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 struct buffer_head *sbh = sbi->s_sbh;
709 handle = jbd2_journal_start(journal, 1);
712 if (jbd2_journal_get_write_access(handle, sbh)) {
713 jbd2_journal_stop(handle);
716 ext4_update_super(sbi->s_sb);
717 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
718 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
719 "superblock detected");
720 clear_buffer_write_io_error(sbh);
721 set_buffer_uptodate(sbh);
724 if (jbd2_journal_dirty_metadata(handle, sbh)) {
725 jbd2_journal_stop(handle);
728 jbd2_journal_stop(handle);
733 * Write through journal failed. Write sb directly to get error info
734 * out and hope for the best.
736 ext4_commit_super(sbi->s_sb);
739 #define ext4_error_ratelimit(sb) \
740 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
743 void __ext4_error(struct super_block *sb, const char *function,
744 unsigned int line, bool force_ro, int error, __u64 block,
745 const char *fmt, ...)
747 struct va_format vaf;
750 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
753 trace_ext4_error(sb, function, line);
754 if (ext4_error_ratelimit(sb)) {
759 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
760 sb->s_id, function, line, current->comm, &vaf);
763 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
766 void __ext4_error_inode(struct inode *inode, const char *function,
767 unsigned int line, ext4_fsblk_t block, int error,
768 const char *fmt, ...)
771 struct va_format vaf;
773 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
776 trace_ext4_error(inode->i_sb, function, line);
777 if (ext4_error_ratelimit(inode->i_sb)) {
782 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
783 "inode #%lu: block %llu: comm %s: %pV\n",
784 inode->i_sb->s_id, function, line, inode->i_ino,
785 block, current->comm, &vaf);
787 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
788 "inode #%lu: comm %s: %pV\n",
789 inode->i_sb->s_id, function, line, inode->i_ino,
790 current->comm, &vaf);
793 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
797 void __ext4_error_file(struct file *file, const char *function,
798 unsigned int line, ext4_fsblk_t block,
799 const char *fmt, ...)
802 struct va_format vaf;
803 struct inode *inode = file_inode(file);
804 char pathname[80], *path;
806 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
809 trace_ext4_error(inode->i_sb, function, line);
810 if (ext4_error_ratelimit(inode->i_sb)) {
811 path = file_path(file, pathname, sizeof(pathname));
819 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
820 "block %llu: comm %s: path %s: %pV\n",
821 inode->i_sb->s_id, function, line, inode->i_ino,
822 block, current->comm, path, &vaf);
825 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
826 "comm %s: path %s: %pV\n",
827 inode->i_sb->s_id, function, line, inode->i_ino,
828 current->comm, path, &vaf);
831 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
835 const char *ext4_decode_error(struct super_block *sb, int errno,
842 errstr = "Corrupt filesystem";
845 errstr = "Filesystem failed CRC";
848 errstr = "IO failure";
851 errstr = "Out of memory";
854 if (!sb || (EXT4_SB(sb)->s_journal &&
855 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
856 errstr = "Journal has aborted";
858 errstr = "Readonly filesystem";
861 /* If the caller passed in an extra buffer for unknown
862 * errors, textualise them now. Else we just return
865 /* Check for truncated error codes... */
866 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
875 /* __ext4_std_error decodes expected errors from journaling functions
876 * automatically and invokes the appropriate error response. */
878 void __ext4_std_error(struct super_block *sb, const char *function,
879 unsigned int line, int errno)
884 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
887 /* Special case: if the error is EROFS, and we're not already
888 * inside a transaction, then there's really no point in logging
890 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
893 if (ext4_error_ratelimit(sb)) {
894 errstr = ext4_decode_error(sb, errno, nbuf);
895 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
896 sb->s_id, function, line, errstr);
899 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
902 void __ext4_msg(struct super_block *sb,
903 const char *prefix, const char *fmt, ...)
905 struct va_format vaf;
908 atomic_inc(&EXT4_SB(sb)->s_msg_count);
909 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
915 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
919 static int ext4_warning_ratelimit(struct super_block *sb)
921 atomic_inc(&EXT4_SB(sb)->s_warning_count);
922 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
926 void __ext4_warning(struct super_block *sb, const char *function,
927 unsigned int line, const char *fmt, ...)
929 struct va_format vaf;
932 if (!ext4_warning_ratelimit(sb))
938 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
939 sb->s_id, function, line, &vaf);
943 void __ext4_warning_inode(const struct inode *inode, const char *function,
944 unsigned int line, const char *fmt, ...)
946 struct va_format vaf;
949 if (!ext4_warning_ratelimit(inode->i_sb))
955 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
956 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
957 function, line, inode->i_ino, current->comm, &vaf);
961 void __ext4_grp_locked_error(const char *function, unsigned int line,
962 struct super_block *sb, ext4_group_t grp,
963 unsigned long ino, ext4_fsblk_t block,
964 const char *fmt, ...)
968 struct va_format vaf;
971 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
974 trace_ext4_error(sb, function, line);
975 if (ext4_error_ratelimit(sb)) {
979 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
980 sb->s_id, function, line, grp);
982 printk(KERN_CONT "inode %lu: ", ino);
984 printk(KERN_CONT "block %llu:",
985 (unsigned long long) block);
986 printk(KERN_CONT "%pV\n", &vaf);
990 if (test_opt(sb, ERRORS_CONT)) {
991 if (test_opt(sb, WARN_ON_ERROR))
993 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
994 if (!bdev_read_only(sb->s_bdev)) {
995 save_error_info(sb, EFSCORRUPTED, ino, block, function,
997 schedule_work(&EXT4_SB(sb)->s_error_work);
1001 ext4_unlock_group(sb, grp);
1002 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1004 * We only get here in the ERRORS_RO case; relocking the group
1005 * may be dangerous, but nothing bad will happen since the
1006 * filesystem will have already been marked read/only and the
1007 * journal has been aborted. We return 1 as a hint to callers
1008 * who might what to use the return value from
1009 * ext4_grp_locked_error() to distinguish between the
1010 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1011 * aggressively from the ext4 function in question, with a
1012 * more appropriate error code.
1014 ext4_lock_group(sb, grp);
1018 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1022 struct ext4_sb_info *sbi = EXT4_SB(sb);
1023 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1024 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1027 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1028 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1031 percpu_counter_sub(&sbi->s_freeclusters_counter,
1035 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1036 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1041 count = ext4_free_inodes_count(sb, gdp);
1042 percpu_counter_sub(&sbi->s_freeinodes_counter,
1048 void ext4_update_dynamic_rev(struct super_block *sb)
1050 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1052 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1056 "updating to rev %d because of new feature flag, "
1057 "running e2fsck is recommended",
1060 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1061 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1062 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1063 /* leave es->s_feature_*compat flags alone */
1064 /* es->s_uuid will be set by e2fsck if empty */
1067 * The rest of the superblock fields should be zero, and if not it
1068 * means they are likely already in use, so leave them alone. We
1069 * can leave it up to e2fsck to clean up any inconsistencies there.
1074 * Open the external journal device
1076 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1078 struct block_device *bdev;
1080 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1086 ext4_msg(sb, KERN_ERR,
1087 "failed to open journal device unknown-block(%u,%u) %ld",
1088 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1093 * Release the journal device
1095 static void ext4_blkdev_put(struct block_device *bdev)
1097 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1100 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1102 struct block_device *bdev;
1103 bdev = sbi->s_journal_bdev;
1105 ext4_blkdev_put(bdev);
1106 sbi->s_journal_bdev = NULL;
1110 static inline struct inode *orphan_list_entry(struct list_head *l)
1112 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1115 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1117 struct list_head *l;
1119 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1120 le32_to_cpu(sbi->s_es->s_last_orphan));
1122 printk(KERN_ERR "sb_info orphan list:\n");
1123 list_for_each(l, &sbi->s_orphan) {
1124 struct inode *inode = orphan_list_entry(l);
1126 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1127 inode->i_sb->s_id, inode->i_ino, inode,
1128 inode->i_mode, inode->i_nlink,
1129 NEXT_ORPHAN(inode));
1134 static int ext4_quota_off(struct super_block *sb, int type);
1136 static inline void ext4_quota_off_umount(struct super_block *sb)
1140 /* Use our quota_off function to clear inode flags etc. */
1141 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1142 ext4_quota_off(sb, type);
1146 * This is a helper function which is used in the mount/remount
1147 * codepaths (which holds s_umount) to fetch the quota file name.
1149 static inline char *get_qf_name(struct super_block *sb,
1150 struct ext4_sb_info *sbi,
1153 return rcu_dereference_protected(sbi->s_qf_names[type],
1154 lockdep_is_held(&sb->s_umount));
1157 static inline void ext4_quota_off_umount(struct super_block *sb)
1162 static void ext4_put_super(struct super_block *sb)
1164 struct ext4_sb_info *sbi = EXT4_SB(sb);
1165 struct ext4_super_block *es = sbi->s_es;
1166 struct buffer_head **group_desc;
1167 struct flex_groups **flex_groups;
1171 ext4_unregister_li_request(sb);
1172 ext4_quota_off_umount(sb);
1174 flush_work(&sbi->s_error_work);
1175 destroy_workqueue(sbi->rsv_conversion_wq);
1178 * Unregister sysfs before destroying jbd2 journal.
1179 * Since we could still access attr_journal_task attribute via sysfs
1180 * path which could have sbi->s_journal->j_task as NULL
1182 ext4_unregister_sysfs(sb);
1184 if (sbi->s_journal) {
1185 aborted = is_journal_aborted(sbi->s_journal);
1186 err = jbd2_journal_destroy(sbi->s_journal);
1187 sbi->s_journal = NULL;
1188 if ((err < 0) && !aborted) {
1189 ext4_abort(sb, -err, "Couldn't clean up the journal");
1193 ext4_es_unregister_shrinker(sbi);
1194 del_timer_sync(&sbi->s_err_report);
1195 ext4_release_system_zone(sb);
1196 ext4_mb_release(sb);
1197 ext4_ext_release(sb);
1199 if (!sb_rdonly(sb) && !aborted) {
1200 ext4_clear_feature_journal_needs_recovery(sb);
1201 es->s_state = cpu_to_le16(sbi->s_mount_state);
1204 ext4_commit_super(sb);
1207 group_desc = rcu_dereference(sbi->s_group_desc);
1208 for (i = 0; i < sbi->s_gdb_count; i++)
1209 brelse(group_desc[i]);
1211 flex_groups = rcu_dereference(sbi->s_flex_groups);
1213 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1214 kvfree(flex_groups[i]);
1215 kvfree(flex_groups);
1218 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1219 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1220 percpu_counter_destroy(&sbi->s_dirs_counter);
1221 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1222 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1223 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1225 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1226 kfree(get_qf_name(sb, sbi, i));
1229 /* Debugging code just in case the in-memory inode orphan list
1230 * isn't empty. The on-disk one can be non-empty if we've
1231 * detected an error and taken the fs readonly, but the
1232 * in-memory list had better be clean by this point. */
1233 if (!list_empty(&sbi->s_orphan))
1234 dump_orphan_list(sb, sbi);
1235 ASSERT(list_empty(&sbi->s_orphan));
1237 sync_blockdev(sb->s_bdev);
1238 invalidate_bdev(sb->s_bdev);
1239 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1241 * Invalidate the journal device's buffers. We don't want them
1242 * floating about in memory - the physical journal device may
1243 * hotswapped, and it breaks the `ro-after' testing code.
1245 sync_blockdev(sbi->s_journal_bdev);
1246 invalidate_bdev(sbi->s_journal_bdev);
1247 ext4_blkdev_remove(sbi);
1250 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1251 sbi->s_ea_inode_cache = NULL;
1253 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1254 sbi->s_ea_block_cache = NULL;
1256 ext4_stop_mmpd(sbi);
1259 sb->s_fs_info = NULL;
1261 * Now that we are completely done shutting down the
1262 * superblock, we need to actually destroy the kobject.
1264 kobject_put(&sbi->s_kobj);
1265 wait_for_completion(&sbi->s_kobj_unregister);
1266 if (sbi->s_chksum_driver)
1267 crypto_free_shash(sbi->s_chksum_driver);
1268 kfree(sbi->s_blockgroup_lock);
1269 fs_put_dax(sbi->s_daxdev);
1270 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1271 #ifdef CONFIG_UNICODE
1272 utf8_unload(sb->s_encoding);
1277 static struct kmem_cache *ext4_inode_cachep;
1280 * Called inside transaction, so use GFP_NOFS
1282 static struct inode *ext4_alloc_inode(struct super_block *sb)
1284 struct ext4_inode_info *ei;
1286 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1290 inode_set_iversion(&ei->vfs_inode, 1);
1291 spin_lock_init(&ei->i_raw_lock);
1292 INIT_LIST_HEAD(&ei->i_prealloc_list);
1293 atomic_set(&ei->i_prealloc_active, 0);
1294 spin_lock_init(&ei->i_prealloc_lock);
1295 ext4_es_init_tree(&ei->i_es_tree);
1296 rwlock_init(&ei->i_es_lock);
1297 INIT_LIST_HEAD(&ei->i_es_list);
1298 ei->i_es_all_nr = 0;
1299 ei->i_es_shk_nr = 0;
1300 ei->i_es_shrink_lblk = 0;
1301 ei->i_reserved_data_blocks = 0;
1302 spin_lock_init(&(ei->i_block_reservation_lock));
1303 ext4_init_pending_tree(&ei->i_pending_tree);
1305 ei->i_reserved_quota = 0;
1306 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1309 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1310 spin_lock_init(&ei->i_completed_io_lock);
1312 ei->i_datasync_tid = 0;
1313 atomic_set(&ei->i_unwritten, 0);
1314 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1315 ext4_fc_init_inode(&ei->vfs_inode);
1316 mutex_init(&ei->i_fc_lock);
1317 return &ei->vfs_inode;
1320 static int ext4_drop_inode(struct inode *inode)
1322 int drop = generic_drop_inode(inode);
1325 drop = fscrypt_drop_inode(inode);
1327 trace_ext4_drop_inode(inode, drop);
1331 static void ext4_free_in_core_inode(struct inode *inode)
1333 fscrypt_free_inode(inode);
1334 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1335 pr_warn("%s: inode %ld still in fc list",
1336 __func__, inode->i_ino);
1338 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1341 static void ext4_destroy_inode(struct inode *inode)
1343 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1344 ext4_msg(inode->i_sb, KERN_ERR,
1345 "Inode %lu (%p): orphan list check failed!",
1346 inode->i_ino, EXT4_I(inode));
1347 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1348 EXT4_I(inode), sizeof(struct ext4_inode_info),
1354 static void init_once(void *foo)
1356 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1358 INIT_LIST_HEAD(&ei->i_orphan);
1359 init_rwsem(&ei->xattr_sem);
1360 init_rwsem(&ei->i_data_sem);
1361 init_rwsem(&ei->i_mmap_sem);
1362 inode_init_once(&ei->vfs_inode);
1363 ext4_fc_init_inode(&ei->vfs_inode);
1366 static int __init init_inodecache(void)
1368 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1369 sizeof(struct ext4_inode_info), 0,
1370 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1372 offsetof(struct ext4_inode_info, i_data),
1373 sizeof_field(struct ext4_inode_info, i_data),
1375 if (ext4_inode_cachep == NULL)
1380 static void destroy_inodecache(void)
1383 * Make sure all delayed rcu free inodes are flushed before we
1387 kmem_cache_destroy(ext4_inode_cachep);
1390 void ext4_clear_inode(struct inode *inode)
1393 invalidate_inode_buffers(inode);
1395 ext4_discard_preallocations(inode, 0);
1396 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1398 if (EXT4_I(inode)->jinode) {
1399 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1400 EXT4_I(inode)->jinode);
1401 jbd2_free_inode(EXT4_I(inode)->jinode);
1402 EXT4_I(inode)->jinode = NULL;
1404 fscrypt_put_encryption_info(inode);
1405 fsverity_cleanup_inode(inode);
1408 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1409 u64 ino, u32 generation)
1411 struct inode *inode;
1414 * Currently we don't know the generation for parent directory, so
1415 * a generation of 0 means "accept any"
1417 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1419 return ERR_CAST(inode);
1420 if (generation && inode->i_generation != generation) {
1422 return ERR_PTR(-ESTALE);
1428 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1429 int fh_len, int fh_type)
1431 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1432 ext4_nfs_get_inode);
1435 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1436 int fh_len, int fh_type)
1438 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1439 ext4_nfs_get_inode);
1442 static int ext4_nfs_commit_metadata(struct inode *inode)
1444 struct writeback_control wbc = {
1445 .sync_mode = WB_SYNC_ALL
1448 trace_ext4_nfs_commit_metadata(inode);
1449 return ext4_write_inode(inode, &wbc);
1453 * Try to release metadata pages (indirect blocks, directories) which are
1454 * mapped via the block device. Since these pages could have journal heads
1455 * which would prevent try_to_free_buffers() from freeing them, we must use
1456 * jbd2 layer's try_to_free_buffers() function to release them.
1458 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1461 journal_t *journal = EXT4_SB(sb)->s_journal;
1463 WARN_ON(PageChecked(page));
1464 if (!page_has_buffers(page))
1467 return jbd2_journal_try_to_free_buffers(journal, page);
1469 return try_to_free_buffers(page);
1472 #ifdef CONFIG_FS_ENCRYPTION
1473 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1475 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1476 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1479 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1482 handle_t *handle = fs_data;
1483 int res, res2, credits, retries = 0;
1486 * Encrypting the root directory is not allowed because e2fsck expects
1487 * lost+found to exist and be unencrypted, and encrypting the root
1488 * directory would imply encrypting the lost+found directory as well as
1489 * the filename "lost+found" itself.
1491 if (inode->i_ino == EXT4_ROOT_INO)
1494 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1497 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1500 res = ext4_convert_inline_data(inode);
1505 * If a journal handle was specified, then the encryption context is
1506 * being set on a new inode via inheritance and is part of a larger
1507 * transaction to create the inode. Otherwise the encryption context is
1508 * being set on an existing inode in its own transaction. Only in the
1509 * latter case should the "retry on ENOSPC" logic be used.
