =
6002 ext4_get_group_desc(sb, g, NULL);
6004 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6005 ext4_msg(sb, KERN_ERR,
6006 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6007 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6008 le16_to_cpu(gdp->bg_checksum));
6015 * If we have an unprocessed orphan list hanging
6016 * around from a previously readonly bdev mount,
6017 * require a full umount/remount for now.
6019 if (es->s_last_orphan) {
6020 ext4_msg(sb, KERN_WARNING, "Couldn't "
6021 "remount RDWR because of unprocessed "
6022 "orphan inode list. Please "
6023 "umount/remount instead");
6029 * Mounting a RDONLY partition read-write, so reread
6030 * and store the current valid flag. (It may have
6031 * been changed by e2fsck since we originally mounted
6034 if (sbi->s_journal) {
6035 err = ext4_clear_journal_err(sb, es);
6039 sbi->s_mount_state = le16_to_cpu(es->s_state);
6041 err = ext4_setup_super(sb, es, 0);
6045 sb->s_flags &= ~SB_RDONLY;
6046 if (ext4_has_feature_mmp(sb))
6047 if (ext4_multi_mount_protect(sb,
6048 le64_to_cpu(es->s_mmp_block))) {
6057 * Reinitialize lazy itable initialization thread based on
6060 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6061 ext4_unregister_li_request(sb);
6063 ext4_group_t first_not_zeroed;
6064 first_not_zeroed = ext4_has_uninit_itable(sb);
6065 ext4_register_li_request(sb, first_not_zeroed);
6069 * Handle creation of system zone data early because it can fail.
6070 * Releasing of existing data is done when we are sure remount will
6073 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6074 err = ext4_setup_system_zone(sb);
6079 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6080 err = ext4_commit_super(sb);
6086 /* Release old quota file names */
6087 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6088 kfree(old_opts.s_qf_names[i]);
6090 if (sb_any_quota_suspended(sb))
6091 dquot_resume(sb, -1);
6092 else if (ext4_has_feature_quota(sb)) {
6093 err = ext4_enable_quotas(sb);
6099 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6100 ext4_release_system_zone(sb);
6103 * Some options can be enabled by ext4 and/or by VFS mount flag
6104 * either way we need to make sure it matches in both *flags and
6105 * s_flags. Copy those selected flags from s_flags to *flags
6107 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6109 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6110 orig_data, ext4_quota_mode(sb));
6115 sb->s_flags = old_sb_flags;
6116 sbi->s_mount_opt = old_opts.s_mount_opt;
6117 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6118 sbi->s_resuid = old_opts.s_resuid;
6119 sbi->s_resgid = old_opts.s_resgid;
6120 sbi->s_commit_interval = old_opts.s_commit_interval;
6121 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6122 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6123 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6124 ext4_release_system_zone(sb);
6126 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6127 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6128 to_free[i] = get_qf_name(sb, sbi, i);
6129 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6132 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6140 static int ext4_statfs_project(struct super_block *sb,
6141 kprojid_t projid, struct kstatfs *buf)
6144 struct dquot *dquot;
6148 qid = make_kqid_projid(projid);
6149 dquot = dqget(sb, qid);
6151 return PTR_ERR(dquot);
6152 spin_lock(&dquot->dq_dqb_lock);
6154 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6155 dquot->dq_dqb.dqb_bhardlimit);
6156 limit >>= sb->s_blocksize_bits;
6158 if (limit && buf->f_blocks > limit) {
6159 curblock = (dquot->dq_dqb.dqb_curspace +
6160 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6161 buf->f_blocks = limit;
6162 buf->f_bfree = buf->f_bavail =
6163 (buf->f_blocks > curblock) ?
