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 void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 if (trylock_buffer(bh)) {
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
199 if (buffer_uptodate(bh))
207 * This works like __bread_gfp() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
212 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
216 struct buffer_head *bh;
219 bh = sb_getblk_gfp(sb, block, gfp);
221 return ERR_PTR(-ENOMEM);
222 if (ext4_buffer_uptodate(bh))
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
233 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
239 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
245 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
255 static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
258 if (!ext4_has_feature_metadata_csum(sb))
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
264 static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
273 return cpu_to_le32(csum);
276 static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
279 if (!ext4_has_metadata_csum(sb))
282 return es->s_checksum == ext4_superblock_csum(sb, es);
285 void ext4_superblock_csum_set(struct super_block *sb)
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 if (!ext4_has_metadata_csum(sb))
292 es->s_checksum = ext4_superblock_csum(sb, es);
295 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
296 struct ext4_group_desc *bg)
298 return le32_to_cpu(bg->bg_block_bitmap_lo) |
299 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
300 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
303 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
304 struct ext4_group_desc *bg)
306 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
307 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
308 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
311 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
312 struct ext4_group_desc *bg)
314 return le32_to_cpu(bg->bg_inode_table_lo) |
315 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
316 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
319 __u32 ext4_free_group_clusters(struct super_block *sb,
320 struct ext4_group_desc *bg)
322 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
323 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
324 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
327 __u32 ext4_free_inodes_count(struct super_block *sb,
328 struct ext4_group_desc *bg)
330 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
332 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
335 __u32 ext4_used_dirs_count(struct super_block *sb,
336 struct ext4_group_desc *bg)
338 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
340 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
343 __u32 ext4_itable_unused_count(struct super_block *sb,
344 struct ext4_group_desc *bg)
346 return le16_to_cpu(bg->bg_itable_unused_lo) |
347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
348 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
351 void ext4_block_bitmap_set(struct super_block *sb,
352 struct ext4_group_desc *bg, ext4_fsblk_t blk)
354 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
355 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
356 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
359 void ext4_inode_bitmap_set(struct super_block *sb,
360 struct ext4_group_desc *bg, ext4_fsblk_t blk)
362 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
363 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
364 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
367 void ext4_inode_table_set(struct super_block *sb,
368 struct ext4_group_desc *bg, ext4_fsblk_t blk)
370 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
371 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
372 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
375 void ext4_free_group_clusters_set(struct super_block *sb,
376 struct ext4_group_desc *bg, __u32 count)
378 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
379 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
380 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
383 void ext4_free_inodes_set(struct super_block *sb,
384 struct ext4_group_desc *bg, __u32 count)
386 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
391 void ext4_used_dirs_set(struct super_block *sb,
392 struct ext4_group_desc *bg, __u32 count)
394 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
399 void ext4_itable_unused_set(struct super_block *sb,
400 struct ext4_group_desc *bg, __u32 count)
402 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
404 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
407 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
409 now = clamp_val(now, 0, (1ull << 40) - 1);
411 *lo = cpu_to_le32(lower_32_bits(now));
412 *hi = upper_32_bits(now);
415 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
417 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
419 #define ext4_update_tstamp(es, tstamp) \
420 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
421 ktime_get_real_seconds())
422 #define ext4_get_tstamp(es, tstamp) \
423 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
426 * The del_gendisk() function uninitializes the disk-specific data
427 * structures, including the bdi structure, without telling anyone
428 * else. Once this happens, any attempt to call mark_buffer_dirty()
429 * (for example, by ext4_commit_super), will cause a kernel OOPS.
430 * This is a kludge to prevent these oops until we can put in a proper
431 * hook in del_gendisk() to inform the VFS and file system layers.
433 static int block_device_ejected(struct super_block *sb)
435 struct inode *bd_inode = sb->s_bdev->bd_inode;
436 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
438 return bdi->dev == NULL;
441 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
443 struct super_block *sb = journal->j_private;
444 struct ext4_sb_info *sbi = EXT4_SB(sb);
445 int error = is_journal_aborted(journal);
446 struct ext4_journal_cb_entry *jce;
448 BUG_ON(txn->t_state == T_FINISHED);
450 ext4_process_freed_data(sb, txn->t_tid);
452 spin_lock(&sbi->s_md_lock);
453 while (!list_empty(&txn->t_private_list)) {
454 jce = list_entry(txn->t_private_list.next,
455 struct ext4_journal_cb_entry, jce_list);
456 list_del_init(&jce->jce_list);
457 spin_unlock(&sbi->s_md_lock);
458 jce->jce_func(sb, jce, error);
459 spin_lock(&sbi->s_md_lock);
461 spin_unlock(&sbi->s_md_lock);
465 * This writepage callback for write_cache_pages()
466 * takes care of a few cases after page cleaning.
468 * write_cache_pages() already checks for dirty pages
469 * and calls clear_page_dirty_for_io(), which we want,
470 * to write protect the pages.
472 * However, we may have to redirty a page (see below.)
474 static int ext4_journalled_writepage_callback(struct page *page,
475 struct writeback_control *wbc,
478 transaction_t *transaction = (transaction_t *) data;
479 struct buffer_head *bh, *head;
480 struct journal_head *jh;
482 bh = head = page_buffers(page);
485 * We have to redirty a page in these cases:
486 * 1) If buffer is dirty, it means the page was dirty because it
487 * contains a buffer that needs checkpointing. So the dirty bit
488 * needs to be preserved so that checkpointing writes the buffer
490 * 2) If buffer is not part of the committing transaction
491 * (we may have just accidentally come across this buffer because
492 * inode range tracking is not exact) or if the currently running
493 * transaction already contains this buffer as well, dirty bit
494 * needs to be preserved so that the buffer gets writeprotected
495 * properly on running transaction's commit.
498 if (buffer_dirty(bh) ||
499 (jh && (jh->b_transaction != transaction ||
500 jh->b_next_transaction))) {
501 redirty_page_for_writepage(wbc, page);
504 } while ((bh = bh->b_this_page) != head);
507 return AOP_WRITEPAGE_ACTIVATE;
510 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
512 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
513 struct writeback_control wbc = {
514 .sync_mode = WB_SYNC_ALL,
515 .nr_to_write = LONG_MAX,
516 .range_start = jinode->i_dirty_start,
517 .range_end = jinode->i_dirty_end,
520 return write_cache_pages(mapping, &wbc,
521 ext4_journalled_writepage_callback,
522 jinode->i_transaction);
525 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
529 if (ext4_should_journal_data(jinode->i_vfs_inode))
530 ret = ext4_journalled_submit_inode_data_buffers(jinode);
532 ret = jbd2_journal_submit_inode_data_buffers(jinode);
537 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
541 if (!ext4_should_journal_data(jinode->i_vfs_inode))
542 ret = jbd2_journal_finish_inode_data_buffers(jinode);
547 static bool system_going_down(void)
549 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
550 || system_state == SYSTEM_RESTART;
553 struct ext4_err_translation {
558 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
560 static struct ext4_err_translation err_translation[] = {
561 EXT4_ERR_TRANSLATE(EIO),
562 EXT4_ERR_TRANSLATE(ENOMEM),
563 EXT4_ERR_TRANSLATE(EFSBADCRC),
564 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
565 EXT4_ERR_TRANSLATE(ENOSPC),
566 EXT4_ERR_TRANSLATE(ENOKEY),
567 EXT4_ERR_TRANSLATE(EROFS),
568 EXT4_ERR_TRANSLATE(EFBIG),
569 EXT4_ERR_TRANSLATE(EEXIST),
570 EXT4_ERR_TRANSLATE(ERANGE),
571 EXT4_ERR_TRANSLATE(EOVERFLOW),
572 EXT4_ERR_TRANSLATE(EBUSY),
573 EXT4_ERR_TRANSLATE(ENOTDIR),
574 EXT4_ERR_TRANSLATE(ENOTEMPTY),
575 EXT4_ERR_TRANSLATE(ESHUTDOWN),
576 EXT4_ERR_TRANSLATE(EFAULT),
579 static int ext4_errno_to_code(int errno)
583 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
584 if (err_translation[i].errno == errno)
585 return err_translation[i].code;
586 return EXT4_ERR_UNKNOWN;
589 static void save_error_info(struct super_block *sb, int error,
590 __u32 ino, __u64 block,
591 const char *func, unsigned int line)
593 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 /* We default to EFSCORRUPTED error... */
597 error = EFSCORRUPTED;
599 spin_lock(&sbi->s_error_lock);
600 sbi->s_add_error_count++;
601 sbi->s_last_error_code = error;
602 sbi->s_last_error_line = line;
603 sbi->s_last_error_ino = ino;
604 sbi->s_last_error_block = block;
605 sbi->s_last_error_func = func;
606 sbi->s_last_error_time = ktime_get_real_seconds();
607 if (!sbi->s_first_error_time) {
608 sbi->s_first_error_code = error;
609 sbi->s_first_error_line = line;
610 sbi->s_first_error_ino = ino;
611 sbi->s_first_error_block = block;
612 sbi->s_first_error_func = func;
613 sbi->s_first_error_time = sbi->s_last_error_time;
615 spin_unlock(&sbi->s_error_lock);
618 /* Deal with the reporting of failure conditions on a filesystem such as
619 * inconsistencies detected or read IO failures.
621 * On ext2, we can store the error state of the filesystem in the
622 * superblock. That is not possible on ext4, because we may have other
623 * write ordering constraints on the superblock which prevent us from
624 * writing it out straight away; and given that the journal is about to
625 * be aborted, we can't rely on the current, or future, transactions to
626 * write out the superblock safely.
628 * We'll just use the jbd2_journal_abort() error code to record an error in
629 * the journal instead. On recovery, the journal will complain about
630 * that error until we've noted it down and cleared it.
632 * If force_ro is set, we unconditionally force the filesystem into an
633 * ABORT|READONLY state, unless the error response on the fs has been set to
634 * panic in which case we take the easy way out and panic immediately. This is
635 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
636 * at a critical moment in log management.
638 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
639 __u32 ino, __u64 block,
640 const char *func, unsigned int line)
642 journal_t *journal = EXT4_SB(sb)->s_journal;
643 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
645 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
646 if (test_opt(sb, WARN_ON_ERROR))
649 if (!continue_fs && !sb_rdonly(sb)) {
650 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
652 jbd2_journal_abort(journal, -EIO);
655 if (!bdev_read_only(sb->s_bdev)) {
656 save_error_info(sb, error, ino, block, func, line);
658 * In case the fs should keep running, we need to writeout
659 * superblock through the journal. Due to lock ordering
660 * constraints, it may not be safe to do it right here so we
661 * defer superblock flushing to a workqueue.
664 schedule_work(&EXT4_SB(sb)->s_error_work);
666 ext4_commit_super(sb);
670 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
671 * could panic during 'reboot -f' as the underlying device got already
674 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
675 panic("EXT4-fs (device %s): panic forced after error\n",
679 if (sb_rdonly(sb) || continue_fs)
682 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
684 * Make sure updated value of ->s_mount_flags will be visible before
688 sb->s_flags |= SB_RDONLY;
691 static void flush_stashed_error_work(struct work_struct *work)
693 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
695 journal_t *journal = sbi->s_journal;
699 * If the journal is still running, we have to write out superblock
700 * through the journal to avoid collisions of other journalled sb
703 * We use directly jbd2 functions here to avoid recursing back into
704 * ext4 error handling code during handling of previous errors.
706 if (!sb_rdonly(sbi->s_sb) && journal) {
707 struct buffer_head *sbh = sbi->s_sbh;
708 handle = jbd2_journal_start(journal, 1);
711 if (jbd2_journal_get_write_access(handle, sbh)) {
712 jbd2_journal_stop(handle);
715 ext4_update_super(sbi->s_sb);
716 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
717 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
718 "superblock detected");
719 clear_buffer_write_io_error(sbh);
720 set_buffer_uptodate(sbh);
723 if (jbd2_journal_dirty_metadata(handle, sbh)) {
724 jbd2_journal_stop(handle);
727 jbd2_journal_stop(handle);
728 ext4_notify_error_sysfs(sbi);
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);
737 ext4_notify_error_sysfs(sbi);
740 #define ext4_error_ratelimit(sb) \
741 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
744 void __ext4_error(struct super_block *sb, const char *function,
745 unsigned int line, bool force_ro, int error, __u64 block,
746 const char *fmt, ...)
748 struct va_format vaf;
751 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
754 trace_ext4_error(sb, function, line);
755 if (ext4_error_ratelimit(sb)) {
760 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
761 sb->s_id, function, line, current->comm, &vaf);
764 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
767 void __ext4_error_inode(struct inode *inode, const char *function,
768 unsigned int line, ext4_fsblk_t block, int error,
769 const char *fmt, ...)
772 struct va_format vaf;
774 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
777 trace_ext4_error(inode->i_sb, function, line);
778 if (ext4_error_ratelimit(inode->i_sb)) {
783 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
784 "inode #%lu: block %llu: comm %s: %pV\n",
785 inode->i_sb->s_id, function, line, inode->i_ino,
786 block, current->comm, &vaf);
788 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
789 "inode #%lu: comm %s: %pV\n",
790 inode->i_sb->s_id, function, line, inode->i_ino,
791 current->comm, &vaf);
794 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
798 void __ext4_error_file(struct file *file, const char *function,
799 unsigned int line, ext4_fsblk_t block,
800 const char *fmt, ...)
803 struct va_format vaf;
804 struct inode *inode = file_inode(file);
805 char pathname[80], *path;
807 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
810 trace_ext4_error(inode->i_sb, function, line);
811 if (ext4_error_ratelimit(inode->i_sb)) {
812 path = file_path(file, pathname, sizeof(pathname));
820 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
821 "block %llu: comm %s: path %s: %pV\n",
822 inode->i_sb->s_id, function, line, inode->i_ino,
823 block, current->comm, path, &vaf);
826 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
827 "comm %s: path %s: %pV\n",
828 inode->i_sb->s_id, function, line, inode->i_ino,
829 current->comm, path, &vaf);
832 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
836 const char *ext4_decode_error(struct super_block *sb, int errno,
843 errstr = "Corrupt filesystem";
846 errstr = "Filesystem failed CRC";
849 errstr = "IO failure";
852 errstr = "Out of memory";
855 if (!sb || (EXT4_SB(sb)->s_journal &&
856 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
857 errstr = "Journal has aborted";
859 errstr = "Readonly filesystem";
862 /* If the caller passed in an extra buffer for unknown
863 * errors, textualise them now. Else we just return
866 /* Check for truncated error codes... */
867 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
876 /* __ext4_std_error decodes expected errors from journaling functions
877 * automatically and invokes the appropriate error response. */
879 void __ext4_std_error(struct super_block *sb, const char *function,
880 unsigned int line, int errno)
885 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
888 /* Special case: if the error is EROFS, and we're not already
889 * inside a transaction, then there's really no point in logging
891 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
894 if (ext4_error_ratelimit(sb)) {
895 errstr = ext4_decode_error(sb, errno, nbuf);
896 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
897 sb->s_id, function, line, errstr);
900 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
903 void __ext4_msg(struct super_block *sb,
904 const char *prefix, const char *fmt, ...)
906 struct va_format vaf;
909 atomic_inc(&EXT4_SB(sb)->s_msg_count);
910 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
916 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
920 static int ext4_warning_ratelimit(struct super_block *sb)
922 atomic_inc(&EXT4_SB(sb)->s_warning_count);
923 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
927 void __ext4_warning(struct super_block *sb, const char *function,
928 unsigned int line, const char *fmt, ...)
930 struct va_format vaf;
933 if (!ext4_warning_ratelimit(sb))
939 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
940 sb->s_id, function, line, &vaf);
944 void __ext4_warning_inode(const struct inode *inode, const char *function,
945 unsigned int line, const char *fmt, ...)
947 struct va_format vaf;
950 if (!ext4_warning_ratelimit(inode->i_sb))
956 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
957 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
958 function, line, inode->i_ino, current->comm, &vaf);
962 void __ext4_grp_locked_error(const char *function, unsigned int line,
963 struct super_block *sb, ext4_group_t grp,
964 unsigned long ino, ext4_fsblk_t block,
965 const char *fmt, ...)
969 struct va_format vaf;
972 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
975 trace_ext4_error(sb, function, line);
976 if (ext4_error_ratelimit(sb)) {
980 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
981 sb->s_id, function, line, grp);
983 printk(KERN_CONT "inode %lu: ", ino);
985 printk(KERN_CONT "block %llu:",
986 (unsigned long long) block);
987 printk(KERN_CONT "%pV\n", &vaf);
991 if (test_opt(sb, ERRORS_CONT)) {
992 if (test_opt(sb, WARN_ON_ERROR))
994 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
995 if (!bdev_read_only(sb->s_bdev)) {
996 save_error_info(sb, EFSCORRUPTED, ino, block, function,
998 schedule_work(&EXT4_SB(sb)->s_error_work);
1002 ext4_unlock_group(sb, grp);
1003 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1005 * We only get here in the ERRORS_RO case; relocking the group
1006 * may be dangerous, but nothing bad will happen since the
1007 * filesystem will have already been marked read/only and the
1008 * journal has been aborted. We return 1 as a hint to callers
1009 * who might what to use the return value from
1010 * ext4_grp_locked_error() to distinguish between the
1011 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1012 * aggressively from the ext4 function in question, with a
1013 * more appropriate error code.
1015 ext4_lock_group(sb, grp);
1019 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1023 struct ext4_sb_info *sbi = EXT4_SB(sb);
1024 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1025 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1028 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1029 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1032 percpu_counter_sub(&sbi->s_freeclusters_counter,
1036 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1037 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1042 count = ext4_free_inodes_count(sb, gdp);
1043 percpu_counter_sub(&sbi->s_freeinodes_counter,
1049 void ext4_update_dynamic_rev(struct super_block *sb)
1051 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1053 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1057 "updating to rev %d because of new feature flag, "
1058 "running e2fsck is recommended",
1061 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1062 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1063 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1064 /* leave es->s_feature_*compat flags alone */
1065 /* es->s_uuid will be set by e2fsck if empty */
1068 * The rest of the superblock fields should be zero, and if not it
1069 * means they are likely already in use, so leave them alone. We
1070 * can leave it up to e2fsck to clean up any inconsistencies there.
