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);
93 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
94 * -> page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_lock
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
104 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
108 * sb_start_write -> i_mutex -> mmap_lock
109 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
112 * transaction start -> page lock(s) -> i_data_sem (rw)
115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
116 static struct file_system_type ext2_fs_type = {
117 .owner = THIS_MODULE,
120 .kill_sb = kill_block_super,
121 .fs_flags = FS_REQUIRES_DEV,
123 MODULE_ALIAS_FS("ext2");
124 MODULE_ALIAS("ext2");
125 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127 #define IS_EXT2_SB(sb) (0)
131 static struct file_system_type ext3_fs_type = {
132 .owner = THIS_MODULE,
135 .kill_sb = kill_block_super,
136 .fs_flags = FS_REQUIRES_DEV,
138 MODULE_ALIAS_FS("ext3");
139 MODULE_ALIAS("ext3");
140 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
147 * buffer's verified bit is no longer valid after reading from
148 * disk again due to write out error, clear it to make sure we
149 * recheck the buffer contents.
151 clear_buffer_verified(bh);
153 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
155 submit_bh(REQ_OP_READ, op_flags, bh);
158 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
161 BUG_ON(!buffer_locked(bh));
163 if (ext4_buffer_uptodate(bh)) {
167 __ext4_read_bh(bh, op_flags, end_io);
170 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
172 BUG_ON(!buffer_locked(bh));
174 if (ext4_buffer_uptodate(bh)) {
179 __ext4_read_bh(bh, op_flags, end_io);
182 if (buffer_uptodate(bh))
187 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
189 if (trylock_buffer(bh)) {
191 return ext4_read_bh(bh, op_flags, NULL);
192 ext4_read_bh_nowait(bh, op_flags, NULL);
197 if (buffer_uptodate(bh))
205 * This works like __bread_gfp() except it uses ERR_PTR for error
206 * returns. Currently with sb_bread it's impossible to distinguish
207 * between ENOMEM and EIO situations (since both result in a NULL
210 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
211 sector_t block, int op_flags,
214 struct buffer_head *bh;
217 bh = sb_getblk_gfp(sb, block, gfp);
219 return ERR_PTR(-ENOMEM);
220 if (ext4_buffer_uptodate(bh))
223 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
231 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
234 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
237 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
240 return __ext4_sb_bread_gfp(sb, block, 0, 0);
243 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
245 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
248 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
253 static int ext4_verify_csum_type(struct super_block *sb,
254 struct ext4_super_block *es)
256 if (!ext4_has_feature_metadata_csum(sb))
259 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
262 static __le32 ext4_superblock_csum(struct super_block *sb,
263 struct ext4_super_block *es)
265 struct ext4_sb_info *sbi = EXT4_SB(sb);
266 int offset = offsetof(struct ext4_super_block, s_checksum);
269 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
271 return cpu_to_le32(csum);
274 static int ext4_superblock_csum_verify(struct super_block *sb,
275 struct ext4_super_block *es)
277 if (!ext4_has_metadata_csum(sb))
280 return es->s_checksum == ext4_superblock_csum(sb, es);
283 void ext4_superblock_csum_set(struct super_block *sb)
285 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
287 if (!ext4_has_metadata_csum(sb))
290 es->s_checksum = ext4_superblock_csum(sb, es);
293 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
294 struct ext4_group_desc *bg)
296 return le32_to_cpu(bg->bg_block_bitmap_lo) |
297 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
298 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
301 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
302 struct ext4_group_desc *bg)
304 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
305 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
306 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
309 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
310 struct ext4_group_desc *bg)
312 return le32_to_cpu(bg->bg_inode_table_lo) |
313 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
314 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
317 __u32 ext4_free_group_clusters(struct super_block *sb,
318 struct ext4_group_desc *bg)
320 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
321 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
322 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
325 __u32 ext4_free_inodes_count(struct super_block *sb,
326 struct ext4_group_desc *bg)
328 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
329 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
330 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
333 __u32 ext4_used_dirs_count(struct super_block *sb,
334 struct ext4_group_desc *bg)
336 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
337 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
338 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
341 __u32 ext4_itable_unused_count(struct super_block *sb,
342 struct ext4_group_desc *bg)
344 return le16_to_cpu(bg->bg_itable_unused_lo) |
345 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
346 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
349 void ext4_block_bitmap_set(struct super_block *sb,
350 struct ext4_group_desc *bg, ext4_fsblk_t blk)
352 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
353 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
354 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
357 void ext4_inode_bitmap_set(struct super_block *sb,
358 struct ext4_group_desc *bg, ext4_fsblk_t blk)
360 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
361 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
362 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
365 void ext4_inode_table_set(struct super_block *sb,
366 struct ext4_group_desc *bg, ext4_fsblk_t blk)
368 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
369 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
370 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
373 void ext4_free_group_clusters_set(struct super_block *sb,
374 struct ext4_group_desc *bg, __u32 count)
376 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
377 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
378 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
381 void ext4_free_inodes_set(struct super_block *sb,
382 struct ext4_group_desc *bg, __u32 count)
384 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
385 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
386 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
389 void ext4_used_dirs_set(struct super_block *sb,
390 struct ext4_group_desc *bg, __u32 count)
392 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
393 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
394 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
397 void ext4_itable_unused_set(struct super_block *sb,
398 struct ext4_group_desc *bg, __u32 count)
400 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
401 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
402 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
405 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
407 now = clamp_val(now, 0, (1ull << 40) - 1);
409 *lo = cpu_to_le32(lower_32_bits(now));
410 *hi = upper_32_bits(now);
413 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
415 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
417 #define ext4_update_tstamp(es, tstamp) \
418 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
419 ktime_get_real_seconds())
420 #define ext4_get_tstamp(es, tstamp) \
421 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
424 * The del_gendisk() function uninitializes the disk-specific data
425 * structures, including the bdi structure, without telling anyone
426 * else. Once this happens, any attempt to call mark_buffer_dirty()
427 * (for example, by ext4_commit_super), will cause a kernel OOPS.
428 * This is a kludge to prevent these oops until we can put in a proper
429 * hook in del_gendisk() to inform the VFS and file system layers.
431 static int block_device_ejected(struct super_block *sb)
433 struct inode *bd_inode = sb->s_bdev->bd_inode;
434 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
436 return bdi->dev == NULL;
439 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
441 struct super_block *sb = journal->j_private;
442 struct ext4_sb_info *sbi = EXT4_SB(sb);
443 int error = is_journal_aborted(journal);
444 struct ext4_journal_cb_entry *jce;
446 BUG_ON(txn->t_state == T_FINISHED);
448 ext4_process_freed_data(sb, txn->t_tid);
450 spin_lock(&sbi->s_md_lock);
451 while (!list_empty(&txn->t_private_list)) {
452 jce = list_entry(txn->t_private_list.next,
453 struct ext4_journal_cb_entry, jce_list);
454 list_del_init(&jce->jce_list);
455 spin_unlock(&sbi->s_md_lock);
456 jce->jce_func(sb, jce, error);
457 spin_lock(&sbi->s_md_lock);
459 spin_unlock(&sbi->s_md_lock);
463 * This writepage callback for write_cache_pages()
464 * takes care of a few cases after page cleaning.
466 * write_cache_pages() already checks for dirty pages
467 * and calls clear_page_dirty_for_io(), which we want,
468 * to write protect the pages.
470 * However, we may have to redirty a page (see below.)
472 static int ext4_journalled_writepage_callback(struct page *page,
473 struct writeback_control *wbc,
476 transaction_t *transaction = (transaction_t *) data;
477 struct buffer_head *bh, *head;
478 struct journal_head *jh;
480 bh = head = page_buffers(page);
483 * We have to redirty a page in these cases:
484 * 1) If buffer is dirty, it means the page was dirty because it
485 * contains a buffer that needs checkpointing. So the dirty bit
486 * needs to be preserved so that checkpointing writes the buffer
488 * 2) If buffer is not part of the committing transaction
489 * (we may have just accidentally come across this buffer because
490 * inode range tracking is not exact) or if the currently running
491 * transaction already contains this buffer as well, dirty bit
492 * needs to be preserved so that the buffer gets writeprotected
493 * properly on running transaction's commit.
496 if (buffer_dirty(bh) ||
497 (jh && (jh->b_transaction != transaction ||
498 jh->b_next_transaction))) {
499 redirty_page_for_writepage(wbc, page);
502 } while ((bh = bh->b_this_page) != head);
505 return AOP_WRITEPAGE_ACTIVATE;
508 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
510 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
511 struct writeback_control wbc = {
512 .sync_mode = WB_SYNC_ALL,
513 .nr_to_write = LONG_MAX,
514 .range_start = jinode->i_dirty_start,
515 .range_end = jinode->i_dirty_end,
518 return write_cache_pages(mapping, &wbc,
519 ext4_journalled_writepage_callback,
520 jinode->i_transaction);
523 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
527 if (ext4_should_journal_data(jinode->i_vfs_inode))
528 ret = ext4_journalled_submit_inode_data_buffers(jinode);
530 ret = jbd2_journal_submit_inode_data_buffers(jinode);
535 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
539 if (!ext4_should_journal_data(jinode->i_vfs_inode))
540 ret = jbd2_journal_finish_inode_data_buffers(jinode);
545 static bool system_going_down(void)
547 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
548 || system_state == SYSTEM_RESTART;
551 struct ext4_err_translation {
556 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
558 static struct ext4_err_translation err_translation[] = {
559 EXT4_ERR_TRANSLATE(EIO),
560 EXT4_ERR_TRANSLATE(ENOMEM),
561 EXT4_ERR_TRANSLATE(EFSBADCRC),
562 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
563 EXT4_ERR_TRANSLATE(ENOSPC),
564 EXT4_ERR_TRANSLATE(ENOKEY),
565 EXT4_ERR_TRANSLATE(EROFS),
566 EXT4_ERR_TRANSLATE(EFBIG),
567 EXT4_ERR_TRANSLATE(EEXIST),
568 EXT4_ERR_TRANSLATE(ERANGE),
569 EXT4_ERR_TRANSLATE(EOVERFLOW),
570 EXT4_ERR_TRANSLATE(EBUSY),
571 EXT4_ERR_TRANSLATE(ENOTDIR),
572 EXT4_ERR_TRANSLATE(ENOTEMPTY),
573 EXT4_ERR_TRANSLATE(ESHUTDOWN),
574 EXT4_ERR_TRANSLATE(EFAULT),
577 static int ext4_errno_to_code(int errno)
581 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
582 if (err_translation[i].errno == errno)
583 return err_translation[i].code;
584 return EXT4_ERR_UNKNOWN;
587 static void save_error_info(struct super_block *sb, int error,
588 __u32 ino, __u64 block,
589 const char *func, unsigned int line)
591 struct ext4_sb_info *sbi = EXT4_SB(sb);
593 /* We default to EFSCORRUPTED error... */
595 error = EFSCORRUPTED;
597 spin_lock(&sbi->s_error_lock);
598 sbi->s_add_error_count++;
599 sbi->s_last_error_code = error;
600 sbi->s_last_error_line = line;
601 sbi->s_last_error_ino = ino;
602 sbi->s_last_error_block = block;
603 sbi->s_last_error_func = func;
604 sbi->s_last_error_time = ktime_get_real_seconds();
605 if (!sbi->s_first_error_time) {
606 sbi->s_first_error_code = error;
607 sbi->s_first_error_line = line;
608 sbi->s_first_error_ino = ino;
609 sbi->s_first_error_block = block;
610 sbi->s_first_error_func = func;
611 sbi->s_first_error_time = sbi->s_last_error_time;
613 spin_unlock(&sbi->s_error_lock);
616 /* Deal with the reporting of failure conditions on a filesystem such as
617 * inconsistencies detected or read IO failures.
619 * On ext2, we can store the error state of the filesystem in the
620 * superblock. That is not possible on ext4, because we may have other
621 * write ordering constraints on the superblock which prevent us from
622 * writing it out straight away; and given that the journal is about to
623 * be aborted, we can't rely on the current, or future, transactions to
624 * write out the superblock safely.
626 * We'll just use the jbd2_journal_abort() error code to record an error in
627 * the journal instead. On recovery, the journal will complain about
628 * that error until we've noted it down and cleared it.
630 * If force_ro is set, we unconditionally force the filesystem into an
631 * ABORT|READONLY state, unless the error response on the fs has been set to
632 * panic in which case we take the easy way out and panic immediately. This is
633 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
634 * at a critical moment in log management.
636 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
637 __u32 ino, __u64 block,
638 const char *func, unsigned int line)
640 journal_t *journal = EXT4_SB(sb)->s_journal;
641 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
643 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
644 if (test_opt(sb, WARN_ON_ERROR))
647 if (!continue_fs && !sb_rdonly(sb)) {
648 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
650 jbd2_journal_abort(journal, -EIO);
653 if (!bdev_read_only(sb->s_bdev)) {
654 save_error_info(sb, error, ino, block, func, line);
656 * In case the fs should keep running, we need to writeout
657 * superblock through the journal. Due to lock ordering
658 * constraints, it may not be safe to do it right here so we
659 * defer superblock flushing to a workqueue.
662 schedule_work(&EXT4_SB(sb)->s_error_work);
664 ext4_commit_super(sb);
668 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
669 * could panic during 'reboot -f' as the underlying device got already
672 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
673 panic("EXT4-fs (device %s): panic forced after error\n",
677 if (sb_rdonly(sb) || continue_fs)
680 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
682 * Make sure updated value of ->s_mount_flags will be visible before
686 sb->s_flags |= SB_RDONLY;
689 static void flush_stashed_error_work(struct work_struct *work)
691 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
693 journal_t *journal = sbi->s_journal;
697 * If the journal is still running, we have to write out superblock
698 * through the journal to avoid collisions of other journalled sb
701 * We use directly jbd2 functions here to avoid recursing back into
702 * ext4 error handling code during handling of previous errors.
704 if (!sb_rdonly(sbi->s_sb) && journal) {
705 struct buffer_head *sbh = sbi->s_sbh;
706 handle = jbd2_journal_start(journal, 1);
709 if (jbd2_journal_get_write_access(handle, sbh)) {
710 jbd2_journal_stop(handle);
713 ext4_update_super(sbi->s_sb);
714 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
715 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
716 "superblock detected");
717 clear_buffer_write_io_error(sbh);
718 set_buffer_uptodate(sbh);
721 if (jbd2_journal_dirty_metadata(handle, sbh)) {
722 jbd2_journal_stop(handle);
725 jbd2_journal_stop(handle);
726 ext4_notify_error_sysfs(sbi);
731 * Write through journal failed. Write sb directly to get error info
732 * out and hope for the best.
734 ext4_commit_super(sbi->s_sb);
735 ext4_notify_error_sysfs(sbi);
738 #define ext4_error_ratelimit(sb) \
739 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
742 void __ext4_error(struct super_block *sb, const char *function,
743 unsigned int line, bool force_ro, int error, __u64 block,
744 const char *fmt, ...)
746 struct va_format vaf;
749 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
752 trace_ext4_error(sb, function, line);
753 if (ext4_error_ratelimit(sb)) {
758 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
759 sb->s_id, function, line, current->comm, &vaf);
762 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
765 void __ext4_error_inode(struct inode *inode, const char *function,
766 unsigned int line, ext4_fsblk_t block, int error,
767 const char *fmt, ...)
770 struct va_format vaf;
772 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
775 trace_ext4_error(inode->i_sb, function, line);
776 if (ext4_error_ratelimit(inode->i_sb)) {
781 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
782 "inode #%lu: block %llu: comm %s: %pV\n",
783 inode->i_sb->s_id, function, line, inode->i_ino,
784 block, current->comm, &vaf);
786 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
787 "inode #%lu: comm %s: %pV\n",
788 inode->i_sb->s_id, function, line, inode->i_ino,
789 current->comm, &vaf);
792 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
796 void __ext4_error_file(struct file *file, const char *function,
797 unsigned int line, ext4_fsblk_t block,
798 const char *fmt, ...)
801 struct va_format vaf;
802 struct inode *inode = file_inode(file);
803 char pathname[80], *path;
805 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
808 trace_ext4_error(inode->i_sb, function, line);
809 if (ext4_error_ratelimit(inode->i_sb)) {
810 path = file_path(file, pathname, sizeof(pathname));
818 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
819 "block %llu: comm %s: path %s: %pV\n",
820 inode->i_sb->s_id, function, line, inode->i_ino,
821 block, current->comm, path, &vaf);
824 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
825 "comm %s: path %s: %pV\n",
826 inode->i_sb->s_id, function, line, inode->i_ino,
827 current->comm, path, &vaf);
830 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
834 const char *ext4_decode_error(struct super_block *sb, int errno,
841 errstr = "Corrupt filesystem";
844 errstr = "Filesystem failed CRC";
847 errstr = "IO failure";
850 errstr = "Out of memory";
853 if (!sb || (EXT4_SB(sb)->s_journal &&
854 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
855 errstr = "Journal has aborted";
857 errstr = "Readonly filesystem";
860 /* If the caller passed in an extra buffer for unknown
861 * errors, textualise them now. Else we just return
864 /* Check for truncated error codes... */
865 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
874 /* __ext4_std_error decodes expected errors from journaling functions
875 * automatically and invokes the appropriate error response. */
877 void __ext4_std_error(struct super_block *sb, const char *function,
878 unsigned int line, int errno)
883 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
886 /* Special case: if the error is EROFS, and we're not already
887 * inside a transaction, then there's really no point in logging
889 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
892 if (ext4_error_ratelimit(sb)) {
893 errstr = ext4_decode_error(sb, errno, nbuf);
894 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
895 sb->s_id, function, line, errstr);
898 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
901 void __ext4_msg(struct super_block *sb,
902 const char *prefix, const char *fmt, ...)