6164 (buf->f_blocks - curblock) : 0;
6167 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6168 dquot->dq_dqb.dqb_ihardlimit);
6169 if (limit && buf->f_files > limit) {
6170 buf->f_files = limit;
6172 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6173 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6176 spin_unlock(&dquot->dq_dqb_lock);
6182 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6184 struct super_block *sb = dentry->d_sb;
6185 struct ext4_sb_info *sbi = EXT4_SB(sb);
6186 struct ext4_super_block *es = sbi->s_es;
6187 ext4_fsblk_t overhead = 0, resv_blocks;
6189 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6191 if (!test_opt(sb, MINIX_DF))
6192 overhead = sbi->s_overhead;
6194 buf->f_type = EXT4_SUPER_MAGIC;
6195 buf->f_bsize = sb->s_blocksize;
6196 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6197 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6198 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6199 /* prevent underflow in case that few free space is available */
6200 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6201 buf->f_bavail = buf->f_bfree -
6202 (ext4_r_blocks_count(es) + resv_blocks);
6203 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6205 buf->f_files = le32_to_cpu(es->s_inodes_count);
6206 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6207 buf->f_namelen = EXT4_NAME_LEN;
6208 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6211 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6212 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6213 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6222 * Helper functions so that transaction is started before we acquire dqio_sem
6223 * to keep correct lock ordering of transaction > dqio_sem
6225 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6227 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6230 static int ext4_write_dquot(struct dquot *dquot)
6234 struct inode *inode;
6236 inode = dquot_to_inode(dquot);
6237 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6238 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6240 return PTR_ERR(handle);
6241 ret = dquot_commit(dquot);
6242 err = ext4_journal_stop(handle);
6248 static int ext4_acquire_dquot(struct dquot *dquot)
6253 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6254 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6256 return PTR_ERR(handle);
6257 ret = dquot_acquire(dquot);
6258 err = ext4_journal_stop(handle);
6264 static int ext4_release_dquot(struct dquot *dquot)
6269 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6270 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6271 if (IS_ERR(handle)) {
6272 /* Release dquot anyway to avoid endless cycle in dqput() */
6273 dquot_release(dquot);
6274 return PTR_ERR(handle);
6276 ret = dquot_release(dquot);
6277 err = ext4_journal_stop(handle);
6283 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6285 struct super_block *sb = dquot->dq_sb;
6287 if (ext4_is_quota_journalled(sb)) {
6288 dquot_mark_dquot_dirty(dquot);
6289 return ext4_write_dquot(dquot);
6291 return dquot_mark_dquot_dirty(dquot);
6295 static int ext4_write_info(struct super_block *sb, int type)
6300 /* Data block + inode block */
6301 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6303 return PTR_ERR(handle);
6304 ret = dquot_commit_info(sb, type);
6305 err = ext4_journal_stop(handle);
6312 * Turn on quotas during mount time - we need to find
6313 * the quota file and such...
6315 static int ext4_quota_on_mount(struct super_block *sb, int type)
6317 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6318 EXT4_SB(sb)->s_jquota_fmt, type);
6321 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6323 struct ext4_inode_info *ei = EXT4_I(inode);
6325 /* The first argument of lockdep_set_subclass has to be
6326 * *exactly* the same as the argument to init_rwsem() --- in
6327 * this case, in init_once() --- or lockdep gets unhappy
6328 * because the name of the lock is set using the
6329 * stringification of the argument to init_rwsem().
6331 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6332 lockdep_set_subclass(&ei->i_data_sem, subclass);
6336 * Standard function to be called on quota_on
6338 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6339 const struct path *path)
6343 if (!test_opt(sb, QUOTA))
6346 /* Quotafile not on the same filesystem? */
6347 if (path->dentry->d_sb != sb)
6350 /* Quota already enabled for this file? */
6351 if (IS_NOQUOTA(d_inode(path->dentry)))
6354 /* Journaling quota? */
6355 if (EXT4_SB(sb)->s_qf_names[type]) {
6356 /* Quotafile not in fs root? */
6357 if (path->dentry->d_parent != sb->s_root)
6358 ext4_msg(sb, KERN_WARNING,
6359 "Quota file not on filesystem root. "
6360 "Journaled quota will not work");
6361 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6364 * Clear the flag just in case mount options changed since
6367 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6371 * When we journal data on quota file, we have to flush journal to see
6372 * all updates to the file when we bypass pagecache...
6374 if (EXT4_SB(sb)->s_journal &&
6375 ext4_should_journal_data(d_inode(path->dentry))) {
6377 * We don't need to lock updates but journal_flush() could
6378 * otherwise be livelocked...