1075 * Open the external journal device
1077 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1079 struct block_device *bdev;
1081 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1087 ext4_msg(sb, KERN_ERR,
1088 "failed to open journal device unknown-block(%u,%u) %ld",
1089 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1094 * Release the journal device
1096 static void ext4_blkdev_put(struct block_device *bdev)
1098 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1101 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1103 struct block_device *bdev;
1104 bdev = sbi->s_journal_bdev;
1106 ext4_blkdev_put(bdev);
1107 sbi->s_journal_bdev = NULL;
1111 static inline struct inode *orphan_list_entry(struct list_head *l)
1113 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1116 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1118 struct list_head *l;
1120 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1121 le32_to_cpu(sbi->s_es->s_last_orphan));
1123 printk(KERN_ERR "sb_info orphan list:\n");
1124 list_for_each(l, &sbi->s_orphan) {
1125 struct inode *inode = orphan_list_entry(l);
1127 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1128 inode->i_sb->s_id, inode->i_ino, inode,
1129 inode->i_mode, inode->i_nlink,
1130 NEXT_ORPHAN(inode));
1135 static int ext4_quota_off(struct super_block *sb, int type);
1137 static inline void ext4_quota_off_umount(struct super_block *sb)
1141 /* Use our quota_off function to clear inode flags etc. */
1142 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1143 ext4_quota_off(sb, type);
1147 * This is a helper function which is used in the mount/remount
1148 * codepaths (which holds s_umount) to fetch the quota file name.
1150 static inline char *get_qf_name(struct super_block *sb,
1151 struct ext4_sb_info *sbi,
1154 return rcu_dereference_protected(sbi->s_qf_names[type],
1155 lockdep_is_held(&sb->s_umount));
1158 static inline void ext4_quota_off_umount(struct super_block *sb)
1163 static void ext4_put_super(struct super_block *sb)
1165 struct ext4_sb_info *sbi = EXT4_SB(sb);
1166 struct ext4_super_block *es = sbi->s_es;
1167 struct buffer_head **group_desc;
1168 struct flex_groups **flex_groups;
1172 ext4_unregister_li_request(sb);
1173 ext4_quota_off_umount(sb);
1175 flush_work(&sbi->s_error_work);
1176 destroy_workqueue(sbi->rsv_conversion_wq);
1179 * Unregister sysfs before destroying jbd2 journal.
1180 * Since we could still access attr_journal_task attribute via sysfs
1181 * path which could have sbi->s_journal->j_task as NULL
1183 ext4_unregister_sysfs(sb);
1185 if (sbi->s_journal) {
1186 aborted = is_journal_aborted(sbi->s_journal);
1187 err = jbd2_journal_destroy(sbi->s_journal);
1188 sbi->s_journal = NULL;
1189 if ((err < 0) && !aborted) {
1190 ext4_abort(sb, -err, "Couldn't clean up the journal");
1194 ext4_es_unregister_shrinker(sbi);
1195 del_timer_sync(&sbi->s_err_report);
1196 ext4_release_system_zone(sb);
1197 ext4_mb_release(sb);
1198 ext4_ext_release(sb);
1200 if (!sb_rdonly(sb) && !aborted) {
1201 ext4_clear_feature_journal_needs_recovery(sb);
1202 es->s_state = cpu_to_le16(sbi->s_mount_state);
1205 ext4_commit_super(sb);
1208 group_desc = rcu_dereference(sbi->s_group_desc);
1209 for (i = 0; i < sbi->s_gdb_count; i++)
1210 brelse(group_desc[i]);
1212 flex_groups = rcu_dereference(sbi->s_flex_groups);
1214 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1215 kvfree(flex_groups[i]);
1216 kvfree(flex_groups);
1219 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1220 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1221 percpu_counter_destroy(&sbi->s_dirs_counter);
1222 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1223 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1224 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1226 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1227 kfree(get_qf_name(sb, sbi, i));
1230 /* Debugging code just in case the in-memory inode orphan list
1231 * isn't empty. The on-disk one can be non-empty if we've
1232 * detected an error and taken the fs readonly, but the
1233 * in-memory list had better be clean by this point. */
1234 if (!list_empty(&sbi->s_orphan))
1235 dump_orphan_list(sb, sbi);
1236 ASSERT(list_empty(&sbi->s_orphan));
1238 sync_blockdev(sb->s_bdev);
1239 invalidate_bdev(sb->s_bdev);
1240 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1242 * Invalidate the journal device's buffers. We don't want them
1243 * floating about in memory - the physical journal device may
1244 * hotswapped, and it breaks the `ro-after' testing code.
1246 sync_blockdev(sbi->s_journal_bdev);
1247 invalidate_bdev(sbi->s_journal_bdev);
1248 ext4_blkdev_remove(sbi);
1251 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1252 sbi->s_ea_inode_cache = NULL;
1254 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1255 sbi->s_ea_block_cache = NULL;
1257 ext4_stop_mmpd(sbi);
1260 sb->s_fs_info = NULL;
1262 * Now that we are completely done shutting down the
1263 * superblock, we need to actually destroy the kobject.
1265 kobject_put(&sbi->s_kobj);
1266 wait_for_completion(&sbi->s_kobj_unregister);
1267 if (sbi->s_chksum_driver)
1268 crypto_free_shash(sbi->s_chksum_driver);
1269 kfree(sbi->s_blockgroup_lock);
1270 fs_put_dax(sbi->s_daxdev);
1271 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1272 #ifdef CONFIG_UNICODE
1273 utf8_unload(sb->s_encoding);
1278 static struct kmem_cache *ext4_inode_cachep;
1281 * Called inside transaction, so use GFP_NOFS
1283 static struct inode *ext4_alloc_inode(struct super_block *sb)
1285 struct ext4_inode_info *ei;
1287 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1291 inode_set_iversion(&ei->vfs_inode, 1);
1292 spin_lock_init(&ei->i_raw_lock);
1293 INIT_LIST_HEAD(&ei->i_prealloc_list);
1294 atomic_set(&ei->i_prealloc_active, 0);
1295 spin_lock_init(&ei->i_prealloc_lock);
1296 ext4_es_init_tree(&ei->i_es_tree);
1297 rwlock_init(&ei->i_es_lock);
1298 INIT_LIST_HEAD(&ei->i_es_list);
1299 ei->i_es_all_nr = 0;
1300 ei->i_es_shk_nr = 0;
1301 ei->i_es_shrink_lblk = 0;
1302 ei->i_reserved_data_blocks = 0;
1303 spin_lock_init(&(ei->i_block_reservation_lock));
1304 ext4_init_pending_tree(&ei->i_pending_tree);
1306 ei->i_reserved_quota = 0;
1307 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1310 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1311 spin_lock_init(&ei->i_completed_io_lock);
1313 ei->i_datasync_tid = 0;
1314 atomic_set(&ei->i_unwritten, 0);
1315 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1316 ext4_fc_init_inode(&ei->vfs_inode);
1317 mutex_init(&ei->i_fc_lock);
1318 return &ei->vfs_inode;
1321 static int ext4_drop_inode(struct inode *inode)
1323 int drop = generic_drop_inode(inode);
1326 drop = fscrypt_drop_inode(inode);
1328 trace_ext4_drop_inode(inode, drop);
1332 static void ext4_free_in_core_inode(struct inode *inode)
1334 fscrypt_free_inode(inode);
1335 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1336 pr_warn("%s: inode %ld still in fc list",
1337 __func__, inode->i_ino);
1339 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1342 static void ext4_destroy_inode(struct inode *inode)
1344 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1345 ext4_msg(inode->i_sb, KERN_ERR,
1346 "Inode %lu (%p): orphan list check failed!",
1347 inode->i_ino, EXT4_I(inode));
1348 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1349 EXT4_I(inode), sizeof(struct ext4_inode_info),
1355 static void init_once(void *foo)
1357 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1359 INIT_LIST_HEAD(&ei->i_orphan);
1360 init_rwsem(&ei->xattr_sem);
1361 init_rwsem(&ei->i_data_sem);
1362 init_rwsem(&ei->i_mmap_sem);
1363 inode_init_once(&ei->vfs_inode);
1364 ext4_fc_init_inode(&ei->vfs_inode);
1367 static int __init init_inodecache(void)
1369 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1370 sizeof(struct ext4_inode_info), 0,
1371 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1373 offsetof(struct ext4_inode_info, i_data),
1374 sizeof_field(struct ext4_inode_info, i_data),
1376 if (ext4_inode_cachep == NULL)
1381 static void destroy_inodecache(void)
1384 * Make sure all delayed rcu free inodes are flushed before we
1388 kmem_cache_destroy(ext4_inode_cachep);
1391 void ext4_clear_inode(struct inode *inode)
1394 invalidate_inode_buffers(inode);
1396 ext4_discard_preallocations(inode, 0);
1397 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1399 if (EXT4_I(inode)->jinode) {
1400 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1401 EXT4_I(inode)->jinode);
1402 jbd2_free_inode(EXT4_I(inode)->jinode);
1403 EXT4_I(inode)->jinode = NULL;
1405 fscrypt_put_encryption_info(inode);
1406 fsverity_cleanup_inode(inode);
1409 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1410 u64 ino, u32 generation)
1412 struct inode *inode;
1415 * Currently we don't know the generation for parent directory, so
1416 * a generation of 0 means "accept any"
1418 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1420 return ERR_CAST(inode);
1421 if (generation && inode->i_generation != generation) {
1423 return ERR_PTR(-ESTALE);
1429 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1430 int fh_len, int fh_type)
1432 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1433 ext4_nfs_get_inode);
1436 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1437 int fh_len, int fh_type)
1439 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1440 ext4_nfs_get_inode);
1443 static int ext4_nfs_commit_metadata(struct inode *inode)
1445 struct writeback_control wbc = {
1446 .sync_mode = WB_SYNC_ALL
1449 trace_ext4_nfs_commit_metadata(inode);
1450 return ext4_write_inode(inode, &wbc);
1453 #ifdef CONFIG_FS_ENCRYPTION
1454 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1456 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1457 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1460 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1463 handle_t *handle = fs_data;
1464 int res, res2, credits, retries = 0;
1467 * Encrypting the root directory is not allowed because e2fsck expects
1468 * lost+found to exist and be unencrypted, and encrypting the root
1469 * directory would imply encrypting the lost+found directory as well as
1470 * the filename "lost+found" itself.
1472 if (inode->i_ino == EXT4_ROOT_INO)
1475 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1478 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1481 res = ext4_convert_inline_data(inode);
1486 * If a journal handle was specified, then the encryption context is
1487 * being set on a new inode via inheritance and is part of a larger
1488 * transaction to create the inode. Otherwise the encryption context is
1489 * being set on an existing inode in its own transaction. Only in the
1490 * latter case should the "retry on ENOSPC" logic be used.
1494 res = ext4_xattr_set_handle(handle, inode,
1495 EXT4_XATTR_INDEX_ENCRYPTION,
1496 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1499 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1500 ext4_clear_inode_state(inode,
1501 EXT4_STATE_MAY_INLINE_DATA);
1503 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1504 * S_DAX may be disabled
1506 ext4_set_inode_flags(inode, false);
1511 res = dquot_initialize(inode);
1515 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1520 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1522 return PTR_ERR(handle);
1524 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1525 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1528 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1530 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1531 * S_DAX may be disabled
1533 ext4_set_inode_flags(inode, false);
1534 res = ext4_mark_inode_dirty(handle, inode);
1536 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1538 res2 = ext4_journal_stop(handle);
1540 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1547 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1549 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1552 static bool ext4_has_stable_inodes(struct super_block *sb)
1554 return ext4_has_feature_stable_inodes(sb);
1557 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1558 int *ino_bits_ret, int *lblk_bits_ret)
1560 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1561 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1564 static const struct fscrypt_operations ext4_cryptops = {
1565 .key_prefix = "ext4:",
1566 .get_context = ext4_get_context,
1567 .set_context = ext4_set_context,
1568 .get_dummy_policy = ext4_get_dummy_policy,
1569 .empty_dir = ext4_empty_dir,
1570 .max_namelen = EXT4_NAME_LEN,
1571 .has_stable_inodes = ext4_has_stable_inodes,
1572 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1577 static const char * const quotatypes[] = INITQFNAMES;
1578 #define QTYPE2NAME(t) (quotatypes[t])
1580 static int ext4_write_dquot(struct dquot *dquot);
1581 static int ext4_acquire_dquot(struct dquot *dquot);
1582 static int ext4_release_dquot(struct dquot *dquot);
1583 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1584 static int ext4_write_info(struct super_block *sb, int type);
1585 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1586 const struct path *path);
1587 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1588 size_t len, loff_t off);
1589 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1590 const char *data, size_t len, loff_t off);
1591 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1592 unsigned int flags);
1594 static struct dquot **ext4_get_dquots(struct inode *inode)
1596 return EXT4_I(inode)->i_dquot;
1599 static const struct dquot_operations ext4_quota_operations = {
1600 .get_reserved_space = ext4_get_reserved_space,
1601 .write_dquot = ext4_write_dquot,
1602 .acquire_dquot = ext4_acquire_dquot,
1603 .release_dquot = ext4_release_dquot,
1604 .mark_dirty = ext4_mark_dquot_dirty,
1605 .write_info = ext4_write_info,
1606 .alloc_dquot = dquot_alloc,
1607 .destroy_dquot = dquot_destroy,
1608 .get_projid = ext4_get_projid,
1609 .get_inode_usage = ext4_get_inode_usage,
1610 .get_next_id = dquot_get_next_id,
1613 static const struct quotactl_ops ext4_qctl_operations = {
1614 .quota_on = ext4_quota_on,
1615 .quota_off = ext4_quota_off,
1616 .quota_sync = dquot_quota_sync,
1617 .get_state = dquot_get_state,
1618 .set_info = dquot_set_dqinfo,
1619 .get_dqblk = dquot_get_dqblk,
1620 .set_dqblk = dquot_set_dqblk,
1621 .get_nextdqblk = dquot_get_next_dqblk,
1625 static const struct super_operations ext4_sops = {
1626 .alloc_inode = ext4_alloc_inode,
1627 .free_inode = ext4_free_in_core_inode,
1628 .destroy_inode = ext4_destroy_inode,
1629 .write_inode = ext4_write_inode,
1630 .dirty_inode = ext4_dirty_inode,
1631 .drop_inode = ext4_drop_inode,
1632 .evict_inode = ext4_evict_inode,
1633 .put_super = ext4_put_super,
1634 .sync_fs = ext4_sync_fs,
1635 .freeze_fs = ext4_freeze,
1636 .unfreeze_fs = ext4_unfreeze,
1637 .statfs = ext4_statfs,
1638 .remount_fs = ext4_remount,
1639 .show_options = ext4_show_options,
1641 .quota_read = ext4_quota_read,
1642 .quota_write = ext4_quota_write,
1643 .get_dquots = ext4_get_dquots,
1647 static const struct export_operations ext4_export_ops = {
1648 .fh_to_dentry = ext4_fh_to_dentry,
1649 .fh_to_parent = ext4_fh_to_parent,
1650 .get_parent = ext4_get_parent,
1651 .commit_metadata = ext4_nfs_commit_metadata,
1655 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1656 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1657 Opt_nouid32, Opt_debug, Opt_removed,
1658 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1659 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1660 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1661 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1662 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1663 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1665 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1666 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1667 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1668 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1669 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1670 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1671 Opt_nowarn_on_error, Opt_mblk_io_submit,
1672 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1673 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1674 Opt_inode_readahead_blks, Opt_journal_ioprio,
1675 Opt_dioread_nolock, Opt_dioread_lock,
1676 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1677 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1678 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1679 #ifdef CONFIG_EXT4_DEBUG
1680 Opt_fc_debug_max_replay, Opt_fc_debug_force
1684 static const match_table_t tokens = {
1685 {Opt_bsd_df, "bsddf"},
1686 {Opt_minix_df, "minixdf"},
1687 {Opt_grpid, "grpid"},
1688 {Opt_grpid, "bsdgroups"},
1689 {Opt_nogrpid, "nogrpid"},
1690 {Opt_nogrpid, "sysvgroups"},
1691 {Opt_resgid, "resgid=%u"},
1692 {Opt_resuid, "resuid=%u"},
1694 {Opt_err_cont, "errors=continue"},
1695 {Opt_err_panic, "errors=panic"},
1696 {Opt_err_ro, "errors=remount-ro"},
1697 {Opt_nouid32, "nouid32"},
1698 {Opt_debug, "debug"},
1699 {Opt_removed, "oldalloc"},
1700 {Opt_removed, "orlov"},
1701 {Opt_user_xattr, "user_xattr"},
1702 {Opt_nouser_xattr, "nouser_xattr"},
1704 {Opt_noacl, "noacl"},
1705 {Opt_noload, "norecovery"},
1706 {Opt_noload, "noload"},
1707 {Opt_removed, "nobh"},
1708 {Opt_removed, "bh"},
1709 {Opt_commit, "commit=%u"},
1710 {Opt_min_batch_time, "min_batch_time=%u"},
1711 {Opt_max_batch_time, "max_batch_time=%u"},
1712 {Opt_journal_dev, "journal_dev=%u"},
1713 {Opt_journal_path, "journal_path=%s"},
1714 {Opt_journal_checksum, "journal_checksum"},
1715 {Opt_nojournal_checksum, "nojournal_checksum"},
1716 {Opt_journal_async_commit, "journal_async_commit"},
1717 {Opt_abort, "abort"},
1718 {Opt_data_journal, "data=journal"},
1719 {Opt_data_ordered, "data=ordered"},
1720 {Opt_data_writeback, "data=writeback"},
1721 {Opt_data_err_abort, "data_err=abort"},
1722 {Opt_data_err_ignore, "data_err=ignore"},
1723 {Opt_offusrjquota, "usrjquota="},
1724 {Opt_usrjquota, "usrjquota=%s"},
1725 {Opt_offgrpjquota, "grpjquota="},
1726 {Opt_grpjquota, "grpjquota=%s"},
1727 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1728 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1729 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1730 {Opt_grpquota, "grpquota"},
1731 {Opt_noquota, "noquota"},
1732 {Opt_quota, "quota"},
1733 {Opt_usrquota, "usrquota"},
1734 {Opt_prjquota, "prjquota"},
1735 {Opt_barrier, "barrier=%u"},
1736 {Opt_barrier, "barrier"},
1737 {Opt_nobarrier, "nobarrier"},
1738 {Opt_i_version, "i_version"},
1740 {Opt_dax_always, "dax=always"},
1741 {Opt_dax_inode, "dax=inode"},
1742 {Opt_dax_never, "dax=never"},
1743 {Opt_stripe, "stripe=%u"},
1744 {Opt_delalloc, "delalloc"},
1745 {Opt_warn_on_error, "warn_on_error"},
1746 {Opt_nowarn_on_error, "nowarn_on_error"},
1747 {Opt_lazytime, "lazytime"},
1748 {Opt_nolazytime, "nolazytime"},
1749 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1750 {Opt_nodelalloc, "nodelalloc"},
1751 {Opt_removed, "mblk_io_submit"},
1752 {Opt_removed, "nomblk_io_submit"},
1753 {Opt_block_validity, "block_validity"},
1754 {Opt_noblock_validity, "noblock_validity"},
1755 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1756 {Opt_journal_ioprio, "journal_ioprio=%u"},
1757 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1758 {Opt_auto_da_alloc, "auto_da_alloc"},