904 struct va_format vaf;
907 atomic_inc(&EXT4_SB(sb)->s_msg_count);
908 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
914 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
918 static int ext4_warning_ratelimit(struct super_block *sb)
920 atomic_inc(&EXT4_SB(sb)->s_warning_count);
921 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
925 void __ext4_warning(struct super_block *sb, const char *function,
926 unsigned int line, const char *fmt, ...)
928 struct va_format vaf;
931 if (!ext4_warning_ratelimit(sb))
937 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
938 sb->s_id, function, line, &vaf);
942 void __ext4_warning_inode(const struct inode *inode, const char *function,
943 unsigned int line, const char *fmt, ...)
945 struct va_format vaf;
948 if (!ext4_warning_ratelimit(inode->i_sb))
954 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
955 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
956 function, line, inode->i_ino, current->comm, &vaf);
960 void __ext4_grp_locked_error(const char *function, unsigned int line,
961 struct super_block *sb, ext4_group_t grp,
962 unsigned long ino, ext4_fsblk_t block,
963 const char *fmt, ...)
967 struct va_format vaf;
970 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
973 trace_ext4_error(sb, function, line);
974 if (ext4_error_ratelimit(sb)) {
978 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
979 sb->s_id, function, line, grp);
981 printk(KERN_CONT "inode %lu: ", ino);
983 printk(KERN_CONT "block %llu:",
984 (unsigned long long) block);
985 printk(KERN_CONT "%pV\n", &vaf);
989 if (test_opt(sb, ERRORS_CONT)) {
990 if (test_opt(sb, WARN_ON_ERROR))
992 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
993 if (!bdev_read_only(sb->s_bdev)) {
994 save_error_info(sb, EFSCORRUPTED, ino, block, function,
996 schedule_work(&EXT4_SB(sb)->s_error_work);
1000 ext4_unlock_group(sb, grp);
1001 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1003 * We only get here in the ERRORS_RO case; relocking the group
1004 * may be dangerous, but nothing bad will happen since the
1005 * filesystem will have already been marked read/only and the
1006 * journal has been aborted. We return 1 as a hint to callers
1007 * who might what to use the return value from
1008 * ext4_grp_locked_error() to distinguish between the
1009 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1010 * aggressively from the ext4 function in question, with a
1011 * more appropriate error code.
1013 ext4_lock_group(sb, grp);
1017 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1021 struct ext4_sb_info *sbi = EXT4_SB(sb);
1022 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1023 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1026 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1027 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1030 percpu_counter_sub(&sbi->s_freeclusters_counter,
1034 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1035 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1040 count = ext4_free_inodes_count(sb, gdp);
1041 percpu_counter_sub(&sbi->s_freeinodes_counter,
1047 void ext4_update_dynamic_rev(struct super_block *sb)
1049 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1051 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1055 "updating to rev %d because of new feature flag, "
1056 "running e2fsck is recommended",
1059 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1060 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1061 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1062 /* leave es->s_feature_*compat flags alone */
1063 /* es->s_uuid will be set by e2fsck if empty */
1066 * The rest of the superblock fields should be zero, and if not it
1067 * means they are likely already in use, so leave them alone. We
1068 * can leave it up to e2fsck to clean up any inconsistencies there.
1073 * Open the external journal device
1075 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1077 struct block_device *bdev;
1079 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1085 ext4_msg(sb, KERN_ERR,
1086 "failed to open journal device unknown-block(%u,%u) %ld",
1087 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1092 * Release the journal device
1094 static void ext4_blkdev_put(struct block_device *bdev)
1096 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1099 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1101 struct block_device *bdev;
1102 bdev = sbi->s_journal_bdev;
1104 ext4_blkdev_put(bdev);
1105 sbi->s_journal_bdev = NULL;
1109 static inline struct inode *orphan_list_entry(struct list_head *l)
1111 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1114 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1116 struct list_head *l;
1118 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1119 le32_to_cpu(sbi->s_es->s_last_orphan));
1121 printk(KERN_ERR "sb_info orphan list:\n");
1122 list_for_each(l, &sbi->s_orphan) {
1123 struct inode *inode = orphan_list_entry(l);
1125 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1126 inode->i_sb->s_id, inode->i_ino, inode,
1127 inode->i_mode, inode->i_nlink,
1128 NEXT_ORPHAN(inode));
1133 static int ext4_quota_off(struct super_block *sb, int type);
1135 static inline void ext4_quota_off_umount(struct super_block *sb)
1139 /* Use our quota_off function to clear inode flags etc. */
1140 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1141 ext4_quota_off(sb, type);
1145 * This is a helper function which is used in the mount/remount
1146 * codepaths (which holds s_umount) to fetch the quota file name.
1148 static inline char *get_qf_name(struct super_block *sb,
1149 struct ext4_sb_info *sbi,
1152 return rcu_dereference_protected(sbi->s_qf_names[type],
1153 lockdep_is_held(&sb->s_umount));
1156 static inline void ext4_quota_off_umount(struct super_block *sb)
1161 static void ext4_put_super(struct super_block *sb)
1163 struct ext4_sb_info *sbi = EXT4_SB(sb);
1164 struct ext4_super_block *es = sbi->s_es;
1165 struct buffer_head **group_desc;
1166 struct flex_groups **flex_groups;
1170 ext4_unregister_li_request(sb);
1171 ext4_quota_off_umount(sb);
1173 flush_work(&sbi->s_error_work);
1174 destroy_workqueue(sbi->rsv_conversion_wq);
1175 ext4_release_orphan_info(sb);
1178 * Unregister sysfs before destroying jbd2 journal.
1179 * Since we could still access attr_journal_task attribute via sysfs
1180 * path which could have sbi->s_journal->j_task as NULL
1182 ext4_unregister_sysfs(sb);
1184 if (sbi->s_journal) {
1185 aborted = is_journal_aborted(sbi->s_journal);
1186 err = jbd2_journal_destroy(sbi->s_journal);
1187 sbi->s_journal = NULL;
1188 if ((err < 0) && !aborted) {
1189 ext4_abort(sb, -err, "Couldn't clean up the journal");
1193 ext4_es_unregister_shrinker(sbi);
1194 del_timer_sync(&sbi->s_err_report);
1195 ext4_release_system_zone(sb);
1196 ext4_mb_release(sb);
1197 ext4_ext_release(sb);
1199 if (!sb_rdonly(sb) && !aborted) {
1200 ext4_clear_feature_journal_needs_recovery(sb);
1201 ext4_clear_feature_orphan_present(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 inode_init_once(&ei->vfs_inode);
1363 ext4_fc_init_inode(&ei->vfs_inode);
1366 static int __init init_inodecache(void)
1368 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1369 sizeof(struct ext4_inode_info), 0,
1370 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1372 offsetof(struct ext4_inode_info, i_data),
1373 sizeof_field(struct ext4_inode_info, i_data),
1375 if (ext4_inode_cachep == NULL)
1380 static void destroy_inodecache(void)
1383 * Make sure all delayed rcu free inodes are flushed before we
1387 kmem_cache_destroy(ext4_inode_cachep);
1390 void ext4_clear_inode(struct inode *inode)
1393 invalidate_inode_buffers(inode);
1395 ext4_discard_preallocations(inode, 0);
1396 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1398 if (EXT4_I(inode)->jinode) {
1399 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1400 EXT4_I(inode)->jinode);
1401 jbd2_free_inode(EXT4_I(inode)->jinode);
1402 EXT4_I(inode)->jinode = NULL;
1404 fscrypt_put_encryption_info(inode);
1405 fsverity_cleanup_inode(inode);
1408 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1409 u64 ino, u32 generation)
1411 struct inode *inode;
1414 * Currently we don't know the generation for parent directory, so
1415 * a generation of 0 means "accept any"
1417 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1419 return ERR_CAST(inode);
1420 if (generation && inode->i_generation != generation) {
1422 return ERR_PTR(-ESTALE);
1428 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1429 int fh_len, int fh_type)
1431 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1432 ext4_nfs_get_inode);
1435 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1436 int fh_len, int fh_type)
1438 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1439 ext4_nfs_get_inode);
1442 static int ext4_nfs_commit_metadata(struct inode *inode)
1444 struct writeback_control wbc = {
1445 .sync_mode = WB_SYNC_ALL
1448 trace_ext4_nfs_commit_metadata(inode);
1449 return ext4_write_inode(inode, &wbc);
1452 #ifdef CONFIG_FS_ENCRYPTION
1453 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1455 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1456 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1459 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1462 handle_t *handle = fs_data;
1463 int res, res2, credits, retries = 0;
1466 * Encrypting the root directory is not allowed because e2fsck expects
1467 * lost+found to exist and be unencrypted, and encrypting the root
1468 * directory would imply encrypting the lost+found directory as well as
1469 * the filename "lost+found" itself.
1471 if (inode->i_ino == EXT4_ROOT_INO)
1474 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1477 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1480 res = ext4_convert_inline_data(inode);
1485 * If a journal handle was specified, then the encryption context is
1486 * being set on a new inode via inheritance and is part of a larger
1487 * transaction to create the inode. Otherwise the encryption context is
1488 * being set on an existing inode in its own transaction. Only in the
1489 * latter case should the "retry on ENOSPC" logic be used.
1493 res = ext4_xattr_set_handle(handle, inode,
1494 EXT4_XATTR_INDEX_ENCRYPTION,
1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1498 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1499 ext4_clear_inode_state(inode,
1500 EXT4_STATE_MAY_INLINE_DATA);
1502 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1503 * S_DAX may be disabled
1505 ext4_set_inode_flags(inode, false);
1510 res = dquot_initialize(inode);
1514 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1519 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1521 return PTR_ERR(handle);
1523 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1524 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1527 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1529 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1530 * S_DAX may be disabled
1532 ext4_set_inode_flags(inode, false);
1533 res = ext4_mark_inode_dirty(handle, inode);
1535 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1537 res2 = ext4_journal_stop(handle);
1539 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1546 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1548 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1551 static bool ext4_has_stable_inodes(struct super_block *sb)
1553 return ext4_has_feature_stable_inodes(sb);
1556 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1557 int *ino_bits_ret, int *lblk_bits_ret)
1559 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1560 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1563 static const struct fscrypt_operations ext4_cryptops = {
1564 .key_prefix = "ext4:",
1565 .get_context = ext4_get_context,
1566 .set_context = ext4_set_context,
1567 .get_dummy_policy = ext4_get_dummy_policy,
1568 .empty_dir = ext4_empty_dir,
1569 .max_namelen = EXT4_NAME_LEN,
1570 .has_stable_inodes = ext4_has_stable_inodes,
1571 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1576 static const char * const quotatypes[] = INITQFNAMES;
1577 #define QTYPE2NAME(t) (quotatypes[t])
1579 static int ext4_write_dquot(struct dquot *dquot);
1580 static int ext4_acquire_dquot(struct dquot *dquot);
1581 static int ext4_release_dquot(struct dquot *dquot);
1582 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1583 static int ext4_write_info(struct super_block *sb, int type);
1584 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1585 const struct path *path);
1586 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1587 size_t len, loff_t off);
1588 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1589 const char *data, size_t len, loff_t off);
1590 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1591 unsigned int flags);
1593 static struct dquot **ext4_get_dquots(struct inode *inode)
1595 return EXT4_I(inode)->i_dquot;
1598 static const struct dquot_operations ext4_quota_operations = {
1599 .get_reserved_space = ext4_get_reserved_space,
1600 .write_dquot = ext4_write_dquot,
1601 .acquire_dquot = ext4_acquire_dquot,
1602 .release_dquot = ext4_release_dquot,
1603 .mark_dirty = ext4_mark_dquot_dirty,
1604 .write_info = ext4_write_info,
1605 .alloc_dquot = dquot_alloc,
1606 .destroy_dquot = dquot_destroy,
1607 .get_projid = ext4_get_projid,
1608 .get_inode_usage = ext4_get_inode_usage,
1609 .get_next_id = dquot_get_next_id,
1612 static const struct quotactl_ops ext4_qctl_operations = {
1613 .quota_on = ext4_quota_on,
1614 .quota_off = ext4_quota_off,
1615 .quota_sync = dquot_quota_sync,
1616 .get_state = dquot_get_state,
1617 .set_info = dquot_set_dqinfo,
1618 .get_dqblk = dquot_get_dqblk,
1619 .set_dqblk = dquot_set_dqblk,
1620 .get_nextdqblk = dquot_get_next_dqblk,
1624 static const struct super_operations ext4_sops = {
1625 .alloc_inode = ext4_alloc_inode,
1626 .free_inode = ext4_free_in_core_inode,
1627 .destroy_inode = ext4_destroy_inode,
1628 .write_inode = ext4_write_inode,
1629 .dirty_inode = ext4_dirty_inode,
1630 .drop_inode = ext4_drop_inode,
1631 .evict_inode = ext4_evict_inode,
1632 .put_super = ext4_put_super,
1633 .sync_fs = ext4_sync_fs,
1634 .freeze_fs = ext4_freeze,
1635 .unfreeze_fs = ext4_unfreeze,
1636 .statfs = ext4_statfs,
1637 .remount_fs = ext4_remount,
1638 .show_options = ext4_show_options,
1640 .quota_read = ext4_quota_read,
1641 .quota_write = ext4_quota_write,
1642 .get_dquots = ext4_get_dquots,
1646 static const struct export_operations ext4_export_ops = {
1647 .fh_to_dentry = ext4_fh_to_dentry,
1648 .fh_to_parent = ext4_fh_to_parent,
1649 .get_parent = ext4_get_parent,
1650 .commit_metadata = ext4_nfs_commit_metadata,
1654 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1655 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1656 Opt_nouid32, Opt_debug, Opt_removed,
1657 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1658 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1659 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1660 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1661 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1662 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1664 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1665 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1666 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1667 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1668 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1669 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1670 Opt_nowarn_on_error, Opt_mblk_io_submit,
1671 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1672 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1673 Opt_inode_readahead_blks, Opt_journal_ioprio,
1674 Opt_dioread_nolock, Opt_dioread_lock,
1675 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1676 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1677 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1678 #ifdef CONFIG_EXT4_DEBUG
1679 Opt_fc_debug_max_replay, Opt_fc_debug_force
1683 static const match_table_t tokens = {
1684 {Opt_bsd_df, "bsddf"},
1685 {Opt_minix_df, "minixdf"},
1686 {Opt_grpid, "grpid"},
1687 {Opt_grpid, "bsdgroups"},
1688 {Opt_nogrpid, "nogrpid"},
1689 {Opt_nogrpid, "sysvgroups"},
1690 {Opt_resgid, "resgid=%u"},
1691 {Opt_resuid, "resuid=%u"},
1693 {Opt_err_cont, "errors=continue"},
1694 {Opt_err_panic, "errors=panic"},
1695 {Opt_err_ro, "errors=remount-ro"},
1696 {Opt_nouid32, "nouid32"},
1697 {Opt_debug, "debug"},
1698 {Opt_removed, "oldalloc"},
1699 {Opt_removed, "orlov"},
1700 {Opt_user_xattr, "user_xattr"},
1701 {Opt_nouser_xattr, "nouser_xattr"},
1703 {Opt_noacl, "noacl"},
1704 {Opt_noload, "norecovery"},
1705 {Opt_noload, "noload"},
1706 {Opt_removed, "nobh"},
1707 {Opt_removed, "bh"},
1708 {Opt_commit, "commit=%u"},
1709 {Opt_min_batch_time, "min_batch_time=%u"},
1710 {Opt_max_batch_time, "max_batch_time=%u"},
1711 {Opt_journal_dev, "journal_dev=%u"},
1712 {Opt_journal_path, "journal_path=%s"},
1713 {Opt_journal_checksum, "journal_checksum"},
1714 {Opt_nojournal_checksum, "nojournal_checksum"},
1715 {Opt_journal_async_commit, "journal_async_commit"},
1716 {Opt_abort, "abort"},
1717 {Opt_data_journal, "data=journal"},
1718 {Opt_data_ordered, "data=ordered"},
1719 {Opt_data_writeback, "data=writeback"},
1720 {Opt_data_err_abort, "data_err=abort"},
1721 {Opt_data_err_ignore, "data_err=ignore"},
1722 {Opt_offusrjquota, "usrjquota="},
1723 {Opt_usrjquota, "usrjquota=%s"},
1724 {Opt_offgrpjquota, "grpjquota="},
1725 {Opt_grpjquota, "grpjquota=%s"},
1726 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1727 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1728 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1729 {Opt_grpquota, "grpquota"},
1730 {Opt_noquota, "noquota"},
1731 {Opt_quota, "quota"},
1732 {Opt_usrquota, "usrquota"},
1733 {Opt_prjquota, "prjquota"},
1734 {Opt_barrier, "barrier=%u"},
1735 {Opt_barrier, "barrier"},
1736 {Opt_nobarrier, "nobarrier"},
1737 {Opt_i_version, "i_version"},
1739 {Opt_dax_always, "dax=always"},
1740 {Opt_dax_inode, "dax=inode"},
1741 {Opt_dax_never, "dax=never"},
1742 {Opt_stripe, "stripe=%u"},
1743 {Opt_delalloc, "delalloc"},
1744 {Opt_warn_on_error, "warn_on_error"},
1745 {Opt_nowarn_on_error, "nowarn_on_error"},
1746 {Opt_lazytime, "lazytime"},
1747 {Opt_nolazytime, "nolazytime"},
1748 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1749 {Opt_nodelalloc, "nodelalloc"},
1750 {Opt_removed, "mblk_io_submit"},
1751 {Opt_removed, "nomblk_io_submit"},
1752 {Opt_block_validity, "block_validity"},
1753 {Opt_noblock_validity, "noblock_validity"},
1754 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1755 {Opt_journal_ioprio, "journal_ioprio=%u"},
1756 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1757 {Opt_auto_da_alloc, "auto_da_alloc"},
1758 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1759 {Opt_dioread_nolock, "dioread_nolock"},
1760 {Opt_dioread_lock, "nodioread_nolock"},
1761 {Opt_dioread_lock, "dioread_lock"},
1762 {Opt_discard, "discard"},
1763 {Opt_nodiscard, "nodiscard"},
1764 {Opt_init_itable, "init_itable=%u"},