1513 res = ext4_xattr_set_handle(handle, inode,
1514 EXT4_XATTR_INDEX_ENCRYPTION,
1515 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1518 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1519 ext4_clear_inode_state(inode,
1520 EXT4_STATE_MAY_INLINE_DATA);
1522 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1523 * S_DAX may be disabled
1525 ext4_set_inode_flags(inode, false);
1530 res = dquot_initialize(inode);
1534 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1539 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1541 return PTR_ERR(handle);
1543 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1544 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1547 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1549 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1550 * S_DAX may be disabled
1552 ext4_set_inode_flags(inode, false);
1553 res = ext4_mark_inode_dirty(handle, inode);
1555 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1557 res2 = ext4_journal_stop(handle);
1559 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1566 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1568 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1571 static bool ext4_has_stable_inodes(struct super_block *sb)
1573 return ext4_has_feature_stable_inodes(sb);
1576 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1577 int *ino_bits_ret, int *lblk_bits_ret)
1579 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1580 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1583 static const struct fscrypt_operations ext4_cryptops = {
1584 .key_prefix = "ext4:",
1585 .get_context = ext4_get_context,
1586 .set_context = ext4_set_context,
1587 .get_dummy_policy = ext4_get_dummy_policy,
1588 .empty_dir = ext4_empty_dir,
1589 .max_namelen = EXT4_NAME_LEN,
1590 .has_stable_inodes = ext4_has_stable_inodes,
1591 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1596 static const char * const quotatypes[] = INITQFNAMES;
1597 #define QTYPE2NAME(t) (quotatypes[t])
1599 static int ext4_write_dquot(struct dquot *dquot);
1600 static int ext4_acquire_dquot(struct dquot *dquot);
1601 static int ext4_release_dquot(struct dquot *dquot);
1602 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1603 static int ext4_write_info(struct super_block *sb, int type);
1604 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1605 const struct path *path);
1606 static int ext4_quota_on_mount(struct super_block *sb, int type);
1607 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1608 size_t len, loff_t off);
1609 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1610 const char *data, size_t len, loff_t off);
1611 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1612 unsigned int flags);
1613 static int ext4_enable_quotas(struct super_block *sb);
1615 static struct dquot **ext4_get_dquots(struct inode *inode)
1617 return EXT4_I(inode)->i_dquot;
1620 static const struct dquot_operations ext4_quota_operations = {
1621 .get_reserved_space = ext4_get_reserved_space,
1622 .write_dquot = ext4_write_dquot,
1623 .acquire_dquot = ext4_acquire_dquot,
1624 .release_dquot = ext4_release_dquot,
1625 .mark_dirty = ext4_mark_dquot_dirty,
1626 .write_info = ext4_write_info,
1627 .alloc_dquot = dquot_alloc,
1628 .destroy_dquot = dquot_destroy,
1629 .get_projid = ext4_get_projid,
1630 .get_inode_usage = ext4_get_inode_usage,
1631 .get_next_id = dquot_get_next_id,
1634 static const struct quotactl_ops ext4_qctl_operations = {
1635 .quota_on = ext4_quota_on,
1636 .quota_off = ext4_quota_off,
1637 .quota_sync = dquot_quota_sync,
1638 .get_state = dquot_get_state,
1639 .set_info = dquot_set_dqinfo,
1640 .get_dqblk = dquot_get_dqblk,
1641 .set_dqblk = dquot_set_dqblk,
1642 .get_nextdqblk = dquot_get_next_dqblk,
1646 static const struct super_operations ext4_sops = {
1647 .alloc_inode = ext4_alloc_inode,
1648 .free_inode = ext4_free_in_core_inode,
1649 .destroy_inode = ext4_destroy_inode,
1650 .write_inode = ext4_write_inode,
1651 .dirty_inode = ext4_dirty_inode,
1652 .drop_inode = ext4_drop_inode,
1653 .evict_inode = ext4_evict_inode,
1654 .put_super = ext4_put_super,
1655 .sync_fs = ext4_sync_fs,
1656 .freeze_fs = ext4_freeze,
1657 .unfreeze_fs = ext4_unfreeze,
1658 .statfs = ext4_statfs,
1659 .remount_fs = ext4_remount,
1660 .show_options = ext4_show_options,
1662 .quota_read = ext4_quota_read,
1663 .quota_write = ext4_quota_write,
1664 .get_dquots = ext4_get_dquots,
1666 .bdev_try_to_free_page = bdev_try_to_free_page,
1669 static const struct export_operations ext4_export_ops = {
1670 .fh_to_dentry = ext4_fh_to_dentry,
1671 .fh_to_parent = ext4_fh_to_parent,
1672 .get_parent = ext4_get_parent,
1673 .commit_metadata = ext4_nfs_commit_metadata,
1677 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1678 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1679 Opt_nouid32, Opt_debug, Opt_removed,
1680 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1681 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1682 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1683 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1684 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1685 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1687 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1688 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1689 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1690 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1691 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1692 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1693 Opt_nowarn_on_error, Opt_mblk_io_submit,
1694 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1695 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1696 Opt_inode_readahead_blks, Opt_journal_ioprio,
1697 Opt_dioread_nolock, Opt_dioread_lock,
1698 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1699 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1700 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1701 #ifdef CONFIG_EXT4_DEBUG
1702 Opt_fc_debug_max_replay, Opt_fc_debug_force
1706 static const match_table_t tokens = {
1707 {Opt_bsd_df, "bsddf"},
1708 {Opt_minix_df, "minixdf"},
1709 {Opt_grpid, "grpid"},
1710 {Opt_grpid, "bsdgroups"},
1711 {Opt_nogrpid, "nogrpid"},
1712 {Opt_nogrpid, "sysvgroups"},
1713 {Opt_resgid, "resgid=%u"},
1714 {Opt_resuid, "resuid=%u"},
1716 {Opt_err_cont, "errors=continue"},
1717 {Opt_err_panic, "errors=panic"},
1718 {Opt_err_ro, "errors=remount-ro"},
1719 {Opt_nouid32, "nouid32"},
1720 {Opt_debug, "debug"},
1721 {Opt_removed, "oldalloc"},
1722 {Opt_removed, "orlov"},
1723 {Opt_user_xattr, "user_xattr"},
1724 {Opt_nouser_xattr, "nouser_xattr"},
1726 {Opt_noacl, "noacl"},
1727 {Opt_noload, "norecovery"},
1728 {Opt_noload, "noload"},
1729 {Opt_removed, "nobh"},
1730 {Opt_removed, "bh"},
1731 {Opt_commit, "commit=%u"},
1732 {Opt_min_batch_time, "min_batch_time=%u"},
1733 {Opt_max_batch_time, "max_batch_time=%u"},
1734 {Opt_journal_dev, "journal_dev=%u"},
1735 {Opt_journal_path, "journal_path=%s"},
1736 {Opt_journal_checksum, "journal_checksum"},
1737 {Opt_nojournal_checksum, "nojournal_checksum"},
1738 {Opt_journal_async_commit, "journal_async_commit"},
1739 {Opt_abort, "abort"},
1740 {Opt_data_journal, "data=journal"},
1741 {Opt_data_ordered, "data=ordered"},
1742 {Opt_data_writeback, "data=writeback"},
1743 {Opt_data_err_abort, "data_err=abort"},
1744 {Opt_data_err_ignore, "data_err=ignore"},
1745 {Opt_offusrjquota, "usrjquota="},
1746 {Opt_usrjquota, "usrjquota=%s"},
1747 {Opt_offgrpjquota, "grpjquota="},
1748 {Opt_grpjquota, "grpjquota=%s"},
1749 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1750 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1751 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1752 {Opt_grpquota, "grpquota"},
1753 {Opt_noquota, "noquota"},
1754 {Opt_quota, "quota"},
1755 {Opt_usrquota, "usrquota"},
1756 {Opt_prjquota, "prjquota"},
1757 {Opt_barrier, "barrier=%u"},
1758 {Opt_barrier, "barrier"},
1759 {Opt_nobarrier, "nobarrier"},
1760 {Opt_i_version, "i_version"},
1762 {Opt_dax_always, "dax=always"},
1763 {Opt_dax_inode, "dax=inode"},
1764 {Opt_dax_never, "dax=never"},
1765 {Opt_stripe, "stripe=%u"},
1766 {Opt_delalloc, "delalloc"},
1767 {Opt_warn_on_error, "warn_on_error"},
1768 {Opt_nowarn_on_error, "nowarn_on_error"},
1769 {Opt_lazytime, "lazytime"},
1770 {Opt_nolazytime, "nolazytime"},
1771 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1772 {Opt_nodelalloc, "nodelalloc"},
1773 {Opt_removed, "mblk_io_submit"},
1774 {Opt_removed, "nomblk_io_submit"},
1775 {Opt_block_validity, "block_validity"},
1776 {Opt_noblock_validity, "noblock_validity"},
1777 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1778 {Opt_journal_ioprio, "journal_ioprio=%u"},
1779 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1780 {Opt_auto_da_alloc, "auto_da_alloc"},
1781 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1782 {Opt_dioread_nolock, "dioread_nolock"},
1783 {Opt_dioread_lock, "nodioread_nolock"},
1784 {Opt_dioread_lock, "dioread_lock"},
1785 {Opt_discard, "discard"},
1786 {Opt_nodiscard, "nodiscard"},
1787 {Opt_init_itable, "init_itable=%u"},
1788 {Opt_init_itable, "init_itable"},
1789 {Opt_noinit_itable, "noinit_itable"},
1790 #ifdef CONFIG_EXT4_DEBUG
1791 {Opt_fc_debug_force, "fc_debug_force"},
1792 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1794 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1795 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1796 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1797 {Opt_inlinecrypt, "inlinecrypt"},
1798 {Opt_nombcache, "nombcache"},
1799 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1800 {Opt_removed, "prefetch_block_bitmaps"},
1801 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1802 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1803 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1804 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1805 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1806 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1807 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1811 static ext4_fsblk_t get_sb_block(void **data)
1813 ext4_fsblk_t sb_block;
1814 char *options = (char *) *data;
1816 if (!options || strncmp(options, "sb=", 3) != 0)
1817 return 1; /* Default location */
1820 /* TODO: use simple_strtoll with >32bit ext4 */
1821 sb_block = simple_strtoul(options, &options, 0);
1822 if (*options && *options != ',') {
1823 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1827 if (*options == ',')
1829 *data = (void *) options;
1834 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1835 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1837 static const char deprecated_msg[] =
1838 "Mount option \"%s\" will be removed by %s\n"
1839 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1842 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1844 struct ext4_sb_info *sbi = EXT4_SB(sb);
1845 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1848 if (sb_any_quota_loaded(sb) && !old_qname) {
1849 ext4_msg(sb, KERN_ERR,
1850 "Cannot change journaled "
1851 "quota options when quota turned on");
1854 if (ext4_has_feature_quota(sb)) {
1855 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1856 "ignored when QUOTA feature is enabled");
1859 qname = match_strdup(args);
1861 ext4_msg(sb, KERN_ERR,
1862 "Not enough memory for storing quotafile name");
1866 if (strcmp(old_qname, qname) == 0)
1869 ext4_msg(sb, KERN_ERR,
1870 "%s quota file already specified",
1874 if (strchr(qname, '/')) {
1875 ext4_msg(sb, KERN_ERR,
1876 "quotafile must be on filesystem root");
1879 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1887 static int clear_qf_name(struct super_block *sb, int qtype)
1890 struct ext4_sb_info *sbi = EXT4_SB(sb);
1891 char *old_qname = get_qf_name(sb, sbi, qtype);
1893 if (sb_any_quota_loaded(sb) && old_qname) {
1894 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1895 " when quota turned on");
1898 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1905 #define MOPT_SET 0x0001
1906 #define MOPT_CLEAR 0x0002
1907 #define MOPT_NOSUPPORT 0x0004
1908 #define MOPT_EXPLICIT 0x0008
1909 #define MOPT_CLEAR_ERR 0x0010
1910 #define MOPT_GTE0 0x0020
1913 #define MOPT_QFMT 0x0040
1915 #define MOPT_Q MOPT_NOSUPPORT
1916 #define MOPT_QFMT MOPT_NOSUPPORT
1918 #define MOPT_DATAJ 0x0080
1919 #define MOPT_NO_EXT2 0x0100
1920 #define MOPT_NO_EXT3 0x0200
1921 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1922 #define MOPT_STRING 0x0400
1923 #define MOPT_SKIP 0x0800
1924 #define MOPT_2 0x1000
1926 static const struct mount_opts {
1930 } ext4_mount_opts[] = {
1931 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1932 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1933 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1934 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1935 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1936 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1937 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1938 MOPT_EXT4_ONLY | MOPT_SET},
1939 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1940 MOPT_EXT4_ONLY | MOPT_CLEAR},
1941 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1942 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1943 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1944 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1945 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1946 MOPT_EXT4_ONLY | MOPT_CLEAR},
1947 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1948 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1949 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1950 MOPT_EXT4_ONLY | MOPT_CLEAR},
1951 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1952 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1953 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1954 EXT4_MOUNT_JOURNAL_CHECKSUM),
1955 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1956 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1957 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1958 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1959 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1960 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1962 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1964 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1965 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1966 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1967 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1968 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1969 {Opt_commit, 0, MOPT_GTE0},
1970 {Opt_max_batch_time, 0, MOPT_GTE0},
1971 {Opt_min_batch_time, 0, MOPT_GTE0},
1972 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1973 {Opt_init_itable, 0, MOPT_GTE0},
1974 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1975 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1976 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1977 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1978 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1979 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1980 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1981 {Opt_stripe, 0, MOPT_GTE0},
1982 {Opt_resuid, 0, MOPT_GTE0},
1983 {Opt_resgid, 0, MOPT_GTE0},
1984 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1985 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1986 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1987 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1988 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1989 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1990 MOPT_NO_EXT2 | MOPT_DATAJ},
1991 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1992 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1993 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1994 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1995 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1997 {Opt_acl, 0, MOPT_NOSUPPORT},
1998 {Opt_noacl, 0, MOPT_NOSUPPORT},
2000 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
2001 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
2002 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
2003 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2004 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2006 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2008 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2010 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2011 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2012 MOPT_CLEAR | MOPT_Q},
2013 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2014 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
2015 {Opt_offusrjquota, 0, MOPT_Q},
2016 {Opt_offgrpjquota, 0, MOPT_Q},
2017 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2018 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2019 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2020 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2021 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2022 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2023 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2025 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2026 #ifdef CONFIG_EXT4_DEBUG
2027 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2028 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2029 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2034 #ifdef CONFIG_UNICODE
2035 static const struct ext4_sb_encodings {
2039 } ext4_sb_encoding_map[] = {
2040 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2043 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2044 const struct ext4_sb_encodings **encoding,
2047 __u16 magic = le16_to_cpu(es->s_encoding);
2050 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2051 if (magic == ext4_sb_encoding_map[i].magic)
2054 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2057 *encoding = &ext4_sb_encoding_map[i];
2058 *flags = le16_to_cpu(es->s_encoding_flags);
2064 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2066 const substring_t *arg,
2069 #ifdef CONFIG_FS_ENCRYPTION
2070 struct ext4_sb_info *sbi = EXT4_SB(sb);
2074 * This mount option is just for testing, and it's not worthwhile to
2075 * implement the extra complexity (e.g. RCU protection) that would be
2076 * needed to allow it to be set or changed during remount. We do allow
2077 * it to be specified during remount, but only if there is no change.