1759 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1760 {Opt_dioread_nolock, "dioread_nolock"},
1761 {Opt_dioread_lock, "nodioread_nolock"},
1762 {Opt_dioread_lock, "dioread_lock"},
1763 {Opt_discard, "discard"},
1764 {Opt_nodiscard, "nodiscard"},
1765 {Opt_init_itable, "init_itable=%u"},
1766 {Opt_init_itable, "init_itable"},
1767 {Opt_noinit_itable, "noinit_itable"},
1768 #ifdef CONFIG_EXT4_DEBUG
1769 {Opt_fc_debug_force, "fc_debug_force"},
1770 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1772 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1773 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1774 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1775 {Opt_inlinecrypt, "inlinecrypt"},
1776 {Opt_nombcache, "nombcache"},
1777 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1778 {Opt_removed, "prefetch_block_bitmaps"},
1779 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1780 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1781 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1782 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1783 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1784 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1785 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1789 static ext4_fsblk_t get_sb_block(void **data)
1791 ext4_fsblk_t sb_block;
1792 char *options = (char *) *data;
1794 if (!options || strncmp(options, "sb=", 3) != 0)
1795 return 1; /* Default location */
1798 /* TODO: use simple_strtoll with >32bit ext4 */
1799 sb_block = simple_strtoul(options, &options, 0);
1800 if (*options && *options != ',') {
1801 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1805 if (*options == ',')
1807 *data = (void *) options;
1812 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1813 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1815 static const char deprecated_msg[] =
1816 "Mount option \"%s\" will be removed by %s\n"
1817 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1820 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1822 struct ext4_sb_info *sbi = EXT4_SB(sb);
1823 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1826 if (sb_any_quota_loaded(sb) && !old_qname) {
1827 ext4_msg(sb, KERN_ERR,
1828 "Cannot change journaled "
1829 "quota options when quota turned on");
1832 if (ext4_has_feature_quota(sb)) {
1833 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1834 "ignored when QUOTA feature is enabled");
1837 qname = match_strdup(args);
1839 ext4_msg(sb, KERN_ERR,
1840 "Not enough memory for storing quotafile name");
1844 if (strcmp(old_qname, qname) == 0)
1847 ext4_msg(sb, KERN_ERR,
1848 "%s quota file already specified",
1852 if (strchr(qname, '/')) {
1853 ext4_msg(sb, KERN_ERR,
1854 "quotafile must be on filesystem root");
1857 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1865 static int clear_qf_name(struct super_block *sb, int qtype)
1868 struct ext4_sb_info *sbi = EXT4_SB(sb);
1869 char *old_qname = get_qf_name(sb, sbi, qtype);
1871 if (sb_any_quota_loaded(sb) && old_qname) {
1872 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1873 " when quota turned on");
1876 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1883 #define MOPT_SET 0x0001
1884 #define MOPT_CLEAR 0x0002
1885 #define MOPT_NOSUPPORT 0x0004
1886 #define MOPT_EXPLICIT 0x0008
1887 #define MOPT_CLEAR_ERR 0x0010
1888 #define MOPT_GTE0 0x0020
1891 #define MOPT_QFMT 0x0040
1893 #define MOPT_Q MOPT_NOSUPPORT
1894 #define MOPT_QFMT MOPT_NOSUPPORT
1896 #define MOPT_DATAJ 0x0080
1897 #define MOPT_NO_EXT2 0x0100
1898 #define MOPT_NO_EXT3 0x0200
1899 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1900 #define MOPT_STRING 0x0400
1901 #define MOPT_SKIP 0x0800
1902 #define MOPT_2 0x1000
1904 static const struct mount_opts {
1908 } ext4_mount_opts[] = {
1909 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1910 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1911 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1912 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1913 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1914 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1915 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1916 MOPT_EXT4_ONLY | MOPT_SET},
1917 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1918 MOPT_EXT4_ONLY | MOPT_CLEAR},
1919 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1920 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1921 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1922 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1923 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1924 MOPT_EXT4_ONLY | MOPT_CLEAR},
1925 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1926 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1927 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1928 MOPT_EXT4_ONLY | MOPT_CLEAR},
1929 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1930 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1931 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1932 EXT4_MOUNT_JOURNAL_CHECKSUM),
1933 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1934 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1935 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1936 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1937 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1938 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1940 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1942 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1943 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1944 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1945 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1946 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1947 {Opt_commit, 0, MOPT_GTE0},
1948 {Opt_max_batch_time, 0, MOPT_GTE0},
1949 {Opt_min_batch_time, 0, MOPT_GTE0},
1950 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1951 {Opt_init_itable, 0, MOPT_GTE0},
1952 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1953 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1954 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1955 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1956 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1957 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1958 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1959 {Opt_stripe, 0, MOPT_GTE0},
1960 {Opt_resuid, 0, MOPT_GTE0},
1961 {Opt_resgid, 0, MOPT_GTE0},
1962 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1963 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1964 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1965 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1966 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1967 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1968 MOPT_NO_EXT2 | MOPT_DATAJ},
1969 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1970 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1971 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1972 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1973 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1975 {Opt_acl, 0, MOPT_NOSUPPORT},
1976 {Opt_noacl, 0, MOPT_NOSUPPORT},
1978 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1979 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1980 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1981 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1982 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1984 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1986 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1988 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1989 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1990 MOPT_CLEAR | MOPT_Q},
1991 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1992 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1993 {Opt_offusrjquota, 0, MOPT_Q},
1994 {Opt_offgrpjquota, 0, MOPT_Q},
1995 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1996 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1997 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1998 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1999 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2000 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2001 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2003 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2004 #ifdef CONFIG_EXT4_DEBUG
2005 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2006 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2007 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2012 #ifdef CONFIG_UNICODE
2013 static const struct ext4_sb_encodings {
2017 } ext4_sb_encoding_map[] = {
2018 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2021 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2022 const struct ext4_sb_encodings **encoding,
2025 __u16 magic = le16_to_cpu(es->s_encoding);
2028 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2029 if (magic == ext4_sb_encoding_map[i].magic)
2032 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2035 *encoding = &ext4_sb_encoding_map[i];
2036 *flags = le16_to_cpu(es->s_encoding_flags);
2042 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2044 const substring_t *arg,
2047 #ifdef CONFIG_FS_ENCRYPTION
2048 struct ext4_sb_info *sbi = EXT4_SB(sb);
2052 * This mount option is just for testing, and it's not worthwhile to
2053 * implement the extra complexity (e.g. RCU protection) that would be
2054 * needed to allow it to be set or changed during remount. We do allow
2055 * it to be specified during remount, but only if there is no change.
2057 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2058 ext4_msg(sb, KERN_WARNING,
2059 "Can't set test_dummy_encryption on remount");
2062 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2063 &sbi->s_dummy_enc_policy);
2066 ext4_msg(sb, KERN_WARNING,
2067 "Can't change test_dummy_encryption on remount");
2068 else if (err == -EINVAL)
2069 ext4_msg(sb, KERN_WARNING,
2070 "Value of option \"%s\" is unrecognized", opt);
2072 ext4_msg(sb, KERN_WARNING,
2073 "Error processing option \"%s\" [%d]",
2077 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2079 ext4_msg(sb, KERN_WARNING,
2080 "Test dummy encryption mount option ignored");
2085 struct ext4_parsed_options {
2086 unsigned long journal_devnum;
2087 unsigned int journal_ioprio;
2088 int mb_optimize_scan;
2091 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2092 substring_t *args, struct ext4_parsed_options *parsed_opts,
2095 struct ext4_sb_info *sbi = EXT4_SB(sb);
2096 const struct mount_opts *m;
2102 if (token == Opt_usrjquota)
2103 return set_qf_name(sb, USRQUOTA, &args[0]);
2104 else if (token == Opt_grpjquota)
2105 return set_qf_name(sb, GRPQUOTA, &args[0]);
2106 else if (token == Opt_offusrjquota)
2107 return clear_qf_name(sb, USRQUOTA);
2108 else if (token == Opt_offgrpjquota)
2109 return clear_qf_name(sb, GRPQUOTA);
2113 case Opt_nouser_xattr:
2114 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2117 return 1; /* handled by get_sb_block() */
2119 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2122 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2125 sb->s_flags |= SB_I_VERSION;
2128 sb->s_flags |= SB_LAZYTIME;
2130 case Opt_nolazytime:
2131 sb->s_flags &= ~SB_LAZYTIME;
2133 case Opt_inlinecrypt:
2134 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2135 sb->s_flags |= SB_INLINECRYPT;
2137 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2142 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2143 if (token == m->token)
2146 if (m->token == Opt_err) {
2147 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2148 "or missing value", opt);
2152 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2153 ext4_msg(sb, KERN_ERR,
2154 "Mount option \"%s\" incompatible with ext2", opt);
2157 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2158 ext4_msg(sb, KERN_ERR,
2159 "Mount option \"%s\" incompatible with ext3", opt);
2163 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2165 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2167 if (m->flags & MOPT_EXPLICIT) {
2168 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2169 set_opt2(sb, EXPLICIT_DELALLOC);
2170 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2171 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2175 if (m->flags & MOPT_CLEAR_ERR)
2176 clear_opt(sb, ERRORS_MASK);
2177 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2178 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2179 "options when quota turned on");
2183 if (m->flags & MOPT_NOSUPPORT) {
2184 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2185 } else if (token == Opt_commit) {
2187 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2188 else if (arg > INT_MAX / HZ) {
2189 ext4_msg(sb, KERN_ERR,
2190 "Invalid commit interval %d, "
2191 "must be smaller than %d",
2195 sbi->s_commit_interval = HZ * arg;
2196 } else if (token == Opt_debug_want_extra_isize) {
2199 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2200 ext4_msg(sb, KERN_ERR,
2201 "Invalid want_extra_isize %d", arg);
2204 sbi->s_want_extra_isize = arg;
2205 } else if (token == Opt_max_batch_time) {
2206 sbi->s_max_batch_time = arg;
2207 } else if (token == Opt_min_batch_time) {
2208 sbi->s_min_batch_time = arg;
2209 } else if (token == Opt_inode_readahead_blks) {
2210 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2211 ext4_msg(sb, KERN_ERR,
2212 "EXT4-fs: inode_readahead_blks must be "
2213 "0 or a power of 2 smaller than 2^31");
2216 sbi->s_inode_readahead_blks = arg;
2217 } else if (token == Opt_init_itable) {
2218 set_opt(sb, INIT_INODE_TABLE);
2220 arg = EXT4_DEF_LI_WAIT_MULT;
2221 sbi->s_li_wait_mult = arg;
2222 } else if (token == Opt_max_dir_size_kb) {
2223 sbi->s_max_dir_size_kb = arg;
2224 #ifdef CONFIG_EXT4_DEBUG
2225 } else if (token == Opt_fc_debug_max_replay) {
2226 sbi->s_fc_debug_max_replay = arg;
2228 } else if (token == Opt_stripe) {
2229 sbi->s_stripe = arg;
2230 } else if (token == Opt_resuid) {
2231 uid = make_kuid(current_user_ns(), arg);
2232 if (!uid_valid(uid)) {
2233 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2236 sbi->s_resuid = uid;
2237 } else if (token == Opt_resgid) {
2238 gid = make_kgid(current_user_ns(), arg);
2239 if (!gid_valid(gid)) {
2240 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2243 sbi->s_resgid = gid;
2244 } else if (token == Opt_journal_dev) {
2246 ext4_msg(sb, KERN_ERR,
2247 "Cannot specify journal on remount");
2250 parsed_opts->journal_devnum = arg;
2251 } else if (token == Opt_journal_path) {
2253 struct inode *journal_inode;
2258 ext4_msg(sb, KERN_ERR,
2259 "Cannot specify journal on remount");
2262 journal_path = match_strdup(&args[0]);
2263 if (!journal_path) {
2264 ext4_msg(sb, KERN_ERR, "error: could not dup "
2265 "journal device string");
2269 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2271 ext4_msg(sb, KERN_ERR, "error: could not find "
2272 "journal device path: error %d", error);
2273 kfree(journal_path);
2277 journal_inode = d_inode(path.dentry);
2278 if (!S_ISBLK(journal_inode->i_mode)) {
2279 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2280 "is not a block device", journal_path);
2282 kfree(journal_path);
2286 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2288 kfree(journal_path);
2289 } else if (token == Opt_journal_ioprio) {
2291 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2295 parsed_opts->journal_ioprio =
2296 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2297 } else if (token == Opt_test_dummy_encryption) {
2298 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2300 } else if (m->flags & MOPT_DATAJ) {
2302 if (!sbi->s_journal)
2303 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2304 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2305 ext4_msg(sb, KERN_ERR,
2306 "Cannot change data mode on remount");
2310 clear_opt(sb, DATA_FLAGS);
2311 sbi->s_mount_opt |= m->mount_opt;
2314 } else if (m->flags & MOPT_QFMT) {
2315 if (sb_any_quota_loaded(sb) &&
2316 sbi->s_jquota_fmt != m->mount_opt) {
2317 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2318 "quota options when quota turned on");
2321 if (ext4_has_feature_quota(sb)) {
2322 ext4_msg(sb, KERN_INFO,
2323 "Quota format mount options ignored "
2324 "when QUOTA feature is enabled");
2327 sbi->s_jquota_fmt = m->mount_opt;
2329 } else if (token == Opt_dax || token == Opt_dax_always ||
2330 token == Opt_dax_inode || token == Opt_dax_never) {
2331 #ifdef CONFIG_FS_DAX
2334 case Opt_dax_always:
2336 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2337 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2338 fail_dax_change_remount:
2339 ext4_msg(sb, KERN_ERR, "can't change "
2340 "dax mount option while remounting");
2344 (test_opt(sb, DATA_FLAGS) ==
2345 EXT4_MOUNT_JOURNAL_DATA)) {
2346 ext4_msg(sb, KERN_ERR, "can't mount with "
2347 "both data=journal and dax");
2350 ext4_msg(sb, KERN_WARNING,
2351 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2352 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2353 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2357 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2358 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2359 goto fail_dax_change_remount;
2360 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2361 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2365 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2366 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2367 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2368 goto fail_dax_change_remount;
2369 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2370 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2371 /* Strictly for printing options */
2372 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2376 ext4_msg(sb, KERN_INFO, "dax option not supported");
2377 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2378 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2381 } else if (token == Opt_data_err_abort) {
2382 sbi->s_mount_opt |= m->mount_opt;
2383 } else if (token == Opt_data_err_ignore) {
2384 sbi->s_mount_opt &= ~m->mount_opt;
2385 } else if (token == Opt_mb_optimize_scan) {
2386 if (arg != 0 && arg != 1) {
2387 ext4_msg(sb, KERN_WARNING,
2388 "mb_optimize_scan should be set to 0 or 1.");
2391 parsed_opts->mb_optimize_scan = arg;
2395 if (m->flags & MOPT_CLEAR)
2397 else if (unlikely(!(m->flags & MOPT_SET))) {
2398 ext4_msg(sb, KERN_WARNING,
2399 "buggy handling of option %s", opt);
2403 if (m->flags & MOPT_2) {
2405 sbi->s_mount_opt2 |= m->mount_opt;
2407 sbi->s_mount_opt2 &= ~m->mount_opt;
2410 sbi->s_mount_opt |= m->mount_opt;
2412 sbi->s_mount_opt &= ~m->mount_opt;
2418 static int parse_options(char *options, struct super_block *sb,
2419 struct ext4_parsed_options *ret_opts,
2422 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2423 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2424 substring_t args[MAX_OPT_ARGS];
2430 while ((p = strsep(&options, ",")) != NULL) {
2434 * Initialize args struct so we know whether arg was
2435 * found; some options take optional arguments.
2437 args[0].to = args[0].from = NULL;
2438 token = match_token(p, tokens, args);
2439 if (handle_mount_opt(sb, p, token, args, ret_opts,
2445 * We do the test below only for project quotas. 'usrquota' and
2446 * 'grpquota' mount options are allowed even without quota feature
2447 * to support legacy quotas in quota files.