1765 {Opt_init_itable, "init_itable"},
1766 {Opt_noinit_itable, "noinit_itable"},
1767 #ifdef CONFIG_EXT4_DEBUG
1768 {Opt_fc_debug_force, "fc_debug_force"},
1769 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1771 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1772 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1773 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1774 {Opt_inlinecrypt, "inlinecrypt"},
1775 {Opt_nombcache, "nombcache"},
1776 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1777 {Opt_removed, "prefetch_block_bitmaps"},
1778 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1779 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1780 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1781 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1782 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1783 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1784 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1788 static ext4_fsblk_t get_sb_block(void **data)
1790 ext4_fsblk_t sb_block;
1791 char *options = (char *) *data;
1793 if (!options || strncmp(options, "sb=", 3) != 0)
1794 return 1; /* Default location */
1797 /* TODO: use simple_strtoll with >32bit ext4 */
1798 sb_block = simple_strtoul(options, &options, 0);
1799 if (*options && *options != ',') {
1800 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1804 if (*options == ',')
1806 *data = (void *) options;
1811 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1812 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1814 static const char deprecated_msg[] =
1815 "Mount option \"%s\" will be removed by %s\n"
1816 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1819 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1821 struct ext4_sb_info *sbi = EXT4_SB(sb);
1822 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1825 if (sb_any_quota_loaded(sb) && !old_qname) {
1826 ext4_msg(sb, KERN_ERR,
1827 "Cannot change journaled "
1828 "quota options when quota turned on");
1831 if (ext4_has_feature_quota(sb)) {
1832 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1833 "ignored when QUOTA feature is enabled");
1836 qname = match_strdup(args);
1838 ext4_msg(sb, KERN_ERR,
1839 "Not enough memory for storing quotafile name");
1843 if (strcmp(old_qname, qname) == 0)
1846 ext4_msg(sb, KERN_ERR,
1847 "%s quota file already specified",
1851 if (strchr(qname, '/')) {
1852 ext4_msg(sb, KERN_ERR,
1853 "quotafile must be on filesystem root");
1856 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1864 static int clear_qf_name(struct super_block *sb, int qtype)
1867 struct ext4_sb_info *sbi = EXT4_SB(sb);
1868 char *old_qname = get_qf_name(sb, sbi, qtype);
1870 if (sb_any_quota_loaded(sb) && old_qname) {
1871 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1872 " when quota turned on");
1875 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1882 #define MOPT_SET 0x0001
1883 #define MOPT_CLEAR 0x0002
1884 #define MOPT_NOSUPPORT 0x0004
1885 #define MOPT_EXPLICIT 0x0008
1886 #define MOPT_CLEAR_ERR 0x0010
1887 #define MOPT_GTE0 0x0020
1890 #define MOPT_QFMT 0x0040
1892 #define MOPT_Q MOPT_NOSUPPORT
1893 #define MOPT_QFMT MOPT_NOSUPPORT
1895 #define MOPT_DATAJ 0x0080
1896 #define MOPT_NO_EXT2 0x0100
1897 #define MOPT_NO_EXT3 0x0200
1898 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1899 #define MOPT_STRING 0x0400
1900 #define MOPT_SKIP 0x0800
1901 #define MOPT_2 0x1000
1903 static const struct mount_opts {
1907 } ext4_mount_opts[] = {
1908 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1909 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1910 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1911 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1912 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1913 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1914 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1915 MOPT_EXT4_ONLY | MOPT_SET},
1916 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1917 MOPT_EXT4_ONLY | MOPT_CLEAR},
1918 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1919 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1920 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1921 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1922 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1923 MOPT_EXT4_ONLY | MOPT_CLEAR},
1924 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1925 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1926 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1927 MOPT_EXT4_ONLY | MOPT_CLEAR},
1928 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1929 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1930 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1931 EXT4_MOUNT_JOURNAL_CHECKSUM),
1932 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1933 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1934 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1935 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1936 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1937 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1939 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1941 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1942 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1943 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1944 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1945 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1946 {Opt_commit, 0, MOPT_GTE0},
1947 {Opt_max_batch_time, 0, MOPT_GTE0},
1948 {Opt_min_batch_time, 0, MOPT_GTE0},
1949 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1950 {Opt_init_itable, 0, MOPT_GTE0},
1951 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1952 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1953 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1954 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1955 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1956 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1957 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1958 {Opt_stripe, 0, MOPT_GTE0},
1959 {Opt_resuid, 0, MOPT_GTE0},
1960 {Opt_resgid, 0, MOPT_GTE0},
1961 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1962 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1963 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1964 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1965 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1966 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1967 MOPT_NO_EXT2 | MOPT_DATAJ},
1968 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1969 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1970 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1971 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1972 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1974 {Opt_acl, 0, MOPT_NOSUPPORT},
1975 {Opt_noacl, 0, MOPT_NOSUPPORT},
1977 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1978 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1979 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1980 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1981 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1983 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1985 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1987 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1988 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1989 MOPT_CLEAR | MOPT_Q},
1990 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1991 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1992 {Opt_offusrjquota, 0, MOPT_Q},
1993 {Opt_offgrpjquota, 0, MOPT_Q},
1994 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1995 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1996 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1997 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1998 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1999 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2000 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2002 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2003 #ifdef CONFIG_EXT4_DEBUG
2004 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2005 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2006 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2011 #ifdef CONFIG_UNICODE
2012 static const struct ext4_sb_encodings {
2016 } ext4_sb_encoding_map[] = {
2017 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2020 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2021 const struct ext4_sb_encodings **encoding,
2024 __u16 magic = le16_to_cpu(es->s_encoding);
2027 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2028 if (magic == ext4_sb_encoding_map[i].magic)
2031 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2034 *encoding = &ext4_sb_encoding_map[i];
2035 *flags = le16_to_cpu(es->s_encoding_flags);
2041 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2043 const substring_t *arg,
2046 #ifdef CONFIG_FS_ENCRYPTION
2047 struct ext4_sb_info *sbi = EXT4_SB(sb);
2051 * This mount option is just for testing, and it's not worthwhile to
2052 * implement the extra complexity (e.g. RCU protection) that would be
2053 * needed to allow it to be set or changed during remount. We do allow
2054 * it to be specified during remount, but only if there is no change.
2056 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2057 ext4_msg(sb, KERN_WARNING,
2058 "Can't set test_dummy_encryption on remount");
2061 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2062 &sbi->s_dummy_enc_policy);
2065 ext4_msg(sb, KERN_WARNING,
2066 "Can't change test_dummy_encryption on remount");
2067 else if (err == -EINVAL)
2068 ext4_msg(sb, KERN_WARNING,
2069 "Value of option \"%s\" is unrecognized", opt);
2071 ext4_msg(sb, KERN_WARNING,
2072 "Error processing option \"%s\" [%d]",
2076 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2078 ext4_msg(sb, KERN_WARNING,
2079 "Test dummy encryption mount option ignored");
2084 struct ext4_parsed_options {
2085 unsigned long journal_devnum;
2086 unsigned int journal_ioprio;
2087 int mb_optimize_scan;
2090 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2091 substring_t *args, struct ext4_parsed_options *parsed_opts,
2094 struct ext4_sb_info *sbi = EXT4_SB(sb);
2095 const struct mount_opts *m;
2101 if (token == Opt_usrjquota)
2102 return set_qf_name(sb, USRQUOTA, &args[0]);
2103 else if (token == Opt_grpjquota)
2104 return set_qf_name(sb, GRPQUOTA, &args[0]);
2105 else if (token == Opt_offusrjquota)
2106 return clear_qf_name(sb, USRQUOTA);
2107 else if (token == Opt_offgrpjquota)
2108 return clear_qf_name(sb, GRPQUOTA);
2112 case Opt_nouser_xattr:
2113 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2116 return 1; /* handled by get_sb_block() */
2118 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2121 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2124 sb->s_flags |= SB_I_VERSION;
2127 sb->s_flags |= SB_LAZYTIME;
2129 case Opt_nolazytime:
2130 sb->s_flags &= ~SB_LAZYTIME;
2132 case Opt_inlinecrypt:
2133 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2134 sb->s_flags |= SB_INLINECRYPT;
2136 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2141 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2142 if (token == m->token)
2145 if (m->token == Opt_err) {
2146 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2147 "or missing value", opt);
2151 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2152 ext4_msg(sb, KERN_ERR,
2153 "Mount option \"%s\" incompatible with ext2", opt);
2156 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2157 ext4_msg(sb, KERN_ERR,
2158 "Mount option \"%s\" incompatible with ext3", opt);
2162 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2164 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2166 if (m->flags & MOPT_EXPLICIT) {
2167 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2168 set_opt2(sb, EXPLICIT_DELALLOC);
2169 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2170 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2174 if (m->flags & MOPT_CLEAR_ERR)
2175 clear_opt(sb, ERRORS_MASK);
2176 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2177 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2178 "options when quota turned on");
2182 if (m->flags & MOPT_NOSUPPORT) {
2183 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2184 } else if (token == Opt_commit) {
2186 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2187 else if (arg > INT_MAX / HZ) {
2188 ext4_msg(sb, KERN_ERR,
2189 "Invalid commit interval %d, "
2190 "must be smaller than %d",
2194 sbi->s_commit_interval = HZ * arg;
2195 } else if (token == Opt_debug_want_extra_isize) {
2198 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2199 ext4_msg(sb, KERN_ERR,
2200 "Invalid want_extra_isize %d", arg);
2203 sbi->s_want_extra_isize = arg;
2204 } else if (token == Opt_max_batch_time) {
2205 sbi->s_max_batch_time = arg;
2206 } else if (token == Opt_min_batch_time) {
2207 sbi->s_min_batch_time = arg;
2208 } else if (token == Opt_inode_readahead_blks) {
2209 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2210 ext4_msg(sb, KERN_ERR,
2211 "EXT4-fs: inode_readahead_blks must be "
2212 "0 or a power of 2 smaller than 2^31");
2215 sbi->s_inode_readahead_blks = arg;
2216 } else if (token == Opt_init_itable) {
2217 set_opt(sb, INIT_INODE_TABLE);
2219 arg = EXT4_DEF_LI_WAIT_MULT;
2220 sbi->s_li_wait_mult = arg;
2221 } else if (token == Opt_max_dir_size_kb) {
2222 sbi->s_max_dir_size_kb = arg;
2223 #ifdef CONFIG_EXT4_DEBUG
2224 } else if (token == Opt_fc_debug_max_replay) {
2225 sbi->s_fc_debug_max_replay = arg;
2227 } else if (token == Opt_stripe) {
2228 sbi->s_stripe = arg;
2229 } else if (token == Opt_resuid) {
2230 uid = make_kuid(current_user_ns(), arg);
2231 if (!uid_valid(uid)) {
2232 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2235 sbi->s_resuid = uid;
2236 } else if (token == Opt_resgid) {
2237 gid = make_kgid(current_user_ns(), arg);
2238 if (!gid_valid(gid)) {
2239 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2242 sbi->s_resgid = gid;
2243 } else if (token == Opt_journal_dev) {
2245 ext4_msg(sb, KERN_ERR,
2246 "Cannot specify journal on remount");
2249 parsed_opts->journal_devnum = arg;
2250 } else if (token == Opt_journal_path) {
2252 struct inode *journal_inode;
2257 ext4_msg(sb, KERN_ERR,
2258 "Cannot specify journal on remount");
2261 journal_path = match_strdup(&args[0]);
2262 if (!journal_path) {
2263 ext4_msg(sb, KERN_ERR, "error: could not dup "
2264 "journal device string");
2268 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2270 ext4_msg(sb, KERN_ERR, "error: could not find "
2271 "journal device path: error %d", error);
2272 kfree(journal_path);
2276 journal_inode = d_inode(path.dentry);
2277 if (!S_ISBLK(journal_inode->i_mode)) {
2278 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2279 "is not a block device", journal_path);
2281 kfree(journal_path);
2285 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2287 kfree(journal_path);
2288 } else if (token == Opt_journal_ioprio) {
2290 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2294 parsed_opts->journal_ioprio =
2295 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2296 } else if (token == Opt_test_dummy_encryption) {
2297 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2299 } else if (m->flags & MOPT_DATAJ) {
2301 if (!sbi->s_journal)
2302 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2303 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2304 ext4_msg(sb, KERN_ERR,
2305 "Cannot change data mode on remount");
2309 clear_opt(sb, DATA_FLAGS);
2310 sbi->s_mount_opt |= m->mount_opt;
2313 } else if (m->flags & MOPT_QFMT) {
2314 if (sb_any_quota_loaded(sb) &&
2315 sbi->s_jquota_fmt != m->mount_opt) {
2316 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2317 "quota options when quota turned on");
2320 if (ext4_has_feature_quota(sb)) {
2321 ext4_msg(sb, KERN_INFO,
2322 "Quota format mount options ignored "
2323 "when QUOTA feature is enabled");
2326 sbi->s_jquota_fmt = m->mount_opt;
2328 } else if (token == Opt_dax || token == Opt_dax_always ||
2329 token == Opt_dax_inode || token == Opt_dax_never) {
2330 #ifdef CONFIG_FS_DAX
2333 case Opt_dax_always:
2335 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2336 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2337 fail_dax_change_remount:
2338 ext4_msg(sb, KERN_ERR, "can't change "
2339 "dax mount option while remounting");
2343 (test_opt(sb, DATA_FLAGS) ==
2344 EXT4_MOUNT_JOURNAL_DATA)) {
2345 ext4_msg(sb, KERN_ERR, "can't mount with "
2346 "both data=journal and dax");
2349 ext4_msg(sb, KERN_WARNING,
2350 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2351 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2352 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2356 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2357 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2358 goto fail_dax_change_remount;
2359 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2360 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2364 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2365 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2366 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2367 goto fail_dax_change_remount;
2368 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2369 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2370 /* Strictly for printing options */
2371 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2375 ext4_msg(sb, KERN_INFO, "dax option not supported");
2376 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2377 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2380 } else if (token == Opt_data_err_abort) {
2381 sbi->s_mount_opt |= m->mount_opt;
2382 } else if (token == Opt_data_err_ignore) {
2383 sbi->s_mount_opt &= ~m->mount_opt;
2384 } else if (token == Opt_mb_optimize_scan) {
2385 if (arg != 0 && arg != 1) {
2386 ext4_msg(sb, KERN_WARNING,
2387 "mb_optimize_scan should be set to 0 or 1.");
2390 parsed_opts->mb_optimize_scan = arg;
2394 if (m->flags & MOPT_CLEAR)
2396 else if (unlikely(!(m->flags & MOPT_SET))) {
2397 ext4_msg(sb, KERN_WARNING,
2398 "buggy handling of option %s", opt);
2402 if (m->flags & MOPT_2) {
2404 sbi->s_mount_opt2 |= m->mount_opt;
2406 sbi->s_mount_opt2 &= ~m->mount_opt;
2409 sbi->s_mount_opt |= m->mount_opt;
2411 sbi->s_mount_opt &= ~m->mount_opt;
2417 static int parse_options(char *options, struct super_block *sb,
2418 struct ext4_parsed_options *ret_opts,
2421 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2422 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2423 substring_t args[MAX_OPT_ARGS];
2429 while ((p = strsep(&options, ",")) != NULL) {
2433 * Initialize args struct so we know whether arg was
2434 * found; some options take optional arguments.
2436 args[0].to = args[0].from = NULL;
2437 token = match_token(p, tokens, args);
2438 if (handle_mount_opt(sb, p, token, args, ret_opts,
2444 * We do the test below only for project quotas. 'usrquota' and
2445 * 'grpquota' mount options are allowed even without quota feature
2446 * to support legacy quotas in quota files.