2079 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2080 ext4_msg(sb, KERN_WARNING,
2081 "Can't set test_dummy_encryption on remount");
2084 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2085 &sbi->s_dummy_enc_policy);
2088 ext4_msg(sb, KERN_WARNING,
2089 "Can't change test_dummy_encryption on remount");
2090 else if (err == -EINVAL)
2091 ext4_msg(sb, KERN_WARNING,
2092 "Value of option \"%s\" is unrecognized", opt);
2094 ext4_msg(sb, KERN_WARNING,
2095 "Error processing option \"%s\" [%d]",
2099 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2101 ext4_msg(sb, KERN_WARNING,
2102 "Test dummy encryption mount option ignored");
2107 struct ext4_parsed_options {
2108 unsigned long journal_devnum;
2109 unsigned int journal_ioprio;
2110 int mb_optimize_scan;
2113 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2114 substring_t *args, struct ext4_parsed_options *parsed_opts,
2117 struct ext4_sb_info *sbi = EXT4_SB(sb);
2118 const struct mount_opts *m;
2124 if (token == Opt_usrjquota)
2125 return set_qf_name(sb, USRQUOTA, &args[0]);
2126 else if (token == Opt_grpjquota)
2127 return set_qf_name(sb, GRPQUOTA, &args[0]);
2128 else if (token == Opt_offusrjquota)
2129 return clear_qf_name(sb, USRQUOTA);
2130 else if (token == Opt_offgrpjquota)
2131 return clear_qf_name(sb, GRPQUOTA);
2135 case Opt_nouser_xattr:
2136 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2139 return 1; /* handled by get_sb_block() */
2141 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2144 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2147 sb->s_flags |= SB_I_VERSION;
2150 sb->s_flags |= SB_LAZYTIME;
2152 case Opt_nolazytime:
2153 sb->s_flags &= ~SB_LAZYTIME;
2155 case Opt_inlinecrypt:
2156 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2157 sb->s_flags |= SB_INLINECRYPT;
2159 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2164 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2165 if (token == m->token)
2168 if (m->token == Opt_err) {
2169 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2170 "or missing value", opt);
2174 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2175 ext4_msg(sb, KERN_ERR,
2176 "Mount option \"%s\" incompatible with ext2", opt);
2179 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2180 ext4_msg(sb, KERN_ERR,
2181 "Mount option \"%s\" incompatible with ext3", opt);
2185 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2187 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2189 if (m->flags & MOPT_EXPLICIT) {
2190 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2191 set_opt2(sb, EXPLICIT_DELALLOC);
2192 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2193 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2197 if (m->flags & MOPT_CLEAR_ERR)
2198 clear_opt(sb, ERRORS_MASK);
2199 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2200 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2201 "options when quota turned on");
2205 if (m->flags & MOPT_NOSUPPORT) {
2206 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2207 } else if (token == Opt_commit) {
2209 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2210 else if (arg > INT_MAX / HZ) {
2211 ext4_msg(sb, KERN_ERR,
2212 "Invalid commit interval %d, "
2213 "must be smaller than %d",
2217 sbi->s_commit_interval = HZ * arg;
2218 } else if (token == Opt_debug_want_extra_isize) {
2221 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2222 ext4_msg(sb, KERN_ERR,
2223 "Invalid want_extra_isize %d", arg);
2226 sbi->s_want_extra_isize = arg;
2227 } else if (token == Opt_max_batch_time) {
2228 sbi->s_max_batch_time = arg;
2229 } else if (token == Opt_min_batch_time) {
2230 sbi->s_min_batch_time = arg;
2231 } else if (token == Opt_inode_readahead_blks) {
2232 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2233 ext4_msg(sb, KERN_ERR,
2234 "EXT4-fs: inode_readahead_blks must be "
2235 "0 or a power of 2 smaller than 2^31");
2238 sbi->s_inode_readahead_blks = arg;
2239 } else if (token == Opt_init_itable) {
2240 set_opt(sb, INIT_INODE_TABLE);
2242 arg = EXT4_DEF_LI_WAIT_MULT;
2243 sbi->s_li_wait_mult = arg;
2244 } else if (token == Opt_max_dir_size_kb) {
2245 sbi->s_max_dir_size_kb = arg;
2246 #ifdef CONFIG_EXT4_DEBUG
2247 } else if (token == Opt_fc_debug_max_replay) {
2248 sbi->s_fc_debug_max_replay = arg;
2250 } else if (token == Opt_stripe) {
2251 sbi->s_stripe = arg;
2252 } else if (token == Opt_resuid) {
2253 uid = make_kuid(current_user_ns(), arg);
2254 if (!uid_valid(uid)) {
2255 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2258 sbi->s_resuid = uid;
2259 } else if (token == Opt_resgid) {
2260 gid = make_kgid(current_user_ns(), arg);
2261 if (!gid_valid(gid)) {
2262 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2265 sbi->s_resgid = gid;
2266 } else if (token == Opt_journal_dev) {
2268 ext4_msg(sb, KERN_ERR,
2269 "Cannot specify journal on remount");
2272 parsed_opts->journal_devnum = arg;
2273 } else if (token == Opt_journal_path) {
2275 struct inode *journal_inode;
2280 ext4_msg(sb, KERN_ERR,
2281 "Cannot specify journal on remount");
2284 journal_path = match_strdup(&args[0]);
2285 if (!journal_path) {
2286 ext4_msg(sb, KERN_ERR, "error: could not dup "
2287 "journal device string");
2291 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2293 ext4_msg(sb, KERN_ERR, "error: could not find "
2294 "journal device path: error %d", error);
2295 kfree(journal_path);
2299 journal_inode = d_inode(path.dentry);
2300 if (!S_ISBLK(journal_inode->i_mode)) {
2301 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2302 "is not a block device", journal_path);
2304 kfree(journal_path);
2308 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2310 kfree(journal_path);
2311 } else if (token == Opt_journal_ioprio) {
2313 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2317 parsed_opts->journal_ioprio =
2318 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2319 } else if (token == Opt_test_dummy_encryption) {
2320 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2322 } else if (m->flags & MOPT_DATAJ) {
2324 if (!sbi->s_journal)
2325 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2326 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2327 ext4_msg(sb, KERN_ERR,
2328 "Cannot change data mode on remount");
2332 clear_opt(sb, DATA_FLAGS);
2333 sbi->s_mount_opt |= m->mount_opt;
2336 } else if (m->flags & MOPT_QFMT) {
2337 if (sb_any_quota_loaded(sb) &&
2338 sbi->s_jquota_fmt != m->mount_opt) {
2339 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2340 "quota options when quota turned on");
2343 if (ext4_has_feature_quota(sb)) {
2344 ext4_msg(sb, KERN_INFO,
2345 "Quota format mount options ignored "
2346 "when QUOTA feature is enabled");
2349 sbi->s_jquota_fmt = m->mount_opt;
2351 } else if (token == Opt_dax || token == Opt_dax_always ||
2352 token == Opt_dax_inode || token == Opt_dax_never) {
2353 #ifdef CONFIG_FS_DAX
2356 case Opt_dax_always:
2358 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2359 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2360 fail_dax_change_remount:
2361 ext4_msg(sb, KERN_ERR, "can't change "
2362 "dax mount option while remounting");
2366 (test_opt(sb, DATA_FLAGS) ==
2367 EXT4_MOUNT_JOURNAL_DATA)) {
2368 ext4_msg(sb, KERN_ERR, "can't mount with "
2369 "both data=journal and dax");
2372 ext4_msg(sb, KERN_WARNING,
2373 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2374 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2375 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2379 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2380 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2381 goto fail_dax_change_remount;
2382 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2383 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2387 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2388 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2389 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2390 goto fail_dax_change_remount;
2391 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2392 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2393 /* Strictly for printing options */
2394 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2398 ext4_msg(sb, KERN_INFO, "dax option not supported");
2399 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2400 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2403 } else if (token == Opt_data_err_abort) {
2404 sbi->s_mount_opt |= m->mount_opt;
2405 } else if (token == Opt_data_err_ignore) {
2406 sbi->s_mount_opt &= ~m->mount_opt;
2407 } else if (token == Opt_mb_optimize_scan) {
2408 if (arg != 0 && arg != 1) {
2409 ext4_msg(sb, KERN_WARNING,
2410 "mb_optimize_scan should be set to 0 or 1.");
2413 parsed_opts->mb_optimize_scan = arg;
2417 if (m->flags & MOPT_CLEAR)
2419 else if (unlikely(!(m->flags & MOPT_SET))) {
2420 ext4_msg(sb, KERN_WARNING,
2421 "buggy handling of option %s", opt);
2425 if (m->flags & MOPT_2) {
2427 sbi->s_mount_opt2 |= m->mount_opt;
2429 sbi->s_mount_opt2 &= ~m->mount_opt;
2432 sbi->s_mount_opt |= m->mount_opt;
2434 sbi->s_mount_opt &= ~m->mount_opt;
2440 static int parse_options(char *options, struct super_block *sb,
2441 struct ext4_parsed_options *ret_opts,
2444 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2445 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2446 substring_t args[MAX_OPT_ARGS];
2452 while ((p = strsep(&options, ",")) != NULL) {
2456 * Initialize args struct so we know whether arg was
2457 * found; some options take optional arguments.
2459 args[0].to = args[0].from = NULL;
2460 token = match_token(p, tokens, args);
2461 if (handle_mount_opt(sb, p, token, args, ret_opts,
2467 * We do the test below only for project quotas. 'usrquota' and
2468 * 'grpquota' mount options are allowed even without quota feature
2469 * to support legacy quotas in quota files.
2471 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2472 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2473 "Cannot enable project quota enforcement.");
2476 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2477 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2478 if (usr_qf_name || grp_qf_name) {
2479 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2480 clear_opt(sb, USRQUOTA);
2482 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2483 clear_opt(sb, GRPQUOTA);
2485 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2486 ext4_msg(sb, KERN_ERR, "old and new quota "
2491 if (!sbi->s_jquota_fmt) {
2492 ext4_msg(sb, KERN_ERR, "journaled quota format "
2498 if (test_opt(sb, DIOREAD_NOLOCK)) {
2500 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2501 if (blocksize < PAGE_SIZE)
2502 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2503 "experimental mount option 'dioread_nolock' "
2504 "for blocksize < PAGE_SIZE");
2509 static inline void ext4_show_quota_options(struct seq_file *seq,
2510 struct super_block *sb)
2512 #if defined(CONFIG_QUOTA)
2513 struct ext4_sb_info *sbi = EXT4_SB(sb);
2514 char *usr_qf_name, *grp_qf_name;
2516 if (sbi->s_jquota_fmt) {
2519 switch (sbi->s_jquota_fmt) {
2530 seq_printf(seq, ",jqfmt=%s", fmtname);
2534 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2535 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2537 seq_show_option(seq, "usrjquota", usr_qf_name);
2539 seq_show_option(seq, "grpjquota", grp_qf_name);
2544 static const char *token2str(int token)
2546 const struct match_token *t;
2548 for (t = tokens; t->token != Opt_err; t++)
2549 if (t->token == token && !strchr(t->pattern, '='))
2556 * - it's set to a non-default value OR
2557 * - if the per-sb default is different from the global default
2559 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2562 struct ext4_sb_info *sbi = EXT4_SB(sb);
2563 struct ext4_super_block *es = sbi->s_es;
2564 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2565 const struct mount_opts *m;
2566 char sep = nodefs ? '\n' : ',';
2568 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2569 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2571 if (sbi->s_sb_block != 1)
2572 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2574 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2575 int want_set = m->flags & MOPT_SET;
2576 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2577 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2579 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2580 continue; /* skip if same as the default */
2582 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2583 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2584 continue; /* select Opt_noFoo vs Opt_Foo */
2585 SEQ_OPTS_PRINT("%s", token2str(m->token));
2588 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2589 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2590 SEQ_OPTS_PRINT("resuid=%u",
2591 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2592 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2593 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2594 SEQ_OPTS_PRINT("resgid=%u",
2595 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2596 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2597 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2598 SEQ_OPTS_PUTS("errors=remount-ro");
2599 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2600 SEQ_OPTS_PUTS("errors=continue");
2601 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2602 SEQ_OPTS_PUTS("errors=panic");
2603 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2604 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2605 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2606 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2607 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2608 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2609 if (sb->s_flags & SB_I_VERSION)
2610 SEQ_OPTS_PUTS("i_version");
2611 if (nodefs || sbi->s_stripe)
2612 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2613 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2614 (sbi->s_mount_opt ^ def_mount_opt)) {
2615 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2616 SEQ_OPTS_PUTS("data=journal");
2617 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2618 SEQ_OPTS_PUTS("data=ordered");
2619 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2620 SEQ_OPTS_PUTS("data=writeback");
2623 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2624 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2625 sbi->s_inode_readahead_blks);
2627 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2628 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2629 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2630 if (nodefs || sbi->s_max_dir_size_kb)
2631 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2632 if (test_opt(sb, DATA_ERR_ABORT))
2633 SEQ_OPTS_PUTS("data_err=abort");
2635 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2637 if (sb->s_flags & SB_INLINECRYPT)
2638 SEQ_OPTS_PUTS("inlinecrypt");
2640 if (test_opt(sb, DAX_ALWAYS)) {
2642 SEQ_OPTS_PUTS("dax");
2644 SEQ_OPTS_PUTS("dax=always");
2645 } else if (test_opt2(sb, DAX_NEVER)) {
2646 SEQ_OPTS_PUTS("dax=never");
2647 } else if (test_opt2(sb, DAX_INODE)) {
2648 SEQ_OPTS_PUTS("dax=inode");
2650 ext4_show_quota_options(seq, sb);
2654 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2656 return _ext4_show_options(seq, root->d_sb, 0);
2659 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2661 struct super_block *sb = seq->private;
2664 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2665 rc = _ext4_show_options(seq, sb, 1);
2666 seq_puts(seq, "\n");
2670 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2673 struct ext4_sb_info *sbi = EXT4_SB(sb);
2676 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2677 ext4_msg(sb, KERN_ERR, "revision level too high, "
2678 "forcing read-only mode");
2684 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2685 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2686 "running e2fsck is recommended");
2687 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2688 ext4_msg(sb, KERN_WARNING,
2689 "warning: mounting fs with errors, "
2690 "running e2fsck is recommended");
2691 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2692 le16_to_cpu(es->s_mnt_count) >=
2693 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2694 ext4_msg(sb, KERN_WARNING,
2695 "warning: maximal mount count reached, "
2696 "running e2fsck is recommended");
2697 else if (le32_to_cpu(es->s_checkinterval) &&
2698 (ext4_get_tstamp(es, s_lastcheck) +
2699 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2700 ext4_msg(sb, KERN_WARNING,
2701 "warning: checktime reached, "
2702 "running e2fsck is recommended");
2703 if (!sbi->s_journal)
2704 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2705 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2706 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2707 le16_add_cpu(&es->s_mnt_count, 1);
2708 ext4_update_tstamp(es, s_mtime);
2710 ext4_set_feature_journal_needs_recovery(sb);
2712 err = ext4_commit_super(sb);
2714 if (test_opt(sb, DEBUG))
2715 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2716 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2718 sbi->s_groups_count,
2719 EXT4_BLOCKS_PER_GROUP(sb),
2720 EXT4_INODES_PER_GROUP(sb),
2721 sbi->s_mount_opt, sbi->s_mount_opt2);
2723 cleancache_init_fs(sb);
2727 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2729 struct ext4_sb_info *sbi = EXT4_SB(sb);
2730 struct flex_groups **old_groups, **new_groups;
2733 if (!sbi->s_log_groups_per_flex)
2736 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2737 if (size <= sbi->s_flex_groups_allocated)
2740 new_groups = kvzalloc(roundup_pow_of_two(size *
2741 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2743 ext4_msg(sb, KERN_ERR,
2744 "not enough memory for %d flex group pointers", size);
2747 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2748 new_groups[i] = kvzalloc(roundup_pow_of_two(
2749 sizeof(struct flex_groups)),
2751 if (!new_groups[i]) {
2752 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2753 kvfree(new_groups[j]);
2755 ext4_msg(sb, KERN_ERR,
2756 "not enough memory for %d flex groups", size);
2761 old_groups = rcu_dereference(sbi->s_flex_groups);
2763 memcpy(new_groups, old_groups,
2764 (sbi->s_flex_groups_allocated *
2765 sizeof(struct flex_groups *)));
2767 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2768 sbi->s_flex_groups_allocated = size;
2770 ext4_kvfree_array_rcu(old_groups);
2774 static int ext4_fill_flex_info(struct super_block *sb)
2776 struct ext4_sb_info *sbi = EXT4_SB(sb);
2777 struct ext4_group_desc *gdp = NULL;
2778 struct flex_groups *fg;
2779 ext4_group_t flex_group;
2782 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2783 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2784 sbi->s_log_groups_per_flex = 0;
2788 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2792 for (i = 0; i < sbi->s_groups_count; i++) {
2793 gdp = ext4_get_group_desc(sb, i, NULL);
2795 flex_group = ext4_flex_group(sbi, i);
2796 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2797 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2798 atomic64_add(ext4_free_group_clusters(sb, gdp),
2799 &fg->free_clusters);
2800 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2808 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2809 struct ext4_group_desc *gdp)
2811 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2813 __le32 le_group = cpu_to_le32(block_group);
2814 struct ext4_sb_info *sbi = EXT4_SB(sb);
2816 if (ext4_has_metadata_csum(sbi->s_sb)) {
2817 /* Use new metadata_csum algorithm */
2819 __u16 dummy_csum = 0;
2821 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2823 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2824 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2825 sizeof(dummy_csum));
2826 offset += sizeof(dummy_csum);
2827 if (offset < sbi->s_desc_size)
2828 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2829 sbi->s_desc_size - offset);
2831 crc = csum32 & 0xFFFF;
2835 /* old crc16 code */
2836 if (!ext4_has_feature_gdt_csum(sb))
2839 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2840 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2841 crc = crc16(crc, (__u8 *)gdp, offset);
2842 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2843 /* for checksum of struct ext4_group_desc do the rest...*/
2844 if (ext4_has_feature_64bit(sb) &&
2845 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2846 crc = crc16(crc, (__u8 *)gdp + offset,
2847 le16_to_cpu(sbi->s_es->s_desc_size) -
2851 return cpu_to_le16(crc);
2854 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2855 struct ext4_group_desc *gdp)
2857 if (ext4_has_group_desc_csum(sb) &&
2858 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2864 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2865 struct ext4_group_desc *gdp)
2867 if (!ext4_has_group_desc_csum(sb))
2869 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2872 /* Called at mount-time, super-block is locked */
2873 static int ext4_check_descriptors(struct super_block *sb,
2874 ext4_fsblk_t sb_block,
2875 ext4_group_t *first_not_zeroed)
2877 struct ext4_sb_info *sbi = EXT4_SB(sb);
2878 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2879 ext4_fsblk_t last_block;
2880 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2881 ext4_fsblk_t block_bitmap;
2882 ext4_fsblk_t inode_bitmap;
2883 ext4_fsblk_t inode_table;
2884 int flexbg_flag = 0;
2885 ext4_group_t i, grp = sbi->s_groups_count;
2887 if (ext4_has_feature_flex_bg(sb))
2890 ext4_debug("Checking group descriptors");
2892 for (i = 0; i < sbi->s_groups_count; i++) {
2893 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2895 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2896 last_block = ext4_blocks_count(sbi->s_es) - 1;
2898 last_block = first_block +
2899 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2901 if ((grp == sbi->s_groups_count) &&
2902 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2905 block_bitmap = ext4_block_bitmap(sb, gdp);
2906 if (block_bitmap == sb_block) {
2907 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908 "Block bitmap for group %u overlaps "
2913 if (block_bitmap >= sb_block + 1 &&
2914 block_bitmap <= last_bg_block) {
2915 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2916 "Block bitmap for group %u overlaps "
2917 "block group descriptors", i);
2921 if (block_bitmap < first_block || block_bitmap > last_block) {
2922 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2923 "Block bitmap for group %u not in group "
2924 "(block %llu)!", i, block_bitmap);
2927 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2928 if (inode_bitmap == sb_block) {
2929 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2930 "Inode bitmap for group %u overlaps "
2935 if (inode_bitmap >= sb_block + 1 &&
2936 inode_bitmap <= last_bg_block) {
2937 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2938 "Inode bitmap for group %u overlaps "
2939 "block group descriptors", i);
2943 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2944 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2945 "Inode bitmap for group %u not in group "
2946 "(block %llu)!", i, inode_bitmap);
2949 inode_table = ext4_inode_table(sb, gdp);
2950 if (inode_table == sb_block) {
2951 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2952 "Inode table for group %u overlaps "
2957 if (inode_table >= sb_block + 1 &&
2958 inode_table <= last_bg_block) {
2959 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2960 "Inode table for group %u overlaps "
2961 "block group descriptors", i);
2965 if (inode_table < first_block ||
2966 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2967 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2968 "Inode table for group %u not in group "
2969 "(block %llu)!", i, inode_table);
2972 ext4_lock_group(sb, i);
2973 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2974 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2975 "Checksum for group %u failed (%u!=%u)",
2976 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2977 gdp)), le16_to_cpu(gdp->bg_checksum));
2978 if (!sb_rdonly(sb)) {
2979 ext4_unlock_group(sb, i);
2983 ext4_unlock_group(sb, i);
2985 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2987 if (NULL != first_not_zeroed)
2988 *first_not_zeroed = grp;
2992 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2993 * the superblock) which were deleted from all directories, but held open by
2994 * a process at the time of a crash. We walk the list and try to delete these
2995 * inodes at recovery time (only with a read-write filesystem).