6380 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6381 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6382 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6387 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6388 err = dquot_quota_on(sb, type, format_id, path);
6390 lockdep_set_quota_inode(path->dentry->d_inode,
6393 struct inode *inode = d_inode(path->dentry);
6397 * Set inode flags to prevent userspace from messing with quota
6398 * files. If this fails, we return success anyway since quotas
6399 * are already enabled and this is not a hard failure.
6402 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6405 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6406 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6407 S_NOATIME | S_IMMUTABLE);
6408 err = ext4_mark_inode_dirty(handle, inode);
6409 ext4_journal_stop(handle);
6411 inode_unlock(inode);
6416 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6420 struct inode *qf_inode;
6421 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6422 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6423 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6424 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6427 BUG_ON(!ext4_has_feature_quota(sb));
6429 if (!qf_inums[type])
6432 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6433 if (IS_ERR(qf_inode)) {
6434 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6435 return PTR_ERR(qf_inode);
6438 /* Don't account quota for quota files to avoid recursion */
6439 qf_inode->i_flags |= S_NOQUOTA;
6440 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6441 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6443 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6449 /* Enable usage tracking for all quota types. */
6450 static int ext4_enable_quotas(struct super_block *sb)
6453 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6454 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6455 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6456 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6458 bool quota_mopt[EXT4_MAXQUOTAS] = {
6459 test_opt(sb, USRQUOTA),
6460 test_opt(sb, GRPQUOTA),
6461 test_opt(sb, PRJQUOTA),
6464 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6465 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6466 if (qf_inums[type]) {
6467 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6468 DQUOT_USAGE_ENABLED |
6469 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6472 "Failed to enable quota tracking "
6473 "(type=%d, err=%d). Please run "
6474 "e2fsck to fix.", type, err);
6475 for (type--; type >= 0; type--)
6476 dquot_quota_off(sb, type);
6485 static int ext4_quota_off(struct super_block *sb, int type)
6487 struct inode *inode = sb_dqopt(sb)->files[type];
6491 /* Force all delayed allocation blocks to be allocated.
6492 * Caller already holds s_umount sem */
6493 if (test_opt(sb, DELALLOC))
6494 sync_filesystem(sb);
6496 if (!inode || !igrab(inode))
6499 err = dquot_quota_off(sb, type);
6500 if (err || ext4_has_feature_quota(sb))
6505 * Update modification times of quota files when userspace can
6506 * start looking at them. If we fail, we return success anyway since
6507 * this is not a hard failure and quotas are already disabled.
6509 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6510 if (IS_ERR(handle)) {
6511 err = PTR_ERR(handle);
6514 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6515 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6516 inode->i_mtime = inode->i_ctime = current_time(inode);
6517 err = ext4_mark_inode_dirty(handle, inode);
6518 ext4_journal_stop(handle);
6520 inode_unlock(inode);
6522 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6526 return dquot_quota_off(sb, type);
6529 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6530 * acquiring the locks... As quota files are never truncated and quota code
6531 * itself serializes the operations (and no one else should touch the files)
6532 * we don't have to be afraid of races */
6533 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6534 size_t len, loff_t off)
6536 struct inode *inode = sb_dqopt(sb)->files[type];
6537 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6538 int offset = off & (sb->s_blocksize - 1);
6541 struct buffer_head *bh;
6542 loff_t i_size = i_size_read(inode);
6546 if (off+len > i_size)
6549 while (toread > 0) {
6550 tocopy = sb->s_blocksize - offset < toread ?
6551 sb->s_blocksize - offset : toread;
6552 bh = ext4_bread(NULL, inode, blk, 0);
6555 if (!bh) /* A hole? */
6556 memset(data, 0, tocopy);
6558 memcpy(data, bh->b_data+offset, tocopy);
6568 /* Write to quotafile (we know the transaction is already started and has
6569 * enough credits) */
6570 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6571 const char *data, size_t len, loff_t off)
6573 struct inode *inode = sb_dqopt(sb)->files[type];
6574 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6575 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6577 struct buffer_head *bh;
6578 handle_t *handle = journal_current_handle();
6580 if (EXT4_SB(sb)->s_journal && !handle) {
6581 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6582 " cancelled because transaction is not started",
6583 (unsigned long long)off, (unsigned long long)len);
6587 * Since we account only one data block in transaction credits,
6588 * then it is impossible to cross a block boundary.