2448 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2449 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2450 "Cannot enable project quota enforcement.");
2453 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2454 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2455 if (usr_qf_name || grp_qf_name) {
2456 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2457 clear_opt(sb, USRQUOTA);
2459 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2460 clear_opt(sb, GRPQUOTA);
2462 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2463 ext4_msg(sb, KERN_ERR, "old and new quota "
2468 if (!sbi->s_jquota_fmt) {
2469 ext4_msg(sb, KERN_ERR, "journaled quota format "
2475 if (test_opt(sb, DIOREAD_NOLOCK)) {
2477 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2478 if (blocksize < PAGE_SIZE)
2479 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2480 "experimental mount option 'dioread_nolock' "
2481 "for blocksize < PAGE_SIZE");
2486 static inline void ext4_show_quota_options(struct seq_file *seq,
2487 struct super_block *sb)
2489 #if defined(CONFIG_QUOTA)
2490 struct ext4_sb_info *sbi = EXT4_SB(sb);
2491 char *usr_qf_name, *grp_qf_name;
2493 if (sbi->s_jquota_fmt) {
2496 switch (sbi->s_jquota_fmt) {
2507 seq_printf(seq, ",jqfmt=%s", fmtname);
2511 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2512 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2514 seq_show_option(seq, "usrjquota", usr_qf_name);
2516 seq_show_option(seq, "grpjquota", grp_qf_name);
2521 static const char *token2str(int token)
2523 const struct match_token *t;
2525 for (t = tokens; t->token != Opt_err; t++)
2526 if (t->token == token && !strchr(t->pattern, '='))
2533 * - it's set to a non-default value OR
2534 * - if the per-sb default is different from the global default
2536 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2539 struct ext4_sb_info *sbi = EXT4_SB(sb);
2540 struct ext4_super_block *es = sbi->s_es;
2541 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2542 const struct mount_opts *m;
2543 char sep = nodefs ? '\n' : ',';
2545 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2546 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2548 if (sbi->s_sb_block != 1)
2549 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2551 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2552 int want_set = m->flags & MOPT_SET;
2553 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2554 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2556 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2557 continue; /* skip if same as the default */
2559 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2560 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2561 continue; /* select Opt_noFoo vs Opt_Foo */
2562 SEQ_OPTS_PRINT("%s", token2str(m->token));
2565 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2566 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2567 SEQ_OPTS_PRINT("resuid=%u",
2568 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2569 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2570 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2571 SEQ_OPTS_PRINT("resgid=%u",
2572 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2573 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2574 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2575 SEQ_OPTS_PUTS("errors=remount-ro");
2576 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2577 SEQ_OPTS_PUTS("errors=continue");
2578 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2579 SEQ_OPTS_PUTS("errors=panic");
2580 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2581 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2582 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2583 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2584 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2585 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2586 if (sb->s_flags & SB_I_VERSION)
2587 SEQ_OPTS_PUTS("i_version");
2588 if (nodefs || sbi->s_stripe)
2589 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2590 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2591 (sbi->s_mount_opt ^ def_mount_opt)) {
2592 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2593 SEQ_OPTS_PUTS("data=journal");
2594 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2595 SEQ_OPTS_PUTS("data=ordered");
2596 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2597 SEQ_OPTS_PUTS("data=writeback");
2600 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2601 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2602 sbi->s_inode_readahead_blks);
2604 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2605 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2606 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2607 if (nodefs || sbi->s_max_dir_size_kb)
2608 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2609 if (test_opt(sb, DATA_ERR_ABORT))
2610 SEQ_OPTS_PUTS("data_err=abort");
2612 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2614 if (sb->s_flags & SB_INLINECRYPT)
2615 SEQ_OPTS_PUTS("inlinecrypt");
2617 if (test_opt(sb, DAX_ALWAYS)) {
2619 SEQ_OPTS_PUTS("dax");
2621 SEQ_OPTS_PUTS("dax=always");
2622 } else if (test_opt2(sb, DAX_NEVER)) {
2623 SEQ_OPTS_PUTS("dax=never");
2624 } else if (test_opt2(sb, DAX_INODE)) {
2625 SEQ_OPTS_PUTS("dax=inode");
2627 ext4_show_quota_options(seq, sb);
2631 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2633 return _ext4_show_options(seq, root->d_sb, 0);
2636 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2638 struct super_block *sb = seq->private;
2641 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2642 rc = _ext4_show_options(seq, sb, 1);
2643 seq_puts(seq, "\n");
2647 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2650 struct ext4_sb_info *sbi = EXT4_SB(sb);
2653 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2654 ext4_msg(sb, KERN_ERR, "revision level too high, "
2655 "forcing read-only mode");
2661 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2662 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2663 "running e2fsck is recommended");
2664 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2665 ext4_msg(sb, KERN_WARNING,
2666 "warning: mounting fs with errors, "
2667 "running e2fsck is recommended");
2668 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2669 le16_to_cpu(es->s_mnt_count) >=
2670 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2671 ext4_msg(sb, KERN_WARNING,
2672 "warning: maximal mount count reached, "
2673 "running e2fsck is recommended");
2674 else if (le32_to_cpu(es->s_checkinterval) &&
2675 (ext4_get_tstamp(es, s_lastcheck) +
2676 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2677 ext4_msg(sb, KERN_WARNING,
2678 "warning: checktime reached, "
2679 "running e2fsck is recommended");
2680 if (!sbi->s_journal)
2681 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2682 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2683 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2684 le16_add_cpu(&es->s_mnt_count, 1);
2685 ext4_update_tstamp(es, s_mtime);
2686 if (sbi->s_journal) {
2687 ext4_set_feature_journal_needs_recovery(sb);
2688 if (ext4_has_feature_orphan_file(sb))
2689 ext4_set_feature_orphan_present(sb);
2692 err = ext4_commit_super(sb);
2694 if (test_opt(sb, DEBUG))
2695 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2696 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2698 sbi->s_groups_count,
2699 EXT4_BLOCKS_PER_GROUP(sb),
2700 EXT4_INODES_PER_GROUP(sb),
2701 sbi->s_mount_opt, sbi->s_mount_opt2);
2703 cleancache_init_fs(sb);
2707 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2709 struct ext4_sb_info *sbi = EXT4_SB(sb);
2710 struct flex_groups **old_groups, **new_groups;
2713 if (!sbi->s_log_groups_per_flex)
2716 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2717 if (size <= sbi->s_flex_groups_allocated)
2720 new_groups = kvzalloc(roundup_pow_of_two(size *
2721 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2723 ext4_msg(sb, KERN_ERR,
2724 "not enough memory for %d flex group pointers", size);
2727 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2728 new_groups[i] = kvzalloc(roundup_pow_of_two(
2729 sizeof(struct flex_groups)),
2731 if (!new_groups[i]) {
2732 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2733 kvfree(new_groups[j]);
2735 ext4_msg(sb, KERN_ERR,
2736 "not enough memory for %d flex groups", size);
2741 old_groups = rcu_dereference(sbi->s_flex_groups);
2743 memcpy(new_groups, old_groups,
2744 (sbi->s_flex_groups_allocated *
2745 sizeof(struct flex_groups *)));
2747 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2748 sbi->s_flex_groups_allocated = size;
2750 ext4_kvfree_array_rcu(old_groups);
2754 static int ext4_fill_flex_info(struct super_block *sb)
2756 struct ext4_sb_info *sbi = EXT4_SB(sb);
2757 struct ext4_group_desc *gdp = NULL;
2758 struct flex_groups *fg;
2759 ext4_group_t flex_group;
2762 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2763 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2764 sbi->s_log_groups_per_flex = 0;
2768 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2772 for (i = 0; i < sbi->s_groups_count; i++) {
2773 gdp = ext4_get_group_desc(sb, i, NULL);
2775 flex_group = ext4_flex_group(sbi, i);
2776 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2777 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2778 atomic64_add(ext4_free_group_clusters(sb, gdp),
2779 &fg->free_clusters);
2780 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2788 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2789 struct ext4_group_desc *gdp)
2791 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2793 __le32 le_group = cpu_to_le32(block_group);
2794 struct ext4_sb_info *sbi = EXT4_SB(sb);
2796 if (ext4_has_metadata_csum(sbi->s_sb)) {
2797 /* Use new metadata_csum algorithm */
2799 __u16 dummy_csum = 0;
2801 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2803 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2804 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2805 sizeof(dummy_csum));
2806 offset += sizeof(dummy_csum);
2807 if (offset < sbi->s_desc_size)
2808 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2809 sbi->s_desc_size - offset);
2811 crc = csum32 & 0xFFFF;
2815 /* old crc16 code */
2816 if (!ext4_has_feature_gdt_csum(sb))
2819 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2820 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2821 crc = crc16(crc, (__u8 *)gdp, offset);
2822 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2823 /* for checksum of struct ext4_group_desc do the rest...*/
2824 if (ext4_has_feature_64bit(sb) &&
2825 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2826 crc = crc16(crc, (__u8 *)gdp + offset,
2827 le16_to_cpu(sbi->s_es->s_desc_size) -
2831 return cpu_to_le16(crc);
2834 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2835 struct ext4_group_desc *gdp)
2837 if (ext4_has_group_desc_csum(sb) &&
2838 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2844 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2845 struct ext4_group_desc *gdp)
2847 if (!ext4_has_group_desc_csum(sb))
2849 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2852 /* Called at mount-time, super-block is locked */
2853 static int ext4_check_descriptors(struct super_block *sb,
2854 ext4_fsblk_t sb_block,
2855 ext4_group_t *first_not_zeroed)
2857 struct ext4_sb_info *sbi = EXT4_SB(sb);
2858 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2859 ext4_fsblk_t last_block;
2860 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2861 ext4_fsblk_t block_bitmap;
2862 ext4_fsblk_t inode_bitmap;
2863 ext4_fsblk_t inode_table;
2864 int flexbg_flag = 0;
2865 ext4_group_t i, grp = sbi->s_groups_count;
2867 if (ext4_has_feature_flex_bg(sb))
2870 ext4_debug("Checking group descriptors");
2872 for (i = 0; i < sbi->s_groups_count; i++) {
2873 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2875 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2876 last_block = ext4_blocks_count(sbi->s_es) - 1;
2878 last_block = first_block +
2879 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2881 if ((grp == sbi->s_groups_count) &&
2882 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2885 block_bitmap = ext4_block_bitmap(sb, gdp);
2886 if (block_bitmap == sb_block) {
2887 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2888 "Block bitmap for group %u overlaps "
2893 if (block_bitmap >= sb_block + 1 &&
2894 block_bitmap <= last_bg_block) {
2895 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2896 "Block bitmap for group %u overlaps "
2897 "block group descriptors", i);
2901 if (block_bitmap < first_block || block_bitmap > last_block) {
2902 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2903 "Block bitmap for group %u not in group "
2904 "(block %llu)!", i, block_bitmap);
2907 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2908 if (inode_bitmap == sb_block) {
2909 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2910 "Inode bitmap for group %u overlaps "
2915 if (inode_bitmap >= sb_block + 1 &&
2916 inode_bitmap <= last_bg_block) {
2917 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2918 "Inode bitmap for group %u overlaps "
2919 "block group descriptors", i);
2923 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2925 "Inode bitmap for group %u not in group "
2926 "(block %llu)!", i, inode_bitmap);
2929 inode_table = ext4_inode_table(sb, gdp);
2930 if (inode_table == sb_block) {
2931 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2932 "Inode table for group %u overlaps "
2937 if (inode_table >= sb_block + 1 &&
2938 inode_table <= last_bg_block) {
2939 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2940 "Inode table for group %u overlaps "
2941 "block group descriptors", i);
2945 if (inode_table < first_block ||
2946 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2947 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2948 "Inode table for group %u not in group "
2949 "(block %llu)!", i, inode_table);
2952 ext4_lock_group(sb, i);
2953 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2954 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2955 "Checksum for group %u failed (%u!=%u)",
2956 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2957 gdp)), le16_to_cpu(gdp->bg_checksum));
2958 if (!sb_rdonly(sb)) {
2959 ext4_unlock_group(sb, i);
2963 ext4_unlock_group(sb, i);
2965 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2967 if (NULL != first_not_zeroed)
2968 *first_not_zeroed = grp;
2973 * Maximal extent format file size.
2974 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2975 * extent format containers, within a sector_t, and within i_blocks
2976 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2977 * so that won't be a limiting factor.
2979 * However there is other limiting factor. We do store extents in the form
2980 * of starting block and length, hence the resulting length of the extent
2981 * covering maximum file size must fit into on-disk format containers as
2982 * well. Given that length is always by 1 unit bigger than max unit (because
2983 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2985 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2987 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2990 loff_t upper_limit = MAX_LFS_FILESIZE;
2992 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2994 if (!has_huge_files) {
2995 upper_limit = (1LL << 32) - 1;
2997 /* total blocks in file system block size */
2998 upper_limit >>= (blkbits - 9);
2999 upper_limit <<= blkbits;
3003 * 32-bit extent-start container, ee_block. We lower the maxbytes
3004 * by one fs block, so ee_len can cover the extent of maximum file
3007 res = (1LL << 32) - 1;
3010 /* Sanity check against vm- & vfs- imposed limits */
3011 if (res > upper_limit)
3018 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3019 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3020 * We need to be 1 filesystem block less than the 2^48 sector limit.
3022 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3024 loff_t res = EXT4_NDIR_BLOCKS;
3027 /* This is calculated to be the largest file size for a dense, block
3028 * mapped file such that the file's total number of 512-byte sectors,
3029 * including data and all indirect blocks, does not exceed (2^48 - 1).
3031 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3032 * number of 512-byte sectors of the file.
3035 if (!has_huge_files) {
3037 * !has_huge_files or implies that the inode i_block field
3038 * represents total file blocks in 2^32 512-byte sectors ==
3039 * size of vfs inode i_blocks * 8
3041 upper_limit = (1LL << 32) - 1;
3043 /* total blocks in file system block size */
3044 upper_limit >>= (bits - 9);
3048 * We use 48 bit ext4_inode i_blocks
3049 * With EXT4_HUGE_FILE_FL set the i_blocks
3050 * represent total number of blocks in
3051 * file system block size
3053 upper_limit = (1LL << 48) - 1;
3057 /* indirect blocks */
3059 /* double indirect blocks */
3060 meta_blocks += 1 + (1LL << (bits-2));
3061 /* tripple indirect blocks */
3062 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3064 upper_limit -= meta_blocks;
3065 upper_limit <<= bits;
3067 res += 1LL << (bits-2);
3068 res += 1LL << (2*(bits-2));
3069 res += 1LL << (3*(bits-2));
3071 if (res > upper_limit)
3074 if (res > MAX_LFS_FILESIZE)
3075 res = MAX_LFS_FILESIZE;
3080 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3081 ext4_fsblk_t logical_sb_block, int nr)
3083 struct ext4_sb_info *sbi = EXT4_SB(sb);
3084 ext4_group_t bg, first_meta_bg;
3087 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3089 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3090 return logical_sb_block + nr + 1;
3091 bg = sbi->s_desc_per_block * nr;
3092 if (ext4_bg_has_super(sb, bg))
3096 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3097 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3098 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3101 if (sb->s_blocksize == 1024 && nr == 0 &&
3102 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3105 return (has_super + ext4_group_first_block_no(sb, bg));
3109 * ext4_get_stripe_size: Get the stripe size.
3110 * @sbi: In memory super block info
3112 * If we have specified it via mount option, then
3113 * use the mount option value. If the value specified at mount time is
3114 * greater than the blocks per group use the super block value.
3115 * If the super block value is greater than blocks per group return 0.
3116 * Allocator needs it be less than blocks per group.
3119 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3121 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3122 unsigned long stripe_width =
3123 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3126 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3127 ret = sbi->s_stripe;
3128 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3130 else if (stride && stride <= sbi->s_blocks_per_group)
3136 * If the stripe width is 1, this makes no sense and
3137 * we set it to 0 to turn off stripe handling code.
3146 * Check whether this filesystem can be mounted based on
3147 * the features present and the RDONLY/RDWR mount requested.
3148 * Returns 1 if this filesystem can be mounted as requested,
3149 * 0 if it cannot be.
3151 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3153 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3154 ext4_msg(sb, KERN_ERR,
3155 "Couldn't mount because of "
3156 "unsupported optional features (%x)",
3157 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3158 ~EXT4_FEATURE_INCOMPAT_SUPP));
3162 #ifndef CONFIG_UNICODE
3163 if (ext4_has_feature_casefold(sb)) {
3164 ext4_msg(sb, KERN_ERR,
3165 "Filesystem with casefold feature cannot be "
3166 "mounted without CONFIG_UNICODE");
3174 if (ext4_has_feature_readonly(sb)) {
3175 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3176 sb->s_flags |= SB_RDONLY;
3180 /* Check that feature set is OK for a read-write mount */
3181 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3182 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3183 "unsupported optional features (%x)",
3184 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3185 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3188 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3189 ext4_msg(sb, KERN_ERR,
3190 "Can't support bigalloc feature without "
3191 "extents feature\n");
3195 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3196 if (!readonly && (ext4_has_feature_quota(sb) ||
3197 ext4_has_feature_project(sb))) {
3198 ext4_msg(sb, KERN_ERR,
3199 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3202 #endif /* CONFIG_QUOTA */
3207 * This function is called once a day if we have errors logged
3208 * on the file system
3210 static void print_daily_error_info(struct timer_list *t)
3212 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3213 struct super_block *sb = sbi->s_sb;
3214 struct ext4_super_block *es = sbi->s_es;
3216 if (es->s_error_count)
3217 /* fsck newer than v1.41.13 is needed to clean this condition. */
3218 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3219 le32_to_cpu(es->s_error_count));
3220 if (es->s_first_error_time) {
3221 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3223 ext4_get_tstamp(es, s_first_error_time),
3224 (int) sizeof(es->s_first_error_func),
3225 es->s_first_error_func,
3226 le32_to_cpu(es->s_first_error_line));
3227 if (es->s_first_error_ino)
3228 printk(KERN_CONT ": inode %u",
3229 le32_to_cpu(es->s_first_error_ino));
3230 if (es->s_first_error_block)
3231 printk(KERN_CONT ": block %llu", (unsigned long long)
3232 le64_to_cpu(es->s_first_error_block));
3233 printk(KERN_CONT "\n");
3235 if (es->s_last_error_time) {
3236 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3238 ext4_get_tstamp(es, s_last_error_time),
3239 (int) sizeof(es->s_last_error_func),
3240 es->s_last_error_func,
3241 le32_to_cpu(es->s_last_error_line));
3242 if (es->s_last_error_ino)
3243 printk(KERN_CONT ": inode %u",
3244 le32_to_cpu(es->s_last_error_ino));
3245 if (es->s_last_error_block)
3246 printk(KERN_CONT ": block %llu", (unsigned long long)
3247 le64_to_cpu(es->s_last_error_block));
3248 printk(KERN_CONT "\n");
3250 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3253 /* Find next suitable group and run ext4_init_inode_table */
3254 static int ext4_run_li_request(struct ext4_li_request *elr)
3256 struct ext4_group_desc *gdp = NULL;
3257 struct super_block *sb = elr->lr_super;
3258 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3259 ext4_group_t group = elr->lr_next_group;
3260 unsigned long timeout = 0;
3261 unsigned int prefetch_ios = 0;
3264 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3265 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3266 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3268 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3270 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3272 if (group >= elr->lr_next_group) {
3274 if (elr->lr_first_not_zeroed != ngroups &&
3275 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3276 elr->lr_next_group = elr->lr_first_not_zeroed;
3277 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3284 for (; group < ngroups; group++) {
3285 gdp = ext4_get_group_desc(sb, group, NULL);
3291 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3295 if (group >= ngroups)
3300 ret = ext4_init_inode_table(sb, group,
3301 elr->lr_timeout ? 0 : 1);
3302 trace_ext4_lazy_itable_init(sb, group);
3303 if (elr->lr_timeout == 0) {
3304 timeout = (jiffies - timeout) *
3305 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3306 elr->lr_timeout = timeout;
3308 elr->lr_next_sched = jiffies + elr->lr_timeout;
3309 elr->lr_next_group = group + 1;
3315 * Remove lr_request from the list_request and free the
3316 * request structure. Should be called with li_list_mtx held
3318 static void ext4_remove_li_request(struct ext4_li_request *elr)
3323 list_del(&elr->lr_request);
3324 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3328 static void ext4_unregister_li_request(struct super_block *sb)
3330 mutex_lock(&ext4_li_mtx);
3331 if (!ext4_li_info) {
3332 mutex_unlock(&ext4_li_mtx);
3336 mutex_lock(&ext4_li_info->li_list_mtx);
3337 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3338 mutex_unlock(&ext4_li_info->li_list_mtx);
3339 mutex_unlock(&ext4_li_mtx);
3342 static struct task_struct *ext4_lazyinit_task;
3345 * This is the function where ext4lazyinit thread lives. It walks
3346 * through the request list searching for next scheduled filesystem.