2997 * In order to keep the orphan inode chain consistent during traversal (in
2998 * case of crash during recovery), we link each inode into the superblock
2999 * orphan list_head and handle it the same way as an inode deletion during
3000 * normal operation (which journals the operations for us).
3002 * We only do an iget() and an iput() on each inode, which is very safe if we
3003 * accidentally point at an in-use or already deleted inode. The worst that
3004 * can happen in this case is that we get a "bit already cleared" message from
3005 * ext4_free_inode(). The only reason we would point at a wrong inode is if
3006 * e2fsck was run on this filesystem, and it must have already done the orphan
3007 * inode cleanup for us, so we can safely abort without any further action.
3009 static void ext4_orphan_cleanup(struct super_block *sb,
3010 struct ext4_super_block *es)
3012 unsigned int s_flags = sb->s_flags;
3013 int ret, nr_orphans = 0, nr_truncates = 0;
3015 int quota_update = 0;
3018 if (!es->s_last_orphan) {
3019 jbd_debug(4, "no orphan inodes to clean up\n");
3023 if (bdev_read_only(sb->s_bdev)) {
3024 ext4_msg(sb, KERN_ERR, "write access "
3025 "unavailable, skipping orphan cleanup");
3029 /* Check if feature set would not allow a r/w mount */
3030 if (!ext4_feature_set_ok(sb, 0)) {
3031 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3032 "unknown ROCOMPAT features");
3036 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3037 /* don't clear list on RO mount w/ errors */
3038 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
3039 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
3040 "clearing orphan list.\n");
3041 es->s_last_orphan = 0;
3043 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3047 if (s_flags & SB_RDONLY) {
3048 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
3049 sb->s_flags &= ~SB_RDONLY;
3053 * Turn on quotas which were not enabled for read-only mounts if
3054 * filesystem has quota feature, so that they are updated correctly.
3056 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3057 int ret = ext4_enable_quotas(sb);
3062 ext4_msg(sb, KERN_ERR,
3063 "Cannot turn on quotas: error %d", ret);
3066 /* Turn on journaled quotas used for old sytle */
3067 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3068 if (EXT4_SB(sb)->s_qf_names[i]) {
3069 int ret = ext4_quota_on_mount(sb, i);
3074 ext4_msg(sb, KERN_ERR,
3075 "Cannot turn on journaled "
3076 "quota: type %d: error %d", i, ret);
3081 while (es->s_last_orphan) {
3082 struct inode *inode;
3085 * We may have encountered an error during cleanup; if
3086 * so, skip the rest.
3088 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3089 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3090 es->s_last_orphan = 0;
3094 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3095 if (IS_ERR(inode)) {
3096 es->s_last_orphan = 0;
3100 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3101 dquot_initialize(inode);
3102 if (inode->i_nlink) {
3103 if (test_opt(sb, DEBUG))
3104 ext4_msg(sb, KERN_DEBUG,
3105 "%s: truncating inode %lu to %lld bytes",
3106 __func__, inode->i_ino, inode->i_size);
3107 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3108 inode->i_ino, inode->i_size);
3110 truncate_inode_pages(inode->i_mapping, inode->i_size);
3111 ret = ext4_truncate(inode);
3114 * We need to clean up the in-core orphan list
3115 * manually if ext4_truncate() failed to get a
3116 * transaction handle.
3118 ext4_orphan_del(NULL, inode);
3119 ext4_std_error(inode->i_sb, ret);
3121 inode_unlock(inode);
3124 if (test_opt(sb, DEBUG))
3125 ext4_msg(sb, KERN_DEBUG,
3126 "%s: deleting unreferenced inode %lu",
3127 __func__, inode->i_ino);
3128 jbd_debug(2, "deleting unreferenced inode %lu\n",
3132 iput(inode); /* The delete magic happens here! */
3135 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3138 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3139 PLURAL(nr_orphans));
3141 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3142 PLURAL(nr_truncates));
3144 /* Turn off quotas if they were enabled for orphan cleanup */
3146 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3147 if (sb_dqopt(sb)->files[i])
3148 dquot_quota_off(sb, i);
3152 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3156 * Maximal extent format file size.
3157 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3158 * extent format containers, within a sector_t, and within i_blocks
3159 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3160 * so that won't be a limiting factor.
3162 * However there is other limiting factor. We do store extents in the form
3163 * of starting block and length, hence the resulting length of the extent
3164 * covering maximum file size must fit into on-disk format containers as
3165 * well. Given that length is always by 1 unit bigger than max unit (because
3166 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3168 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3170 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3173 loff_t upper_limit = MAX_LFS_FILESIZE;
3175 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3177 if (!has_huge_files) {
3178 upper_limit = (1LL << 32) - 1;
3180 /* total blocks in file system block size */
3181 upper_limit >>= (blkbits - 9);
3182 upper_limit <<= blkbits;
3186 * 32-bit extent-start container, ee_block. We lower the maxbytes
3187 * by one fs block, so ee_len can cover the extent of maximum file
3190 res = (1LL << 32) - 1;
3193 /* Sanity check against vm- & vfs- imposed limits */
3194 if (res > upper_limit)
3201 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3202 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3203 * We need to be 1 filesystem block less than the 2^48 sector limit.
3205 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3207 loff_t res = EXT4_NDIR_BLOCKS;
3210 /* This is calculated to be the largest file size for a dense, block
3211 * mapped file such that the file's total number of 512-byte sectors,
3212 * including data and all indirect blocks, does not exceed (2^48 - 1).
3214 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3215 * number of 512-byte sectors of the file.
3218 if (!has_huge_files) {
3220 * !has_huge_files or implies that the inode i_block field
3221 * represents total file blocks in 2^32 512-byte sectors ==
3222 * size of vfs inode i_blocks * 8
3224 upper_limit = (1LL << 32) - 1;
3226 /* total blocks in file system block size */
3227 upper_limit >>= (bits - 9);
3231 * We use 48 bit ext4_inode i_blocks
3232 * With EXT4_HUGE_FILE_FL set the i_blocks
3233 * represent total number of blocks in
3234 * file system block size
3236 upper_limit = (1LL << 48) - 1;
3240 /* indirect blocks */
3242 /* double indirect blocks */
3243 meta_blocks += 1 + (1LL << (bits-2));
3244 /* tripple indirect blocks */
3245 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3247 upper_limit -= meta_blocks;
3248 upper_limit <<= bits;
3250 res += 1LL << (bits-2);
3251 res += 1LL << (2*(bits-2));
3252 res += 1LL << (3*(bits-2));
3254 if (res > upper_limit)
3257 if (res > MAX_LFS_FILESIZE)
3258 res = MAX_LFS_FILESIZE;
3263 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3264 ext4_fsblk_t logical_sb_block, int nr)
3266 struct ext4_sb_info *sbi = EXT4_SB(sb);
3267 ext4_group_t bg, first_meta_bg;
3270 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3272 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3273 return logical_sb_block + nr + 1;
3274 bg = sbi->s_desc_per_block * nr;
3275 if (ext4_bg_has_super(sb, bg))
3279 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3280 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3281 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3284 if (sb->s_blocksize == 1024 && nr == 0 &&
3285 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3288 return (has_super + ext4_group_first_block_no(sb, bg));
3292 * ext4_get_stripe_size: Get the stripe size.
3293 * @sbi: In memory super block info
3295 * If we have specified it via mount option, then
3296 * use the mount option value. If the value specified at mount time is
3297 * greater than the blocks per group use the super block value.
3298 * If the super block value is greater than blocks per group return 0.
3299 * Allocator needs it be less than blocks per group.
3302 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3304 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3305 unsigned long stripe_width =
3306 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3309 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3310 ret = sbi->s_stripe;
3311 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3313 else if (stride && stride <= sbi->s_blocks_per_group)
3319 * If the stripe width is 1, this makes no sense and
3320 * we set it to 0 to turn off stripe handling code.
3329 * Check whether this filesystem can be mounted based on
3330 * the features present and the RDONLY/RDWR mount requested.
3331 * Returns 1 if this filesystem can be mounted as requested,
3332 * 0 if it cannot be.
3334 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3336 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3337 ext4_msg(sb, KERN_ERR,
3338 "Couldn't mount because of "
3339 "unsupported optional features (%x)",
3340 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3341 ~EXT4_FEATURE_INCOMPAT_SUPP));
3345 #ifndef CONFIG_UNICODE
3346 if (ext4_has_feature_casefold(sb)) {
3347 ext4_msg(sb, KERN_ERR,
3348 "Filesystem with casefold feature cannot be "
3349 "mounted without CONFIG_UNICODE");
3357 if (ext4_has_feature_readonly(sb)) {
3358 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3359 sb->s_flags |= SB_RDONLY;
3363 /* Check that feature set is OK for a read-write mount */
3364 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3365 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3366 "unsupported optional features (%x)",
3367 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3368 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3371 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3372 ext4_msg(sb, KERN_ERR,
3373 "Can't support bigalloc feature without "
3374 "extents feature\n");
3378 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3379 if (!readonly && (ext4_has_feature_quota(sb) ||
3380 ext4_has_feature_project(sb))) {
3381 ext4_msg(sb, KERN_ERR,
3382 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3385 #endif /* CONFIG_QUOTA */
3390 * This function is called once a day if we have errors logged
3391 * on the file system
3393 static void print_daily_error_info(struct timer_list *t)
3395 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3396 struct super_block *sb = sbi->s_sb;
3397 struct ext4_super_block *es = sbi->s_es;
3399 if (es->s_error_count)
3400 /* fsck newer than v1.41.13 is needed to clean this condition. */
3401 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3402 le32_to_cpu(es->s_error_count));
3403 if (es->s_first_error_time) {
3404 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3406 ext4_get_tstamp(es, s_first_error_time),
3407 (int) sizeof(es->s_first_error_func),
3408 es->s_first_error_func,
3409 le32_to_cpu(es->s_first_error_line));
3410 if (es->s_first_error_ino)
3411 printk(KERN_CONT ": inode %u",
3412 le32_to_cpu(es->s_first_error_ino));
3413 if (es->s_first_error_block)
3414 printk(KERN_CONT ": block %llu", (unsigned long long)
3415 le64_to_cpu(es->s_first_error_block));
3416 printk(KERN_CONT "\n");
3418 if (es->s_last_error_time) {
3419 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3421 ext4_get_tstamp(es, s_last_error_time),
3422 (int) sizeof(es->s_last_error_func),
3423 es->s_last_error_func,
3424 le32_to_cpu(es->s_last_error_line));
3425 if (es->s_last_error_ino)
3426 printk(KERN_CONT ": inode %u",
3427 le32_to_cpu(es->s_last_error_ino));
3428 if (es->s_last_error_block)
3429 printk(KERN_CONT ": block %llu", (unsigned long long)
3430 le64_to_cpu(es->s_last_error_block));
3431 printk(KERN_CONT "\n");
3433 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3436 /* Find next suitable group and run ext4_init_inode_table */
3437 static int ext4_run_li_request(struct ext4_li_request *elr)
3439 struct ext4_group_desc *gdp = NULL;
3440 struct super_block *sb = elr->lr_super;
3441 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3442 ext4_group_t group = elr->lr_next_group;
3443 unsigned long timeout = 0;
3444 unsigned int prefetch_ios = 0;
3447 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3448 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3449 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3451 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3453 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3455 if (group >= elr->lr_next_group) {
3457 if (elr->lr_first_not_zeroed != ngroups &&
3458 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3459 elr->lr_next_group = elr->lr_first_not_zeroed;
3460 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3467 for (; group < ngroups; group++) {
3468 gdp = ext4_get_group_desc(sb, group, NULL);
3474 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3478 if (group >= ngroups)
3483 ret = ext4_init_inode_table(sb, group,
3484 elr->lr_timeout ? 0 : 1);
3485 trace_ext4_lazy_itable_init(sb, group);
3486 if (elr->lr_timeout == 0) {
3487 timeout = (jiffies - timeout) *
3488 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3489 elr->lr_timeout = timeout;
3491 elr->lr_next_sched = jiffies + elr->lr_timeout;
3492 elr->lr_next_group = group + 1;
3498 * Remove lr_request from the list_request and free the
3499 * request structure. Should be called with li_list_mtx held
3501 static void ext4_remove_li_request(struct ext4_li_request *elr)
3506 list_del(&elr->lr_request);
3507 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3511 static void ext4_unregister_li_request(struct super_block *sb)
3513 mutex_lock(&ext4_li_mtx);
3514 if (!ext4_li_info) {
3515 mutex_unlock(&ext4_li_mtx);
3519 mutex_lock(&ext4_li_info->li_list_mtx);
3520 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3521 mutex_unlock(&ext4_li_info->li_list_mtx);
3522 mutex_unlock(&ext4_li_mtx);
3525 static struct task_struct *ext4_lazyinit_task;
3528 * This is the function where ext4lazyinit thread lives. It walks
3529 * through the request list searching for next scheduled filesystem.
3530 * When such a fs is found, run the lazy initialization request
3531 * (ext4_rn_li_request) and keep track of the time spend in this
3532 * function. Based on that time we compute next schedule time of
3533 * the request. When walking through the list is complete, compute
3534 * next waking time and put itself into sleep.
3536 static int ext4_lazyinit_thread(void *arg)
3538 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3539 struct list_head *pos, *n;
3540 struct ext4_li_request *elr;
3541 unsigned long next_wakeup, cur;
3543 BUG_ON(NULL == eli);
3547 next_wakeup = MAX_JIFFY_OFFSET;
3549 mutex_lock(&eli->li_list_mtx);
3550 if (list_empty(&eli->li_request_list)) {
3551 mutex_unlock(&eli->li_list_mtx);
3554 list_for_each_safe(pos, n, &eli->li_request_list) {
3557 elr = list_entry(pos, struct ext4_li_request,
3560 if (time_before(jiffies, elr->lr_next_sched)) {
3561 if (time_before(elr->lr_next_sched, next_wakeup))
3562 next_wakeup = elr->lr_next_sched;
3565 if (down_read_trylock(&elr->lr_super->s_umount)) {
3566 if (sb_start_write_trylock(elr->lr_super)) {
3569 * We hold sb->s_umount, sb can not
3570 * be removed from the list, it is
3571 * now safe to drop li_list_mtx
3573 mutex_unlock(&eli->li_list_mtx);
3574 err = ext4_run_li_request(elr);
3575 sb_end_write(elr->lr_super);
3576 mutex_lock(&eli->li_list_mtx);
3579 up_read((&elr->lr_super->s_umount));
3581 /* error, remove the lazy_init job */
3583 ext4_remove_li_request(elr);
3587 elr->lr_next_sched = jiffies +
3589 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3591 if (time_before(elr->lr_next_sched, next_wakeup))
3592 next_wakeup = elr->lr_next_sched;
3594 mutex_unlock(&eli->li_list_mtx);
3599 if ((time_after_eq(cur, next_wakeup)) ||
3600 (MAX_JIFFY_OFFSET == next_wakeup)) {
3605 schedule_timeout_interruptible(next_wakeup - cur);
3607 if (kthread_should_stop()) {
3608 ext4_clear_request_list();
3615 * It looks like the request list is empty, but we need
3616 * to check it under the li_list_mtx lock, to prevent any
3617 * additions into it, and of course we should lock ext4_li_mtx
3618 * to atomically free the list and ext4_li_info, because at
3619 * this point another ext4 filesystem could be registering
3622 mutex_lock(&ext4_li_mtx);
3623 mutex_lock(&eli->li_list_mtx);
3624 if (!list_empty(&eli->li_request_list)) {
3625 mutex_unlock(&eli->li_list_mtx);
3626 mutex_unlock(&ext4_li_mtx);
3629 mutex_unlock(&eli->li_list_mtx);
3630 kfree(ext4_li_info);
3631 ext4_li_info = NULL;
3632 mutex_unlock(&ext4_li_mtx);
3637 static void ext4_clear_request_list(void)
3639 struct list_head *pos, *n;
3640 struct ext4_li_request *elr;
3642 mutex_lock(&ext4_li_info->li_list_mtx);
3643 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3644 elr = list_entry(pos, struct ext4_li_request,
3646 ext4_remove_li_request(elr);
3648 mutex_unlock(&ext4_li_info->li_list_mtx);
3651 static int ext4_run_lazyinit_thread(void)
3653 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3654 ext4_li_info, "ext4lazyinit");
3655 if (IS_ERR(ext4_lazyinit_task)) {
3656 int err = PTR_ERR(ext4_lazyinit_task);
3657 ext4_clear_request_list();
3658 kfree(ext4_li_info);
3659 ext4_li_info = NULL;
3660 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3661 "initialization thread\n",
3665 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3670 * Check whether it make sense to run itable init. thread or not.