2449 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2450 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2451 "Cannot enable project quota enforcement.");
2454 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2455 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2456 if (usr_qf_name || grp_qf_name) {
2457 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2458 clear_opt(sb, USRQUOTA);
2460 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2461 clear_opt(sb, GRPQUOTA);
2463 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2464 ext4_msg(sb, KERN_ERR, "old and new quota "
2469 if (!sbi->s_jquota_fmt) {
2470 ext4_msg(sb, KERN_ERR, "journaled quota format "
2476 if (test_opt(sb, DIOREAD_NOLOCK)) {
2478 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2479 if (blocksize < PAGE_SIZE)
2480 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2481 "experimental mount option 'dioread_nolock' "
2482 "for blocksize < PAGE_SIZE");
2487 static inline void ext4_show_quota_options(struct seq_file *seq,
2488 struct super_block *sb)
2490 #if defined(CONFIG_QUOTA)
2491 struct ext4_sb_info *sbi = EXT4_SB(sb);
2492 char *usr_qf_name, *grp_qf_name;
2494 if (sbi->s_jquota_fmt) {
2497 switch (sbi->s_jquota_fmt) {
2508 seq_printf(seq, ",jqfmt=%s", fmtname);
2512 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2513 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2515 seq_show_option(seq, "usrjquota", usr_qf_name);
2517 seq_show_option(seq, "grpjquota", grp_qf_name);
2522 static const char *token2str(int token)
2524 const struct match_token *t;
2526 for (t = tokens; t->token != Opt_err; t++)
2527 if (t->token == token && !strchr(t->pattern, '='))
2534 * - it's set to a non-default value OR
2535 * - if the per-sb default is different from the global default
2537 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2540 struct ext4_sb_info *sbi = EXT4_SB(sb);
2541 struct ext4_super_block *es = sbi->s_es;
2542 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2543 const struct mount_opts *m;
2544 char sep = nodefs ? '\n' : ',';
2546 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2547 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2549 if (sbi->s_sb_block != 1)
2550 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2552 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2553 int want_set = m->flags & MOPT_SET;
2554 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2555 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2557 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2558 continue; /* skip if same as the default */
2560 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2561 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2562 continue; /* select Opt_noFoo vs Opt_Foo */
2563 SEQ_OPTS_PRINT("%s", token2str(m->token));
2566 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2567 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2568 SEQ_OPTS_PRINT("resuid=%u",
2569 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2570 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2571 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2572 SEQ_OPTS_PRINT("resgid=%u",
2573 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2574 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2575 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2576 SEQ_OPTS_PUTS("errors=remount-ro");
2577 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2578 SEQ_OPTS_PUTS("errors=continue");
2579 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2580 SEQ_OPTS_PUTS("errors=panic");
2581 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2582 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2583 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2584 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2585 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2586 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2587 if (sb->s_flags & SB_I_VERSION)
2588 SEQ_OPTS_PUTS("i_version");
2589 if (nodefs || sbi->s_stripe)
2590 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2591 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2592 (sbi->s_mount_opt ^ def_mount_opt)) {
2593 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2594 SEQ_OPTS_PUTS("data=journal");
2595 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2596 SEQ_OPTS_PUTS("data=ordered");
2597 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2598 SEQ_OPTS_PUTS("data=writeback");
2601 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2602 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2603 sbi->s_inode_readahead_blks);
2605 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2606 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2607 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2608 if (nodefs || sbi->s_max_dir_size_kb)
2609 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2610 if (test_opt(sb, DATA_ERR_ABORT))
2611 SEQ_OPTS_PUTS("data_err=abort");
2613 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2615 if (sb->s_flags & SB_INLINECRYPT)
2616 SEQ_OPTS_PUTS("inlinecrypt");
2618 if (test_opt(sb, DAX_ALWAYS)) {
2620 SEQ_OPTS_PUTS("dax");
2622 SEQ_OPTS_PUTS("dax=always");
2623 } else if (test_opt2(sb, DAX_NEVER)) {
2624 SEQ_OPTS_PUTS("dax=never");
2625 } else if (test_opt2(sb, DAX_INODE)) {
2626 SEQ_OPTS_PUTS("dax=inode");
2628 ext4_show_quota_options(seq, sb);
2632 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2634 return _ext4_show_options(seq, root->d_sb, 0);
2637 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2639 struct super_block *sb = seq->private;
2642 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2643 rc = _ext4_show_options(seq, sb, 1);
2644 seq_puts(seq, "\n");
2648 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2651 struct ext4_sb_info *sbi = EXT4_SB(sb);
2654 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2655 ext4_msg(sb, KERN_ERR, "revision level too high, "
2656 "forcing read-only mode");
2662 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2663 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2664 "running e2fsck is recommended");
2665 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2666 ext4_msg(sb, KERN_WARNING,
2667 "warning: mounting fs with errors, "
2668 "running e2fsck is recommended");
2669 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2670 le16_to_cpu(es->s_mnt_count) >=
2671 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2672 ext4_msg(sb, KERN_WARNING,
2673 "warning: maximal mount count reached, "
2674 "running e2fsck is recommended");
2675 else if (le32_to_cpu(es->s_checkinterval) &&
2676 (ext4_get_tstamp(es, s_lastcheck) +
2677 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2678 ext4_msg(sb, KERN_WARNING,
2679 "warning: checktime reached, "
2680 "running e2fsck is recommended");
2681 if (!sbi->s_journal)
2682 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2683 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2684 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2685 le16_add_cpu(&es->s_mnt_count, 1);
2686 ext4_update_tstamp(es, s_mtime);
2688 ext4_set_feature_journal_needs_recovery(sb);
2690 err = ext4_commit_super(sb);
2692 if (test_opt(sb, DEBUG))
2693 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2694 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2696 sbi->s_groups_count,
2697 EXT4_BLOCKS_PER_GROUP(sb),
2698 EXT4_INODES_PER_GROUP(sb),
2699 sbi->s_mount_opt, sbi->s_mount_opt2);
2701 cleancache_init_fs(sb);
2705 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2707 struct ext4_sb_info *sbi = EXT4_SB(sb);
2708 struct flex_groups **old_groups, **new_groups;
2711 if (!sbi->s_log_groups_per_flex)
2714 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2715 if (size <= sbi->s_flex_groups_allocated)
2718 new_groups = kvzalloc(roundup_pow_of_two(size *
2719 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2721 ext4_msg(sb, KERN_ERR,
2722 "not enough memory for %d flex group pointers", size);
2725 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2726 new_groups[i] = kvzalloc(roundup_pow_of_two(
2727 sizeof(struct flex_groups)),
2729 if (!new_groups[i]) {
2730 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2731 kvfree(new_groups[j]);
2733 ext4_msg(sb, KERN_ERR,
2734 "not enough memory for %d flex groups", size);
2739 old_groups = rcu_dereference(sbi->s_flex_groups);
2741 memcpy(new_groups, old_groups,
2742 (sbi->s_flex_groups_allocated *
2743 sizeof(struct flex_groups *)));
2745 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2746 sbi->s_flex_groups_allocated = size;
2748 ext4_kvfree_array_rcu(old_groups);
2752 static int ext4_fill_flex_info(struct super_block *sb)
2754 struct ext4_sb_info *sbi = EXT4_SB(sb);
2755 struct ext4_group_desc *gdp = NULL;
2756 struct flex_groups *fg;
2757 ext4_group_t flex_group;
2760 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2761 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2762 sbi->s_log_groups_per_flex = 0;
2766 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2770 for (i = 0; i < sbi->s_groups_count; i++) {
2771 gdp = ext4_get_group_desc(sb, i, NULL);
2773 flex_group = ext4_flex_group(sbi, i);
2774 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2775 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2776 atomic64_add(ext4_free_group_clusters(sb, gdp),
2777 &fg->free_clusters);
2778 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2786 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2787 struct ext4_group_desc *gdp)
2789 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2791 __le32 le_group = cpu_to_le32(block_group);
2792 struct ext4_sb_info *sbi = EXT4_SB(sb);
2794 if (ext4_has_metadata_csum(sbi->s_sb)) {
2795 /* Use new metadata_csum algorithm */
2797 __u16 dummy_csum = 0;
2799 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2801 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2802 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2803 sizeof(dummy_csum));
2804 offset += sizeof(dummy_csum);
2805 if (offset < sbi->s_desc_size)
2806 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2807 sbi->s_desc_size - offset);
2809 crc = csum32 & 0xFFFF;
2813 /* old crc16 code */
2814 if (!ext4_has_feature_gdt_csum(sb))
2817 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2818 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2819 crc = crc16(crc, (__u8 *)gdp, offset);
2820 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2821 /* for checksum of struct ext4_group_desc do the rest...*/
2822 if (ext4_has_feature_64bit(sb) &&
2823 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2824 crc = crc16(crc, (__u8 *)gdp + offset,
2825 le16_to_cpu(sbi->s_es->s_desc_size) -
2829 return cpu_to_le16(crc);
2832 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2833 struct ext4_group_desc *gdp)
2835 if (ext4_has_group_desc_csum(sb) &&
2836 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2842 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2843 struct ext4_group_desc *gdp)
2845 if (!ext4_has_group_desc_csum(sb))
2847 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2850 /* Called at mount-time, super-block is locked */
2851 static int ext4_check_descriptors(struct super_block *sb,
2852 ext4_fsblk_t sb_block,
2853 ext4_group_t *first_not_zeroed)
2855 struct ext4_sb_info *sbi = EXT4_SB(sb);
2856 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2857 ext4_fsblk_t last_block;
2858 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2859 ext4_fsblk_t block_bitmap;
2860 ext4_fsblk_t inode_bitmap;
2861 ext4_fsblk_t inode_table;
2862 int flexbg_flag = 0;
2863 ext4_group_t i, grp = sbi->s_groups_count;
2865 if (ext4_has_feature_flex_bg(sb))
2868 ext4_debug("Checking group descriptors");
2870 for (i = 0; i < sbi->s_groups_count; i++) {
2871 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2873 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2874 last_block = ext4_blocks_count(sbi->s_es) - 1;
2876 last_block = first_block +
2877 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2879 if ((grp == sbi->s_groups_count) &&
2880 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2883 block_bitmap = ext4_block_bitmap(sb, gdp);
2884 if (block_bitmap == sb_block) {
2885 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2886 "Block bitmap for group %u overlaps "
2891 if (block_bitmap >= sb_block + 1 &&
2892 block_bitmap <= last_bg_block) {
2893 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2894 "Block bitmap for group %u overlaps "
2895 "block group descriptors", i);
2899 if (block_bitmap < first_block || block_bitmap > last_block) {
2900 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901 "Block bitmap for group %u not in group "
2902 "(block %llu)!", i, block_bitmap);
2905 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2906 if (inode_bitmap == sb_block) {
2907 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908 "Inode bitmap for group %u overlaps "
2913 if (inode_bitmap >= sb_block + 1 &&
2914 inode_bitmap <= last_bg_block) {
2915 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2916 "Inode bitmap for group %u overlaps "
2917 "block group descriptors", i);
2921 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2922 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2923 "Inode bitmap for group %u not in group "
2924 "(block %llu)!", i, inode_bitmap);
2927 inode_table = ext4_inode_table(sb, gdp);
2928 if (inode_table == sb_block) {
2929 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2930 "Inode table for group %u overlaps "
2935 if (inode_table >= sb_block + 1 &&
2936 inode_table <= last_bg_block) {
2937 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2938 "Inode table for group %u overlaps "
2939 "block group descriptors", i);
2943 if (inode_table < first_block ||
2944 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2945 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2946 "Inode table for group %u not in group "
2947 "(block %llu)!", i, inode_table);
2950 ext4_lock_group(sb, i);
2951 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2952 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2953 "Checksum for group %u failed (%u!=%u)",
2954 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2955 gdp)), le16_to_cpu(gdp->bg_checksum));
2956 if (!sb_rdonly(sb)) {
2957 ext4_unlock_group(sb, i);
2961 ext4_unlock_group(sb, i);
2963 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2965 if (NULL != first_not_zeroed)
2966 *first_not_zeroed = grp;
2971 * Maximal extent format file size.
2972 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2973 * extent format containers, within a sector_t, and within i_blocks
2974 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2975 * so that won't be a limiting factor.
2977 * However there is other limiting factor. We do store extents in the form
2978 * of starting block and length, hence the resulting length of the extent
2979 * covering maximum file size must fit into on-disk format containers as
2980 * well. Given that length is always by 1 unit bigger than max unit (because
2981 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2983 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2985 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2988 loff_t upper_limit = MAX_LFS_FILESIZE;
2990 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2992 if (!has_huge_files) {
2993 upper_limit = (1LL << 32) - 1;
2995 /* total blocks in file system block size */
2996 upper_limit >>= (blkbits - 9);
2997 upper_limit <<= blkbits;
3001 * 32-bit extent-start container, ee_block. We lower the maxbytes
3002 * by one fs block, so ee_len can cover the extent of maximum file
3005 res = (1LL << 32) - 1;
3008 /* Sanity check against vm- & vfs- imposed limits */
3009 if (res > upper_limit)
3016 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3017 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3018 * We need to be 1 filesystem block less than the 2^48 sector limit.
3020 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3022 loff_t res = EXT4_NDIR_BLOCKS;
3025 /* This is calculated to be the largest file size for a dense, block
3026 * mapped file such that the file's total number of 512-byte sectors,
3027 * including data and all indirect blocks, does not exceed (2^48 - 1).
3029 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3030 * number of 512-byte sectors of the file.
3033 if (!has_huge_files) {
3035 * !has_huge_files or implies that the inode i_block field
3036 * represents total file blocks in 2^32 512-byte sectors ==
3037 * size of vfs inode i_blocks * 8
3039 upper_limit = (1LL << 32) - 1;
3041 /* total blocks in file system block size */
3042 upper_limit >>= (bits - 9);
3046 * We use 48 bit ext4_inode i_blocks
3047 * With EXT4_HUGE_FILE_FL set the i_blocks
3048 * represent total number of blocks in
3049 * file system block size
3051 upper_limit = (1LL << 48) - 1;
3055 /* indirect blocks */
3057 /* double indirect blocks */
3058 meta_blocks += 1 + (1LL << (bits-2));
3059 /* tripple indirect blocks */
3060 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3062 upper_limit -= meta_blocks;
3063 upper_limit <<= bits;
3065 res += 1LL << (bits-2);
3066 res += 1LL << (2*(bits-2));
3067 res += 1LL << (3*(bits-2));
3069 if (res > upper_limit)
3072 if (res > MAX_LFS_FILESIZE)
3073 res = MAX_LFS_FILESIZE;
3078 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3079 ext4_fsblk_t logical_sb_block, int nr)
3081 struct ext4_sb_info *sbi = EXT4_SB(sb);
3082 ext4_group_t bg, first_meta_bg;
3085 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3087 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3088 return logical_sb_block + nr + 1;
3089 bg = sbi->s_desc_per_block * nr;
3090 if (ext4_bg_has_super(sb, bg))
3094 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3095 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3096 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3099 if (sb->s_blocksize == 1024 && nr == 0 &&
3100 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3103 return (has_super + ext4_group_first_block_no(sb, bg));
3107 * ext4_get_stripe_size: Get the stripe size.
3108 * @sbi: In memory super block info
3110 * If we have specified it via mount option, then
3111 * use the mount option value. If the value specified at mount time is
3112 * greater than the blocks per group use the super block value.
3113 * If the super block value is greater than blocks per group return 0.
3114 * Allocator needs it be less than blocks per group.
3117 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3119 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3120 unsigned long stripe_width =
3121 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3124 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3125 ret = sbi->s_stripe;
3126 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3128 else if (stride && stride <= sbi->s_blocks_per_group)
3134 * If the stripe width is 1, this makes no sense and
3135 * we set it to 0 to turn off stripe handling code.
3144 * Check whether this filesystem can be mounted based on
3145 * the features present and the RDONLY/RDWR mount requested.
3146 * Returns 1 if this filesystem can be mounted as requested,
3147 * 0 if it cannot be.
3149 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3151 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3152 ext4_msg(sb, KERN_ERR,
3153 "Couldn't mount because of "
3154 "unsupported optional features (%x)",
3155 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3156 ~EXT4_FEATURE_INCOMPAT_SUPP));
3160 #ifndef CONFIG_UNICODE
3161 if (ext4_has_feature_casefold(sb)) {
3162 ext4_msg(sb, KERN_ERR,
3163 "Filesystem with casefold feature cannot be "
3164 "mounted without CONFIG_UNICODE");
3172 if (ext4_has_feature_readonly(sb)) {
3173 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3174 sb->s_flags |= SB_RDONLY;
3178 /* Check that feature set is OK for a read-write mount */
3179 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3180 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3181 "unsupported optional features (%x)",
3182 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3183 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3186 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3187 ext4_msg(sb, KERN_ERR,
3188 "Can't support bigalloc feature without "
3189 "extents feature\n");
3193 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3194 if (!readonly && (ext4_has_feature_quota(sb) ||
3195 ext4_has_feature_project(sb))) {
3196 ext4_msg(sb, KERN_ERR,
3197 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3200 #endif /* CONFIG_QUOTA */
3205 * This function is called once a day if we have errors logged
3206 * on the file system
3208 static void print_daily_error_info(struct timer_list *t)
3210 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3211 struct super_block *sb = sbi->s_sb;
3212 struct ext4_super_block *es = sbi->s_es;
3214 if (es->s_error_count)
3215 /* fsck newer than v1.41.13 is needed to clean this condition. */
3216 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3217 le32_to_cpu(es->s_error_count));
3218 if (es->s_first_error_time) {
3219 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3221 ext4_get_tstamp(es, s_first_error_time),
3222 (int) sizeof(es->s_first_error_func),
3223 es->s_first_error_func,
3224 le32_to_cpu(es->s_first_error_line));
3225 if (es->s_first_error_ino)
3226 printk(KERN_CONT ": inode %u",
3227 le32_to_cpu(es->s_first_error_ino));
3228 if (es->s_first_error_block)
3229 printk(KERN_CONT ": block %llu", (unsigned long long)
3230 le64_to_cpu(es->s_first_error_block));
3231 printk(KERN_CONT "\n");
3233 if (es->s_last_error_time) {
3234 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3236 ext4_get_tstamp(es, s_last_error_time),
3237 (int) sizeof(es->s_last_error_func),
3238 es->s_last_error_func,
3239 le32_to_cpu(es->s_last_error_line));
3240 if (es->s_last_error_ino)
3241 printk(KERN_CONT ": inode %u",
3242 le32_to_cpu(es->s_last_error_ino));
3243 if (es->s_last_error_block)
3244 printk(KERN_CONT ": block %llu", (unsigned long long)
3245 le64_to_cpu(es->s_last_error_block));
3246 printk(KERN_CONT "\n");
3248 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3251 /* Find next suitable group and run ext4_init_inode_table */
3252 static int ext4_run_li_request(struct ext4_li_request *elr)
3254 struct ext4_group_desc *gdp = NULL;
3255 struct super_block *sb = elr->lr_super;
3256 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3257 ext4_group_t group = elr->lr_next_group;
3258 unsigned long timeout = 0;
3259 unsigned int prefetch_ios = 0;
3262 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3263 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3264 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3266 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3268 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3270 if (group >= elr->lr_next_group) {
3272 if (elr->lr_first_not_zeroed != ngroups &&
3273 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3274 elr->lr_next_group = elr->lr_first_not_zeroed;
3275 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3282 for (; group < ngroups; group++) {
3283 gdp = ext4_get_group_desc(sb, group, NULL);
3289 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3293 if (group >= ngroups)
3298 ret = ext4_init_inode_table(sb, group,
3299 elr->lr_timeout ? 0 : 1);
3300 trace_ext4_lazy_itable_init(sb, group);
3301 if (elr->lr_timeout == 0) {
3302 timeout = (jiffies - timeout) *
3303 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3304 elr->lr_timeout = timeout;
3306 elr->lr_next_sched = jiffies + elr->lr_timeout;
3307 elr->lr_next_group = group + 1;
3313 * Remove lr_request from the list_request and free the
3314 * request structure. Should be called with li_list_mtx held
3316 static void ext4_remove_li_request(struct ext4_li_request *elr)
3321 list_del(&elr->lr_request);
3322 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3326 static void ext4_unregister_li_request(struct super_block *sb)
3328 mutex_lock(&ext4_li_mtx);
3329 if (!ext4_li_info) {
3330 mutex_unlock(&ext4_li_mtx);
3334 mutex_lock(&ext4_li_info->li_list_mtx);
3335 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3336 mutex_unlock(&ext4_li_info->li_list_mtx);
3337 mutex_unlock(&ext4_li_mtx);
3340 static struct task_struct *ext4_lazyinit_task;
3343 * This is the function where ext4lazyinit thread lives. It walks
3344 * through the request list searching for next scheduled filesystem.
3345 * When such a fs is found, run the lazy initialization request
3346 * (ext4_rn_li_request) and keep track of the time spend in this
3347 * function. Based on that time we compute next schedule time of
3348 * the request. When walking through the list is complete, compute
3349 * next waking time and put itself into sleep.