3347 * When such a fs is found, run the lazy initialization request
3348 * (ext4_rn_li_request) and keep track of the time spend in this
3349 * function. Based on that time we compute next schedule time of
3350 * the request. When walking through the list is complete, compute
3351 * next waking time and put itself into sleep.
3353 static int ext4_lazyinit_thread(void *arg)
3355 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3356 struct list_head *pos, *n;
3357 struct ext4_li_request *elr;
3358 unsigned long next_wakeup, cur;
3360 BUG_ON(NULL == eli);
3364 next_wakeup = MAX_JIFFY_OFFSET;
3366 mutex_lock(&eli->li_list_mtx);
3367 if (list_empty(&eli->li_request_list)) {
3368 mutex_unlock(&eli->li_list_mtx);
3371 list_for_each_safe(pos, n, &eli->li_request_list) {
3374 elr = list_entry(pos, struct ext4_li_request,
3377 if (time_before(jiffies, elr->lr_next_sched)) {
3378 if (time_before(elr->lr_next_sched, next_wakeup))
3379 next_wakeup = elr->lr_next_sched;
3382 if (down_read_trylock(&elr->lr_super->s_umount)) {
3383 if (sb_start_write_trylock(elr->lr_super)) {
3386 * We hold sb->s_umount, sb can not
3387 * be removed from the list, it is
3388 * now safe to drop li_list_mtx
3390 mutex_unlock(&eli->li_list_mtx);
3391 err = ext4_run_li_request(elr);
3392 sb_end_write(elr->lr_super);
3393 mutex_lock(&eli->li_list_mtx);
3396 up_read((&elr->lr_super->s_umount));
3398 /* error, remove the lazy_init job */
3400 ext4_remove_li_request(elr);
3404 elr->lr_next_sched = jiffies +
3406 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3408 if (time_before(elr->lr_next_sched, next_wakeup))
3409 next_wakeup = elr->lr_next_sched;
3411 mutex_unlock(&eli->li_list_mtx);
3416 if ((time_after_eq(cur, next_wakeup)) ||
3417 (MAX_JIFFY_OFFSET == next_wakeup)) {
3422 schedule_timeout_interruptible(next_wakeup - cur);
3424 if (kthread_should_stop()) {
3425 ext4_clear_request_list();
3432 * It looks like the request list is empty, but we need
3433 * to check it under the li_list_mtx lock, to prevent any
3434 * additions into it, and of course we should lock ext4_li_mtx
3435 * to atomically free the list and ext4_li_info, because at
3436 * this point another ext4 filesystem could be registering
3439 mutex_lock(&ext4_li_mtx);
3440 mutex_lock(&eli->li_list_mtx);
3441 if (!list_empty(&eli->li_request_list)) {
3442 mutex_unlock(&eli->li_list_mtx);
3443 mutex_unlock(&ext4_li_mtx);
3446 mutex_unlock(&eli->li_list_mtx);
3447 kfree(ext4_li_info);
3448 ext4_li_info = NULL;
3449 mutex_unlock(&ext4_li_mtx);
3454 static void ext4_clear_request_list(void)
3456 struct list_head *pos, *n;
3457 struct ext4_li_request *elr;
3459 mutex_lock(&ext4_li_info->li_list_mtx);
3460 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3461 elr = list_entry(pos, struct ext4_li_request,
3463 ext4_remove_li_request(elr);
3465 mutex_unlock(&ext4_li_info->li_list_mtx);
3468 static int ext4_run_lazyinit_thread(void)
3470 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3471 ext4_li_info, "ext4lazyinit");
3472 if (IS_ERR(ext4_lazyinit_task)) {
3473 int err = PTR_ERR(ext4_lazyinit_task);
3474 ext4_clear_request_list();
3475 kfree(ext4_li_info);
3476 ext4_li_info = NULL;
3477 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3478 "initialization thread\n",
3482 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3487 * Check whether it make sense to run itable init. thread or not.
3488 * If there is at least one uninitialized inode table, return
3489 * corresponding group number, else the loop goes through all
3490 * groups and return total number of groups.
3492 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3494 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3495 struct ext4_group_desc *gdp = NULL;
3497 if (!ext4_has_group_desc_csum(sb))
3500 for (group = 0; group < ngroups; group++) {
3501 gdp = ext4_get_group_desc(sb, group, NULL);
3505 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3512 static int ext4_li_info_new(void)
3514 struct ext4_lazy_init *eli = NULL;
3516 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3520 INIT_LIST_HEAD(&eli->li_request_list);
3521 mutex_init(&eli->li_list_mtx);
3523 eli->li_state |= EXT4_LAZYINIT_QUIT;
3530 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3533 struct ext4_li_request *elr;
3535 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3540 elr->lr_first_not_zeroed = start;
3541 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3542 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3543 elr->lr_next_group = start;
3545 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3549 * Randomize first schedule time of the request to
3550 * spread the inode table initialization requests
3553 elr->lr_next_sched = jiffies + (prandom_u32() %
3554 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3558 int ext4_register_li_request(struct super_block *sb,
3559 ext4_group_t first_not_zeroed)
3561 struct ext4_sb_info *sbi = EXT4_SB(sb);
3562 struct ext4_li_request *elr = NULL;
3563 ext4_group_t ngroups = sbi->s_groups_count;
3566 mutex_lock(&ext4_li_mtx);
3567 if (sbi->s_li_request != NULL) {
3569 * Reset timeout so it can be computed again, because
3570 * s_li_wait_mult might have changed.
3572 sbi->s_li_request->lr_timeout = 0;
3576 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3577 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3578 !test_opt(sb, INIT_INODE_TABLE)))
3581 elr = ext4_li_request_new(sb, first_not_zeroed);
3587 if (NULL == ext4_li_info) {
3588 ret = ext4_li_info_new();
3593 mutex_lock(&ext4_li_info->li_list_mtx);
3594 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3595 mutex_unlock(&ext4_li_info->li_list_mtx);
3597 sbi->s_li_request = elr;
3599 * set elr to NULL here since it has been inserted to
3600 * the request_list and the removal and free of it is
3601 * handled by ext4_clear_request_list from now on.
3605 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3606 ret = ext4_run_lazyinit_thread();
3611 mutex_unlock(&ext4_li_mtx);
3618 * We do not need to lock anything since this is called on
3621 static void ext4_destroy_lazyinit_thread(void)
3624 * If thread exited earlier
3625 * there's nothing to be done.
3627 if (!ext4_li_info || !ext4_lazyinit_task)
3630 kthread_stop(ext4_lazyinit_task);
3633 static int set_journal_csum_feature_set(struct super_block *sb)
3636 int compat, incompat;
3637 struct ext4_sb_info *sbi = EXT4_SB(sb);
3639 if (ext4_has_metadata_csum(sb)) {
3640 /* journal checksum v3 */
3642 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3644 /* journal checksum v1 */
3645 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3649 jbd2_journal_clear_features(sbi->s_journal,
3650 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3651 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3652 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3653 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3654 ret = jbd2_journal_set_features(sbi->s_journal,
3656 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3658 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3659 ret = jbd2_journal_set_features(sbi->s_journal,
3662 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3663 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3665 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3666 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3673 * Note: calculating the overhead so we can be compatible with
3674 * historical BSD practice is quite difficult in the face of
3675 * clusters/bigalloc. This is because multiple metadata blocks from
3676 * different block group can end up in the same allocation cluster.
3677 * Calculating the exact overhead in the face of clustered allocation
3678 * requires either O(all block bitmaps) in memory or O(number of block
3679 * groups**2) in time. We will still calculate the superblock for
3680 * older file systems --- and if we come across with a bigalloc file
3681 * system with zero in s_overhead_clusters the estimate will be close to
3682 * correct especially for very large cluster sizes --- but for newer
3683 * file systems, it's better to calculate this figure once at mkfs
3684 * time, and store it in the superblock. If the superblock value is
3685 * present (even for non-bigalloc file systems), we will use it.
3687 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3690 struct ext4_sb_info *sbi = EXT4_SB(sb);
3691 struct ext4_group_desc *gdp;
3692 ext4_fsblk_t first_block, last_block, b;
3693 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3694 int s, j, count = 0;
3696 if (!ext4_has_feature_bigalloc(sb))
3697 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3698 sbi->s_itb_per_group + 2);
3700 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3701 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3702 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3703 for (i = 0; i < ngroups; i++) {
3704 gdp = ext4_get_group_desc(sb, i, NULL);
3705 b = ext4_block_bitmap(sb, gdp);
3706 if (b >= first_block && b <= last_block) {
3707 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3710 b = ext4_inode_bitmap(sb, gdp);
3711 if (b >= first_block && b <= last_block) {
3712 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3715 b = ext4_inode_table(sb, gdp);
3716 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3717 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3718 int c = EXT4_B2C(sbi, b - first_block);
3719 ext4_set_bit(c, buf);
3725 if (ext4_bg_has_super(sb, grp)) {
3726 ext4_set_bit(s++, buf);
3729 j = ext4_bg_num_gdb(sb, grp);
3730 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3731 ext4_error(sb, "Invalid number of block group "
3732 "descriptor blocks: %d", j);
3733 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3737 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3741 return EXT4_CLUSTERS_PER_GROUP(sb) -
3742 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3746 * Compute the overhead and stash it in sbi->s_overhead
3748 int ext4_calculate_overhead(struct super_block *sb)
3750 struct ext4_sb_info *sbi = EXT4_SB(sb);
3751 struct ext4_super_block *es = sbi->s_es;
3752 struct inode *j_inode;
3753 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3754 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3755 ext4_fsblk_t overhead = 0;
3756 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3762 * Compute the overhead (FS structures). This is constant
3763 * for a given filesystem unless the number of block groups
3764 * changes so we cache the previous value until it does.
3768 * All of the blocks before first_data_block are overhead
3770 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3773 * Add the overhead found in each block group
3775 for (i = 0; i < ngroups; i++) {
3778 blks = count_overhead(sb, i, buf);
3781 memset(buf, 0, PAGE_SIZE);
3786 * Add the internal journal blocks whether the journal has been
3789 if (sbi->s_journal && !sbi->s_journal_bdev)
3790 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3791 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3792 /* j_inum for internal journal is non-zero */
3793 j_inode = ext4_get_journal_inode(sb, j_inum);
3795 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3796 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3799 ext4_msg(sb, KERN_ERR, "can't get journal size");
3802 sbi->s_overhead = overhead;
3804 free_page((unsigned long) buf);
3808 static void ext4_set_resv_clusters(struct super_block *sb)
3810 ext4_fsblk_t resv_clusters;
3811 struct ext4_sb_info *sbi = EXT4_SB(sb);
3814 * There's no need to reserve anything when we aren't using extents.
3815 * The space estimates are exact, there are no unwritten extents,
3816 * hole punching doesn't need new metadata... This is needed especially
3817 * to keep ext2/3 backward compatibility.
3819 if (!ext4_has_feature_extents(sb))
3822 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3823 * This should cover the situations where we can not afford to run
3824 * out of space like for example punch hole, or converting
3825 * unwritten extents in delalloc path. In most cases such
3826 * allocation would require 1, or 2 blocks, higher numbers are
3829 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3830 sbi->s_cluster_bits);
3832 do_div(resv_clusters, 50);
3833 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3835 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3838 static const char *ext4_quota_mode(struct super_block *sb)
3841 if (!ext4_quota_capable(sb))
3844 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3845 return "journalled";
3853 static void ext4_setup_csum_trigger(struct super_block *sb,
3854 enum ext4_journal_trigger_type type,
3856 struct jbd2_buffer_trigger_type *type,
3857 struct buffer_head *bh,
3861 struct ext4_sb_info *sbi = EXT4_SB(sb);
3863 sbi->s_journal_triggers[type].sb = sb;
3864 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
3867 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3869 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3870 char *orig_data = kstrdup(data, GFP_KERNEL);
3871 struct buffer_head *bh, **group_desc;
3872 struct ext4_super_block *es = NULL;
3873 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3874 struct flex_groups **flex_groups;
3876 ext4_fsblk_t sb_block = get_sb_block(&data);
3877 ext4_fsblk_t logical_sb_block;
3878 unsigned long offset = 0;
3879 unsigned long def_mount_opts;
3883 int blocksize, clustersize;
3884 unsigned int db_count;
3886 int needs_recovery, has_huge_files;
3889 ext4_group_t first_not_zeroed;
3890 struct ext4_parsed_options parsed_opts;
3892 /* Set defaults for the variables that will be set during parsing */
3893 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3894 parsed_opts.journal_devnum = 0;
3895 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
3897 if ((data && !orig_data) || !sbi)
3900 sbi->s_daxdev = dax_dev;
3901 sbi->s_blockgroup_lock =
3902 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3903 if (!sbi->s_blockgroup_lock)
3906 sb->s_fs_info = sbi;
3908 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3909 sbi->s_sb_block = sb_block;
3910 sbi->s_sectors_written_start =
3911 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
3913 /* Cleanup superblock name */
3914 strreplace(sb->s_id, '/', '!');
3916 /* -EINVAL is default */
3918 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3920 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3925 * The ext4 superblock will not be buffer aligned for other than 1kB
3926 * block sizes. We need to calculate the offset from buffer start.
3928 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3929 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3930 offset = do_div(logical_sb_block, blocksize);
3932 logical_sb_block = sb_block;
3935 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
3937 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3942 * Note: s_es must be initialized as soon as possible because
3943 * some ext4 macro-instructions depend on its value
3945 es = (struct ext4_super_block *) (bh->b_data + offset);
3947 sb->s_magic = le16_to_cpu(es->s_magic);
3948 if (sb->s_magic != EXT4_SUPER_MAGIC)
3950 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3952 /* Warn if metadata_csum and gdt_csum are both set. */
3953 if (ext4_has_feature_metadata_csum(sb) &&
3954 ext4_has_feature_gdt_csum(sb))
3955 ext4_warning(sb, "metadata_csum and uninit_bg are "
3956 "redundant flags; please run fsck.");
3958 /* Check for a known checksum algorithm */
3959 if (!ext4_verify_csum_type(sb, es)) {
3960 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3961 "unknown checksum algorithm.");
3965 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
3966 ext4_orphan_file_block_trigger);
3968 /* Load the checksum driver */
3969 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3970 if (IS_ERR(sbi->s_chksum_driver)) {
3971 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3972 ret = PTR_ERR(sbi->s_chksum_driver);
3973 sbi->s_chksum_driver = NULL;
3977 /* Check superblock checksum */
3978 if (!ext4_superblock_csum_verify(sb, es)) {
3979 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3980 "invalid superblock checksum. Run e2fsck?");
3986 /* Precompute checksum seed for all metadata */
3987 if (ext4_has_feature_csum_seed(sb))
3988 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3989 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3990 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3991 sizeof(es->s_uuid));
3993 /* Set defaults before we parse the mount options */
3994 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3995 set_opt(sb, INIT_INODE_TABLE);
3996 if (def_mount_opts & EXT4_DEFM_DEBUG)
3998 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4000 if (def_mount_opts & EXT4_DEFM_UID16)
4001 set_opt(sb, NO_UID32);
4002 /* xattr user namespace & acls are now defaulted on */
4003 set_opt(sb, XATTR_USER);
4004 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4005 set_opt(sb, POSIX_ACL);
4007 if (ext4_has_feature_fast_commit(sb))
4008 set_opt2(sb, JOURNAL_FAST_COMMIT);
4009 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4010 if (ext4_has_metadata_csum(sb))
4011 set_opt(sb, JOURNAL_CHECKSUM);
4013 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4014 set_opt(sb, JOURNAL_DATA);
4015 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4016 set_opt(sb, ORDERED_DATA);
4017 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4018 set_opt(sb, WRITEBACK_DATA);
4020 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4021 set_opt(sb, ERRORS_PANIC);
4022 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4023 set_opt(sb, ERRORS_CONT);
4025 set_opt(sb, ERRORS_RO);
4026 /* block_validity enabled by default; disable with noblock_validity */
4027 set_opt(sb, BLOCK_VALIDITY);
4028 if (def_mount_opts & EXT4_DEFM_DISCARD)
4029 set_opt(sb, DISCARD);
4031 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4032 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4033 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4034 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4035 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4037 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4038 set_opt(sb, BARRIER);
4041 * enable delayed allocation by default
4042 * Use -o nodelalloc to turn it off
4044 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4045 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4046 set_opt(sb, DELALLOC);
4049 * set default s_li_wait_mult for lazyinit, for the case there is
4050 * no mount option specified.