3671 * If there is at least one uninitialized inode table, return
3672 * corresponding group number, else the loop goes through all
3673 * groups and return total number of groups.
3675 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3677 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3678 struct ext4_group_desc *gdp = NULL;
3680 if (!ext4_has_group_desc_csum(sb))
3683 for (group = 0; group < ngroups; group++) {
3684 gdp = ext4_get_group_desc(sb, group, NULL);
3688 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3695 static int ext4_li_info_new(void)
3697 struct ext4_lazy_init *eli = NULL;
3699 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3703 INIT_LIST_HEAD(&eli->li_request_list);
3704 mutex_init(&eli->li_list_mtx);
3706 eli->li_state |= EXT4_LAZYINIT_QUIT;
3713 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3716 struct ext4_li_request *elr;
3718 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3723 elr->lr_first_not_zeroed = start;
3724 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3725 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3726 elr->lr_next_group = start;
3728 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3732 * Randomize first schedule time of the request to
3733 * spread the inode table initialization requests
3736 elr->lr_next_sched = jiffies + (prandom_u32() %
3737 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3741 int ext4_register_li_request(struct super_block *sb,
3742 ext4_group_t first_not_zeroed)
3744 struct ext4_sb_info *sbi = EXT4_SB(sb);
3745 struct ext4_li_request *elr = NULL;
3746 ext4_group_t ngroups = sbi->s_groups_count;
3749 mutex_lock(&ext4_li_mtx);
3750 if (sbi->s_li_request != NULL) {
3752 * Reset timeout so it can be computed again, because
3753 * s_li_wait_mult might have changed.
3755 sbi->s_li_request->lr_timeout = 0;
3759 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3760 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3761 !test_opt(sb, INIT_INODE_TABLE)))
3764 elr = ext4_li_request_new(sb, first_not_zeroed);
3770 if (NULL == ext4_li_info) {
3771 ret = ext4_li_info_new();
3776 mutex_lock(&ext4_li_info->li_list_mtx);
3777 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3778 mutex_unlock(&ext4_li_info->li_list_mtx);
3780 sbi->s_li_request = elr;
3782 * set elr to NULL here since it has been inserted to
3783 * the request_list and the removal and free of it is
3784 * handled by ext4_clear_request_list from now on.
3788 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3789 ret = ext4_run_lazyinit_thread();
3794 mutex_unlock(&ext4_li_mtx);
3801 * We do not need to lock anything since this is called on
3804 static void ext4_destroy_lazyinit_thread(void)
3807 * If thread exited earlier
3808 * there's nothing to be done.
3810 if (!ext4_li_info || !ext4_lazyinit_task)
3813 kthread_stop(ext4_lazyinit_task);
3816 static int set_journal_csum_feature_set(struct super_block *sb)
3819 int compat, incompat;
3820 struct ext4_sb_info *sbi = EXT4_SB(sb);
3822 if (ext4_has_metadata_csum(sb)) {
3823 /* journal checksum v3 */
3825 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3827 /* journal checksum v1 */
3828 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3832 jbd2_journal_clear_features(sbi->s_journal,
3833 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3834 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3835 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3836 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3837 ret = jbd2_journal_set_features(sbi->s_journal,
3839 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3841 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3842 ret = jbd2_journal_set_features(sbi->s_journal,
3845 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3846 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3848 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3849 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3856 * Note: calculating the overhead so we can be compatible with
3857 * historical BSD practice is quite difficult in the face of
3858 * clusters/bigalloc. This is because multiple metadata blocks from
3859 * different block group can end up in the same allocation cluster.
3860 * Calculating the exact overhead in the face of clustered allocation
3861 * requires either O(all block bitmaps) in memory or O(number of block
3862 * groups**2) in time. We will still calculate the superblock for
3863 * older file systems --- and if we come across with a bigalloc file
3864 * system with zero in s_overhead_clusters the estimate will be close to
3865 * correct especially for very large cluster sizes --- but for newer
3866 * file systems, it's better to calculate this figure once at mkfs
3867 * time, and store it in the superblock. If the superblock value is
3868 * present (even for non-bigalloc file systems), we will use it.
3870 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3873 struct ext4_sb_info *sbi = EXT4_SB(sb);
3874 struct ext4_group_desc *gdp;
3875 ext4_fsblk_t first_block, last_block, b;
3876 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3877 int s, j, count = 0;
3879 if (!ext4_has_feature_bigalloc(sb))
3880 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3881 sbi->s_itb_per_group + 2);
3883 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3884 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3885 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3886 for (i = 0; i < ngroups; i++) {
3887 gdp = ext4_get_group_desc(sb, i, NULL);
3888 b = ext4_block_bitmap(sb, gdp);
3889 if (b >= first_block && b <= last_block) {
3890 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3893 b = ext4_inode_bitmap(sb, gdp);
3894 if (b >= first_block && b <= last_block) {
3895 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3898 b = ext4_inode_table(sb, gdp);
3899 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3900 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3901 int c = EXT4_B2C(sbi, b - first_block);
3902 ext4_set_bit(c, buf);
3908 if (ext4_bg_has_super(sb, grp)) {
3909 ext4_set_bit(s++, buf);
3912 j = ext4_bg_num_gdb(sb, grp);
3913 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3914 ext4_error(sb, "Invalid number of block group "
3915 "descriptor blocks: %d", j);
3916 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3920 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3924 return EXT4_CLUSTERS_PER_GROUP(sb) -
3925 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3929 * Compute the overhead and stash it in sbi->s_overhead
3931 int ext4_calculate_overhead(struct super_block *sb)
3933 struct ext4_sb_info *sbi = EXT4_SB(sb);
3934 struct ext4_super_block *es = sbi->s_es;
3935 struct inode *j_inode;
3936 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3937 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3938 ext4_fsblk_t overhead = 0;
3939 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3945 * Compute the overhead (FS structures). This is constant
3946 * for a given filesystem unless the number of block groups
3947 * changes so we cache the previous value until it does.
3951 * All of the blocks before first_data_block are overhead
3953 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3956 * Add the overhead found in each block group
3958 for (i = 0; i < ngroups; i++) {
3961 blks = count_overhead(sb, i, buf);
3964 memset(buf, 0, PAGE_SIZE);
3969 * Add the internal journal blocks whether the journal has been
3972 if (sbi->s_journal && !sbi->s_journal_bdev)
3973 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3974 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3975 /* j_inum for internal journal is non-zero */
3976 j_inode = ext4_get_journal_inode(sb, j_inum);
3978 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3979 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3982 ext4_msg(sb, KERN_ERR, "can't get journal size");
3985 sbi->s_overhead = overhead;
3987 free_page((unsigned long) buf);
3991 static void ext4_set_resv_clusters(struct super_block *sb)
3993 ext4_fsblk_t resv_clusters;
3994 struct ext4_sb_info *sbi = EXT4_SB(sb);
3997 * There's no need to reserve anything when we aren't using extents.
3998 * The space estimates are exact, there are no unwritten extents,
3999 * hole punching doesn't need new metadata... This is needed especially
4000 * to keep ext2/3 backward compatibility.
4002 if (!ext4_has_feature_extents(sb))
4005 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4006 * This should cover the situations where we can not afford to run
4007 * out of space like for example punch hole, or converting
4008 * unwritten extents in delalloc path. In most cases such
4009 * allocation would require 1, or 2 blocks, higher numbers are
4012 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4013 sbi->s_cluster_bits);
4015 do_div(resv_clusters, 50);
4016 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4018 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4021 static const char *ext4_quota_mode(struct super_block *sb)
4024 if (!ext4_quota_capable(sb))
4027 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4028 return "journalled";
4036 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
4038 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
4039 char *orig_data = kstrdup(data, GFP_KERNEL);
4040 struct buffer_head *bh, **group_desc;
4041 struct ext4_super_block *es = NULL;
4042 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4043 struct flex_groups **flex_groups;
4045 ext4_fsblk_t sb_block = get_sb_block(&data);
4046 ext4_fsblk_t logical_sb_block;
4047 unsigned long offset = 0;
4048 unsigned long def_mount_opts;
4052 int blocksize, clustersize;
4053 unsigned int db_count;
4055 int needs_recovery, has_huge_files;
4058 ext4_group_t first_not_zeroed;
4059 struct ext4_parsed_options parsed_opts;
4061 /* Set defaults for the variables that will be set during parsing */
4062 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4063 parsed_opts.journal_devnum = 0;
4064 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
4066 if ((data && !orig_data) || !sbi)
4069 sbi->s_daxdev = dax_dev;
4070 sbi->s_blockgroup_lock =
4071 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4072 if (!sbi->s_blockgroup_lock)
4075 sb->s_fs_info = sbi;
4077 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4078 sbi->s_sb_block = sb_block;
4079 sbi->s_sectors_written_start =
4080 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4082 /* Cleanup superblock name */
4083 strreplace(sb->s_id, '/', '!');
4085 /* -EINVAL is default */
4087 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4089 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4094 * The ext4 superblock will not be buffer aligned for other than 1kB
4095 * block sizes. We need to calculate the offset from buffer start.
4097 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4098 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4099 offset = do_div(logical_sb_block, blocksize);
4101 logical_sb_block = sb_block;
4104 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4106 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4111 * Note: s_es must be initialized as soon as possible because
4112 * some ext4 macro-instructions depend on its value
4114 es = (struct ext4_super_block *) (bh->b_data + offset);
4116 sb->s_magic = le16_to_cpu(es->s_magic);
4117 if (sb->s_magic != EXT4_SUPER_MAGIC)
4119 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4121 /* Warn if metadata_csum and gdt_csum are both set. */
4122 if (ext4_has_feature_metadata_csum(sb) &&
4123 ext4_has_feature_gdt_csum(sb))
4124 ext4_warning(sb, "metadata_csum and uninit_bg are "
4125 "redundant flags; please run fsck.");
4127 /* Check for a known checksum algorithm */
4128 if (!ext4_verify_csum_type(sb, es)) {
4129 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4130 "unknown checksum algorithm.");
4135 /* Load the checksum driver */
4136 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4137 if (IS_ERR(sbi->s_chksum_driver)) {
4138 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4139 ret = PTR_ERR(sbi->s_chksum_driver);
4140 sbi->s_chksum_driver = NULL;
4144 /* Check superblock checksum */
4145 if (!ext4_superblock_csum_verify(sb, es)) {
4146 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4147 "invalid superblock checksum. Run e2fsck?");
4153 /* Precompute checksum seed for all metadata */
4154 if (ext4_has_feature_csum_seed(sb))
4155 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4156 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4157 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4158 sizeof(es->s_uuid));
4160 /* Set defaults before we parse the mount options */
4161 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4162 set_opt(sb, INIT_INODE_TABLE);
4163 if (def_mount_opts & EXT4_DEFM_DEBUG)
4165 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4167 if (def_mount_opts & EXT4_DEFM_UID16)
4168 set_opt(sb, NO_UID32);
4169 /* xattr user namespace & acls are now defaulted on */
4170 set_opt(sb, XATTR_USER);
4171 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4172 set_opt(sb, POSIX_ACL);
4174 if (ext4_has_feature_fast_commit(sb))
4175 set_opt2(sb, JOURNAL_FAST_COMMIT);
4176 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4177 if (ext4_has_metadata_csum(sb))
4178 set_opt(sb, JOURNAL_CHECKSUM);
4180 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4181 set_opt(sb, JOURNAL_DATA);
4182 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4183 set_opt(sb, ORDERED_DATA);
4184 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4185 set_opt(sb, WRITEBACK_DATA);
4187 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4188 set_opt(sb, ERRORS_PANIC);
4189 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4190 set_opt(sb, ERRORS_CONT);
4192 set_opt(sb, ERRORS_RO);
4193 /* block_validity enabled by default; disable with noblock_validity */
4194 set_opt(sb, BLOCK_VALIDITY);
4195 if (def_mount_opts & EXT4_DEFM_DISCARD)
4196 set_opt(sb, DISCARD);
4198 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4199 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4200 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4201 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4202 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4204 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4205 set_opt(sb, BARRIER);
4208 * enable delayed allocation by default
4209 * Use -o nodelalloc to turn it off
4211 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4212 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4213 set_opt(sb, DELALLOC);
4216 * set default s_li_wait_mult for lazyinit, for the case there is
4217 * no mount option specified.
4219 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4221 if (le32_to_cpu(es->s_log_block_size) >
4222 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4223 ext4_msg(sb, KERN_ERR,
4224 "Invalid log block size: %u",
4225 le32_to_cpu(es->s_log_block_size));
4228 if (le32_to_cpu(es->s_log_cluster_size) >
4229 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4230 ext4_msg(sb, KERN_ERR,
4231 "Invalid log cluster size: %u",
4232 le32_to_cpu(es->s_log_cluster_size));
4236 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4238 if (blocksize == PAGE_SIZE)
4239 set_opt(sb, DIOREAD_NOLOCK);
4241 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4242 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4243 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4245 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4246 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4247 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4248 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4252 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4253 (!is_power_of_2(sbi->s_inode_size)) ||
4254 (sbi->s_inode_size > blocksize)) {
4255 ext4_msg(sb, KERN_ERR,
4256 "unsupported inode size: %d",
4258 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4262 * i_atime_extra is the last extra field available for
4263 * [acm]times in struct ext4_inode. Checking for that
4264 * field should suffice to ensure we have extra space
4267 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4268 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4269 sb->s_time_gran = 1;
4270 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4272 sb->s_time_gran = NSEC_PER_SEC;
4273 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4275 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4277 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4278 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4279 EXT4_GOOD_OLD_INODE_SIZE;
4280 if (ext4_has_feature_extra_isize(sb)) {
4281 unsigned v, max = (sbi->s_inode_size -
4282 EXT4_GOOD_OLD_INODE_SIZE);
4284 v = le16_to_cpu(es->s_want_extra_isize);
4286 ext4_msg(sb, KERN_ERR,
4287 "bad s_want_extra_isize: %d", v);
4290 if (sbi->s_want_extra_isize < v)
4291 sbi->s_want_extra_isize = v;
4293 v = le16_to_cpu(es->s_min_extra_isize);
4295 ext4_msg(sb, KERN_ERR,
4296 "bad s_min_extra_isize: %d", v);
4299 if (sbi->s_want_extra_isize < v)
4300 sbi->s_want_extra_isize = v;
4304 if (sbi->s_es->s_mount_opts[0]) {
4305 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4306 sizeof(sbi->s_es->s_mount_opts),
4310 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4311 ext4_msg(sb, KERN_WARNING,
4312 "failed to parse options in superblock: %s",
4315 kfree(s_mount_opts);
4317 sbi->s_def_mount_opt = sbi->s_mount_opt;
4318 if (!parse_options((char *) data, sb, &parsed_opts, 0))
4321 #ifdef CONFIG_UNICODE
4322 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4323 const struct ext4_sb_encodings *encoding_info;
4324 struct unicode_map *encoding;
4325 __u16 encoding_flags;
4327 if (ext4_sb_read_encoding(es, &encoding_info,
4329 ext4_msg(sb, KERN_ERR,
4330 "Encoding requested by superblock is unknown");
4334 encoding = utf8_load(encoding_info->version);
4335 if (IS_ERR(encoding)) {
4336 ext4_msg(sb, KERN_ERR,
4337 "can't mount with superblock charset: %s-%s "
4338 "not supported by the kernel. flags: 0x%x.",
4339 encoding_info->name, encoding_info->version,
4343 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4344 "%s-%s with flags 0x%hx", encoding_info->name,
4345 encoding_info->version?:"\b", encoding_flags);
4347 sb->s_encoding = encoding;
4348 sb->s_encoding_flags = encoding_flags;
4352 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4353 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4354 /* can't mount with both data=journal and dioread_nolock. */
4355 clear_opt(sb, DIOREAD_NOLOCK);
4356 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4357 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4358 ext4_msg(sb, KERN_ERR, "can't mount with "
4359 "both data=journal and delalloc");
4362 if (test_opt(sb, DAX_ALWAYS)) {
4363 ext4_msg(sb, KERN_ERR, "can't mount with "
4364 "both data=journal and dax");
4367 if (ext4_has_feature_encrypt(sb)) {
4368 ext4_msg(sb, KERN_WARNING,
4369 "encrypted files will use data=ordered "
4370 "instead of data journaling mode");
4372 if (test_opt(sb, DELALLOC))
4373 clear_opt(sb, DELALLOC);
4375 sb->s_iflags |= SB_I_CGROUPWB;
4378 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4379 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4381 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4382 (ext4_has_compat_features(sb) ||
4383 ext4_has_ro_compat_features(sb) ||
4384 ext4_has_incompat_features(sb)))
4385 ext4_msg(sb, KERN_WARNING,
4386 "feature flags set on rev 0 fs, "
4387 "running e2fsck is recommended");
4389 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4390 set_opt2(sb, HURD_COMPAT);
4391 if (ext4_has_feature_64bit(sb)) {
4392 ext4_msg(sb, KERN_ERR,
4393 "The Hurd can't support 64-bit file systems");
4398 * ea_inode feature uses l_i_version field which is not
4399 * available in HURD_COMPAT mode.
4401 if (ext4_has_feature_ea_inode(sb)) {
4402 ext4_msg(sb, KERN_ERR,
4403 "ea_inode feature is not supported for Hurd");
4408 if (IS_EXT2_SB(sb)) {
4409 if (ext2_feature_set_ok(sb))
4410 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4411 "using the ext4 subsystem");
4414 * If we're probing be silent, if this looks like
4415 * it's actually an ext[34] filesystem.
4417 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4419 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4420 "to feature incompatibilities");
4425 if (IS_EXT3_SB(sb)) {
4426 if (ext3_feature_set_ok(sb))
4427 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4428 "using the ext4 subsystem");
4431 * If we're probing be silent, if this looks like
4432 * it's actually an ext4 filesystem.