3351 static int ext4_lazyinit_thread(void *arg)
3353 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3354 struct list_head *pos, *n;
3355 struct ext4_li_request *elr;
3356 unsigned long next_wakeup, cur;
3358 BUG_ON(NULL == eli);
3362 next_wakeup = MAX_JIFFY_OFFSET;
3364 mutex_lock(&eli->li_list_mtx);
3365 if (list_empty(&eli->li_request_list)) {
3366 mutex_unlock(&eli->li_list_mtx);
3369 list_for_each_safe(pos, n, &eli->li_request_list) {
3372 elr = list_entry(pos, struct ext4_li_request,
3375 if (time_before(jiffies, elr->lr_next_sched)) {
3376 if (time_before(elr->lr_next_sched, next_wakeup))
3377 next_wakeup = elr->lr_next_sched;
3380 if (down_read_trylock(&elr->lr_super->s_umount)) {
3381 if (sb_start_write_trylock(elr->lr_super)) {
3384 * We hold sb->s_umount, sb can not
3385 * be removed from the list, it is
3386 * now safe to drop li_list_mtx
3388 mutex_unlock(&eli->li_list_mtx);
3389 err = ext4_run_li_request(elr);
3390 sb_end_write(elr->lr_super);
3391 mutex_lock(&eli->li_list_mtx);
3394 up_read((&elr->lr_super->s_umount));
3396 /* error, remove the lazy_init job */
3398 ext4_remove_li_request(elr);
3402 elr->lr_next_sched = jiffies +
3404 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3406 if (time_before(elr->lr_next_sched, next_wakeup))
3407 next_wakeup = elr->lr_next_sched;
3409 mutex_unlock(&eli->li_list_mtx);
3414 if ((time_after_eq(cur, next_wakeup)) ||
3415 (MAX_JIFFY_OFFSET == next_wakeup)) {
3420 schedule_timeout_interruptible(next_wakeup - cur);
3422 if (kthread_should_stop()) {
3423 ext4_clear_request_list();
3430 * It looks like the request list is empty, but we need
3431 * to check it under the li_list_mtx lock, to prevent any
3432 * additions into it, and of course we should lock ext4_li_mtx
3433 * to atomically free the list and ext4_li_info, because at
3434 * this point another ext4 filesystem could be registering
3437 mutex_lock(&ext4_li_mtx);
3438 mutex_lock(&eli->li_list_mtx);
3439 if (!list_empty(&eli->li_request_list)) {
3440 mutex_unlock(&eli->li_list_mtx);
3441 mutex_unlock(&ext4_li_mtx);
3444 mutex_unlock(&eli->li_list_mtx);
3445 kfree(ext4_li_info);
3446 ext4_li_info = NULL;
3447 mutex_unlock(&ext4_li_mtx);
3452 static void ext4_clear_request_list(void)
3454 struct list_head *pos, *n;
3455 struct ext4_li_request *elr;
3457 mutex_lock(&ext4_li_info->li_list_mtx);
3458 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3459 elr = list_entry(pos, struct ext4_li_request,
3461 ext4_remove_li_request(elr);
3463 mutex_unlock(&ext4_li_info->li_list_mtx);
3466 static int ext4_run_lazyinit_thread(void)
3468 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3469 ext4_li_info, "ext4lazyinit");
3470 if (IS_ERR(ext4_lazyinit_task)) {
3471 int err = PTR_ERR(ext4_lazyinit_task);
3472 ext4_clear_request_list();
3473 kfree(ext4_li_info);
3474 ext4_li_info = NULL;
3475 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3476 "initialization thread\n",
3480 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3485 * Check whether it make sense to run itable init. thread or not.
3486 * If there is at least one uninitialized inode table, return
3487 * corresponding group number, else the loop goes through all
3488 * groups and return total number of groups.
3490 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3492 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3493 struct ext4_group_desc *gdp = NULL;
3495 if (!ext4_has_group_desc_csum(sb))
3498 for (group = 0; group < ngroups; group++) {
3499 gdp = ext4_get_group_desc(sb, group, NULL);
3503 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3510 static int ext4_li_info_new(void)
3512 struct ext4_lazy_init *eli = NULL;
3514 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3518 INIT_LIST_HEAD(&eli->li_request_list);
3519 mutex_init(&eli->li_list_mtx);
3521 eli->li_state |= EXT4_LAZYINIT_QUIT;
3528 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3531 struct ext4_li_request *elr;
3533 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3538 elr->lr_first_not_zeroed = start;
3539 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3540 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3541 elr->lr_next_group = start;
3543 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3547 * Randomize first schedule time of the request to
3548 * spread the inode table initialization requests
3551 elr->lr_next_sched = jiffies + (prandom_u32() %
3552 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3556 int ext4_register_li_request(struct super_block *sb,
3557 ext4_group_t first_not_zeroed)
3559 struct ext4_sb_info *sbi = EXT4_SB(sb);
3560 struct ext4_li_request *elr = NULL;
3561 ext4_group_t ngroups = sbi->s_groups_count;
3564 mutex_lock(&ext4_li_mtx);
3565 if (sbi->s_li_request != NULL) {
3567 * Reset timeout so it can be computed again, because
3568 * s_li_wait_mult might have changed.
3570 sbi->s_li_request->lr_timeout = 0;
3574 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3575 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3576 !test_opt(sb, INIT_INODE_TABLE)))
3579 elr = ext4_li_request_new(sb, first_not_zeroed);
3585 if (NULL == ext4_li_info) {
3586 ret = ext4_li_info_new();
3591 mutex_lock(&ext4_li_info->li_list_mtx);
3592 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3593 mutex_unlock(&ext4_li_info->li_list_mtx);
3595 sbi->s_li_request = elr;
3597 * set elr to NULL here since it has been inserted to
3598 * the request_list and the removal and free of it is
3599 * handled by ext4_clear_request_list from now on.
3603 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3604 ret = ext4_run_lazyinit_thread();
3609 mutex_unlock(&ext4_li_mtx);
3616 * We do not need to lock anything since this is called on
3619 static void ext4_destroy_lazyinit_thread(void)
3622 * If thread exited earlier
3623 * there's nothing to be done.
3625 if (!ext4_li_info || !ext4_lazyinit_task)
3628 kthread_stop(ext4_lazyinit_task);
3631 static int set_journal_csum_feature_set(struct super_block *sb)
3634 int compat, incompat;
3635 struct ext4_sb_info *sbi = EXT4_SB(sb);
3637 if (ext4_has_metadata_csum(sb)) {
3638 /* journal checksum v3 */
3640 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3642 /* journal checksum v1 */
3643 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3647 jbd2_journal_clear_features(sbi->s_journal,
3648 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3649 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3650 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3651 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3652 ret = jbd2_journal_set_features(sbi->s_journal,
3654 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3656 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3657 ret = jbd2_journal_set_features(sbi->s_journal,
3660 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3661 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3663 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3664 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3671 * Note: calculating the overhead so we can be compatible with
3672 * historical BSD practice is quite difficult in the face of
3673 * clusters/bigalloc. This is because multiple metadata blocks from
3674 * different block group can end up in the same allocation cluster.
3675 * Calculating the exact overhead in the face of clustered allocation
3676 * requires either O(all block bitmaps) in memory or O(number of block
3677 * groups**2) in time. We will still calculate the superblock for
3678 * older file systems --- and if we come across with a bigalloc file
3679 * system with zero in s_overhead_clusters the estimate will be close to
3680 * correct especially for very large cluster sizes --- but for newer
3681 * file systems, it's better to calculate this figure once at mkfs
3682 * time, and store it in the superblock. If the superblock value is
3683 * present (even for non-bigalloc file systems), we will use it.
3685 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3688 struct ext4_sb_info *sbi = EXT4_SB(sb);
3689 struct ext4_group_desc *gdp;
3690 ext4_fsblk_t first_block, last_block, b;
3691 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3692 int s, j, count = 0;
3694 if (!ext4_has_feature_bigalloc(sb))
3695 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3696 sbi->s_itb_per_group + 2);
3698 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3699 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3700 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3701 for (i = 0; i < ngroups; i++) {
3702 gdp = ext4_get_group_desc(sb, i, NULL);
3703 b = ext4_block_bitmap(sb, gdp);
3704 if (b >= first_block && b <= last_block) {
3705 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3708 b = ext4_inode_bitmap(sb, gdp);
3709 if (b >= first_block && b <= last_block) {
3710 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3713 b = ext4_inode_table(sb, gdp);
3714 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3715 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3716 int c = EXT4_B2C(sbi, b - first_block);
3717 ext4_set_bit(c, buf);
3723 if (ext4_bg_has_super(sb, grp)) {
3724 ext4_set_bit(s++, buf);
3727 j = ext4_bg_num_gdb(sb, grp);
3728 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3729 ext4_error(sb, "Invalid number of block group "
3730 "descriptor blocks: %d", j);
3731 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3735 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3739 return EXT4_CLUSTERS_PER_GROUP(sb) -
3740 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3744 * Compute the overhead and stash it in sbi->s_overhead
3746 int ext4_calculate_overhead(struct super_block *sb)
3748 struct ext4_sb_info *sbi = EXT4_SB(sb);
3749 struct ext4_super_block *es = sbi->s_es;
3750 struct inode *j_inode;
3751 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3752 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3753 ext4_fsblk_t overhead = 0;
3754 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3760 * Compute the overhead (FS structures). This is constant
3761 * for a given filesystem unless the number of block groups
3762 * changes so we cache the previous value until it does.
3766 * All of the blocks before first_data_block are overhead
3768 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3771 * Add the overhead found in each block group
3773 for (i = 0; i < ngroups; i++) {
3776 blks = count_overhead(sb, i, buf);
3779 memset(buf, 0, PAGE_SIZE);
3784 * Add the internal journal blocks whether the journal has been
3787 if (sbi->s_journal && !sbi->s_journal_bdev)
3788 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3789 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3790 /* j_inum for internal journal is non-zero */
3791 j_inode = ext4_get_journal_inode(sb, j_inum);
3793 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3794 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3797 ext4_msg(sb, KERN_ERR, "can't get journal size");
3800 sbi->s_overhead = overhead;
3802 free_page((unsigned long) buf);
3806 static void ext4_set_resv_clusters(struct super_block *sb)
3808 ext4_fsblk_t resv_clusters;
3809 struct ext4_sb_info *sbi = EXT4_SB(sb);
3812 * There's no need to reserve anything when we aren't using extents.
3813 * The space estimates are exact, there are no unwritten extents,
3814 * hole punching doesn't need new metadata... This is needed especially
3815 * to keep ext2/3 backward compatibility.
3817 if (!ext4_has_feature_extents(sb))
3820 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3821 * This should cover the situations where we can not afford to run
3822 * out of space like for example punch hole, or converting
3823 * unwritten extents in delalloc path. In most cases such
3824 * allocation would require 1, or 2 blocks, higher numbers are
3827 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3828 sbi->s_cluster_bits);
3830 do_div(resv_clusters, 50);
3831 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3833 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3836 static const char *ext4_quota_mode(struct super_block *sb)
3839 if (!ext4_quota_capable(sb))
3842 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3843 return "journalled";
3851 static void ext4_setup_csum_trigger(struct super_block *sb,
3852 enum ext4_journal_trigger_type type,
3854 struct jbd2_buffer_trigger_type *type,
3855 struct buffer_head *bh,
3859 struct ext4_sb_info *sbi = EXT4_SB(sb);
3861 sbi->s_journal_triggers[type].sb = sb;
3862 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
3865 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3867 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3868 char *orig_data = kstrdup(data, GFP_KERNEL);
3869 struct buffer_head *bh, **group_desc;
3870 struct ext4_super_block *es = NULL;
3871 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3872 struct flex_groups **flex_groups;
3874 ext4_fsblk_t sb_block = get_sb_block(&data);
3875 ext4_fsblk_t logical_sb_block;
3876 unsigned long offset = 0;
3877 unsigned long def_mount_opts;
3881 int blocksize, clustersize;
3882 unsigned int db_count;
3884 int needs_recovery, has_huge_files;
3887 ext4_group_t first_not_zeroed;
3888 struct ext4_parsed_options parsed_opts;
3890 /* Set defaults for the variables that will be set during parsing */
3891 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3892 parsed_opts.journal_devnum = 0;
3893 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
3895 if ((data && !orig_data) || !sbi)
3898 sbi->s_daxdev = dax_dev;
3899 sbi->s_blockgroup_lock =
3900 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3901 if (!sbi->s_blockgroup_lock)
3904 sb->s_fs_info = sbi;
3906 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3907 sbi->s_sb_block = sb_block;
3908 sbi->s_sectors_written_start =
3909 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
3911 /* Cleanup superblock name */
3912 strreplace(sb->s_id, '/', '!');
3914 /* -EINVAL is default */
3916 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3918 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3923 * The ext4 superblock will not be buffer aligned for other than 1kB
3924 * block sizes. We need to calculate the offset from buffer start.
3926 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3927 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3928 offset = do_div(logical_sb_block, blocksize);
3930 logical_sb_block = sb_block;
3933 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
3935 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3940 * Note: s_es must be initialized as soon as possible because
3941 * some ext4 macro-instructions depend on its value
3943 es = (struct ext4_super_block *) (bh->b_data + offset);
3945 sb->s_magic = le16_to_cpu(es->s_magic);
3946 if (sb->s_magic != EXT4_SUPER_MAGIC)
3948 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3950 /* Warn if metadata_csum and gdt_csum are both set. */
3951 if (ext4_has_feature_metadata_csum(sb) &&
3952 ext4_has_feature_gdt_csum(sb))
3953 ext4_warning(sb, "metadata_csum and uninit_bg are "
3954 "redundant flags; please run fsck.");
3956 /* Check for a known checksum algorithm */
3957 if (!ext4_verify_csum_type(sb, es)) {
3958 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3959 "unknown checksum algorithm.");
3964 /* Load the checksum driver */
3965 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3966 if (IS_ERR(sbi->s_chksum_driver)) {
3967 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3968 ret = PTR_ERR(sbi->s_chksum_driver);
3969 sbi->s_chksum_driver = NULL;
3973 /* Check superblock checksum */
3974 if (!ext4_superblock_csum_verify(sb, es)) {
3975 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3976 "invalid superblock checksum. Run e2fsck?");
3982 /* Precompute checksum seed for all metadata */
3983 if (ext4_has_feature_csum_seed(sb))
3984 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3985 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3986 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3987 sizeof(es->s_uuid));
3989 /* Set defaults before we parse the mount options */
3990 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3991 set_opt(sb, INIT_INODE_TABLE);
3992 if (def_mount_opts & EXT4_DEFM_DEBUG)
3994 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3996 if (def_mount_opts & EXT4_DEFM_UID16)
3997 set_opt(sb, NO_UID32);
3998 /* xattr user namespace & acls are now defaulted on */
3999 set_opt(sb, XATTR_USER);
4000 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4001 set_opt(sb, POSIX_ACL);
4003 if (ext4_has_feature_fast_commit(sb))
4004 set_opt2(sb, JOURNAL_FAST_COMMIT);
4005 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4006 if (ext4_has_metadata_csum(sb))
4007 set_opt(sb, JOURNAL_CHECKSUM);
4009 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4010 set_opt(sb, JOURNAL_DATA);
4011 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4012 set_opt(sb, ORDERED_DATA);
4013 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4014 set_opt(sb, WRITEBACK_DATA);
4016 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4017 set_opt(sb, ERRORS_PANIC);
4018 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4019 set_opt(sb, ERRORS_CONT);
4021 set_opt(sb, ERRORS_RO);
4022 /* block_validity enabled by default; disable with noblock_validity */
4023 set_opt(sb, BLOCK_VALIDITY);
4024 if (def_mount_opts & EXT4_DEFM_DISCARD)
4025 set_opt(sb, DISCARD);
4027 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4028 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4029 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4030 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4031 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4033 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4034 set_opt(sb, BARRIER);
4037 * enable delayed allocation by default
4038 * Use -o nodelalloc to turn it off
4040 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4041 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4042 set_opt(sb, DELALLOC);
4045 * set default s_li_wait_mult for lazyinit, for the case there is
4046 * no mount option specified.
4048 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4050 if (le32_to_cpu(es->s_log_block_size) >
4051 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4052 ext4_msg(sb, KERN_ERR,
4053 "Invalid log block size: %u",
4054 le32_to_cpu(es->s_log_block_size));
4057 if (le32_to_cpu(es->s_log_cluster_size) >
4058 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4059 ext4_msg(sb, KERN_ERR,
4060 "Invalid log cluster size: %u",
4061 le32_to_cpu(es->s_log_cluster_size));
4065 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4067 if (blocksize == PAGE_SIZE)
4068 set_opt(sb, DIOREAD_NOLOCK);
4070 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4071 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4072 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4074 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4075 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4076 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4077 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4081 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4082 (!is_power_of_2(sbi->s_inode_size)) ||
4083 (sbi->s_inode_size > blocksize)) {
4084 ext4_msg(sb, KERN_ERR,
4085 "unsupported inode size: %d",
4087 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4091 * i_atime_extra is the last extra field available for
4092 * [acm]times in struct ext4_inode. Checking for that
4093 * field should suffice to ensure we have extra space
4096 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4097 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4098 sb->s_time_gran = 1;
4099 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4101 sb->s_time_gran = NSEC_PER_SEC;
4102 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4104 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4106 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4107 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4108 EXT4_GOOD_OLD_INODE_SIZE;
4109 if (ext4_has_feature_extra_isize(sb)) {
4110 unsigned v, max = (sbi->s_inode_size -
4111 EXT4_GOOD_OLD_INODE_SIZE);
4113 v = le16_to_cpu(es->s_want_extra_isize);
4115 ext4_msg(sb, KERN_ERR,
4116 "bad s_want_extra_isize: %d", v);
4119 if (sbi->s_want_extra_isize < v)
4120 sbi->s_want_extra_isize = v;
4122 v = le16_to_cpu(es->s_min_extra_isize);
4124 ext4_msg(sb, KERN_ERR,
4125 "bad s_min_extra_isize: %d", v);
4128 if (sbi->s_want_extra_isize < v)
4129 sbi->s_want_extra_isize = v;
4133 if (sbi->s_es->s_mount_opts[0]) {
4134 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4135 sizeof(sbi->s_es->s_mount_opts),
4139 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4140 ext4_msg(sb, KERN_WARNING,
4141 "failed to parse options in superblock: %s",
4144 kfree(s_mount_opts);
4146 sbi->s_def_mount_opt = sbi->s_mount_opt;
4147 if (!parse_options((char *) data, sb, &parsed_opts, 0))
4150 #ifdef CONFIG_UNICODE
4151 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4152 const struct ext4_sb_encodings *encoding_info;
4153 struct unicode_map *encoding;
4154 __u16 encoding_flags;
4156 if (ext4_sb_read_encoding(es, &encoding_info,
4158 ext4_msg(sb, KERN_ERR,
4159 "Encoding requested by superblock is unknown");
4163 encoding = utf8_load(encoding_info->version);
4164 if (IS_ERR(encoding)) {
4165 ext4_msg(sb, KERN_ERR,
4166 "can't mount with superblock charset: %s-%s "
4167 "not supported by the kernel. flags: 0x%x.",
4168 encoding_info->name, encoding_info->version,
4172 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4173 "%s-%s with flags 0x%hx", encoding_info->name,
4174 encoding_info->version?:"\b", encoding_flags);
4176 sb->s_encoding = encoding;
4177 sb->s_encoding_flags = encoding_flags;
4181 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4182 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4183 /* can't mount with both data=journal and dioread_nolock. */
4184 clear_opt(sb, DIOREAD_NOLOCK);
4185 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4186 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4187 ext4_msg(sb, KERN_ERR, "can't mount with "
4188 "both data=journal and delalloc");
4191 if (test_opt(sb, DAX_ALWAYS)) {
4192 ext4_msg(sb, KERN_ERR, "can't mount with "
4193 "both data=journal and dax");
4196 if (ext4_has_feature_encrypt(sb)) {
4197 ext4_msg(sb, KERN_WARNING,
4198 "encrypted files will use data=ordered "
4199 "instead of data journaling mode");
4201 if (test_opt(sb, DELALLOC))
4202 clear_opt(sb, DELALLOC);
4204 sb->s_iflags |= SB_I_CGROUPWB;
4207 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4208 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4210 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4211 (ext4_has_compat_features(sb) ||
4212 ext4_has_ro_compat_features(sb) ||
4213 ext4_has_incompat_features(sb)))
4214 ext4_msg(sb, KERN_WARNING,
4215 "feature flags set on rev 0 fs, "
4216 "running e2fsck is recommended");
4218 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4219 set_opt2(sb, HURD_COMPAT);
4220 if (ext4_has_feature_64bit(sb)) {
4221 ext4_msg(sb, KERN_ERR,
4222 "The Hurd can't support 64-bit file systems");
4227 * ea_inode feature uses l_i_version field which is not
4228 * available in HURD_COMPAT mode.