4052 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4054 if (le32_to_cpu(es->s_log_block_size) >
4055 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4056 ext4_msg(sb, KERN_ERR,
4057 "Invalid log block size: %u",
4058 le32_to_cpu(es->s_log_block_size));
4061 if (le32_to_cpu(es->s_log_cluster_size) >
4062 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4063 ext4_msg(sb, KERN_ERR,
4064 "Invalid log cluster size: %u",
4065 le32_to_cpu(es->s_log_cluster_size));
4069 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4071 if (blocksize == PAGE_SIZE)
4072 set_opt(sb, DIOREAD_NOLOCK);
4074 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4075 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4076 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4078 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4079 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4080 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4081 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4085 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4086 (!is_power_of_2(sbi->s_inode_size)) ||
4087 (sbi->s_inode_size > blocksize)) {
4088 ext4_msg(sb, KERN_ERR,
4089 "unsupported inode size: %d",
4091 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4095 * i_atime_extra is the last extra field available for
4096 * [acm]times in struct ext4_inode. Checking for that
4097 * field should suffice to ensure we have extra space
4100 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4101 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4102 sb->s_time_gran = 1;
4103 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4105 sb->s_time_gran = NSEC_PER_SEC;
4106 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4108 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4110 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4111 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4112 EXT4_GOOD_OLD_INODE_SIZE;
4113 if (ext4_has_feature_extra_isize(sb)) {
4114 unsigned v, max = (sbi->s_inode_size -
4115 EXT4_GOOD_OLD_INODE_SIZE);
4117 v = le16_to_cpu(es->s_want_extra_isize);
4119 ext4_msg(sb, KERN_ERR,
4120 "bad s_want_extra_isize: %d", v);
4123 if (sbi->s_want_extra_isize < v)
4124 sbi->s_want_extra_isize = v;
4126 v = le16_to_cpu(es->s_min_extra_isize);
4128 ext4_msg(sb, KERN_ERR,
4129 "bad s_min_extra_isize: %d", v);
4132 if (sbi->s_want_extra_isize < v)
4133 sbi->s_want_extra_isize = v;
4137 if (sbi->s_es->s_mount_opts[0]) {
4138 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4139 sizeof(sbi->s_es->s_mount_opts),
4143 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4144 ext4_msg(sb, KERN_WARNING,
4145 "failed to parse options in superblock: %s",
4148 kfree(s_mount_opts);
4150 sbi->s_def_mount_opt = sbi->s_mount_opt;
4151 if (!parse_options((char *) data, sb, &parsed_opts, 0))
4154 #ifdef CONFIG_UNICODE
4155 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4156 const struct ext4_sb_encodings *encoding_info;
4157 struct unicode_map *encoding;
4158 __u16 encoding_flags;
4160 if (ext4_sb_read_encoding(es, &encoding_info,
4162 ext4_msg(sb, KERN_ERR,
4163 "Encoding requested by superblock is unknown");
4167 encoding = utf8_load(encoding_info->version);
4168 if (IS_ERR(encoding)) {
4169 ext4_msg(sb, KERN_ERR,
4170 "can't mount with superblock charset: %s-%s "
4171 "not supported by the kernel. flags: 0x%x.",
4172 encoding_info->name, encoding_info->version,
4176 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4177 "%s-%s with flags 0x%hx", encoding_info->name,
4178 encoding_info->version?:"\b", encoding_flags);
4180 sb->s_encoding = encoding;
4181 sb->s_encoding_flags = encoding_flags;
4185 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4186 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4187 /* can't mount with both data=journal and dioread_nolock. */
4188 clear_opt(sb, DIOREAD_NOLOCK);
4189 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4190 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4191 ext4_msg(sb, KERN_ERR, "can't mount with "
4192 "both data=journal and delalloc");
4195 if (test_opt(sb, DAX_ALWAYS)) {
4196 ext4_msg(sb, KERN_ERR, "can't mount with "
4197 "both data=journal and dax");
4200 if (ext4_has_feature_encrypt(sb)) {
4201 ext4_msg(sb, KERN_WARNING,
4202 "encrypted files will use data=ordered "
4203 "instead of data journaling mode");
4205 if (test_opt(sb, DELALLOC))
4206 clear_opt(sb, DELALLOC);
4208 sb->s_iflags |= SB_I_CGROUPWB;
4211 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4212 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4214 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4215 (ext4_has_compat_features(sb) ||
4216 ext4_has_ro_compat_features(sb) ||
4217 ext4_has_incompat_features(sb)))
4218 ext4_msg(sb, KERN_WARNING,
4219 "feature flags set on rev 0 fs, "
4220 "running e2fsck is recommended");
4222 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4223 set_opt2(sb, HURD_COMPAT);
4224 if (ext4_has_feature_64bit(sb)) {
4225 ext4_msg(sb, KERN_ERR,
4226 "The Hurd can't support 64-bit file systems");
4231 * ea_inode feature uses l_i_version field which is not
4232 * available in HURD_COMPAT mode.
4234 if (ext4_has_feature_ea_inode(sb)) {
4235 ext4_msg(sb, KERN_ERR,
4236 "ea_inode feature is not supported for Hurd");
4241 if (IS_EXT2_SB(sb)) {
4242 if (ext2_feature_set_ok(sb))
4243 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4244 "using the ext4 subsystem");
4247 * If we're probing be silent, if this looks like
4248 * it's actually an ext[34] filesystem.
4250 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4252 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4253 "to feature incompatibilities");
4258 if (IS_EXT3_SB(sb)) {
4259 if (ext3_feature_set_ok(sb))
4260 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4261 "using the ext4 subsystem");
4264 * If we're probing be silent, if this looks like
4265 * it's actually an ext4 filesystem.
4267 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4269 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4270 "to feature incompatibilities");
4276 * Check feature flags regardless of the revision level, since we
4277 * previously didn't change the revision level when setting the flags,
4278 * so there is a chance incompat flags are set on a rev 0 filesystem.
4280 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4283 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4284 ext4_msg(sb, KERN_ERR,
4285 "Number of reserved GDT blocks insanely large: %d",
4286 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4290 if (dax_supported(dax_dev, sb->s_bdev, blocksize, 0,
4291 bdev_nr_sectors(sb->s_bdev)))
4292 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4294 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4295 if (ext4_has_feature_inline_data(sb)) {
4296 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4297 " that may contain inline data");
4300 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4301 ext4_msg(sb, KERN_ERR,
4302 "DAX unsupported by block device.");
4307 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4308 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4309 es->s_encryption_level);
4313 if (sb->s_blocksize != blocksize) {
4315 * bh must be released before kill_bdev(), otherwise
4316 * it won't be freed and its page also. kill_bdev()
4317 * is called by sb_set_blocksize().
4320 /* Validate the filesystem blocksize */
4321 if (!sb_set_blocksize(sb, blocksize)) {
4322 ext4_msg(sb, KERN_ERR, "bad block size %d",
4328 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4329 offset = do_div(logical_sb_block, blocksize);
4330 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4332 ext4_msg(sb, KERN_ERR,
4333 "Can't read superblock on 2nd try");
4338 es = (struct ext4_super_block *)(bh->b_data + offset);
4340 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4341 ext4_msg(sb, KERN_ERR,
4342 "Magic mismatch, very weird!");
4347 has_huge_files = ext4_has_feature_huge_file(sb);
4348 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4350 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4352 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4353 if (ext4_has_feature_64bit(sb)) {
4354 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4355 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4356 !is_power_of_2(sbi->s_desc_size)) {
4357 ext4_msg(sb, KERN_ERR,
4358 "unsupported descriptor size %lu",
4363 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4365 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4366 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4368 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4369 if (sbi->s_inodes_per_block == 0)
4371 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4372 sbi->s_inodes_per_group > blocksize * 8) {
4373 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4374 sbi->s_inodes_per_group);
4377 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4378 sbi->s_inodes_per_block;
4379 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4381 sbi->s_mount_state = le16_to_cpu(es->s_state);
4382 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4383 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4385 for (i = 0; i < 4; i++)
4386 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4387 sbi->s_def_hash_version = es->s_def_hash_version;
4388 if (ext4_has_feature_dir_index(sb)) {
4389 i = le32_to_cpu(es->s_flags);
4390 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4391 sbi->s_hash_unsigned = 3;
4392 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4393 #ifdef __CHAR_UNSIGNED__
4396 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4397 sbi->s_hash_unsigned = 3;
4401 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4406 /* Handle clustersize */
4407 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4408 if (ext4_has_feature_bigalloc(sb)) {
4409 if (clustersize < blocksize) {
4410 ext4_msg(sb, KERN_ERR,
4411 "cluster size (%d) smaller than "
4412 "block size (%d)", clustersize, blocksize);
4415 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4416 le32_to_cpu(es->s_log_block_size);
4417 sbi->s_clusters_per_group =
4418 le32_to_cpu(es->s_clusters_per_group);
4419 if (sbi->s_clusters_per_group > blocksize * 8) {
4420 ext4_msg(sb, KERN_ERR,
4421 "#clusters per group too big: %lu",
4422 sbi->s_clusters_per_group);
4425 if (sbi->s_blocks_per_group !=
4426 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4427 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4428 "clusters per group (%lu) inconsistent",
4429 sbi->s_blocks_per_group,
4430 sbi->s_clusters_per_group);
4434 if (clustersize != blocksize) {
4435 ext4_msg(sb, KERN_ERR,
4436 "fragment/cluster size (%d) != "
4437 "block size (%d)", clustersize, blocksize);
4440 if (sbi->s_blocks_per_group > blocksize * 8) {
4441 ext4_msg(sb, KERN_ERR,
4442 "#blocks per group too big: %lu",
4443 sbi->s_blocks_per_group);
4446 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4447 sbi->s_cluster_bits = 0;
4449 sbi->s_cluster_ratio = clustersize / blocksize;
4451 /* Do we have standard group size of clustersize * 8 blocks ? */
4452 if (sbi->s_blocks_per_group == clustersize << 3)
4453 set_opt2(sb, STD_GROUP_SIZE);
4456 * Test whether we have more sectors than will fit in sector_t,
4457 * and whether the max offset is addressable by the page cache.
4459 err = generic_check_addressable(sb->s_blocksize_bits,
4460 ext4_blocks_count(es));
4462 ext4_msg(sb, KERN_ERR, "filesystem"
4463 " too large to mount safely on this system");
4467 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4470 /* check blocks count against device size */
4471 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4472 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4473 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4474 "exceeds size of device (%llu blocks)",
4475 ext4_blocks_count(es), blocks_count);
4480 * It makes no sense for the first data block to be beyond the end
4481 * of the filesystem.
4483 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4484 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4485 "block %u is beyond end of filesystem (%llu)",
4486 le32_to_cpu(es->s_first_data_block),
4487 ext4_blocks_count(es));
4490 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4491 (sbi->s_cluster_ratio == 1)) {
4492 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4493 "block is 0 with a 1k block and cluster size");
4497 blocks_count = (ext4_blocks_count(es) -
4498 le32_to_cpu(es->s_first_data_block) +
4499 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4500 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4501 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4502 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4503 "(block count %llu, first data block %u, "
4504 "blocks per group %lu)", blocks_count,
4505 ext4_blocks_count(es),
4506 le32_to_cpu(es->s_first_data_block),
4507 EXT4_BLOCKS_PER_GROUP(sb));
4510 sbi->s_groups_count = blocks_count;
4511 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4512 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4513 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4514 le32_to_cpu(es->s_inodes_count)) {
4515 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4516 le32_to_cpu(es->s_inodes_count),
4517 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4521 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4522 EXT4_DESC_PER_BLOCK(sb);
4523 if (ext4_has_feature_meta_bg(sb)) {
4524 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4525 ext4_msg(sb, KERN_WARNING,
4526 "first meta block group too large: %u "
4527 "(group descriptor block count %u)",
4528 le32_to_cpu(es->s_first_meta_bg), db_count);
4532 rcu_assign_pointer(sbi->s_group_desc,
4533 kvmalloc_array(db_count,
4534 sizeof(struct buffer_head *),
4536 if (sbi->s_group_desc == NULL) {
4537 ext4_msg(sb, KERN_ERR, "not enough memory");
4542 bgl_lock_init(sbi->s_blockgroup_lock);
4544 /* Pre-read the descriptors into the buffer cache */
4545 for (i = 0; i < db_count; i++) {
4546 block = descriptor_loc(sb, logical_sb_block, i);
4547 ext4_sb_breadahead_unmovable(sb, block);
4550 for (i = 0; i < db_count; i++) {
4551 struct buffer_head *bh;
4553 block = descriptor_loc(sb, logical_sb_block, i);
4554 bh = ext4_sb_bread_unmovable(sb, block);
4556 ext4_msg(sb, KERN_ERR,
4557 "can't read group descriptor %d", i);
4563 rcu_dereference(sbi->s_group_desc)[i] = bh;
4566 sbi->s_gdb_count = db_count;
4567 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4568 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4569 ret = -EFSCORRUPTED;
4573 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4574 spin_lock_init(&sbi->s_error_lock);
4575 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4577 /* Register extent status tree shrinker */
4578 if (ext4_es_register_shrinker(sbi))
4581 sbi->s_stripe = ext4_get_stripe_size(sbi);
4582 sbi->s_extent_max_zeroout_kb = 32;
4585 * set up enough so that it can read an inode
4587 sb->s_op = &ext4_sops;
4588 sb->s_export_op = &ext4_export_ops;
4589 sb->s_xattr = ext4_xattr_handlers;
4590 #ifdef CONFIG_FS_ENCRYPTION
4591 sb->s_cop = &ext4_cryptops;
4593 #ifdef CONFIG_FS_VERITY
4594 sb->s_vop = &ext4_verityops;
4597 sb->dq_op = &ext4_quota_operations;
4598 if (ext4_has_feature_quota(sb))
4599 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4601 sb->s_qcop = &ext4_qctl_operations;
4602 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4604 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4606 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4607 mutex_init(&sbi->s_orphan_lock);
4609 /* Initialize fast commit stuff */
4610 atomic_set(&sbi->s_fc_subtid, 0);
4611 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4612 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4613 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4614 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4615 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4616 sbi->s_fc_bytes = 0;
4617 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4618 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4619 spin_lock_init(&sbi->s_fc_lock);
4620 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4621 sbi->s_fc_replay_state.fc_regions = NULL;
4622 sbi->s_fc_replay_state.fc_regions_size = 0;
4623 sbi->s_fc_replay_state.fc_regions_used = 0;
4624 sbi->s_fc_replay_state.fc_regions_valid = 0;
4625 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4626 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4627 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4631 needs_recovery = (es->s_last_orphan != 0 ||
4632 ext4_has_feature_orphan_present(sb) ||
4633 ext4_has_feature_journal_needs_recovery(sb));
4635 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4636 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4637 goto failed_mount3a;
4640 * The first inode we look at is the journal inode. Don't try
4641 * root first: it may be modified in the journal!
4643 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4644 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4646 goto failed_mount3a;
4647 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4648 ext4_has_feature_journal_needs_recovery(sb)) {
4649 ext4_msg(sb, KERN_ERR, "required journal recovery "
4650 "suppressed and not mounted read-only");
4651 goto failed_mount_wq;
4653 /* Nojournal mode, all journal mount options are illegal */
4654 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4655 ext4_msg(sb, KERN_ERR, "can't mount with "
4656 "journal_checksum, fs mounted w/o journal");
4657 goto failed_mount_wq;
4659 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4660 ext4_msg(sb, KERN_ERR, "can't mount with "
4661 "journal_async_commit, fs mounted w/o journal");
4662 goto failed_mount_wq;
4664 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4665 ext4_msg(sb, KERN_ERR, "can't mount with "
4666 "commit=%lu, fs mounted w/o journal",
4667 sbi->s_commit_interval / HZ);
4668 goto failed_mount_wq;
4670 if (EXT4_MOUNT_DATA_FLAGS &
4671 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4672 ext4_msg(sb, KERN_ERR, "can't mount with "
4673 "data=, fs mounted w/o journal");
4674 goto failed_mount_wq;
4676 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4677 clear_opt(sb, JOURNAL_CHECKSUM);
4678 clear_opt(sb, DATA_FLAGS);
4679 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4680 sbi->s_journal = NULL;
4685 if (ext4_has_feature_64bit(sb) &&
4686 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4687 JBD2_FEATURE_INCOMPAT_64BIT)) {
4688 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4689 goto failed_mount_wq;
4692 if (!set_journal_csum_feature_set(sb)) {
4693 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4695 goto failed_mount_wq;
4698 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4699 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4700 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4701 ext4_msg(sb, KERN_ERR,
4702 "Failed to set fast commit journal feature");
4703 goto failed_mount_wq;
4706 /* We have now updated the journal if required, so we can
4707 * validate the data journaling mode. */
4708 switch (test_opt(sb, DATA_FLAGS)) {
4710 /* No mode set, assume a default based on the journal
4711 * capabilities: ORDERED_DATA if the journal can
4712 * cope, else JOURNAL_DATA
4714 if (jbd2_journal_check_available_features
4715 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4716 set_opt(sb, ORDERED_DATA);
4717 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4719 set_opt(sb, JOURNAL_DATA);
4720 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4724 case EXT4_MOUNT_ORDERED_DATA:
4725 case EXT4_MOUNT_WRITEBACK_DATA:
4726 if (!jbd2_journal_check_available_features
4727 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4728 ext4_msg(sb, KERN_ERR, "Journal does not support "
4729 "requested data journaling mode");
4730 goto failed_mount_wq;
4737 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4738 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4739 ext4_msg(sb, KERN_ERR, "can't mount with "
4740 "journal_async_commit in data=ordered mode");
4741 goto failed_mount_wq;
4744 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4746 sbi->s_journal->j_submit_inode_data_buffers =
4747 ext4_journal_submit_inode_data_buffers;
4748 sbi->s_journal->j_finish_inode_data_buffers =
4749 ext4_journal_finish_inode_data_buffers;
4752 if (!test_opt(sb, NO_MBCACHE)) {
4753 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4754 if (!sbi->s_ea_block_cache) {
4755 ext4_msg(sb, KERN_ERR,
4756 "Failed to create ea_block_cache");
4757 goto failed_mount_wq;
4760 if (ext4_has_feature_ea_inode(sb)) {
4761 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4762 if (!sbi->s_ea_inode_cache) {
4763 ext4_msg(sb, KERN_ERR,
4764 "Failed to create ea_inode_cache");
4765 goto failed_mount_wq;
4770 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4771 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4772 goto failed_mount_wq;
4775 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4776 !ext4_has_feature_encrypt(sb)) {
4777 ext4_set_feature_encrypt(sb);
4778 ext4_commit_super(sb);
4782 * Get the # of file system overhead blocks from the
4783 * superblock if present.