4434 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4436 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4437 "to feature incompatibilities");
4443 * Check feature flags regardless of the revision level, since we
4444 * previously didn't change the revision level when setting the flags,
4445 * so there is a chance incompat flags are set on a rev 0 filesystem.
4447 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4450 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4451 ext4_msg(sb, KERN_ERR,
4452 "Number of reserved GDT blocks insanely large: %d",
4453 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4457 if (bdev_dax_supported(sb->s_bdev, blocksize))
4458 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4460 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4461 if (ext4_has_feature_inline_data(sb)) {
4462 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4463 " that may contain inline data");
4466 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4467 ext4_msg(sb, KERN_ERR,
4468 "DAX unsupported by block device.");
4473 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4474 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4475 es->s_encryption_level);
4479 if (sb->s_blocksize != blocksize) {
4481 * bh must be released before kill_bdev(), otherwise
4482 * it won't be freed and its page also. kill_bdev()
4483 * is called by sb_set_blocksize().
4486 /* Validate the filesystem blocksize */
4487 if (!sb_set_blocksize(sb, blocksize)) {
4488 ext4_msg(sb, KERN_ERR, "bad block size %d",
4494 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4495 offset = do_div(logical_sb_block, blocksize);
4496 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4498 ext4_msg(sb, KERN_ERR,
4499 "Can't read superblock on 2nd try");
4504 es = (struct ext4_super_block *)(bh->b_data + offset);
4506 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4507 ext4_msg(sb, KERN_ERR,
4508 "Magic mismatch, very weird!");
4513 has_huge_files = ext4_has_feature_huge_file(sb);
4514 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4516 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4518 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4519 if (ext4_has_feature_64bit(sb)) {
4520 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4521 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4522 !is_power_of_2(sbi->s_desc_size)) {
4523 ext4_msg(sb, KERN_ERR,
4524 "unsupported descriptor size %lu",
4529 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4531 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4532 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4534 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4535 if (sbi->s_inodes_per_block == 0)
4537 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4538 sbi->s_inodes_per_group > blocksize * 8) {
4539 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4540 sbi->s_inodes_per_group);
4543 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4544 sbi->s_inodes_per_block;
4545 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4547 sbi->s_mount_state = le16_to_cpu(es->s_state);
4548 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4549 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4551 for (i = 0; i < 4; i++)
4552 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4553 sbi->s_def_hash_version = es->s_def_hash_version;
4554 if (ext4_has_feature_dir_index(sb)) {
4555 i = le32_to_cpu(es->s_flags);
4556 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4557 sbi->s_hash_unsigned = 3;
4558 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4559 #ifdef __CHAR_UNSIGNED__
4562 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4563 sbi->s_hash_unsigned = 3;
4567 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4572 /* Handle clustersize */
4573 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4574 if (ext4_has_feature_bigalloc(sb)) {
4575 if (clustersize < blocksize) {
4576 ext4_msg(sb, KERN_ERR,
4577 "cluster size (%d) smaller than "
4578 "block size (%d)", clustersize, blocksize);
4581 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4582 le32_to_cpu(es->s_log_block_size);
4583 sbi->s_clusters_per_group =
4584 le32_to_cpu(es->s_clusters_per_group);
4585 if (sbi->s_clusters_per_group > blocksize * 8) {
4586 ext4_msg(sb, KERN_ERR,
4587 "#clusters per group too big: %lu",
4588 sbi->s_clusters_per_group);
4591 if (sbi->s_blocks_per_group !=
4592 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4593 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4594 "clusters per group (%lu) inconsistent",
4595 sbi->s_blocks_per_group,
4596 sbi->s_clusters_per_group);
4600 if (clustersize != blocksize) {
4601 ext4_msg(sb, KERN_ERR,
4602 "fragment/cluster size (%d) != "
4603 "block size (%d)", clustersize, blocksize);
4606 if (sbi->s_blocks_per_group > blocksize * 8) {
4607 ext4_msg(sb, KERN_ERR,
4608 "#blocks per group too big: %lu",
4609 sbi->s_blocks_per_group);
4612 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4613 sbi->s_cluster_bits = 0;
4615 sbi->s_cluster_ratio = clustersize / blocksize;
4617 /* Do we have standard group size of clustersize * 8 blocks ? */
4618 if (sbi->s_blocks_per_group == clustersize << 3)
4619 set_opt2(sb, STD_GROUP_SIZE);
4622 * Test whether we have more sectors than will fit in sector_t,
4623 * and whether the max offset is addressable by the page cache.
4625 err = generic_check_addressable(sb->s_blocksize_bits,
4626 ext4_blocks_count(es));
4628 ext4_msg(sb, KERN_ERR, "filesystem"
4629 " too large to mount safely on this system");
4633 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4636 /* check blocks count against device size */
4637 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4638 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4639 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4640 "exceeds size of device (%llu blocks)",
4641 ext4_blocks_count(es), blocks_count);
4646 * It makes no sense for the first data block to be beyond the end
4647 * of the filesystem.
4649 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4650 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4651 "block %u is beyond end of filesystem (%llu)",
4652 le32_to_cpu(es->s_first_data_block),
4653 ext4_blocks_count(es));
4656 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4657 (sbi->s_cluster_ratio == 1)) {
4658 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4659 "block is 0 with a 1k block and cluster size");
4663 blocks_count = (ext4_blocks_count(es) -
4664 le32_to_cpu(es->s_first_data_block) +
4665 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4666 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4667 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4668 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4669 "(block count %llu, first data block %u, "
4670 "blocks per group %lu)", blocks_count,
4671 ext4_blocks_count(es),
4672 le32_to_cpu(es->s_first_data_block),
4673 EXT4_BLOCKS_PER_GROUP(sb));
4676 sbi->s_groups_count = blocks_count;
4677 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4678 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4679 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4680 le32_to_cpu(es->s_inodes_count)) {
4681 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4682 le32_to_cpu(es->s_inodes_count),
4683 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4687 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4688 EXT4_DESC_PER_BLOCK(sb);
4689 if (ext4_has_feature_meta_bg(sb)) {
4690 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4691 ext4_msg(sb, KERN_WARNING,
4692 "first meta block group too large: %u "
4693 "(group descriptor block count %u)",
4694 le32_to_cpu(es->s_first_meta_bg), db_count);
4698 rcu_assign_pointer(sbi->s_group_desc,
4699 kvmalloc_array(db_count,
4700 sizeof(struct buffer_head *),
4702 if (sbi->s_group_desc == NULL) {
4703 ext4_msg(sb, KERN_ERR, "not enough memory");
4708 bgl_lock_init(sbi->s_blockgroup_lock);
4710 /* Pre-read the descriptors into the buffer cache */
4711 for (i = 0; i < db_count; i++) {
4712 block = descriptor_loc(sb, logical_sb_block, i);
4713 ext4_sb_breadahead_unmovable(sb, block);
4716 for (i = 0; i < db_count; i++) {
4717 struct buffer_head *bh;
4719 block = descriptor_loc(sb, logical_sb_block, i);
4720 bh = ext4_sb_bread_unmovable(sb, block);
4722 ext4_msg(sb, KERN_ERR,
4723 "can't read group descriptor %d", i);
4729 rcu_dereference(sbi->s_group_desc)[i] = bh;
4732 sbi->s_gdb_count = db_count;
4733 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4734 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4735 ret = -EFSCORRUPTED;
4739 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4740 spin_lock_init(&sbi->s_error_lock);
4741 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4743 /* Register extent status tree shrinker */
4744 if (ext4_es_register_shrinker(sbi))
4747 sbi->s_stripe = ext4_get_stripe_size(sbi);
4748 sbi->s_extent_max_zeroout_kb = 32;
4751 * set up enough so that it can read an inode
4753 sb->s_op = &ext4_sops;
4754 sb->s_export_op = &ext4_export_ops;
4755 sb->s_xattr = ext4_xattr_handlers;
4756 #ifdef CONFIG_FS_ENCRYPTION
4757 sb->s_cop = &ext4_cryptops;
4759 #ifdef CONFIG_FS_VERITY
4760 sb->s_vop = &ext4_verityops;
4763 sb->dq_op = &ext4_quota_operations;
4764 if (ext4_has_feature_quota(sb))
4765 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4767 sb->s_qcop = &ext4_qctl_operations;
4768 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4770 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4772 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4773 mutex_init(&sbi->s_orphan_lock);
4775 /* Initialize fast commit stuff */
4776 atomic_set(&sbi->s_fc_subtid, 0);
4777 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4778 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4779 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4780 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4781 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4782 sbi->s_fc_bytes = 0;
4783 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4784 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4785 spin_lock_init(&sbi->s_fc_lock);
4786 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4787 sbi->s_fc_replay_state.fc_regions = NULL;
4788 sbi->s_fc_replay_state.fc_regions_size = 0;
4789 sbi->s_fc_replay_state.fc_regions_used = 0;
4790 sbi->s_fc_replay_state.fc_regions_valid = 0;
4791 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4792 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4793 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4797 needs_recovery = (es->s_last_orphan != 0 ||
4798 ext4_has_feature_journal_needs_recovery(sb));
4800 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4801 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4802 goto failed_mount3a;
4805 * The first inode we look at is the journal inode. Don't try
4806 * root first: it may be modified in the journal!
4808 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4809 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4811 goto failed_mount3a;
4812 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4813 ext4_has_feature_journal_needs_recovery(sb)) {
4814 ext4_msg(sb, KERN_ERR, "required journal recovery "
4815 "suppressed and not mounted read-only");
4816 goto failed_mount_wq;
4818 /* Nojournal mode, all journal mount options are illegal */
4819 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4820 ext4_msg(sb, KERN_ERR, "can't mount with "
4821 "journal_checksum, fs mounted w/o journal");
4822 goto failed_mount_wq;
4824 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4825 ext4_msg(sb, KERN_ERR, "can't mount with "
4826 "journal_async_commit, fs mounted w/o journal");
4827 goto failed_mount_wq;
4829 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4830 ext4_msg(sb, KERN_ERR, "can't mount with "
4831 "commit=%lu, fs mounted w/o journal",
4832 sbi->s_commit_interval / HZ);
4833 goto failed_mount_wq;
4835 if (EXT4_MOUNT_DATA_FLAGS &
4836 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4837 ext4_msg(sb, KERN_ERR, "can't mount with "
4838 "data=, fs mounted w/o journal");
4839 goto failed_mount_wq;
4841 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4842 clear_opt(sb, JOURNAL_CHECKSUM);
4843 clear_opt(sb, DATA_FLAGS);
4844 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4845 sbi->s_journal = NULL;
4850 if (ext4_has_feature_64bit(sb) &&
4851 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4852 JBD2_FEATURE_INCOMPAT_64BIT)) {
4853 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4854 goto failed_mount_wq;
4857 if (!set_journal_csum_feature_set(sb)) {
4858 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4860 goto failed_mount_wq;
4863 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4864 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4865 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4866 ext4_msg(sb, KERN_ERR,
4867 "Failed to set fast commit journal feature");
4868 goto failed_mount_wq;
4871 /* We have now updated the journal if required, so we can
4872 * validate the data journaling mode. */
4873 switch (test_opt(sb, DATA_FLAGS)) {
4875 /* No mode set, assume a default based on the journal
4876 * capabilities: ORDERED_DATA if the journal can
4877 * cope, else JOURNAL_DATA
4879 if (jbd2_journal_check_available_features
4880 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4881 set_opt(sb, ORDERED_DATA);
4882 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4884 set_opt(sb, JOURNAL_DATA);
4885 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4889 case EXT4_MOUNT_ORDERED_DATA:
4890 case EXT4_MOUNT_WRITEBACK_DATA:
4891 if (!jbd2_journal_check_available_features
4892 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4893 ext4_msg(sb, KERN_ERR, "Journal does not support "
4894 "requested data journaling mode");
4895 goto failed_mount_wq;
4902 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4903 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4904 ext4_msg(sb, KERN_ERR, "can't mount with "
4905 "journal_async_commit in data=ordered mode");
4906 goto failed_mount_wq;
4909 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4911 sbi->s_journal->j_submit_inode_data_buffers =
4912 ext4_journal_submit_inode_data_buffers;
4913 sbi->s_journal->j_finish_inode_data_buffers =
4914 ext4_journal_finish_inode_data_buffers;
4917 if (!test_opt(sb, NO_MBCACHE)) {
4918 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4919 if (!sbi->s_ea_block_cache) {
4920 ext4_msg(sb, KERN_ERR,
4921 "Failed to create ea_block_cache");
4922 goto failed_mount_wq;
4925 if (ext4_has_feature_ea_inode(sb)) {
4926 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4927 if (!sbi->s_ea_inode_cache) {
4928 ext4_msg(sb, KERN_ERR,
4929 "Failed to create ea_inode_cache");
4930 goto failed_mount_wq;
4935 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4936 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4937 goto failed_mount_wq;
4940 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4941 !ext4_has_feature_encrypt(sb)) {
4942 ext4_set_feature_encrypt(sb);
4943 ext4_commit_super(sb);
4947 * Get the # of file system overhead blocks from the
4948 * superblock if present.
4950 if (es->s_overhead_clusters)
4951 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4953 err = ext4_calculate_overhead(sb);
4955 goto failed_mount_wq;
4959 * The maximum number of concurrent works can be high and
4960 * concurrency isn't really necessary. Limit it to 1.
4962 EXT4_SB(sb)->rsv_conversion_wq =
4963 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4964 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4965 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4971 * The jbd2_journal_load will have done any necessary log recovery,
4972 * so we can safely mount the rest of the filesystem now.
4975 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4977 ext4_msg(sb, KERN_ERR, "get root inode failed");
4978 ret = PTR_ERR(root);
4982 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4983 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4988 sb->s_root = d_make_root(root);
4990 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4995 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4996 if (ret == -EROFS) {
4997 sb->s_flags |= SB_RDONLY;
5000 goto failed_mount4a;
5002 ext4_set_resv_clusters(sb);
5004 if (test_opt(sb, BLOCK_VALIDITY)) {
5005 err = ext4_setup_system_zone(sb);
5007 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5009 goto failed_mount4a;
5012 ext4_fc_replay_cleanup(sb);
5017 * Enable optimize_scan if number of groups is > threshold. This can be
5018 * turned off by passing "mb_optimize_scan=0". This can also be
5019 * turned on forcefully by passing "mb_optimize_scan=1".
5021 if (parsed_opts.mb_optimize_scan == 1)
5022 set_opt2(sb, MB_OPTIMIZE_SCAN);
5023 else if (parsed_opts.mb_optimize_scan == 0)
5024 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5025 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5026 set_opt2(sb, MB_OPTIMIZE_SCAN);
5028 err = ext4_mb_init(sb);
5030 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5036 * We can only set up the journal commit callback once
5037 * mballoc is initialized
5040 sbi->s_journal->j_commit_callback =
5041 ext4_journal_commit_callback;
5043 block = ext4_count_free_clusters(sb);
5044 ext4_free_blocks_count_set(sbi->s_es,
5045 EXT4_C2B(sbi, block));
5046 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5049 unsigned long freei = ext4_count_free_inodes(sb);
5050 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5051 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5055 err = percpu_counter_init(&sbi->s_dirs_counter,
5056 ext4_count_dirs(sb), GFP_KERNEL);
5058 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5061 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5064 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5067 ext4_msg(sb, KERN_ERR, "insufficient memory");
5071 if (ext4_has_feature_flex_bg(sb))
5072 if (!ext4_fill_flex_info(sb)) {
5073 ext4_msg(sb, KERN_ERR,
5074 "unable to initialize "
5075 "flex_bg meta info!");
5080 err = ext4_register_li_request(sb, first_not_zeroed);
5084 err = ext4_register_sysfs(sb);
5089 /* Enable quota usage during mount. */
5090 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5091 err = ext4_enable_quotas(sb);
5095 #endif /* CONFIG_QUOTA */
5098 * Save the original bdev mapping's wb_err value which could be
5099 * used to detect the metadata async write error.
5101 spin_lock_init(&sbi->s_bdev_wb_lock);
5102 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5103 &sbi->s_bdev_wb_err);
5104 sb->s_bdev->bd_super = sb;
5105 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5106 ext4_orphan_cleanup(sb, es);
5107 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5108 if (needs_recovery) {
5109 ext4_msg(sb, KERN_INFO, "recovery complete");
5110 err = ext4_mark_recovery_complete(sb, es);
5114 if (EXT4_SB(sb)->s_journal) {
5115 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5116 descr = " journalled data mode";
5117 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5118 descr = " ordered data mode";
5120 descr = " writeback data mode";
5122 descr = "out journal";
5124 if (test_opt(sb, DISCARD)) {
5125 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5126 if (!blk_queue_discard(q))
5127 ext4_msg(sb, KERN_WARNING,
5128 "mounting with \"discard\" option, but "
5129 "the device does not support discard");
5132 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5133 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5134 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5135 (int) sizeof(sbi->s_es->s_mount_opts),
5136 sbi->s_es->s_mount_opts,
5137 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5138 ext4_quota_mode(sb));
5140 if (es->s_error_count)
5141 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5143 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5144 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5145 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5146 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5147 atomic_set(&sbi->s_warning_count, 0);
5148 atomic_set(&sbi->s_msg_count, 0);
5155 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5159 ext4_unregister_sysfs(sb);
5160 kobject_put(&sbi->s_kobj);
5162 ext4_unregister_li_request(sb);
5164 ext4_mb_release(sb);
5166 flex_groups = rcu_dereference(sbi->s_flex_groups);
5168 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5169 kvfree(flex_groups[i]);
5170 kvfree(flex_groups);
5173 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5174 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5175 percpu_counter_destroy(&sbi->s_dirs_counter);
5176 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5177 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5178 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5180 ext4_ext_release(sb);
5181 ext4_release_system_zone(sb);
5186 ext4_msg(sb, KERN_ERR, "mount failed");
5187 if (EXT4_SB(sb)->rsv_conversion_wq)
5188 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5190 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5191 sbi->s_ea_inode_cache = NULL;
5193 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5194 sbi->s_ea_block_cache = NULL;
5196 if (sbi->s_journal) {
5197 jbd2_journal_destroy(sbi->s_journal);
5198 sbi->s_journal = NULL;
5201 ext4_es_unregister_shrinker(sbi);
5203 flush_work(&sbi->s_error_work);
5204 del_timer_sync(&sbi->s_err_report);
5205 ext4_stop_mmpd(sbi);
5208 group_desc = rcu_dereference(sbi->s_group_desc);
5209 for (i = 0; i < db_count; i++)
5210 brelse(group_desc[i]);
5214 if (sbi->s_chksum_driver)
5215 crypto_free_shash(sbi->s_chksum_driver);
5217 #ifdef CONFIG_UNICODE
5218 utf8_unload(sb->s_encoding);
5222 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5223 kfree(get_qf_name(sb, sbi, i));
5225 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5226 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5228 ext4_blkdev_remove(sbi);
5230 sb->s_fs_info = NULL;
5231 kfree(sbi->s_blockgroup_lock);
5235 fs_put_dax(dax_dev);
5236 return err ? err : ret;
5240 * Setup any per-fs journal parameters now. We'll do this both on
5241 * initial mount, once the journal has been initialised but before we've
5242 * done any recovery; and again on any subsequent remount.