4230 if (ext4_has_feature_ea_inode(sb)) {
4231 ext4_msg(sb, KERN_ERR,
4232 "ea_inode feature is not supported for Hurd");
4237 if (IS_EXT2_SB(sb)) {
4238 if (ext2_feature_set_ok(sb))
4239 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4240 "using the ext4 subsystem");
4243 * If we're probing be silent, if this looks like
4244 * it's actually an ext[34] filesystem.
4246 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4248 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4249 "to feature incompatibilities");
4254 if (IS_EXT3_SB(sb)) {
4255 if (ext3_feature_set_ok(sb))
4256 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4257 "using the ext4 subsystem");
4260 * If we're probing be silent, if this looks like
4261 * it's actually an ext4 filesystem.
4263 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4265 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4266 "to feature incompatibilities");
4272 * Check feature flags regardless of the revision level, since we
4273 * previously didn't change the revision level when setting the flags,
4274 * so there is a chance incompat flags are set on a rev 0 filesystem.
4276 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4279 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4280 ext4_msg(sb, KERN_ERR,
4281 "Number of reserved GDT blocks insanely large: %d",
4282 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4286 if (bdev_dax_supported(sb->s_bdev, blocksize))
4287 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4289 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4290 if (ext4_has_feature_inline_data(sb)) {
4291 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4292 " that may contain inline data");
4295 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4296 ext4_msg(sb, KERN_ERR,
4297 "DAX unsupported by block device.");
4302 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4303 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4304 es->s_encryption_level);
4308 if (sb->s_blocksize != blocksize) {
4310 * bh must be released before kill_bdev(), otherwise
4311 * it won't be freed and its page also. kill_bdev()
4312 * is called by sb_set_blocksize().
4315 /* Validate the filesystem blocksize */
4316 if (!sb_set_blocksize(sb, blocksize)) {
4317 ext4_msg(sb, KERN_ERR, "bad block size %d",
4323 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4324 offset = do_div(logical_sb_block, blocksize);
4325 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4327 ext4_msg(sb, KERN_ERR,
4328 "Can't read superblock on 2nd try");
4333 es = (struct ext4_super_block *)(bh->b_data + offset);
4335 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4336 ext4_msg(sb, KERN_ERR,
4337 "Magic mismatch, very weird!");
4342 has_huge_files = ext4_has_feature_huge_file(sb);
4343 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4345 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4347 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4348 if (ext4_has_feature_64bit(sb)) {
4349 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4350 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4351 !is_power_of_2(sbi->s_desc_size)) {
4352 ext4_msg(sb, KERN_ERR,
4353 "unsupported descriptor size %lu",
4358 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4360 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4361 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4363 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4364 if (sbi->s_inodes_per_block == 0)
4366 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4367 sbi->s_inodes_per_group > blocksize * 8) {
4368 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4369 sbi->s_inodes_per_group);
4372 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4373 sbi->s_inodes_per_block;
4374 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4376 sbi->s_mount_state = le16_to_cpu(es->s_state);
4377 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4378 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4380 for (i = 0; i < 4; i++)
4381 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4382 sbi->s_def_hash_version = es->s_def_hash_version;
4383 if (ext4_has_feature_dir_index(sb)) {
4384 i = le32_to_cpu(es->s_flags);
4385 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4386 sbi->s_hash_unsigned = 3;
4387 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4388 #ifdef __CHAR_UNSIGNED__
4391 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4392 sbi->s_hash_unsigned = 3;
4396 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4401 /* Handle clustersize */
4402 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4403 if (ext4_has_feature_bigalloc(sb)) {
4404 if (clustersize < blocksize) {
4405 ext4_msg(sb, KERN_ERR,
4406 "cluster size (%d) smaller than "
4407 "block size (%d)", clustersize, blocksize);
4410 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4411 le32_to_cpu(es->s_log_block_size);
4412 sbi->s_clusters_per_group =
4413 le32_to_cpu(es->s_clusters_per_group);
4414 if (sbi->s_clusters_per_group > blocksize * 8) {
4415 ext4_msg(sb, KERN_ERR,
4416 "#clusters per group too big: %lu",
4417 sbi->s_clusters_per_group);
4420 if (sbi->s_blocks_per_group !=
4421 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4422 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4423 "clusters per group (%lu) inconsistent",
4424 sbi->s_blocks_per_group,
4425 sbi->s_clusters_per_group);
4429 if (clustersize != blocksize) {
4430 ext4_msg(sb, KERN_ERR,
4431 "fragment/cluster size (%d) != "
4432 "block size (%d)", clustersize, blocksize);
4435 if (sbi->s_blocks_per_group > blocksize * 8) {
4436 ext4_msg(sb, KERN_ERR,
4437 "#blocks per group too big: %lu",
4438 sbi->s_blocks_per_group);
4441 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4442 sbi->s_cluster_bits = 0;
4444 sbi->s_cluster_ratio = clustersize / blocksize;
4446 /* Do we have standard group size of clustersize * 8 blocks ? */
4447 if (sbi->s_blocks_per_group == clustersize << 3)
4448 set_opt2(sb, STD_GROUP_SIZE);
4451 * Test whether we have more sectors than will fit in sector_t,
4452 * and whether the max offset is addressable by the page cache.
4454 err = generic_check_addressable(sb->s_blocksize_bits,
4455 ext4_blocks_count(es));
4457 ext4_msg(sb, KERN_ERR, "filesystem"
4458 " too large to mount safely on this system");
4462 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4465 /* check blocks count against device size */
4466 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4467 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4468 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4469 "exceeds size of device (%llu blocks)",
4470 ext4_blocks_count(es), blocks_count);
4475 * It makes no sense for the first data block to be beyond the end
4476 * of the filesystem.
4478 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4479 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4480 "block %u is beyond end of filesystem (%llu)",
4481 le32_to_cpu(es->s_first_data_block),
4482 ext4_blocks_count(es));
4485 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4486 (sbi->s_cluster_ratio == 1)) {
4487 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4488 "block is 0 with a 1k block and cluster size");
4492 blocks_count = (ext4_blocks_count(es) -
4493 le32_to_cpu(es->s_first_data_block) +
4494 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4495 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4496 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4497 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4498 "(block count %llu, first data block %u, "
4499 "blocks per group %lu)", blocks_count,
4500 ext4_blocks_count(es),
4501 le32_to_cpu(es->s_first_data_block),
4502 EXT4_BLOCKS_PER_GROUP(sb));
4505 sbi->s_groups_count = blocks_count;
4506 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4507 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4508 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4509 le32_to_cpu(es->s_inodes_count)) {
4510 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4511 le32_to_cpu(es->s_inodes_count),
4512 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4516 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4517 EXT4_DESC_PER_BLOCK(sb);
4518 if (ext4_has_feature_meta_bg(sb)) {
4519 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4520 ext4_msg(sb, KERN_WARNING,
4521 "first meta block group too large: %u "
4522 "(group descriptor block count %u)",
4523 le32_to_cpu(es->s_first_meta_bg), db_count);
4527 rcu_assign_pointer(sbi->s_group_desc,
4528 kvmalloc_array(db_count,
4529 sizeof(struct buffer_head *),
4531 if (sbi->s_group_desc == NULL) {
4532 ext4_msg(sb, KERN_ERR, "not enough memory");
4537 bgl_lock_init(sbi->s_blockgroup_lock);
4539 /* Pre-read the descriptors into the buffer cache */
4540 for (i = 0; i < db_count; i++) {
4541 block = descriptor_loc(sb, logical_sb_block, i);
4542 ext4_sb_breadahead_unmovable(sb, block);
4545 for (i = 0; i < db_count; i++) {
4546 struct buffer_head *bh;
4548 block = descriptor_loc(sb, logical_sb_block, i);
4549 bh = ext4_sb_bread_unmovable(sb, block);
4551 ext4_msg(sb, KERN_ERR,
4552 "can't read group descriptor %d", i);
4558 rcu_dereference(sbi->s_group_desc)[i] = bh;
4561 sbi->s_gdb_count = db_count;
4562 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4563 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4564 ret = -EFSCORRUPTED;
4568 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4569 spin_lock_init(&sbi->s_error_lock);
4570 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4572 /* Register extent status tree shrinker */
4573 if (ext4_es_register_shrinker(sbi))
4576 sbi->s_stripe = ext4_get_stripe_size(sbi);
4577 sbi->s_extent_max_zeroout_kb = 32;
4580 * set up enough so that it can read an inode
4582 sb->s_op = &ext4_sops;
4583 sb->s_export_op = &ext4_export_ops;
4584 sb->s_xattr = ext4_xattr_handlers;
4585 #ifdef CONFIG_FS_ENCRYPTION
4586 sb->s_cop = &ext4_cryptops;
4588 #ifdef CONFIG_FS_VERITY
4589 sb->s_vop = &ext4_verityops;
4592 sb->dq_op = &ext4_quota_operations;
4593 if (ext4_has_feature_quota(sb))
4594 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4596 sb->s_qcop = &ext4_qctl_operations;
4597 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4599 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4601 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4602 mutex_init(&sbi->s_orphan_lock);
4604 /* Initialize fast commit stuff */
4605 atomic_set(&sbi->s_fc_subtid, 0);
4606 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4607 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4608 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4609 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4610 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4611 sbi->s_fc_bytes = 0;
4612 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4613 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4614 spin_lock_init(&sbi->s_fc_lock);
4615 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4616 sbi->s_fc_replay_state.fc_regions = NULL;
4617 sbi->s_fc_replay_state.fc_regions_size = 0;
4618 sbi->s_fc_replay_state.fc_regions_used = 0;
4619 sbi->s_fc_replay_state.fc_regions_valid = 0;
4620 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4621 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4622 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4626 needs_recovery = (es->s_last_orphan != 0 ||
4627 ext4_has_feature_journal_needs_recovery(sb));
4629 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4630 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4631 goto failed_mount3a;
4634 * The first inode we look at is the journal inode. Don't try
4635 * root first: it may be modified in the journal!
4637 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4638 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4640 goto failed_mount3a;
4641 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4642 ext4_has_feature_journal_needs_recovery(sb)) {
4643 ext4_msg(sb, KERN_ERR, "required journal recovery "
4644 "suppressed and not mounted read-only");
4645 goto failed_mount_wq;
4647 /* Nojournal mode, all journal mount options are illegal */
4648 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4649 ext4_msg(sb, KERN_ERR, "can't mount with "
4650 "journal_checksum, fs mounted w/o journal");
4651 goto failed_mount_wq;
4653 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4654 ext4_msg(sb, KERN_ERR, "can't mount with "
4655 "journal_async_commit, fs mounted w/o journal");
4656 goto failed_mount_wq;
4658 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4659 ext4_msg(sb, KERN_ERR, "can't mount with "
4660 "commit=%lu, fs mounted w/o journal",
4661 sbi->s_commit_interval / HZ);
4662 goto failed_mount_wq;
4664 if (EXT4_MOUNT_DATA_FLAGS &
4665 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4666 ext4_msg(sb, KERN_ERR, "can't mount with "
4667 "data=, fs mounted w/o journal");
4668 goto failed_mount_wq;
4670 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4671 clear_opt(sb, JOURNAL_CHECKSUM);
4672 clear_opt(sb, DATA_FLAGS);
4673 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4674 sbi->s_journal = NULL;
4679 if (ext4_has_feature_64bit(sb) &&
4680 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4681 JBD2_FEATURE_INCOMPAT_64BIT)) {
4682 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4683 goto failed_mount_wq;
4686 if (!set_journal_csum_feature_set(sb)) {
4687 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4689 goto failed_mount_wq;
4692 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4693 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4694 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4695 ext4_msg(sb, KERN_ERR,
4696 "Failed to set fast commit journal feature");
4697 goto failed_mount_wq;
4700 /* We have now updated the journal if required, so we can
4701 * validate the data journaling mode. */
4702 switch (test_opt(sb, DATA_FLAGS)) {
4704 /* No mode set, assume a default based on the journal
4705 * capabilities: ORDERED_DATA if the journal can
4706 * cope, else JOURNAL_DATA
4708 if (jbd2_journal_check_available_features
4709 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4710 set_opt(sb, ORDERED_DATA);
4711 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4713 set_opt(sb, JOURNAL_DATA);
4714 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4718 case EXT4_MOUNT_ORDERED_DATA:
4719 case EXT4_MOUNT_WRITEBACK_DATA:
4720 if (!jbd2_journal_check_available_features
4721 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4722 ext4_msg(sb, KERN_ERR, "Journal does not support "
4723 "requested data journaling mode");
4724 goto failed_mount_wq;
4731 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4732 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4733 ext4_msg(sb, KERN_ERR, "can't mount with "
4734 "journal_async_commit in data=ordered mode");
4735 goto failed_mount_wq;
4738 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4740 sbi->s_journal->j_submit_inode_data_buffers =
4741 ext4_journal_submit_inode_data_buffers;
4742 sbi->s_journal->j_finish_inode_data_buffers =
4743 ext4_journal_finish_inode_data_buffers;
4746 if (!test_opt(sb, NO_MBCACHE)) {
4747 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4748 if (!sbi->s_ea_block_cache) {
4749 ext4_msg(sb, KERN_ERR,
4750 "Failed to create ea_block_cache");
4751 goto failed_mount_wq;
4754 if (ext4_has_feature_ea_inode(sb)) {
4755 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4756 if (!sbi->s_ea_inode_cache) {
4757 ext4_msg(sb, KERN_ERR,
4758 "Failed to create ea_inode_cache");
4759 goto failed_mount_wq;
4764 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4765 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4766 goto failed_mount_wq;
4769 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4770 !ext4_has_feature_encrypt(sb)) {
4771 ext4_set_feature_encrypt(sb);
4772 ext4_commit_super(sb);
4776 * Get the # of file system overhead blocks from the
4777 * superblock if present.
4779 if (es->s_overhead_clusters)
4780 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4782 err = ext4_calculate_overhead(sb);
4784 goto failed_mount_wq;
4788 * The maximum number of concurrent works can be high and
4789 * concurrency isn't really necessary. Limit it to 1.
4791 EXT4_SB(sb)->rsv_conversion_wq =
4792 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4793 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4794 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4800 * The jbd2_journal_load will have done any necessary log recovery,
4801 * so we can safely mount the rest of the filesystem now.
4804 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4806 ext4_msg(sb, KERN_ERR, "get root inode failed");
4807 ret = PTR_ERR(root);
4811 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4812 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4817 sb->s_root = d_make_root(root);
4819 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4824 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4825 if (ret == -EROFS) {
4826 sb->s_flags |= SB_RDONLY;
4829 goto failed_mount4a;
4831 ext4_set_resv_clusters(sb);
4833 if (test_opt(sb, BLOCK_VALIDITY)) {
4834 err = ext4_setup_system_zone(sb);
4836 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4838 goto failed_mount4a;
4841 ext4_fc_replay_cleanup(sb);
4846 * Enable optimize_scan if number of groups is > threshold. This can be
4847 * turned off by passing "mb_optimize_scan=0". This can also be
4848 * turned on forcefully by passing "mb_optimize_scan=1".
4850 if (parsed_opts.mb_optimize_scan == 1)
4851 set_opt2(sb, MB_OPTIMIZE_SCAN);
4852 else if (parsed_opts.mb_optimize_scan == 0)
4853 clear_opt2(sb, MB_OPTIMIZE_SCAN);
4854 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
4855 set_opt2(sb, MB_OPTIMIZE_SCAN);
4857 err = ext4_mb_init(sb);
4859 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4865 * We can only set up the journal commit callback once
4866 * mballoc is initialized
4869 sbi->s_journal->j_commit_callback =
4870 ext4_journal_commit_callback;
4872 block = ext4_count_free_clusters(sb);
4873 ext4_free_blocks_count_set(sbi->s_es,
4874 EXT4_C2B(sbi, block));
4875 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4878 unsigned long freei = ext4_count_free_inodes(sb);
4879 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4880 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4884 * Update the checksum after updating free space/inode
4885 * counters. Otherwise the superblock can have an incorrect
4886 * checksum in the buffer cache until it is written out and
4887 * e2fsprogs programs trying to open a file system immediately
4888 * after it is mounted can fail.
4890 ext4_superblock_csum_set(sb);
4892 err = percpu_counter_init(&sbi->s_dirs_counter,
4893 ext4_count_dirs(sb), GFP_KERNEL);
4895 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4898 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
4901 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4904 ext4_msg(sb, KERN_ERR, "insufficient memory");
4908 if (ext4_has_feature_flex_bg(sb))
4909 if (!ext4_fill_flex_info(sb)) {
4910 ext4_msg(sb, KERN_ERR,
4911 "unable to initialize "
4912 "flex_bg meta info!");
4917 err = ext4_register_li_request(sb, first_not_zeroed);
4921 err = ext4_register_sysfs(sb);
4926 /* Enable quota usage during mount. */
4927 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4928 err = ext4_enable_quotas(sb);
4932 #endif /* CONFIG_QUOTA */
4935 * Save the original bdev mapping's wb_err value which could be
4936 * used to detect the metadata async write error.