4785 if (es->s_overhead_clusters)
4786 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4788 err = ext4_calculate_overhead(sb);
4790 goto failed_mount_wq;
4794 * The maximum number of concurrent works can be high and
4795 * concurrency isn't really necessary. Limit it to 1.
4797 EXT4_SB(sb)->rsv_conversion_wq =
4798 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4799 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4800 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4806 * The jbd2_journal_load will have done any necessary log recovery,
4807 * so we can safely mount the rest of the filesystem now.
4810 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4812 ext4_msg(sb, KERN_ERR, "get root inode failed");
4813 ret = PTR_ERR(root);
4817 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4818 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4823 sb->s_root = d_make_root(root);
4825 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4830 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4831 if (ret == -EROFS) {
4832 sb->s_flags |= SB_RDONLY;
4835 goto failed_mount4a;
4837 ext4_set_resv_clusters(sb);
4839 if (test_opt(sb, BLOCK_VALIDITY)) {
4840 err = ext4_setup_system_zone(sb);
4842 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4844 goto failed_mount4a;
4847 ext4_fc_replay_cleanup(sb);
4852 * Enable optimize_scan if number of groups is > threshold. This can be
4853 * turned off by passing "mb_optimize_scan=0". This can also be
4854 * turned on forcefully by passing "mb_optimize_scan=1".
4856 if (parsed_opts.mb_optimize_scan == 1)
4857 set_opt2(sb, MB_OPTIMIZE_SCAN);
4858 else if (parsed_opts.mb_optimize_scan == 0)
4859 clear_opt2(sb, MB_OPTIMIZE_SCAN);
4860 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
4861 set_opt2(sb, MB_OPTIMIZE_SCAN);
4863 err = ext4_mb_init(sb);
4865 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4871 * We can only set up the journal commit callback once
4872 * mballoc is initialized
4875 sbi->s_journal->j_commit_callback =
4876 ext4_journal_commit_callback;
4878 block = ext4_count_free_clusters(sb);
4879 ext4_free_blocks_count_set(sbi->s_es,
4880 EXT4_C2B(sbi, block));
4881 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4884 unsigned long freei = ext4_count_free_inodes(sb);
4885 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4886 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4890 * Update the checksum after updating free space/inode
4891 * counters. Otherwise the superblock can have an incorrect
4892 * checksum in the buffer cache until it is written out and
4893 * e2fsprogs programs trying to open a file system immediately
4894 * after it is mounted can fail.
4896 ext4_superblock_csum_set(sb);
4898 err = percpu_counter_init(&sbi->s_dirs_counter,
4899 ext4_count_dirs(sb), GFP_KERNEL);
4901 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4904 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
4907 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4910 ext4_msg(sb, KERN_ERR, "insufficient memory");
4914 if (ext4_has_feature_flex_bg(sb))
4915 if (!ext4_fill_flex_info(sb)) {
4916 ext4_msg(sb, KERN_ERR,
4917 "unable to initialize "
4918 "flex_bg meta info!");
4923 err = ext4_register_li_request(sb, first_not_zeroed);
4927 err = ext4_register_sysfs(sb);
4931 err = ext4_init_orphan_info(sb);
4935 /* Enable quota usage during mount. */
4936 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4937 err = ext4_enable_quotas(sb);
4941 #endif /* CONFIG_QUOTA */
4944 * Save the original bdev mapping's wb_err value which could be
4945 * used to detect the metadata async write error.
4947 spin_lock_init(&sbi->s_bdev_wb_lock);
4948 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4949 &sbi->s_bdev_wb_err);
4950 sb->s_bdev->bd_super = sb;
4951 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4952 ext4_orphan_cleanup(sb, es);
4953 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4954 if (needs_recovery) {
4955 ext4_msg(sb, KERN_INFO, "recovery complete");
4956 err = ext4_mark_recovery_complete(sb, es);
4960 if (EXT4_SB(sb)->s_journal) {
4961 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4962 descr = " journalled data mode";
4963 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4964 descr = " ordered data mode";
4966 descr = " writeback data mode";
4968 descr = "out journal";
4970 if (test_opt(sb, DISCARD)) {
4971 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4972 if (!blk_queue_discard(q))
4973 ext4_msg(sb, KERN_WARNING,
4974 "mounting with \"discard\" option, but "
4975 "the device does not support discard");
4978 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4979 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4980 "Opts: %.*s%s%s. Quota mode: %s.", descr,
4981 (int) sizeof(sbi->s_es->s_mount_opts),
4982 sbi->s_es->s_mount_opts,
4983 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
4984 ext4_quota_mode(sb));
4986 if (es->s_error_count)
4987 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4989 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4990 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4991 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4992 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4993 atomic_set(&sbi->s_warning_count, 0);
4994 atomic_set(&sbi->s_msg_count, 0);
5001 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5005 ext4_release_orphan_info(sb);
5007 ext4_unregister_sysfs(sb);
5008 kobject_put(&sbi->s_kobj);
5010 ext4_unregister_li_request(sb);
5012 ext4_mb_release(sb);
5014 flex_groups = rcu_dereference(sbi->s_flex_groups);
5016 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5017 kvfree(flex_groups[i]);
5018 kvfree(flex_groups);
5021 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5022 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5023 percpu_counter_destroy(&sbi->s_dirs_counter);
5024 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5025 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5026 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5028 ext4_ext_release(sb);
5029 ext4_release_system_zone(sb);
5034 ext4_msg(sb, KERN_ERR, "mount failed");
5035 if (EXT4_SB(sb)->rsv_conversion_wq)
5036 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5038 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5039 sbi->s_ea_inode_cache = NULL;
5041 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5042 sbi->s_ea_block_cache = NULL;
5044 if (sbi->s_journal) {
5045 jbd2_journal_destroy(sbi->s_journal);
5046 sbi->s_journal = NULL;
5049 ext4_es_unregister_shrinker(sbi);
5051 flush_work(&sbi->s_error_work);
5052 del_timer_sync(&sbi->s_err_report);
5053 ext4_stop_mmpd(sbi);
5056 group_desc = rcu_dereference(sbi->s_group_desc);
5057 for (i = 0; i < db_count; i++)
5058 brelse(group_desc[i]);
5062 if (sbi->s_chksum_driver)
5063 crypto_free_shash(sbi->s_chksum_driver);
5065 #ifdef CONFIG_UNICODE
5066 utf8_unload(sb->s_encoding);
5070 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5071 kfree(get_qf_name(sb, sbi, i));
5073 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5074 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5076 ext4_blkdev_remove(sbi);
5078 sb->s_fs_info = NULL;
5079 kfree(sbi->s_blockgroup_lock);
5083 fs_put_dax(dax_dev);
5084 return err ? err : ret;
5088 * Setup any per-fs journal parameters now. We'll do this both on
5089 * initial mount, once the journal has been initialised but before we've
5090 * done any recovery; and again on any subsequent remount.
5092 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5094 struct ext4_sb_info *sbi = EXT4_SB(sb);
5096 journal->j_commit_interval = sbi->s_commit_interval;
5097 journal->j_min_batch_time = sbi->s_min_batch_time;
5098 journal->j_max_batch_time = sbi->s_max_batch_time;
5099 ext4_fc_init(sb, journal);
5101 write_lock(&journal->j_state_lock);
5102 if (test_opt(sb, BARRIER))
5103 journal->j_flags |= JBD2_BARRIER;
5105 journal->j_flags &= ~JBD2_BARRIER;
5106 if (test_opt(sb, DATA_ERR_ABORT))
5107 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5109 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5110 write_unlock(&journal->j_state_lock);
5113 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5114 unsigned int journal_inum)
5116 struct inode *journal_inode;
5119 * Test for the existence of a valid inode on disk. Bad things
5120 * happen if we iget() an unused inode, as the subsequent iput()
5121 * will try to delete it.
5123 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5124 if (IS_ERR(journal_inode)) {
5125 ext4_msg(sb, KERN_ERR, "no journal found");
5128 if (!journal_inode->i_nlink) {
5129 make_bad_inode(journal_inode);
5130 iput(journal_inode);
5131 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5135 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5136 journal_inode, journal_inode->i_size);
5137 if (!S_ISREG(journal_inode->i_mode)) {
5138 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5139 iput(journal_inode);
5142 return journal_inode;
5145 static journal_t *ext4_get_journal(struct super_block *sb,
5146 unsigned int journal_inum)
5148 struct inode *journal_inode;
5151 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5154 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5158 journal = jbd2_journal_init_inode(journal_inode);
5160 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5161 iput(journal_inode);
5164 journal->j_private = sb;
5165 ext4_init_journal_params(sb, journal);
5169 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5172 struct buffer_head *bh;
5176 int hblock, blocksize;
5177 ext4_fsblk_t sb_block;
5178 unsigned long offset;
5179 struct ext4_super_block *es;
5180 struct block_device *bdev;
5182 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5185 bdev = ext4_blkdev_get(j_dev, sb);
5189 blocksize = sb->s_blocksize;
5190 hblock = bdev_logical_block_size(bdev);
5191 if (blocksize < hblock) {
5192 ext4_msg(sb, KERN_ERR,
5193 "blocksize too small for journal device");
5197 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5198 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5199 set_blocksize(bdev, blocksize);
5200 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5201 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5202 "external journal");
5206 es = (struct ext4_super_block *) (bh->b_data + offset);
5207 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5208 !(le32_to_cpu(es->s_feature_incompat) &
5209 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5210 ext4_msg(sb, KERN_ERR, "external journal has "
5216 if ((le32_to_cpu(es->s_feature_ro_compat) &
5217 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5218 es->s_checksum != ext4_superblock_csum(sb, es)) {
5219 ext4_msg(sb, KERN_ERR, "external journal has "
5220 "corrupt superblock");
5225 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5226 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5231 len = ext4_blocks_count(es);
5232 start = sb_block + 1;
5233 brelse(bh); /* we're done with the superblock */
5235 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5236 start, len, blocksize);
5238 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5241 journal->j_private = sb;
5242 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5243 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5246 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5247 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5248 "user (unsupported) - %d",
5249 be32_to_cpu(journal->j_superblock->s_nr_users));
5252 EXT4_SB(sb)->s_journal_bdev = bdev;
5253 ext4_init_journal_params(sb, journal);
5257 jbd2_journal_destroy(journal);
5259 ext4_blkdev_put(bdev);
5263 static int ext4_load_journal(struct super_block *sb,
5264 struct ext4_super_block *es,
5265 unsigned long journal_devnum)
5268 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5271 int really_read_only;
5274 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5275 return -EFSCORRUPTED;
5277 if (journal_devnum &&
5278 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5279 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5280 "numbers have changed");
5281 journal_dev = new_decode_dev(journal_devnum);
5283 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5285 if (journal_inum && journal_dev) {
5286 ext4_msg(sb, KERN_ERR,
5287 "filesystem has both journal inode and journal device!");
5292 journal = ext4_get_journal(sb, journal_inum);
5296 journal = ext4_get_dev_journal(sb, journal_dev);
5301 journal_dev_ro = bdev_read_only(journal->j_dev);
5302 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5304 if (journal_dev_ro && !sb_rdonly(sb)) {
5305 ext4_msg(sb, KERN_ERR,
5306 "journal device read-only, try mounting with '-o ro'");
5312 * Are we loading a blank journal or performing recovery after a
5313 * crash? For recovery, we need to check in advance whether we
5314 * can get read-write access to the device.
5316 if (ext4_has_feature_journal_needs_recovery(sb)) {
5317 if (sb_rdonly(sb)) {
5318 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5319 "required on readonly filesystem");
5320 if (really_read_only) {
5321 ext4_msg(sb, KERN_ERR, "write access "
5322 "unavailable, cannot proceed "
5323 "(try mounting with noload)");
5327 ext4_msg(sb, KERN_INFO, "write access will "
5328 "be enabled during recovery");
5332 if (!(journal->j_flags & JBD2_BARRIER))
5333 ext4_msg(sb, KERN_INFO, "barriers disabled");
5335 if (!ext4_has_feature_journal_needs_recovery(sb))
5336 err = jbd2_journal_wipe(journal, !really_read_only);
5338 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5340 memcpy(save, ((char *) es) +
5341 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5342 err = jbd2_journal_load(journal);
5344 memcpy(((char *) es) + EXT4_S_ERR_START,
5345 save, EXT4_S_ERR_LEN);
5350 ext4_msg(sb, KERN_ERR, "error loading journal");
5354 EXT4_SB(sb)->s_journal = journal;
5355 err = ext4_clear_journal_err(sb, es);
5357 EXT4_SB(sb)->s_journal = NULL;
5358 jbd2_journal_destroy(journal);
5362 if (!really_read_only && journal_devnum &&
5363 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5364 es->s_journal_dev = cpu_to_le32(journal_devnum);
5366 /* Make sure we flush the recovery flag to disk. */
5367 ext4_commit_super(sb);
5373 jbd2_journal_destroy(journal);
5377 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5378 static void ext4_update_super(struct super_block *sb)
5380 struct ext4_sb_info *sbi = EXT4_SB(sb);
5381 struct ext4_super_block *es = sbi->s_es;
5382 struct buffer_head *sbh = sbi->s_sbh;
5386 * If the file system is mounted read-only, don't update the
5387 * superblock write time. This avoids updating the superblock
5388 * write time when we are mounting the root file system
5389 * read/only but we need to replay the journal; at that point,
5390 * for people who are east of GMT and who make their clock
5391 * tick in localtime for Windows bug-for-bug compatibility,
5392 * the clock is set in the future, and this will cause e2fsck
5393 * to complain and force a full file system check.
5395 if (!(sb->s_flags & SB_RDONLY))
5396 ext4_update_tstamp(es, s_wtime);
5397 es->s_kbytes_written =
5398 cpu_to_le64(sbi->s_kbytes_written +
5399 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5400 sbi->s_sectors_written_start) >> 1));
5401 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5402 ext4_free_blocks_count_set(es,
5403 EXT4_C2B(sbi, percpu_counter_sum_positive(
5404 &sbi->s_freeclusters_counter)));
5405 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5406 es->s_free_inodes_count =
5407 cpu_to_le32(percpu_counter_sum_positive(
5408 &sbi->s_freeinodes_counter));
5409 /* Copy error information to the on-disk superblock */
5410 spin_lock(&sbi->s_error_lock);
5411 if (sbi->s_add_error_count > 0) {
5412 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5413 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5414 __ext4_update_tstamp(&es->s_first_error_time,
5415 &es->s_first_error_time_hi,
5416 sbi->s_first_error_time);
5417 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5418 sizeof(es->s_first_error_func));
5419 es->s_first_error_line =
5420 cpu_to_le32(sbi->s_first_error_line);
5421 es->s_first_error_ino =
5422 cpu_to_le32(sbi->s_first_error_ino);
5423 es->s_first_error_block =
5424 cpu_to_le64(sbi->s_first_error_block);
5425 es->s_first_error_errcode =
5426 ext4_errno_to_code(sbi->s_first_error_code);
5428 __ext4_update_tstamp(&es->s_last_error_time,
5429 &es->s_last_error_time_hi,
5430 sbi->s_last_error_time);
5431 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5432 sizeof(es->s_last_error_func));
5433 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5434 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5435 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5436 es->s_last_error_errcode =
5437 ext4_errno_to_code(sbi->s_last_error_code);
5439 * Start the daily error reporting function if it hasn't been
5442 if (!es->s_error_count)
5443 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5444 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5445 sbi->s_add_error_count = 0;
5447 spin_unlock(&sbi->s_error_lock);
5449 ext4_superblock_csum_set(sb);
5453 static int ext4_commit_super(struct super_block *sb)
5455 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5460 if (block_device_ejected(sb))
5463 ext4_update_super(sb);
5465 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5467 * Oh, dear. A previous attempt to write the
5468 * superblock failed. This could happen because the
5469 * USB device was yanked out. Or it could happen to
5470 * be a transient write error and maybe the block will
5471 * be remapped. Nothing we can do but to retry the
5472 * write and hope for the best.
5474 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5475 "superblock detected");
5476 clear_buffer_write_io_error(sbh);
5477 set_buffer_uptodate(sbh);
5479 BUFFER_TRACE(sbh, "marking dirty");
5480 mark_buffer_dirty(sbh);
5481 error = __sync_dirty_buffer(sbh,
5482 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5483 if (buffer_write_io_error(sbh)) {
5484 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5486 clear_buffer_write_io_error(sbh);
5487 set_buffer_uptodate(sbh);
5493 * Have we just finished recovery? If so, and if we are mounting (or
5494 * remounting) the filesystem readonly, then we will end up with a
5495 * consistent fs on disk. Record that fact.