5244 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5246 struct ext4_sb_info *sbi = EXT4_SB(sb);
5248 journal->j_commit_interval = sbi->s_commit_interval;
5249 journal->j_min_batch_time = sbi->s_min_batch_time;
5250 journal->j_max_batch_time = sbi->s_max_batch_time;
5251 ext4_fc_init(sb, journal);
5253 write_lock(&journal->j_state_lock);
5254 if (test_opt(sb, BARRIER))
5255 journal->j_flags |= JBD2_BARRIER;
5257 journal->j_flags &= ~JBD2_BARRIER;
5258 if (test_opt(sb, DATA_ERR_ABORT))
5259 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5261 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5262 write_unlock(&journal->j_state_lock);
5265 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5266 unsigned int journal_inum)
5268 struct inode *journal_inode;
5271 * Test for the existence of a valid inode on disk. Bad things
5272 * happen if we iget() an unused inode, as the subsequent iput()
5273 * will try to delete it.
5275 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5276 if (IS_ERR(journal_inode)) {
5277 ext4_msg(sb, KERN_ERR, "no journal found");
5280 if (!journal_inode->i_nlink) {
5281 make_bad_inode(journal_inode);
5282 iput(journal_inode);
5283 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5287 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5288 journal_inode, journal_inode->i_size);
5289 if (!S_ISREG(journal_inode->i_mode)) {
5290 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5291 iput(journal_inode);
5294 return journal_inode;
5297 static journal_t *ext4_get_journal(struct super_block *sb,
5298 unsigned int journal_inum)
5300 struct inode *journal_inode;
5303 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5306 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5310 journal = jbd2_journal_init_inode(journal_inode);
5312 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5313 iput(journal_inode);
5316 journal->j_private = sb;
5317 ext4_init_journal_params(sb, journal);
5321 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5324 struct buffer_head *bh;
5328 int hblock, blocksize;
5329 ext4_fsblk_t sb_block;
5330 unsigned long offset;
5331 struct ext4_super_block *es;
5332 struct block_device *bdev;
5334 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5337 bdev = ext4_blkdev_get(j_dev, sb);
5341 blocksize = sb->s_blocksize;
5342 hblock = bdev_logical_block_size(bdev);
5343 if (blocksize < hblock) {
5344 ext4_msg(sb, KERN_ERR,
5345 "blocksize too small for journal device");
5349 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5350 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5351 set_blocksize(bdev, blocksize);
5352 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5353 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5354 "external journal");
5358 es = (struct ext4_super_block *) (bh->b_data + offset);
5359 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5360 !(le32_to_cpu(es->s_feature_incompat) &
5361 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5362 ext4_msg(sb, KERN_ERR, "external journal has "
5368 if ((le32_to_cpu(es->s_feature_ro_compat) &
5369 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5370 es->s_checksum != ext4_superblock_csum(sb, es)) {
5371 ext4_msg(sb, KERN_ERR, "external journal has "
5372 "corrupt superblock");
5377 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5378 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5383 len = ext4_blocks_count(es);
5384 start = sb_block + 1;
5385 brelse(bh); /* we're done with the superblock */
5387 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5388 start, len, blocksize);
5390 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5393 journal->j_private = sb;
5394 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5395 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5398 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5399 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5400 "user (unsupported) - %d",
5401 be32_to_cpu(journal->j_superblock->s_nr_users));
5404 EXT4_SB(sb)->s_journal_bdev = bdev;
5405 ext4_init_journal_params(sb, journal);
5409 jbd2_journal_destroy(journal);
5411 ext4_blkdev_put(bdev);
5415 static int ext4_load_journal(struct super_block *sb,
5416 struct ext4_super_block *es,
5417 unsigned long journal_devnum)
5420 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5423 int really_read_only;
5426 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5427 return -EFSCORRUPTED;
5429 if (journal_devnum &&
5430 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5431 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5432 "numbers have changed");
5433 journal_dev = new_decode_dev(journal_devnum);
5435 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5437 if (journal_inum && journal_dev) {
5438 ext4_msg(sb, KERN_ERR,
5439 "filesystem has both journal inode and journal device!");
5444 journal = ext4_get_journal(sb, journal_inum);
5448 journal = ext4_get_dev_journal(sb, journal_dev);
5453 journal_dev_ro = bdev_read_only(journal->j_dev);
5454 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5456 if (journal_dev_ro && !sb_rdonly(sb)) {
5457 ext4_msg(sb, KERN_ERR,
5458 "journal device read-only, try mounting with '-o ro'");
5464 * Are we loading a blank journal or performing recovery after a
5465 * crash? For recovery, we need to check in advance whether we
5466 * can get read-write access to the device.
5468 if (ext4_has_feature_journal_needs_recovery(sb)) {
5469 if (sb_rdonly(sb)) {
5470 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5471 "required on readonly filesystem");
5472 if (really_read_only) {
5473 ext4_msg(sb, KERN_ERR, "write access "
5474 "unavailable, cannot proceed "
5475 "(try mounting with noload)");
5479 ext4_msg(sb, KERN_INFO, "write access will "
5480 "be enabled during recovery");
5484 if (!(journal->j_flags & JBD2_BARRIER))
5485 ext4_msg(sb, KERN_INFO, "barriers disabled");
5487 if (!ext4_has_feature_journal_needs_recovery(sb))
5488 err = jbd2_journal_wipe(journal, !really_read_only);
5490 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5492 memcpy(save, ((char *) es) +
5493 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5494 err = jbd2_journal_load(journal);
5496 memcpy(((char *) es) + EXT4_S_ERR_START,
5497 save, EXT4_S_ERR_LEN);
5502 ext4_msg(sb, KERN_ERR, "error loading journal");
5506 EXT4_SB(sb)->s_journal = journal;
5507 err = ext4_clear_journal_err(sb, es);
5509 EXT4_SB(sb)->s_journal = NULL;
5510 jbd2_journal_destroy(journal);
5514 if (!really_read_only && journal_devnum &&
5515 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5516 es->s_journal_dev = cpu_to_le32(journal_devnum);
5518 /* Make sure we flush the recovery flag to disk. */
5519 ext4_commit_super(sb);
5525 jbd2_journal_destroy(journal);
5529 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5530 static void ext4_update_super(struct super_block *sb)
5532 struct ext4_sb_info *sbi = EXT4_SB(sb);
5533 struct ext4_super_block *es = sbi->s_es;
5534 struct buffer_head *sbh = sbi->s_sbh;
5538 * If the file system is mounted read-only, don't update the
5539 * superblock write time. This avoids updating the superblock
5540 * write time when we are mounting the root file system
5541 * read/only but we need to replay the journal; at that point,
5542 * for people who are east of GMT and who make their clock
5543 * tick in localtime for Windows bug-for-bug compatibility,
5544 * the clock is set in the future, and this will cause e2fsck
5545 * to complain and force a full file system check.
5547 if (!(sb->s_flags & SB_RDONLY))
5548 ext4_update_tstamp(es, s_wtime);
5549 es->s_kbytes_written =
5550 cpu_to_le64(sbi->s_kbytes_written +
5551 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5552 sbi->s_sectors_written_start) >> 1));
5553 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5554 ext4_free_blocks_count_set(es,
5555 EXT4_C2B(sbi, percpu_counter_sum_positive(
5556 &sbi->s_freeclusters_counter)));
5557 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5558 es->s_free_inodes_count =
5559 cpu_to_le32(percpu_counter_sum_positive(
5560 &sbi->s_freeinodes_counter));
5561 /* Copy error information to the on-disk superblock */
5562 spin_lock(&sbi->s_error_lock);
5563 if (sbi->s_add_error_count > 0) {
5564 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5565 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5566 __ext4_update_tstamp(&es->s_first_error_time,
5567 &es->s_first_error_time_hi,
5568 sbi->s_first_error_time);
5569 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5570 sizeof(es->s_first_error_func));
5571 es->s_first_error_line =
5572 cpu_to_le32(sbi->s_first_error_line);
5573 es->s_first_error_ino =
5574 cpu_to_le32(sbi->s_first_error_ino);
5575 es->s_first_error_block =
5576 cpu_to_le64(sbi->s_first_error_block);
5577 es->s_first_error_errcode =
5578 ext4_errno_to_code(sbi->s_first_error_code);
5580 __ext4_update_tstamp(&es->s_last_error_time,
5581 &es->s_last_error_time_hi,
5582 sbi->s_last_error_time);
5583 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5584 sizeof(es->s_last_error_func));
5585 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5586 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5587 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5588 es->s_last_error_errcode =
5589 ext4_errno_to_code(sbi->s_last_error_code);
5591 * Start the daily error reporting function if it hasn't been
5594 if (!es->s_error_count)
5595 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5596 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5597 sbi->s_add_error_count = 0;
5599 spin_unlock(&sbi->s_error_lock);
5601 ext4_superblock_csum_set(sb);
5605 static int ext4_commit_super(struct super_block *sb)
5607 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5612 if (block_device_ejected(sb))
5615 ext4_update_super(sb);
5617 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5619 * Oh, dear. A previous attempt to write the
5620 * superblock failed. This could happen because the
5621 * USB device was yanked out. Or it could happen to
5622 * be a transient write error and maybe the block will
5623 * be remapped. Nothing we can do but to retry the
5624 * write and hope for the best.
5626 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5627 "superblock detected");
5628 clear_buffer_write_io_error(sbh);
5629 set_buffer_uptodate(sbh);
5631 BUFFER_TRACE(sbh, "marking dirty");
5632 mark_buffer_dirty(sbh);
5633 error = __sync_dirty_buffer(sbh,
5634 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5635 if (buffer_write_io_error(sbh)) {
5636 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5638 clear_buffer_write_io_error(sbh);
5639 set_buffer_uptodate(sbh);
5645 * Have we just finished recovery? If so, and if we are mounting (or
5646 * remounting) the filesystem readonly, then we will end up with a
5647 * consistent fs on disk. Record that fact.
5649 static int ext4_mark_recovery_complete(struct super_block *sb,
5650 struct ext4_super_block *es)
5653 journal_t *journal = EXT4_SB(sb)->s_journal;
5655 if (!ext4_has_feature_journal(sb)) {
5656 if (journal != NULL) {
5657 ext4_error(sb, "Journal got removed while the fs was "
5659 return -EFSCORRUPTED;
5663 jbd2_journal_lock_updates(journal);
5664 err = jbd2_journal_flush(journal);
5668 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5669 ext4_clear_feature_journal_needs_recovery(sb);
5670 ext4_commit_super(sb);
5673 jbd2_journal_unlock_updates(journal);
5678 * If we are mounting (or read-write remounting) a filesystem whose journal
5679 * has recorded an error from a previous lifetime, move that error to the
5680 * main filesystem now.
5682 static int ext4_clear_journal_err(struct super_block *sb,
5683 struct ext4_super_block *es)
5689 if (!ext4_has_feature_journal(sb)) {
5690 ext4_error(sb, "Journal got removed while the fs was mounted!");
5691 return -EFSCORRUPTED;
5694 journal = EXT4_SB(sb)->s_journal;
5697 * Now check for any error status which may have been recorded in the
5698 * journal by a prior ext4_error() or ext4_abort()
5701 j_errno = jbd2_journal_errno(journal);
5705 errstr = ext4_decode_error(sb, j_errno, nbuf);
5706 ext4_warning(sb, "Filesystem error recorded "
5707 "from previous mount: %s", errstr);
5708 ext4_warning(sb, "Marking fs in need of filesystem check.");
5710 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5711 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5712 ext4_commit_super(sb);
5714 jbd2_journal_clear_err(journal);
5715 jbd2_journal_update_sb_errno(journal);
5721 * Force the running and committing transactions to commit,
5722 * and wait on the commit.
5724 int ext4_force_commit(struct super_block *sb)
5731 journal = EXT4_SB(sb)->s_journal;
5732 return ext4_journal_force_commit(journal);
5735 static int ext4_sync_fs(struct super_block *sb, int wait)
5739 bool needs_barrier = false;
5740 struct ext4_sb_info *sbi = EXT4_SB(sb);
5742 if (unlikely(ext4_forced_shutdown(sbi)))
5745 trace_ext4_sync_fs(sb, wait);
5746 flush_workqueue(sbi->rsv_conversion_wq);
5748 * Writeback quota in non-journalled quota case - journalled quota has
5751 dquot_writeback_dquots(sb, -1);
5753 * Data writeback is possible w/o journal transaction, so barrier must
5754 * being sent at the end of the function. But we can skip it if
5755 * transaction_commit will do it for us.
5757 if (sbi->s_journal) {
5758 target = jbd2_get_latest_transaction(sbi->s_journal);
5759 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5760 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5761 needs_barrier = true;
5763 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5765 ret = jbd2_log_wait_commit(sbi->s_journal,
5768 } else if (wait && test_opt(sb, BARRIER))
5769 needs_barrier = true;
5770 if (needs_barrier) {
5772 err = blkdev_issue_flush(sb->s_bdev);
5781 * LVM calls this function before a (read-only) snapshot is created. This
5782 * gives us a chance to flush the journal completely and mark the fs clean.
5784 * Note that only this function cannot bring a filesystem to be in a clean
5785 * state independently. It relies on upper layer to stop all data & metadata
5788 static int ext4_freeze(struct super_block *sb)
5796 journal = EXT4_SB(sb)->s_journal;
5799 /* Now we set up the journal barrier. */
5800 jbd2_journal_lock_updates(journal);
5803 * Don't clear the needs_recovery flag if we failed to
5804 * flush the journal.
5806 error = jbd2_journal_flush(journal);
5810 /* Journal blocked and flushed, clear needs_recovery flag. */
5811 ext4_clear_feature_journal_needs_recovery(sb);
5814 error = ext4_commit_super(sb);
5817 /* we rely on upper layer to stop further updates */
5818 jbd2_journal_unlock_updates(journal);
5823 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5824 * flag here, even though the filesystem is not technically dirty yet.
5826 static int ext4_unfreeze(struct super_block *sb)
5828 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5831 if (EXT4_SB(sb)->s_journal) {
5832 /* Reset the needs_recovery flag before the fs is unlocked. */
5833 ext4_set_feature_journal_needs_recovery(sb);
5836 ext4_commit_super(sb);
5841 * Structure to save mount options for ext4_remount's benefit
5843 struct ext4_mount_options {
5844 unsigned long s_mount_opt;
5845 unsigned long s_mount_opt2;
5848 unsigned long s_commit_interval;
5849 u32 s_min_batch_time, s_max_batch_time;
5852 char *s_qf_names[EXT4_MAXQUOTAS];
5856 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5858 struct ext4_super_block *es;
5859 struct ext4_sb_info *sbi = EXT4_SB(sb);
5860 unsigned long old_sb_flags, vfs_flags;
5861 struct ext4_mount_options old_opts;
5862 int enable_quota = 0;
5867 char *to_free[EXT4_MAXQUOTAS];
5869 char *orig_data = kstrdup(data, GFP_KERNEL);
5870 struct ext4_parsed_options parsed_opts;
5872 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5873 parsed_opts.journal_devnum = 0;
5875 if (data && !orig_data)
5878 /* Store the original options */
5879 old_sb_flags = sb->s_flags;
5880 old_opts.s_mount_opt = sbi->s_mount_opt;
5881 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5882 old_opts.s_resuid = sbi->s_resuid;
5883 old_opts.s_resgid = sbi->s_resgid;
5884 old_opts.s_commit_interval = sbi->s_commit_interval;
5885 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5886 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5888 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5889 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5890 if (sbi->s_qf_names[i]) {
5891 char *qf_name = get_qf_name(sb, sbi, i);
5893 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5894 if (!old_opts.s_qf_names[i]) {
5895 for (j = 0; j < i; j++)
5896 kfree(old_opts.s_qf_names[j]);
5901 old_opts.s_qf_names[i] = NULL;
5903 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5904 parsed_opts.journal_ioprio =
5905 sbi->s_journal->j_task->io_context->ioprio;
5908 * Some options can be enabled by ext4 and/or by VFS mount flag
5909 * either way we need to make sure it matches in both *flags and
5910 * s_flags. Copy those selected flags from *flags to s_flags
5912 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5913 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5915 if (!parse_options(data, sb, &parsed_opts, 1)) {
5920 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5921 test_opt(sb, JOURNAL_CHECKSUM)) {
5922 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5923 "during remount not supported; ignoring");
5924 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5927 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5928 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5929 ext4_msg(sb, KERN_ERR, "can't mount with "
5930 "both data=journal and delalloc");
5934 if (test_opt(sb, DIOREAD_NOLOCK)) {
5935 ext4_msg(sb, KERN_ERR, "can't mount with "
5936 "both data=journal and dioread_nolock");
5940 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5941 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5942 ext4_msg(sb, KERN_ERR, "can't mount with "
5943 "journal_async_commit in data=ordered mode");
5949 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5950 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5955 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5956 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5958 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5959 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5963 if (sbi->s_journal) {
5964 ext4_init_journal_params(sb, sbi->s_journal);
5965 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5968 /* Flush outstanding errors before changing fs state */
5969 flush_work(&sbi->s_error_work);
5971 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5972 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5977 if (*flags & SB_RDONLY) {
5978 err = sync_filesystem(sb);
5981 err = dquot_suspend(sb, -1);
5986 * First of all, the unconditional stuff we have to do
5987 * to disable replay of the journal when we next remount
5989 sb->s_flags |= SB_RDONLY;
5992 * OK, test if we are remounting a valid rw partition
5993 * readonly, and if so set the rdonly flag and then
5994 * mark the partition as valid again.