4938 spin_lock_init(&sbi->s_bdev_wb_lock);
4939 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4940 &sbi->s_bdev_wb_err);
4941 sb->s_bdev->bd_super = sb;
4942 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4943 ext4_orphan_cleanup(sb, es);
4944 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4945 if (needs_recovery) {
4946 ext4_msg(sb, KERN_INFO, "recovery complete");
4947 err = ext4_mark_recovery_complete(sb, es);
4951 if (EXT4_SB(sb)->s_journal) {
4952 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4953 descr = " journalled data mode";
4954 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4955 descr = " ordered data mode";
4957 descr = " writeback data mode";
4959 descr = "out journal";
4961 if (test_opt(sb, DISCARD)) {
4962 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4963 if (!blk_queue_discard(q))
4964 ext4_msg(sb, KERN_WARNING,
4965 "mounting with \"discard\" option, but "
4966 "the device does not support discard");
4969 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4970 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4971 "Opts: %.*s%s%s. Quota mode: %s.", descr,
4972 (int) sizeof(sbi->s_es->s_mount_opts),
4973 sbi->s_es->s_mount_opts,
4974 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
4975 ext4_quota_mode(sb));
4977 if (es->s_error_count)
4978 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4980 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4981 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4982 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4983 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4984 atomic_set(&sbi->s_warning_count, 0);
4985 atomic_set(&sbi->s_msg_count, 0);
4992 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4996 ext4_unregister_sysfs(sb);
4997 kobject_put(&sbi->s_kobj);
4999 ext4_unregister_li_request(sb);
5001 ext4_mb_release(sb);
5003 flex_groups = rcu_dereference(sbi->s_flex_groups);
5005 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5006 kvfree(flex_groups[i]);
5007 kvfree(flex_groups);
5010 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5011 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5012 percpu_counter_destroy(&sbi->s_dirs_counter);
5013 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5014 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5015 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5017 ext4_ext_release(sb);
5018 ext4_release_system_zone(sb);
5023 ext4_msg(sb, KERN_ERR, "mount failed");
5024 if (EXT4_SB(sb)->rsv_conversion_wq)
5025 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5027 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5028 sbi->s_ea_inode_cache = NULL;
5030 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5031 sbi->s_ea_block_cache = NULL;
5033 if (sbi->s_journal) {
5034 jbd2_journal_destroy(sbi->s_journal);
5035 sbi->s_journal = NULL;
5038 ext4_es_unregister_shrinker(sbi);
5040 flush_work(&sbi->s_error_work);
5041 del_timer_sync(&sbi->s_err_report);
5042 ext4_stop_mmpd(sbi);
5045 group_desc = rcu_dereference(sbi->s_group_desc);
5046 for (i = 0; i < db_count; i++)
5047 brelse(group_desc[i]);
5051 if (sbi->s_chksum_driver)
5052 crypto_free_shash(sbi->s_chksum_driver);
5054 #ifdef CONFIG_UNICODE
5055 utf8_unload(sb->s_encoding);
5059 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5060 kfree(get_qf_name(sb, sbi, i));
5062 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5063 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5065 ext4_blkdev_remove(sbi);
5067 sb->s_fs_info = NULL;
5068 kfree(sbi->s_blockgroup_lock);
5072 fs_put_dax(dax_dev);
5073 return err ? err : ret;
5077 * Setup any per-fs journal parameters now. We'll do this both on
5078 * initial mount, once the journal has been initialised but before we've
5079 * done any recovery; and again on any subsequent remount.
5081 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5083 struct ext4_sb_info *sbi = EXT4_SB(sb);
5085 journal->j_commit_interval = sbi->s_commit_interval;
5086 journal->j_min_batch_time = sbi->s_min_batch_time;
5087 journal->j_max_batch_time = sbi->s_max_batch_time;
5088 ext4_fc_init(sb, journal);
5090 write_lock(&journal->j_state_lock);
5091 if (test_opt(sb, BARRIER))
5092 journal->j_flags |= JBD2_BARRIER;
5094 journal->j_flags &= ~JBD2_BARRIER;
5095 if (test_opt(sb, DATA_ERR_ABORT))
5096 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5098 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5099 write_unlock(&journal->j_state_lock);
5102 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5103 unsigned int journal_inum)
5105 struct inode *journal_inode;
5108 * Test for the existence of a valid inode on disk. Bad things
5109 * happen if we iget() an unused inode, as the subsequent iput()
5110 * will try to delete it.
5112 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5113 if (IS_ERR(journal_inode)) {
5114 ext4_msg(sb, KERN_ERR, "no journal found");
5117 if (!journal_inode->i_nlink) {
5118 make_bad_inode(journal_inode);
5119 iput(journal_inode);
5120 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5124 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5125 journal_inode, journal_inode->i_size);
5126 if (!S_ISREG(journal_inode->i_mode)) {
5127 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5128 iput(journal_inode);
5131 return journal_inode;
5134 static journal_t *ext4_get_journal(struct super_block *sb,
5135 unsigned int journal_inum)
5137 struct inode *journal_inode;
5140 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5143 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5147 journal = jbd2_journal_init_inode(journal_inode);
5149 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5150 iput(journal_inode);
5153 journal->j_private = sb;
5154 ext4_init_journal_params(sb, journal);
5158 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5161 struct buffer_head *bh;
5165 int hblock, blocksize;
5166 ext4_fsblk_t sb_block;
5167 unsigned long offset;
5168 struct ext4_super_block *es;
5169 struct block_device *bdev;
5171 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5174 bdev = ext4_blkdev_get(j_dev, sb);
5178 blocksize = sb->s_blocksize;
5179 hblock = bdev_logical_block_size(bdev);
5180 if (blocksize < hblock) {
5181 ext4_msg(sb, KERN_ERR,
5182 "blocksize too small for journal device");
5186 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5187 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5188 set_blocksize(bdev, blocksize);
5189 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5190 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5191 "external journal");
5195 es = (struct ext4_super_block *) (bh->b_data + offset);
5196 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5197 !(le32_to_cpu(es->s_feature_incompat) &
5198 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5199 ext4_msg(sb, KERN_ERR, "external journal has "
5205 if ((le32_to_cpu(es->s_feature_ro_compat) &
5206 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5207 es->s_checksum != ext4_superblock_csum(sb, es)) {
5208 ext4_msg(sb, KERN_ERR, "external journal has "
5209 "corrupt superblock");
5214 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5215 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5220 len = ext4_blocks_count(es);
5221 start = sb_block + 1;
5222 brelse(bh); /* we're done with the superblock */
5224 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5225 start, len, blocksize);
5227 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5230 journal->j_private = sb;
5231 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5232 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5235 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5236 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5237 "user (unsupported) - %d",
5238 be32_to_cpu(journal->j_superblock->s_nr_users));
5241 EXT4_SB(sb)->s_journal_bdev = bdev;
5242 ext4_init_journal_params(sb, journal);
5246 jbd2_journal_destroy(journal);
5248 ext4_blkdev_put(bdev);
5252 static int ext4_load_journal(struct super_block *sb,
5253 struct ext4_super_block *es,
5254 unsigned long journal_devnum)
5257 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5260 int really_read_only;
5263 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5264 return -EFSCORRUPTED;
5266 if (journal_devnum &&
5267 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5268 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5269 "numbers have changed");
5270 journal_dev = new_decode_dev(journal_devnum);
5272 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5274 if (journal_inum && journal_dev) {
5275 ext4_msg(sb, KERN_ERR,
5276 "filesystem has both journal inode and journal device!");
5281 journal = ext4_get_journal(sb, journal_inum);
5285 journal = ext4_get_dev_journal(sb, journal_dev);
5290 journal_dev_ro = bdev_read_only(journal->j_dev);
5291 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5293 if (journal_dev_ro && !sb_rdonly(sb)) {
5294 ext4_msg(sb, KERN_ERR,
5295 "journal device read-only, try mounting with '-o ro'");
5301 * Are we loading a blank journal or performing recovery after a
5302 * crash? For recovery, we need to check in advance whether we
5303 * can get read-write access to the device.
5305 if (ext4_has_feature_journal_needs_recovery(sb)) {
5306 if (sb_rdonly(sb)) {
5307 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5308 "required on readonly filesystem");
5309 if (really_read_only) {
5310 ext4_msg(sb, KERN_ERR, "write access "
5311 "unavailable, cannot proceed "
5312 "(try mounting with noload)");
5316 ext4_msg(sb, KERN_INFO, "write access will "
5317 "be enabled during recovery");
5321 if (!(journal->j_flags & JBD2_BARRIER))
5322 ext4_msg(sb, KERN_INFO, "barriers disabled");
5324 if (!ext4_has_feature_journal_needs_recovery(sb))
5325 err = jbd2_journal_wipe(journal, !really_read_only);
5327 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5329 memcpy(save, ((char *) es) +
5330 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5331 err = jbd2_journal_load(journal);
5333 memcpy(((char *) es) + EXT4_S_ERR_START,
5334 save, EXT4_S_ERR_LEN);
5339 ext4_msg(sb, KERN_ERR, "error loading journal");
5343 EXT4_SB(sb)->s_journal = journal;
5344 err = ext4_clear_journal_err(sb, es);
5346 EXT4_SB(sb)->s_journal = NULL;
5347 jbd2_journal_destroy(journal);
5351 if (!really_read_only && journal_devnum &&
5352 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5353 es->s_journal_dev = cpu_to_le32(journal_devnum);
5355 /* Make sure we flush the recovery flag to disk. */
5356 ext4_commit_super(sb);
5362 jbd2_journal_destroy(journal);
5366 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5367 static void ext4_update_super(struct super_block *sb)
5369 struct ext4_sb_info *sbi = EXT4_SB(sb);
5370 struct ext4_super_block *es = sbi->s_es;
5371 struct buffer_head *sbh = sbi->s_sbh;
5375 * If the file system is mounted read-only, don't update the
5376 * superblock write time. This avoids updating the superblock
5377 * write time when we are mounting the root file system
5378 * read/only but we need to replay the journal; at that point,
5379 * for people who are east of GMT and who make their clock
5380 * tick in localtime for Windows bug-for-bug compatibility,
5381 * the clock is set in the future, and this will cause e2fsck
5382 * to complain and force a full file system check.
5384 if (!(sb->s_flags & SB_RDONLY))
5385 ext4_update_tstamp(es, s_wtime);
5386 es->s_kbytes_written =
5387 cpu_to_le64(sbi->s_kbytes_written +
5388 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5389 sbi->s_sectors_written_start) >> 1));
5390 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5391 ext4_free_blocks_count_set(es,
5392 EXT4_C2B(sbi, percpu_counter_sum_positive(
5393 &sbi->s_freeclusters_counter)));
5394 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5395 es->s_free_inodes_count =
5396 cpu_to_le32(percpu_counter_sum_positive(
5397 &sbi->s_freeinodes_counter));
5398 /* Copy error information to the on-disk superblock */
5399 spin_lock(&sbi->s_error_lock);
5400 if (sbi->s_add_error_count > 0) {
5401 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5402 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5403 __ext4_update_tstamp(&es->s_first_error_time,
5404 &es->s_first_error_time_hi,
5405 sbi->s_first_error_time);
5406 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5407 sizeof(es->s_first_error_func));
5408 es->s_first_error_line =
5409 cpu_to_le32(sbi->s_first_error_line);
5410 es->s_first_error_ino =
5411 cpu_to_le32(sbi->s_first_error_ino);
5412 es->s_first_error_block =
5413 cpu_to_le64(sbi->s_first_error_block);
5414 es->s_first_error_errcode =
5415 ext4_errno_to_code(sbi->s_first_error_code);
5417 __ext4_update_tstamp(&es->s_last_error_time,
5418 &es->s_last_error_time_hi,
5419 sbi->s_last_error_time);
5420 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5421 sizeof(es->s_last_error_func));
5422 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5423 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5424 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5425 es->s_last_error_errcode =
5426 ext4_errno_to_code(sbi->s_last_error_code);
5428 * Start the daily error reporting function if it hasn't been
5431 if (!es->s_error_count)
5432 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5433 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5434 sbi->s_add_error_count = 0;
5436 spin_unlock(&sbi->s_error_lock);
5438 ext4_superblock_csum_set(sb);
5442 static int ext4_commit_super(struct super_block *sb)
5444 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5449 if (block_device_ejected(sb))
5452 ext4_update_super(sb);
5454 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5456 * Oh, dear. A previous attempt to write the
5457 * superblock failed. This could happen because the
5458 * USB device was yanked out. Or it could happen to
5459 * be a transient write error and maybe the block will
5460 * be remapped. Nothing we can do but to retry the
5461 * write and hope for the best.
5463 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5464 "superblock detected");
5465 clear_buffer_write_io_error(sbh);
5466 set_buffer_uptodate(sbh);
5468 BUFFER_TRACE(sbh, "marking dirty");
5469 mark_buffer_dirty(sbh);
5470 error = __sync_dirty_buffer(sbh,
5471 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5472 if (buffer_write_io_error(sbh)) {
5473 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5475 clear_buffer_write_io_error(sbh);
5476 set_buffer_uptodate(sbh);
5482 * Have we just finished recovery? If so, and if we are mounting (or
5483 * remounting) the filesystem readonly, then we will end up with a
5484 * consistent fs on disk. Record that fact.
5486 static int ext4_mark_recovery_complete(struct super_block *sb,
5487 struct ext4_super_block *es)
5490 journal_t *journal = EXT4_SB(sb)->s_journal;
5492 if (!ext4_has_feature_journal(sb)) {
5493 if (journal != NULL) {
5494 ext4_error(sb, "Journal got removed while the fs was "
5496 return -EFSCORRUPTED;
5500 jbd2_journal_lock_updates(journal);
5501 err = jbd2_journal_flush(journal, 0);
5505 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5506 ext4_clear_feature_journal_needs_recovery(sb);
5507 ext4_commit_super(sb);
5510 jbd2_journal_unlock_updates(journal);
5515 * If we are mounting (or read-write remounting) a filesystem whose journal
5516 * has recorded an error from a previous lifetime, move that error to the
5517 * main filesystem now.
5519 static int ext4_clear_journal_err(struct super_block *sb,
5520 struct ext4_super_block *es)
5526 if (!ext4_has_feature_journal(sb)) {
5527 ext4_error(sb, "Journal got removed while the fs was mounted!");
5528 return -EFSCORRUPTED;
5531 journal = EXT4_SB(sb)->s_journal;
5534 * Now check for any error status which may have been recorded in the
5535 * journal by a prior ext4_error() or ext4_abort()
5538 j_errno = jbd2_journal_errno(journal);
5542 errstr = ext4_decode_error(sb, j_errno, nbuf);
5543 ext4_warning(sb, "Filesystem error recorded "
5544 "from previous mount: %s", errstr);
5545 ext4_warning(sb, "Marking fs in need of filesystem check.");
5547 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5548 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5549 ext4_commit_super(sb);
5551 jbd2_journal_clear_err(journal);
5552 jbd2_journal_update_sb_errno(journal);
5558 * Force the running and committing transactions to commit,
5559 * and wait on the commit.
5561 int ext4_force_commit(struct super_block *sb)
5568 journal = EXT4_SB(sb)->s_journal;
5569 return ext4_journal_force_commit(journal);
5572 static int ext4_sync_fs(struct super_block *sb, int wait)
5576 bool needs_barrier = false;
5577 struct ext4_sb_info *sbi = EXT4_SB(sb);
5579 if (unlikely(ext4_forced_shutdown(sbi)))
5582 trace_ext4_sync_fs(sb, wait);
5583 flush_workqueue(sbi->rsv_conversion_wq);
5585 * Writeback quota in non-journalled quota case - journalled quota has
5588 dquot_writeback_dquots(sb, -1);
5590 * Data writeback is possible w/o journal transaction, so barrier must
5591 * being sent at the end of the function. But we can skip it if
5592 * transaction_commit will do it for us.
5594 if (sbi->s_journal) {
5595 target = jbd2_get_latest_transaction(sbi->s_journal);
5596 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5597 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5598 needs_barrier = true;
5600 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5602 ret = jbd2_log_wait_commit(sbi->s_journal,
5605 } else if (wait && test_opt(sb, BARRIER))
5606 needs_barrier = true;
5607 if (needs_barrier) {
5609 err = blkdev_issue_flush(sb->s_bdev);
5618 * LVM calls this function before a (read-only) snapshot is created. This
5619 * gives us a chance to flush the journal completely and mark the fs clean.
5621 * Note that only this function cannot bring a filesystem to be in a clean
5622 * state independently. It relies on upper layer to stop all data & metadata
5625 static int ext4_freeze(struct super_block *sb)
5633 journal = EXT4_SB(sb)->s_journal;
5636 /* Now we set up the journal barrier. */
5637 jbd2_journal_lock_updates(journal);
5640 * Don't clear the needs_recovery flag if we failed to
5641 * flush the journal.
5643 error = jbd2_journal_flush(journal, 0);
5647 /* Journal blocked and flushed, clear needs_recovery flag. */
5648 ext4_clear_feature_journal_needs_recovery(sb);
5651 error = ext4_commit_super(sb);
5654 /* we rely on upper layer to stop further updates */
5655 jbd2_journal_unlock_updates(journal);
5660 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5661 * flag here, even though the filesystem is not technically dirty yet.