6590 if (sb->s_blocksize - offset < len) {
6591 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6592 " cancelled because not block aligned",
6593 (unsigned long long)off, (unsigned long long)len);
6598 bh = ext4_bread(handle, inode, blk,
6599 EXT4_GET_BLOCKS_CREATE |
6600 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6601 } while (PTR_ERR(bh) == -ENOSPC &&
6602 ext4_should_retry_alloc(inode->i_sb, &retries));
6607 BUFFER_TRACE(bh, "get write access");
6608 err = ext4_journal_get_write_access(handle, bh);
6614 memcpy(bh->b_data+offset, data, len);
6615 flush_dcache_page(bh->b_page);
6617 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6620 if (inode->i_size < off + len) {
6621 i_size_write(inode, off + len);
6622 EXT4_I(inode)->i_disksize = inode->i_size;
6623 err2 = ext4_mark_inode_dirty(handle, inode);
6624 if (unlikely(err2 && !err))
6627 return err ? err : len;
6631 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6632 const char *dev_name, void *data)
6634 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6637 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6638 static inline void register_as_ext2(void)
6640 int err = register_filesystem(&ext2_fs_type);
6643 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6646 static inline void unregister_as_ext2(void)
6648 unregister_filesystem(&ext2_fs_type);
6651 static inline int ext2_feature_set_ok(struct super_block *sb)
6653 if (ext4_has_unknown_ext2_incompat_features(sb))
6657 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6662 static inline void register_as_ext2(void) { }
6663 static inline void unregister_as_ext2(void) { }
6664 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6667 static inline void register_as_ext3(void)
6669 int err = register_filesystem(&ext3_fs_type);
6672 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6675 static inline void unregister_as_ext3(void)
6677 unregister_filesystem(&ext3_fs_type);
6680 static inline int ext3_feature_set_ok(struct super_block *sb)
6682 if (ext4_has_unknown_ext3_incompat_features(sb))
6684 if (!ext4_has_feature_journal(sb))
6688 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6693 static struct file_system_type ext4_fs_type = {
6694 .owner = THIS_MODULE,
6696 .mount = ext4_mount,
6697 .kill_sb = kill_block_super,
6698 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6700 MODULE_ALIAS_FS("ext4");
6702 /* Shared across all ext4 file systems */
6703 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6705 static int __init ext4_init_fs(void)
6709 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6710 ext4_li_info = NULL;
6712 /* Build-time check for flags consistency */
6713 ext4_check_flag_values();
6715 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6716 init_waitqueue_head(&ext4__ioend_wq[i]);
6718 err = ext4_init_es();
6722 err = ext4_init_pending();
6726 err = ext4_init_post_read_processing();
6730 err = ext4_init_pageio();
6734 err = ext4_init_system_zone();
6738 err = ext4_init_sysfs();
6742 err = ext4_init_mballoc();
6745 err = init_inodecache();
6749 err = ext4_fc_init_dentry_cache();
6755 err = register_filesystem(&ext4_fs_type);
6761 unregister_as_ext2();
6762 unregister_as_ext3();
6764 destroy_inodecache();
6766 ext4_exit_mballoc();
6770 ext4_exit_system_zone();
6774 ext4_exit_post_read_processing();
6776 ext4_exit_pending();
6783 static void __exit ext4_exit_fs(void)
6785 ext4_destroy_lazyinit_thread();
6786 unregister_as_ext2();
6787 unregister_as_ext3();
6788 unregister_filesystem(&ext4_fs_type);
6789 destroy_inodecache();
6790 ext4_exit_mballoc();
6792 ext4_exit_system_zone();
6794 ext4_exit_post_read_processing();
6796 ext4_exit_pending();
6799 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6800 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6801 MODULE_LICENSE("GPL");
6802 MODULE_SOFTDEP("pre: crc32c");
6803 module_init(ext4_init_fs)
6804 module_exit(ext4_exit_fs)