5996 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5997 (sbi->s_mount_state & EXT4_VALID_FS))
5998 es->s_state = cpu_to_le16(sbi->s_mount_state);
6000 if (sbi->s_journal) {
6002 * We let remount-ro finish even if marking fs
6003 * as clean failed...
6005 ext4_mark_recovery_complete(sb, es);
6008 /* Make sure we can mount this feature set readwrite */
6009 if (ext4_has_feature_readonly(sb) ||
6010 !ext4_feature_set_ok(sb, 0)) {
6015 * Make sure the group descriptor checksums
6016 * are sane. If they aren't, refuse to remount r/w.
6018 for (g = 0; g < sbi->s_groups_count; g++) {
6019 struct ext4_group_desc *gdp =
6020 ext4_get_group_desc(sb, g, NULL);
6022 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6023 ext4_msg(sb, KERN_ERR,
6024 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6025 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6026 le16_to_cpu(gdp->bg_checksum));
6033 * If we have an unprocessed orphan list hanging
6034 * around from a previously readonly bdev mount,
6035 * require a full umount/remount for now.
6037 if (es->s_last_orphan) {
6038 ext4_msg(sb, KERN_WARNING, "Couldn't "
6039 "remount RDWR because of unprocessed "
6040 "orphan inode list. Please "
6041 "umount/remount instead");
6047 * Mounting a RDONLY partition read-write, so reread
6048 * and store the current valid flag. (It may have
6049 * been changed by e2fsck since we originally mounted
6052 if (sbi->s_journal) {
6053 err = ext4_clear_journal_err(sb, es);
6057 sbi->s_mount_state = le16_to_cpu(es->s_state);
6059 err = ext4_setup_super(sb, es, 0);
6063 sb->s_flags &= ~SB_RDONLY;
6064 if (ext4_has_feature_mmp(sb))
6065 if (ext4_multi_mount_protect(sb,
6066 le64_to_cpu(es->s_mmp_block))) {
6075 * Reinitialize lazy itable initialization thread based on
6078 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6079 ext4_unregister_li_request(sb);
6081 ext4_group_t first_not_zeroed;
6082 first_not_zeroed = ext4_has_uninit_itable(sb);
6083 ext4_register_li_request(sb, first_not_zeroed);
6087 * Handle creation of system zone data early because it can fail.
6088 * Releasing of existing data is done when we are sure remount will
6091 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6092 err = ext4_setup_system_zone(sb);
6097 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6098 err = ext4_commit_super(sb);
6104 /* Release old quota file names */
6105 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6106 kfree(old_opts.s_qf_names[i]);
6108 if (sb_any_quota_suspended(sb))
6109 dquot_resume(sb, -1);
6110 else if (ext4_has_feature_quota(sb)) {
6111 err = ext4_enable_quotas(sb);
6117 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6118 ext4_release_system_zone(sb);
6120 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6121 ext4_stop_mmpd(sbi);
6124 * Some options can be enabled by ext4 and/or by VFS mount flag
6125 * either way we need to make sure it matches in both *flags and
6126 * s_flags. Copy those selected flags from s_flags to *flags
6128 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6130 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6131 orig_data, ext4_quota_mode(sb));
6136 sb->s_flags = old_sb_flags;
6137 sbi->s_mount_opt = old_opts.s_mount_opt;
6138 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6139 sbi->s_resuid = old_opts.s_resuid;
6140 sbi->s_resgid = old_opts.s_resgid;
6141 sbi->s_commit_interval = old_opts.s_commit_interval;
6142 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6143 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6144 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6145 ext4_release_system_zone(sb);
6147 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6148 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6149 to_free[i] = get_qf_name(sb, sbi, i);
6150 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6153 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6156 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6157 ext4_stop_mmpd(sbi);
6163 static int ext4_statfs_project(struct super_block *sb,
6164 kprojid_t projid, struct kstatfs *buf)
6167 struct dquot *dquot;
6171 qid = make_kqid_projid(projid);
6172 dquot = dqget(sb, qid);
6174 return PTR_ERR(dquot);
6175 spin_lock(&dquot->dq_dqb_lock);
6177 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6178 dquot->dq_dqb.dqb_bhardlimit);
6179 limit >>= sb->s_blocksize_bits;
6181 if (limit && buf->f_blocks > limit) {
6182 curblock = (dquot->dq_dqb.dqb_curspace +
6183 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6184 buf->f_blocks = limit;
6185 buf->f_bfree = buf->f_bavail =
6186 (buf->f_blocks > curblock) ?
6187 (buf->f_blocks - curblock) : 0;
6190 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6191 dquot->dq_dqb.dqb_ihardlimit);
6192 if (limit && buf->f_files > limit) {
6193 buf->f_files = limit;
6195 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6196 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6199 spin_unlock(&dquot->dq_dqb_lock);
6205 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6207 struct super_block *sb = dentry->d_sb;
6208 struct ext4_sb_info *sbi = EXT4_SB(sb);
6209 struct ext4_super_block *es = sbi->s_es;
6210 ext4_fsblk_t overhead = 0, resv_blocks;
6212 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6214 if (!test_opt(sb, MINIX_DF))
6215 overhead = sbi->s_overhead;
6217 buf->f_type = EXT4_SUPER_MAGIC;
6218 buf->f_bsize = sb->s_blocksize;
6219 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6220 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6221 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6222 /* prevent underflow in case that few free space is available */
6223 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6224 buf->f_bavail = buf->f_bfree -
6225 (ext4_r_blocks_count(es) + resv_blocks);
6226 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6228 buf->f_files = le32_to_cpu(es->s_inodes_count);
6229 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6230 buf->f_namelen = EXT4_NAME_LEN;
6231 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6234 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6235 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6236 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6245 * Helper functions so that transaction is started before we acquire dqio_sem
6246 * to keep correct lock ordering of transaction > dqio_sem
6248 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6250 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6253 static int ext4_write_dquot(struct dquot *dquot)
6257 struct inode *inode;
6259 inode = dquot_to_inode(dquot);
6260 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6261 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6263 return PTR_ERR(handle);
6264 ret = dquot_commit(dquot);
6265 err = ext4_journal_stop(handle);
6271 static int ext4_acquire_dquot(struct dquot *dquot)
6276 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6277 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6279 return PTR_ERR(handle);
6280 ret = dquot_acquire(dquot);
6281 err = ext4_journal_stop(handle);
6287 static int ext4_release_dquot(struct dquot *dquot)
6292 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6293 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6294 if (IS_ERR(handle)) {
6295 /* Release dquot anyway to avoid endless cycle in dqput() */
6296 dquot_release(dquot);
6297 return PTR_ERR(handle);
6299 ret = dquot_release(dquot);
6300 err = ext4_journal_stop(handle);
6306 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6308 struct super_block *sb = dquot->dq_sb;
6310 if (ext4_is_quota_journalled(sb)) {
6311 dquot_mark_dquot_dirty(dquot);
6312 return ext4_write_dquot(dquot);
6314 return dquot_mark_dquot_dirty(dquot);
6318 static int ext4_write_info(struct super_block *sb, int type)
6323 /* Data block + inode block */
6324 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6326 return PTR_ERR(handle);
6327 ret = dquot_commit_info(sb, type);
6328 err = ext4_journal_stop(handle);
6335 * Turn on quotas during mount time - we need to find
6336 * the quota file and such...
6338 static int ext4_quota_on_mount(struct super_block *sb, int type)
6340 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6341 EXT4_SB(sb)->s_jquota_fmt, type);
6344 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6346 struct ext4_inode_info *ei = EXT4_I(inode);
6348 /* The first argument of lockdep_set_subclass has to be
6349 * *exactly* the same as the argument to init_rwsem() --- in
6350 * this case, in init_once() --- or lockdep gets unhappy
6351 * because the name of the lock is set using the
6352 * stringification of the argument to init_rwsem().
6354 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6355 lockdep_set_subclass(&ei->i_data_sem, subclass);
6359 * Standard function to be called on quota_on
6361 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6362 const struct path *path)
6366 if (!test_opt(sb, QUOTA))
6369 /* Quotafile not on the same filesystem? */
6370 if (path->dentry->d_sb != sb)
6373 /* Quota already enabled for this file? */
6374 if (IS_NOQUOTA(d_inode(path->dentry)))
6377 /* Journaling quota? */
6378 if (EXT4_SB(sb)->s_qf_names[type]) {
6379 /* Quotafile not in fs root? */
6380 if (path->dentry->d_parent != sb->s_root)
6381 ext4_msg(sb, KERN_WARNING,
6382 "Quota file not on filesystem root. "
6383 "Journaled quota will not work");
6384 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6387 * Clear the flag just in case mount options changed since
6390 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6394 * When we journal data on quota file, we have to flush journal to see
6395 * all updates to the file when we bypass pagecache...
6397 if (EXT4_SB(sb)->s_journal &&
6398 ext4_should_journal_data(d_inode(path->dentry))) {
6400 * We don't need to lock updates but journal_flush() could
6401 * otherwise be livelocked...
6403 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6404 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6405 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6410 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6411 err = dquot_quota_on(sb, type, format_id, path);
6413 lockdep_set_quota_inode(path->dentry->d_inode,
6416 struct inode *inode = d_inode(path->dentry);
6420 * Set inode flags to prevent userspace from messing with quota
6421 * files. If this fails, we return success anyway since quotas
6422 * are already enabled and this is not a hard failure.
6425 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6428 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6429 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6430 S_NOATIME | S_IMMUTABLE);
6431 err = ext4_mark_inode_dirty(handle, inode);
6432 ext4_journal_stop(handle);
6434 inode_unlock(inode);
6439 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6443 struct inode *qf_inode;
6444 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6445 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6446 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6447 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6450 BUG_ON(!ext4_has_feature_quota(sb));
6452 if (!qf_inums[type])
6455 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6456 if (IS_ERR(qf_inode)) {
6457 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6458 return PTR_ERR(qf_inode);
6461 /* Don't account quota for quota files to avoid recursion */
6462 qf_inode->i_flags |= S_NOQUOTA;
6463 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6464 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6466 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6472 /* Enable usage tracking for all quota types. */
6473 static int ext4_enable_quotas(struct super_block *sb)
6476 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6477 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6478 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6479 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6481 bool quota_mopt[EXT4_MAXQUOTAS] = {
6482 test_opt(sb, USRQUOTA),
6483 test_opt(sb, GRPQUOTA),
6484 test_opt(sb, PRJQUOTA),
6487 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6488 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6489 if (qf_inums[type]) {
6490 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6491 DQUOT_USAGE_ENABLED |
6492 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6495 "Failed to enable quota tracking "
6496 "(type=%d, err=%d). Please run "
6497 "e2fsck to fix.", type, err);
6498 for (type--; type >= 0; type--)
6499 dquot_quota_off(sb, type);
6508 static int ext4_quota_off(struct super_block *sb, int type)
6510 struct inode *inode = sb_dqopt(sb)->files[type];
6514 /* Force all delayed allocation blocks to be allocated.
6515 * Caller already holds s_umount sem */
6516 if (test_opt(sb, DELALLOC))
6517 sync_filesystem(sb);
6519 if (!inode || !igrab(inode))
6522 err = dquot_quota_off(sb, type);
6523 if (err || ext4_has_feature_quota(sb))
6528 * Update modification times of quota files when userspace can
6529 * start looking at them. If we fail, we return success anyway since
6530 * this is not a hard failure and quotas are already disabled.
6532 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6533 if (IS_ERR(handle)) {
6534 err = PTR_ERR(handle);
6537 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6538 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6539 inode->i_mtime = inode->i_ctime = current_time(inode);
6540 err = ext4_mark_inode_dirty(handle, inode);
6541 ext4_journal_stop(handle);
6543 inode_unlock(inode);
6545 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6549 return dquot_quota_off(sb, type);
6552 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6553 * acquiring the locks... As quota files are never truncated and quota code
6554 * itself serializes the operations (and no one else should touch the files)
6555 * we don't have to be afraid of races */
6556 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6557 size_t len, loff_t off)
6559 struct inode *inode = sb_dqopt(sb)->files[type];
6560 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6561 int offset = off & (sb->s_blocksize - 1);
6564 struct buffer_head *bh;
6565 loff_t i_size = i_size_read(inode);
6569 if (off+len > i_size)
6572 while (toread > 0) {
6573 tocopy = sb->s_blocksize - offset < toread ?
6574 sb->s_blocksize - offset : toread;
6575 bh = ext4_bread(NULL, inode, blk, 0);
6578 if (!bh) /* A hole? */
6579 memset(data, 0, tocopy);
6581 memcpy(data, bh->b_data+offset, tocopy);
6591 /* Write to quotafile (we know the transaction is already started and has
6592 * enough credits) */
6593 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6594 const char *data, size_t len, loff_t off)
6596 struct inode *inode = sb_dqopt(sb)->files[type];
6597 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6598 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6600 struct buffer_head *bh;
6601 handle_t *handle = journal_current_handle();
6603 if (EXT4_SB(sb)->s_journal && !handle) {
6604 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6605 " cancelled because transaction is not started",
6606 (unsigned long long)off, (unsigned long long)len);
6610 * Since we account only one data block in transaction credits,
6611 * then it is impossible to cross a block boundary.
6613 if (sb->s_blocksize - offset < len) {
6614 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6615 " cancelled because not block aligned",
6616 (unsigned long long)off, (unsigned long long)len);
6621 bh = ext4_bread(handle, inode, blk,
6622 EXT4_GET_BLOCKS_CREATE |
6623 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6624 } while (PTR_ERR(bh) == -ENOSPC &&
6625 ext4_should_retry_alloc(inode->i_sb, &retries));
6630 BUFFER_TRACE(bh, "get write access");
6631 err = ext4_journal_get_write_access(handle, bh);
6637 memcpy(bh->b_data+offset, data, len);
6638 flush_dcache_page(bh->b_page);
6640 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6643 if (inode->i_size < off + len) {
6644 i_size_write(inode, off + len);
6645 EXT4_I(inode)->i_disksize = inode->i_size;
6646 err2 = ext4_mark_inode_dirty(handle, inode);
6647 if (unlikely(err2 && !err))
6650 return err ? err : len;
6654 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6655 const char *dev_name, void *data)
6657 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6660 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6661 static inline void register_as_ext2(void)
6663 int err = register_filesystem(&ext2_fs_type);
6666 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6669 static inline void unregister_as_ext2(void)
6671 unregister_filesystem(&ext2_fs_type);
6674 static inline int ext2_feature_set_ok(struct super_block *sb)
6676 if (ext4_has_unknown_ext2_incompat_features(sb))
6680 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6685 static inline void register_as_ext2(void) { }
6686 static inline void unregister_as_ext2(void) { }
6687 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6690 static inline void register_as_ext3(void)
6692 int err = register_filesystem(&ext3_fs_type);
6695 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6698 static inline void unregister_as_ext3(void)
6700 unregister_filesystem(&ext3_fs_type);
6703 static inline int ext3_feature_set_ok(struct super_block *sb)
6705 if (ext4_has_unknown_ext3_incompat_features(sb))
6707 if (!ext4_has_feature_journal(sb))
6711 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6716 static struct file_system_type ext4_fs_type = {
6717 .owner = THIS_MODULE,
6719 .mount = ext4_mount,
6720 .kill_sb = kill_block_super,
6721 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6723 MODULE_ALIAS_FS("ext4");
6725 /* Shared across all ext4 file systems */
6726 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6728 static int __init ext4_init_fs(void)
6732 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6733 ext4_li_info = NULL;
6735 /* Build-time check for flags consistency */
6736 ext4_check_flag_values();
6738 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6739 init_waitqueue_head(&ext4__ioend_wq[i]);
6741 err = ext4_init_es();
6745 err = ext4_init_pending();
6749 err = ext4_init_post_read_processing();
6753 err = ext4_init_pageio();
6757 err = ext4_init_system_zone();
6761 err = ext4_init_sysfs();
6765 err = ext4_init_mballoc();
6768 err = init_inodecache();
6772 err = ext4_fc_init_dentry_cache();
6778 err = register_filesystem(&ext4_fs_type);
6784 unregister_as_ext2();
6785 unregister_as_ext3();
6787 destroy_inodecache();
6789 ext4_exit_mballoc();
6793 ext4_exit_system_zone();
6797 ext4_exit_post_read_processing();
6799 ext4_exit_pending();
6806 static void __exit ext4_exit_fs(void)
6808 ext4_destroy_lazyinit_thread();
6809 unregister_as_ext2();
6810 unregister_as_ext3();
6811 unregister_filesystem(&ext4_fs_type);
6812 destroy_inodecache();
6813 ext4_exit_mballoc();
6815 ext4_exit_system_zone();
6817 ext4_exit_post_read_processing();
6819 ext4_exit_pending();
6822 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6823 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6824 MODULE_LICENSE("GPL");
6825 MODULE_SOFTDEP("pre: crc32c");
6826 module_init(ext4_init_fs)
6827 module_exit(ext4_exit_fs)