5663 static int ext4_unfreeze(struct super_block *sb)
5665 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5668 if (EXT4_SB(sb)->s_journal) {
5669 /* Reset the needs_recovery flag before the fs is unlocked. */
5670 ext4_set_feature_journal_needs_recovery(sb);
5673 ext4_commit_super(sb);
5678 * Structure to save mount options for ext4_remount's benefit
5680 struct ext4_mount_options {
5681 unsigned long s_mount_opt;
5682 unsigned long s_mount_opt2;
5685 unsigned long s_commit_interval;
5686 u32 s_min_batch_time, s_max_batch_time;
5689 char *s_qf_names[EXT4_MAXQUOTAS];
5693 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5695 struct ext4_super_block *es;
5696 struct ext4_sb_info *sbi = EXT4_SB(sb);
5697 unsigned long old_sb_flags, vfs_flags;
5698 struct ext4_mount_options old_opts;
5699 int enable_quota = 0;
5704 char *to_free[EXT4_MAXQUOTAS];
5706 char *orig_data = kstrdup(data, GFP_KERNEL);
5707 struct ext4_parsed_options parsed_opts;
5709 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5710 parsed_opts.journal_devnum = 0;
5712 if (data && !orig_data)
5715 /* Store the original options */
5716 old_sb_flags = sb->s_flags;
5717 old_opts.s_mount_opt = sbi->s_mount_opt;
5718 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5719 old_opts.s_resuid = sbi->s_resuid;
5720 old_opts.s_resgid = sbi->s_resgid;
5721 old_opts.s_commit_interval = sbi->s_commit_interval;
5722 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5723 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5725 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5726 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5727 if (sbi->s_qf_names[i]) {
5728 char *qf_name = get_qf_name(sb, sbi, i);
5730 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5731 if (!old_opts.s_qf_names[i]) {
5732 for (j = 0; j < i; j++)
5733 kfree(old_opts.s_qf_names[j]);
5738 old_opts.s_qf_names[i] = NULL;
5740 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5741 parsed_opts.journal_ioprio =
5742 sbi->s_journal->j_task->io_context->ioprio;
5745 * Some options can be enabled by ext4 and/or by VFS mount flag
5746 * either way we need to make sure it matches in both *flags and
5747 * s_flags. Copy those selected flags from *flags to s_flags
5749 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5750 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5752 if (!parse_options(data, sb, &parsed_opts, 1)) {
5757 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5758 test_opt(sb, JOURNAL_CHECKSUM)) {
5759 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5760 "during remount not supported; ignoring");
5761 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5764 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5765 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5766 ext4_msg(sb, KERN_ERR, "can't mount with "
5767 "both data=journal and delalloc");
5771 if (test_opt(sb, DIOREAD_NOLOCK)) {
5772 ext4_msg(sb, KERN_ERR, "can't mount with "
5773 "both data=journal and dioread_nolock");
5777 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5778 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5779 ext4_msg(sb, KERN_ERR, "can't mount with "
5780 "journal_async_commit in data=ordered mode");
5786 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5787 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5792 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5793 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5795 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5796 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5800 if (sbi->s_journal) {
5801 ext4_init_journal_params(sb, sbi->s_journal);
5802 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5805 /* Flush outstanding errors before changing fs state */
5806 flush_work(&sbi->s_error_work);
5808 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5809 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5814 if (*flags & SB_RDONLY) {
5815 err = sync_filesystem(sb);
5818 err = dquot_suspend(sb, -1);
5823 * First of all, the unconditional stuff we have to do
5824 * to disable replay of the journal when we next remount
5826 sb->s_flags |= SB_RDONLY;
5829 * OK, test if we are remounting a valid rw partition
5830 * readonly, and if so set the rdonly flag and then
5831 * mark the partition as valid again.
5833 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5834 (sbi->s_mount_state & EXT4_VALID_FS))
5835 es->s_state = cpu_to_le16(sbi->s_mount_state);
5837 if (sbi->s_journal) {
5839 * We let remount-ro finish even if marking fs
5840 * as clean failed...
5842 ext4_mark_recovery_complete(sb, es);
5845 /* Make sure we can mount this feature set readwrite */
5846 if (ext4_has_feature_readonly(sb) ||
5847 !ext4_feature_set_ok(sb, 0)) {
5852 * Make sure the group descriptor checksums
5853 * are sane. If they aren't, refuse to remount r/w.
5855 for (g = 0; g < sbi->s_groups_count; g++) {
5856 struct ext4_group_desc *gdp =
5857 ext4_get_group_desc(sb, g, NULL);
5859 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5860 ext4_msg(sb, KERN_ERR,
5861 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5862 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5863 le16_to_cpu(gdp->bg_checksum));
5870 * If we have an unprocessed orphan list hanging
5871 * around from a previously readonly bdev mount,
5872 * require a full umount/remount for now.
5874 if (es->s_last_orphan) {
5875 ext4_msg(sb, KERN_WARNING, "Couldn't "
5876 "remount RDWR because of unprocessed "
5877 "orphan inode list. Please "
5878 "umount/remount instead");
5884 * Mounting a RDONLY partition read-write, so reread
5885 * and store the current valid flag. (It may have
5886 * been changed by e2fsck since we originally mounted
5889 if (sbi->s_journal) {
5890 err = ext4_clear_journal_err(sb, es);
5894 sbi->s_mount_state = le16_to_cpu(es->s_state);
5896 err = ext4_setup_super(sb, es, 0);
5900 sb->s_flags &= ~SB_RDONLY;
5901 if (ext4_has_feature_mmp(sb))
5902 if (ext4_multi_mount_protect(sb,
5903 le64_to_cpu(es->s_mmp_block))) {
5912 * Reinitialize lazy itable initialization thread based on
5915 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5916 ext4_unregister_li_request(sb);
5918 ext4_group_t first_not_zeroed;
5919 first_not_zeroed = ext4_has_uninit_itable(sb);
5920 ext4_register_li_request(sb, first_not_zeroed);
5924 * Handle creation of system zone data early because it can fail.
5925 * Releasing of existing data is done when we are sure remount will
5928 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5929 err = ext4_setup_system_zone(sb);
5934 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5935 err = ext4_commit_super(sb);
5941 /* Release old quota file names */
5942 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5943 kfree(old_opts.s_qf_names[i]);
5945 if (sb_any_quota_suspended(sb))
5946 dquot_resume(sb, -1);
5947 else if (ext4_has_feature_quota(sb)) {
5948 err = ext4_enable_quotas(sb);
5954 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5955 ext4_release_system_zone(sb);
5957 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
5958 ext4_stop_mmpd(sbi);
5961 * Some options can be enabled by ext4 and/or by VFS mount flag
5962 * either way we need to make sure it matches in both *flags and
5963 * s_flags. Copy those selected flags from s_flags to *flags
5965 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5967 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
5968 orig_data, ext4_quota_mode(sb));
5973 sb->s_flags = old_sb_flags;
5974 sbi->s_mount_opt = old_opts.s_mount_opt;
5975 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5976 sbi->s_resuid = old_opts.s_resuid;
5977 sbi->s_resgid = old_opts.s_resgid;
5978 sbi->s_commit_interval = old_opts.s_commit_interval;
5979 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5980 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5981 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5982 ext4_release_system_zone(sb);
5984 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5985 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5986 to_free[i] = get_qf_name(sb, sbi, i);
5987 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5990 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5993 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
5994 ext4_stop_mmpd(sbi);
6000 static int ext4_statfs_project(struct super_block *sb,
6001 kprojid_t projid, struct kstatfs *buf)
6004 struct dquot *dquot;
6008 qid = make_kqid_projid(projid);
6009 dquot = dqget(sb, qid);
6011 return PTR_ERR(dquot);
6012 spin_lock(&dquot->dq_dqb_lock);
6014 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6015 dquot->dq_dqb.dqb_bhardlimit);
6016 limit >>= sb->s_blocksize_bits;
6018 if (limit && buf->f_blocks > limit) {
6019 curblock = (dquot->dq_dqb.dqb_curspace +
6020 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6021 buf->f_blocks = limit;
6022 buf->f_bfree = buf->f_bavail =
6023 (buf->f_blocks > curblock) ?
6024 (buf->f_blocks - curblock) : 0;
6027 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6028 dquot->dq_dqb.dqb_ihardlimit);
6029 if (limit && buf->f_files > limit) {
6030 buf->f_files = limit;
6032 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6033 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6036 spin_unlock(&dquot->dq_dqb_lock);
6042 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6044 struct super_block *sb = dentry->d_sb;
6045 struct ext4_sb_info *sbi = EXT4_SB(sb);
6046 struct ext4_super_block *es = sbi->s_es;
6047 ext4_fsblk_t overhead = 0, resv_blocks;
6049 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6051 if (!test_opt(sb, MINIX_DF))
6052 overhead = sbi->s_overhead;
6054 buf->f_type = EXT4_SUPER_MAGIC;
6055 buf->f_bsize = sb->s_blocksize;
6056 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6057 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6058 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6059 /* prevent underflow in case that few free space is available */
6060 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6061 buf->f_bavail = buf->f_bfree -
6062 (ext4_r_blocks_count(es) + resv_blocks);
6063 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6065 buf->f_files = le32_to_cpu(es->s_inodes_count);
6066 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6067 buf->f_namelen = EXT4_NAME_LEN;
6068 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6071 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6072 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6073 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6082 * Helper functions so that transaction is started before we acquire dqio_sem
6083 * to keep correct lock ordering of transaction > dqio_sem
6085 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6087 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6090 static int ext4_write_dquot(struct dquot *dquot)
6094 struct inode *inode;
6096 inode = dquot_to_inode(dquot);
6097 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6098 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6100 return PTR_ERR(handle);
6101 ret = dquot_commit(dquot);
6102 err = ext4_journal_stop(handle);
6108 static int ext4_acquire_dquot(struct dquot *dquot)
6113 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6114 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6116 return PTR_ERR(handle);
6117 ret = dquot_acquire(dquot);
6118 err = ext4_journal_stop(handle);
6124 static int ext4_release_dquot(struct dquot *dquot)
6129 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6130 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6131 if (IS_ERR(handle)) {
6132 /* Release dquot anyway to avoid endless cycle in dqput() */
6133 dquot_release(dquot);
6134 return PTR_ERR(handle);
6136 ret = dquot_release(dquot);
6137 err = ext4_journal_stop(handle);
6143 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6145 struct super_block *sb = dquot->dq_sb;
6147 if (ext4_is_quota_journalled(sb)) {
6148 dquot_mark_dquot_dirty(dquot);
6149 return ext4_write_dquot(dquot);
6151 return dquot_mark_dquot_dirty(dquot);
6155 static int ext4_write_info(struct super_block *sb, int type)
6160 /* Data block + inode block */
6161 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6163 return PTR_ERR(handle);
6164 ret = dquot_commit_info(sb, type);
6165 err = ext4_journal_stop(handle);
6171 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6173 struct ext4_inode_info *ei = EXT4_I(inode);
6175 /* The first argument of lockdep_set_subclass has to be
6176 * *exactly* the same as the argument to init_rwsem() --- in
6177 * this case, in init_once() --- or lockdep gets unhappy
6178 * because the name of the lock is set using the
6179 * stringification of the argument to init_rwsem().
6181 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6182 lockdep_set_subclass(&ei->i_data_sem, subclass);
6186 * Standard function to be called on quota_on
6188 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6189 const struct path *path)
6193 if (!test_opt(sb, QUOTA))
6196 /* Quotafile not on the same filesystem? */
6197 if (path->dentry->d_sb != sb)
6200 /* Quota already enabled for this file? */
6201 if (IS_NOQUOTA(d_inode(path->dentry)))
6204 /* Journaling quota? */
6205 if (EXT4_SB(sb)->s_qf_names[type]) {
6206 /* Quotafile not in fs root? */
6207 if (path->dentry->d_parent != sb->s_root)
6208 ext4_msg(sb, KERN_WARNING,
6209 "Quota file not on filesystem root. "
6210 "Journaled quota will not work");
6211 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6214 * Clear the flag just in case mount options changed since
6217 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6221 * When we journal data on quota file, we have to flush journal to see
6222 * all updates to the file when we bypass pagecache...
6224 if (EXT4_SB(sb)->s_journal &&
6225 ext4_should_journal_data(d_inode(path->dentry))) {
6227 * We don't need to lock updates but journal_flush() could
6228 * otherwise be livelocked...
6230 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6231 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6232 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6237 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6238 err = dquot_quota_on(sb, type, format_id, path);
6240 lockdep_set_quota_inode(path->dentry->d_inode,
6243 struct inode *inode = d_inode(path->dentry);
6247 * Set inode flags to prevent userspace from messing with quota
6248 * files. If this fails, we return success anyway since quotas
6249 * are already enabled and this is not a hard failure.
6252 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6255 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6256 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6257 S_NOATIME | S_IMMUTABLE);
6258 err = ext4_mark_inode_dirty(handle, inode);
6259 ext4_journal_stop(handle);
6261 inode_unlock(inode);
6266 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6270 struct inode *qf_inode;
6271 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6272 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6273 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6274 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6277 BUG_ON(!ext4_has_feature_quota(sb));
6279 if (!qf_inums[type])
6282 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6283 if (IS_ERR(qf_inode)) {
6284 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6285 return PTR_ERR(qf_inode);
6288 /* Don't account quota for quota files to avoid recursion */
6289 qf_inode->i_flags |= S_NOQUOTA;
6290 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6291 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6293 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6299 /* Enable usage tracking for all quota types. */
6300 int ext4_enable_quotas(struct super_block *sb)
6303 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6304 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6305 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6306 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6308 bool quota_mopt[EXT4_MAXQUOTAS] = {
6309 test_opt(sb, USRQUOTA),
6310 test_opt(sb, GRPQUOTA),
6311 test_opt(sb, PRJQUOTA),
6314 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6315 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6316 if (qf_inums[type]) {
6317 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6318 DQUOT_USAGE_ENABLED |
6319 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6322 "Failed to enable quota tracking "
6323 "(type=%d, err=%d). Please run "
6324 "e2fsck to fix.", type, err);
6325 for (type--; type >= 0; type--)
6326 dquot_quota_off(sb, type);
6335 static int ext4_quota_off(struct super_block *sb, int type)
6337 struct inode *inode = sb_dqopt(sb)->files[type];
6341 /* Force all delayed allocation blocks to be allocated.
6342 * Caller already holds s_umount sem */
6343 if (test_opt(sb, DELALLOC))
6344 sync_filesystem(sb);
6346 if (!inode || !igrab(inode))
6349 err = dquot_quota_off(sb, type);
6350 if (err || ext4_has_feature_quota(sb))
6355 * Update modification times of quota files when userspace can
6356 * start looking at them. If we fail, we return success anyway since
6357 * this is not a hard failure and quotas are already disabled.
6359 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6360 if (IS_ERR(handle)) {
6361 err = PTR_ERR(handle);
6364 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6365 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6366 inode->i_mtime = inode->i_ctime = current_time(inode);
6367 err = ext4_mark_inode_dirty(handle, inode);
6368 ext4_journal_stop(handle);
6370 inode_unlock(inode);
6372 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6376 return dquot_quota_off(sb, type);
6379 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6380 * acquiring the locks... As quota files are never truncated and quota code
6381 * itself serializes the operations (and no one else should touch the files)
6382 * we don't have to be afraid of races */
6383 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6384 size_t len, loff_t off)
6386 struct inode *inode = sb_dqopt(sb)->files[type];
6387 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6388 int offset = off & (sb->s_blocksize - 1);
6391 struct buffer_head *bh;
6392 loff_t i_size = i_size_read(inode);
6396 if (off+len > i_size)
6399 while (toread > 0) {
6400 tocopy = sb->s_blocksize - offset < toread ?
6401 sb->s_blocksize - offset : toread;
6402 bh = ext4_bread(NULL, inode, blk, 0);
6405 if (!bh) /* A hole? */
6406 memset(data, 0, tocopy);
6408 memcpy(data, bh->b_data+offset, tocopy);
6418 /* Write to quotafile (we know the transaction is already started and has
6419 * enough credits) */
6420 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6421 const char *data, size_t len, loff_t off)
6423 struct inode *inode = sb_dqopt(sb)->files[type];
6424 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6425 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6427 struct buffer_head *bh;
6428 handle_t *handle = journal_current_handle();
6430 if (EXT4_SB(sb)->s_journal && !handle) {
6431 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6432 " cancelled because transaction is not started",
6433 (unsigned long long)off, (unsigned long long)len);
6437 * Since we account only one data block in transaction credits,
6438 * then it is impossible to cross a block boundary.
6440 if (sb->s_blocksize - offset < len) {
6441 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6442 " cancelled because not block aligned",
6443 (unsigned long long)off, (unsigned long long)len);
6448 bh = ext4_bread(handle, inode, blk,
6449 EXT4_GET_BLOCKS_CREATE |
6450 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6451 } while (PTR_ERR(bh) == -ENOSPC &&
6452 ext4_should_retry_alloc(inode->i_sb, &retries));
6457 BUFFER_TRACE(bh, "get write access");
6458 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6464 memcpy(bh->b_data+offset, data, len);
6465 flush_dcache_page(bh->b_page);
6467 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6470 if (inode->i_size < off + len) {
6471 i_size_write(inode, off + len);
6472 EXT4_I(inode)->i_disksize = inode->i_size;
6473 err2 = ext4_mark_inode_dirty(handle, inode);
6474 if (unlikely(err2 && !err))
6477 return err ? err : len;
6481 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6482 const char *dev_name, void *data)
6484 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6487 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6488 static inline void register_as_ext2(void)
6490 int err = register_filesystem(&ext2_fs_type);
6493 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6496 static inline void unregister_as_ext2(void)
6498 unregister_filesystem(&ext2_fs_type);
6501 static inline int ext2_feature_set_ok(struct super_block *sb)
6503 if (ext4_has_unknown_ext2_incompat_features(sb))
6507 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6512 static inline void register_as_ext2(void) { }
6513 static inline void unregister_as_ext2(void) { }
6514 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6517 static inline void register_as_ext3(void)
6519 int err = register_filesystem(&ext3_fs_type);
6522 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6525 static inline void unregister_as_ext3(void)
6527 unregister_filesystem(&ext3_fs_type);
6530 static inline int ext3_feature_set_ok(struct super_block *sb)
6532 if (ext4_has_unknown_ext3_incompat_features(sb))
6534 if (!ext4_has_feature_journal(sb))
6538 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6543 static struct file_system_type ext4_fs_type = {
6544 .owner = THIS_MODULE,
6546 .mount = ext4_mount,
6547 .kill_sb = kill_block_super,
6548 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6550 MODULE_ALIAS_FS("ext4");
6552 /* Shared across all ext4 file systems */
6553 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6555 static int __init ext4_init_fs(void)
6559 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6560 ext4_li_info = NULL;
6562 /* Build-time check for flags consistency */
6563 ext4_check_flag_values();
6565 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6566 init_waitqueue_head(&ext4__ioend_wq[i]);
6568 err = ext4_init_es();
6572 err = ext4_init_pending();
6576 err = ext4_init_post_read_processing();
6580 err = ext4_init_pageio();
6584 err = ext4_init_system_zone();
6588 err = ext4_init_sysfs();
6592 err = ext4_init_mballoc();
6595 err = init_inodecache();
6599 err = ext4_fc_init_dentry_cache();
6605 err = register_filesystem(&ext4_fs_type);
6611 unregister_as_ext2();
6612 unregister_as_ext3();
6614 destroy_inodecache();
6616 ext4_exit_mballoc();
6620 ext4_exit_system_zone();
6624 ext4_exit_post_read_processing();
6626 ext4_exit_pending();
6633 static void __exit ext4_exit_fs(void)
6635 ext4_destroy_lazyinit_thread();
6636 unregister_as_ext2();
6637 unregister_as_ext3();
6638 unregister_filesystem(&ext4_fs_type);
6639 destroy_inodecache();
6640 ext4_exit_mballoc();
6642 ext4_exit_system_zone();
6644 ext4_exit_post_read_processing();
6646 ext4_exit_pending();
6649 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6650 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6651 MODULE_LICENSE("GPL");
6652 MODULE_SOFTDEP("pre: crc32c");
6653 module_init(ext4_init_fs)
6654 module_exit(ext4_exit_fs)