5497 static int ext4_mark_recovery_complete(struct super_block *sb,
5498 struct ext4_super_block *es)
5501 journal_t *journal = EXT4_SB(sb)->s_journal;
5503 if (!ext4_has_feature_journal(sb)) {
5504 if (journal != NULL) {
5505 ext4_error(sb, "Journal got removed while the fs was "
5507 return -EFSCORRUPTED;
5511 jbd2_journal_lock_updates(journal);
5512 err = jbd2_journal_flush(journal, 0);
5516 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
5517 ext4_has_feature_orphan_present(sb))) {
5518 if (!ext4_orphan_file_empty(sb)) {
5519 ext4_error(sb, "Orphan file not empty on read-only fs.");
5520 err = -EFSCORRUPTED;
5523 ext4_clear_feature_journal_needs_recovery(sb);
5524 ext4_clear_feature_orphan_present(sb);
5525 ext4_commit_super(sb);
5528 jbd2_journal_unlock_updates(journal);
5533 * If we are mounting (or read-write remounting) a filesystem whose journal
5534 * has recorded an error from a previous lifetime, move that error to the
5535 * main filesystem now.
5537 static int ext4_clear_journal_err(struct super_block *sb,
5538 struct ext4_super_block *es)
5544 if (!ext4_has_feature_journal(sb)) {
5545 ext4_error(sb, "Journal got removed while the fs was mounted!");
5546 return -EFSCORRUPTED;
5549 journal = EXT4_SB(sb)->s_journal;
5552 * Now check for any error status which may have been recorded in the
5553 * journal by a prior ext4_error() or ext4_abort()
5556 j_errno = jbd2_journal_errno(journal);
5560 errstr = ext4_decode_error(sb, j_errno, nbuf);
5561 ext4_warning(sb, "Filesystem error recorded "
5562 "from previous mount: %s", errstr);
5563 ext4_warning(sb, "Marking fs in need of filesystem check.");
5565 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5566 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5567 ext4_commit_super(sb);
5569 jbd2_journal_clear_err(journal);
5570 jbd2_journal_update_sb_errno(journal);
5576 * Force the running and committing transactions to commit,
5577 * and wait on the commit.
5579 int ext4_force_commit(struct super_block *sb)
5586 journal = EXT4_SB(sb)->s_journal;
5587 return ext4_journal_force_commit(journal);
5590 static int ext4_sync_fs(struct super_block *sb, int wait)
5594 bool needs_barrier = false;
5595 struct ext4_sb_info *sbi = EXT4_SB(sb);
5597 if (unlikely(ext4_forced_shutdown(sbi)))
5600 trace_ext4_sync_fs(sb, wait);
5601 flush_workqueue(sbi->rsv_conversion_wq);
5603 * Writeback quota in non-journalled quota case - journalled quota has
5606 dquot_writeback_dquots(sb, -1);
5608 * Data writeback is possible w/o journal transaction, so barrier must
5609 * being sent at the end of the function. But we can skip it if
5610 * transaction_commit will do it for us.
5612 if (sbi->s_journal) {
5613 target = jbd2_get_latest_transaction(sbi->s_journal);
5614 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5615 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5616 needs_barrier = true;
5618 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5620 ret = jbd2_log_wait_commit(sbi->s_journal,
5623 } else if (wait && test_opt(sb, BARRIER))
5624 needs_barrier = true;
5625 if (needs_barrier) {
5627 err = blkdev_issue_flush(sb->s_bdev);
5636 * LVM calls this function before a (read-only) snapshot is created. This
5637 * gives us a chance to flush the journal completely and mark the fs clean.
5639 * Note that only this function cannot bring a filesystem to be in a clean
5640 * state independently. It relies on upper layer to stop all data & metadata
5643 static int ext4_freeze(struct super_block *sb)
5651 journal = EXT4_SB(sb)->s_journal;
5654 /* Now we set up the journal barrier. */
5655 jbd2_journal_lock_updates(journal);
5658 * Don't clear the needs_recovery flag if we failed to
5659 * flush the journal.
5661 error = jbd2_journal_flush(journal, 0);
5665 /* Journal blocked and flushed, clear needs_recovery flag. */
5666 ext4_clear_feature_journal_needs_recovery(sb);
5667 if (ext4_orphan_file_empty(sb))
5668 ext4_clear_feature_orphan_present(sb);
5671 error = ext4_commit_super(sb);
5674 /* we rely on upper layer to stop further updates */
5675 jbd2_journal_unlock_updates(journal);
5680 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5681 * flag here, even though the filesystem is not technically dirty yet.
5683 static int ext4_unfreeze(struct super_block *sb)
5685 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5688 if (EXT4_SB(sb)->s_journal) {
5689 /* Reset the needs_recovery flag before the fs is unlocked. */
5690 ext4_set_feature_journal_needs_recovery(sb);
5691 if (ext4_has_feature_orphan_file(sb))
5692 ext4_set_feature_orphan_present(sb);
5695 ext4_commit_super(sb);
5700 * Structure to save mount options for ext4_remount's benefit
5702 struct ext4_mount_options {
5703 unsigned long s_mount_opt;
5704 unsigned long s_mount_opt2;
5707 unsigned long s_commit_interval;
5708 u32 s_min_batch_time, s_max_batch_time;
5711 char *s_qf_names[EXT4_MAXQUOTAS];
5715 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5717 struct ext4_super_block *es;
5718 struct ext4_sb_info *sbi = EXT4_SB(sb);
5719 unsigned long old_sb_flags, vfs_flags;
5720 struct ext4_mount_options old_opts;
5721 int enable_quota = 0;
5726 char *to_free[EXT4_MAXQUOTAS];
5728 char *orig_data = kstrdup(data, GFP_KERNEL);
5729 struct ext4_parsed_options parsed_opts;
5731 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5732 parsed_opts.journal_devnum = 0;
5734 if (data && !orig_data)
5737 /* Store the original options */
5738 old_sb_flags = sb->s_flags;
5739 old_opts.s_mount_opt = sbi->s_mount_opt;
5740 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5741 old_opts.s_resuid = sbi->s_resuid;
5742 old_opts.s_resgid = sbi->s_resgid;
5743 old_opts.s_commit_interval = sbi->s_commit_interval;
5744 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5745 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5747 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5748 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5749 if (sbi->s_qf_names[i]) {
5750 char *qf_name = get_qf_name(sb, sbi, i);
5752 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5753 if (!old_opts.s_qf_names[i]) {
5754 for (j = 0; j < i; j++)
5755 kfree(old_opts.s_qf_names[j]);
5760 old_opts.s_qf_names[i] = NULL;
5762 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5763 parsed_opts.journal_ioprio =
5764 sbi->s_journal->j_task->io_context->ioprio;
5767 * Some options can be enabled by ext4 and/or by VFS mount flag
5768 * either way we need to make sure it matches in both *flags and
5769 * s_flags. Copy those selected flags from *flags to s_flags
5771 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5772 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5774 if (!parse_options(data, sb, &parsed_opts, 1)) {
5779 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5780 test_opt(sb, JOURNAL_CHECKSUM)) {
5781 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5782 "during remount not supported; ignoring");
5783 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5786 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5787 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5788 ext4_msg(sb, KERN_ERR, "can't mount with "
5789 "both data=journal and delalloc");
5793 if (test_opt(sb, DIOREAD_NOLOCK)) {
5794 ext4_msg(sb, KERN_ERR, "can't mount with "
5795 "both data=journal and dioread_nolock");
5799 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5800 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5801 ext4_msg(sb, KERN_ERR, "can't mount with "
5802 "journal_async_commit in data=ordered mode");
5808 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5809 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5814 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5815 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5817 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5818 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5822 if (sbi->s_journal) {
5823 ext4_init_journal_params(sb, sbi->s_journal);
5824 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5827 /* Flush outstanding errors before changing fs state */
5828 flush_work(&sbi->s_error_work);
5830 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5831 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5836 if (*flags & SB_RDONLY) {
5837 err = sync_filesystem(sb);
5840 err = dquot_suspend(sb, -1);
5845 * First of all, the unconditional stuff we have to do
5846 * to disable replay of the journal when we next remount
5848 sb->s_flags |= SB_RDONLY;
5851 * OK, test if we are remounting a valid rw partition
5852 * readonly, and if so set the rdonly flag and then
5853 * mark the partition as valid again.
5855 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5856 (sbi->s_mount_state & EXT4_VALID_FS))
5857 es->s_state = cpu_to_le16(sbi->s_mount_state);
5859 if (sbi->s_journal) {
5861 * We let remount-ro finish even if marking fs
5862 * as clean failed...
5864 ext4_mark_recovery_complete(sb, es);
5867 /* Make sure we can mount this feature set readwrite */
5868 if (ext4_has_feature_readonly(sb) ||
5869 !ext4_feature_set_ok(sb, 0)) {
5874 * Make sure the group descriptor checksums
5875 * are sane. If they aren't, refuse to remount r/w.
5877 for (g = 0; g < sbi->s_groups_count; g++) {
5878 struct ext4_group_desc *gdp =
5879 ext4_get_group_desc(sb, g, NULL);
5881 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5882 ext4_msg(sb, KERN_ERR,
5883 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5884 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5885 le16_to_cpu(gdp->bg_checksum));
5892 * If we have an unprocessed orphan list hanging
5893 * around from a previously readonly bdev mount,
5894 * require a full umount/remount for now.
5896 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
5897 ext4_msg(sb, KERN_WARNING, "Couldn't "
5898 "remount RDWR because of unprocessed "
5899 "orphan inode list. Please "
5900 "umount/remount instead");
5906 * Mounting a RDONLY partition read-write, so reread
5907 * and store the current valid flag. (It may have
5908 * been changed by e2fsck since we originally mounted
5911 if (sbi->s_journal) {
5912 err = ext4_clear_journal_err(sb, es);
5916 sbi->s_mount_state = le16_to_cpu(es->s_state);
5918 err = ext4_setup_super(sb, es, 0);
5922 sb->s_flags &= ~SB_RDONLY;
5923 if (ext4_has_feature_mmp(sb))
5924 if (ext4_multi_mount_protect(sb,
5925 le64_to_cpu(es->s_mmp_block))) {
5934 * Reinitialize lazy itable initialization thread based on
5937 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5938 ext4_unregister_li_request(sb);
5940 ext4_group_t first_not_zeroed;
5941 first_not_zeroed = ext4_has_uninit_itable(sb);
5942 ext4_register_li_request(sb, first_not_zeroed);
5946 * Handle creation of system zone data early because it can fail.
5947 * Releasing of existing data is done when we are sure remount will
5950 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5951 err = ext4_setup_system_zone(sb);
5956 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5957 err = ext4_commit_super(sb);
5963 /* Release old quota file names */
5964 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5965 kfree(old_opts.s_qf_names[i]);
5967 if (sb_any_quota_suspended(sb))
5968 dquot_resume(sb, -1);
5969 else if (ext4_has_feature_quota(sb)) {
5970 err = ext4_enable_quotas(sb);
5976 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5977 ext4_release_system_zone(sb);
5979 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
5980 ext4_stop_mmpd(sbi);
5983 * Some options can be enabled by ext4 and/or by VFS mount flag
5984 * either way we need to make sure it matches in both *flags and
5985 * s_flags. Copy those selected flags from s_flags to *flags
5987 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5989 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
5990 orig_data, ext4_quota_mode(sb));
5995 sb->s_flags = old_sb_flags;
5996 sbi->s_mount_opt = old_opts.s_mount_opt;
5997 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5998 sbi->s_resuid = old_opts.s_resuid;
5999 sbi->s_resgid = old_opts.s_resgid;
6000 sbi->s_commit_interval = old_opts.s_commit_interval;
6001 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6002 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6003 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6004 ext4_release_system_zone(sb);
6006 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6007 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6008 to_free[i] = get_qf_name(sb, sbi, i);
6009 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6012 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6015 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6016 ext4_stop_mmpd(sbi);
6022 static int ext4_statfs_project(struct super_block *sb,
6023 kprojid_t projid, struct kstatfs *buf)
6026 struct dquot *dquot;
6030 qid = make_kqid_projid(projid);
6031 dquot = dqget(sb, qid);
6033 return PTR_ERR(dquot);
6034 spin_lock(&dquot->dq_dqb_lock);
6036 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6037 dquot->dq_dqb.dqb_bhardlimit);
6038 limit >>= sb->s_blocksize_bits;
6040 if (limit && buf->f_blocks > limit) {
6041 curblock = (dquot->dq_dqb.dqb_curspace +
6042 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6043 buf->f_blocks = limit;
6044 buf->f_bfree = buf->f_bavail =
6045 (buf->f_blocks > curblock) ?
6046 (buf->f_blocks - curblock) : 0;
6049 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6050 dquot->dq_dqb.dqb_ihardlimit);
6051 if (limit && buf->f_files > limit) {
6052 buf->f_files = limit;
6054 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6055 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6058 spin_unlock(&dquot->dq_dqb_lock);
6064 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6066 struct super_block *sb = dentry->d_sb;
6067 struct ext4_sb_info *sbi = EXT4_SB(sb);
6068 struct ext4_super_block *es = sbi->s_es;
6069 ext4_fsblk_t overhead = 0, resv_blocks;
6071 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6073 if (!test_opt(sb, MINIX_DF))
6074 overhead = sbi->s_overhead;
6076 buf->f_type = EXT4_SUPER_MAGIC;
6077 buf->f_bsize = sb->s_blocksize;
6078 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6079 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6080 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6081 /* prevent underflow in case that few free space is available */
6082 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6083 buf->f_bavail = buf->f_bfree -
6084 (ext4_r_blocks_count(es) + resv_blocks);
6085 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6087 buf->f_files = le32_to_cpu(es->s_inodes_count);
6088 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6089 buf->f_namelen = EXT4_NAME_LEN;
6090 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6093 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6094 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6095 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6104 * Helper functions so that transaction is started before we acquire dqio_sem
6105 * to keep correct lock ordering of transaction > dqio_sem
6107 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6109 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6112 static int ext4_write_dquot(struct dquot *dquot)
6116 struct inode *inode;
6118 inode = dquot_to_inode(dquot);
6119 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6120 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6122 return PTR_ERR(handle);
6123 ret = dquot_commit(dquot);
6124 err = ext4_journal_stop(handle);
6130 static int ext4_acquire_dquot(struct dquot *dquot)
6135 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6136 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6138 return PTR_ERR(handle);
6139 ret = dquot_acquire(dquot);
6140 err = ext4_journal_stop(handle);
6146 static int ext4_release_dquot(struct dquot *dquot)
6151 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6152 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6153 if (IS_ERR(handle)) {
6154 /* Release dquot anyway to avoid endless cycle in dqput() */
6155 dquot_release(dquot);
6156 return PTR_ERR(handle);
6158 ret = dquot_release(dquot);
6159 err = ext4_journal_stop(handle);
6165 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6167 struct super_block *sb = dquot->dq_sb;
6169 if (ext4_is_quota_journalled(sb)) {
6170 dquot_mark_dquot_dirty(dquot);
6171 return ext4_write_dquot(dquot);
6173 return dquot_mark_dquot_dirty(dquot);
6177 static int ext4_write_info(struct super_block *sb, int type)
6182 /* Data block + inode block */
6183 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6185 return PTR_ERR(handle);
6186 ret = dquot_commit_info(sb, type);
6187 err = ext4_journal_stop(handle);
6193 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6195 struct ext4_inode_info *ei = EXT4_I(inode);
6197 /* The first argument of lockdep_set_subclass has to be
6198 * *exactly* the same as the argument to init_rwsem() --- in
6199 * this case, in init_once() --- or lockdep gets unhappy
6200 * because the name of the lock is set using the
6201 * stringification of the argument to init_rwsem().
6203 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6204 lockdep_set_subclass(&ei->i_data_sem, subclass);
6208 * Standard function to be called on quota_on
6210 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6211 const struct path *path)
6215 if (!test_opt(sb, QUOTA))
6218 /* Quotafile not on the same filesystem? */
6219 if (path->dentry->d_sb != sb)
6222 /* Quota already enabled for this file? */
6223 if (IS_NOQUOTA(d_inode(path->dentry)))
6226 /* Journaling quota? */
6227 if (EXT4_SB(sb)->s_qf_names[type]) {
6228 /* Quotafile not in fs root? */
6229 if (path->dentry->d_parent != sb->s_root)
6230 ext4_msg(sb, KERN_WARNING,
6231 "Quota file not on filesystem root. "
6232 "Journaled quota will not work");
6233 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6236 * Clear the flag just in case mount options changed since
6239 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6243 * When we journal data on quota file, we have to flush journal to see
6244 * all updates to the file when we bypass pagecache...
6246 if (EXT4_SB(sb)->s_journal &&
6247 ext4_should_journal_data(d_inode(path->dentry))) {
6249 * We don't need to lock updates but journal_flush() could
6250 * otherwise be livelocked...
6252 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6253 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6254 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6259 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6260 err = dquot_quota_on(sb, type, format_id, path);
6262 lockdep_set_quota_inode(path->dentry->d_inode,
6265 struct inode *inode = d_inode(path->dentry);
6269 * Set inode flags to prevent userspace from messing with quota
6270 * files. If this fails, we return success anyway since quotas
6271 * are already enabled and this is not a hard failure.
6274 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6277 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6278 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6279 S_NOATIME | S_IMMUTABLE);
6280 err = ext4_mark_inode_dirty(handle, inode);
6281 ext4_journal_stop(handle);
6283 inode_unlock(inode);
6288 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6292 struct inode *qf_inode;
6293 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6294 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6295 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6296 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6299 BUG_ON(!ext4_has_feature_quota(sb));
6301 if (!qf_inums[type])
6304 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6305 if (IS_ERR(qf_inode)) {
6306 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6307 return PTR_ERR(qf_inode);
6310 /* Don't account quota for quota files to avoid recursion */
6311 qf_inode->i_flags |= S_NOQUOTA;
6312 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6313 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6315 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6321 /* Enable usage tracking for all quota types. */
6322 int ext4_enable_quotas(struct super_block *sb)
6325 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6326 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6327 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6328 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6330 bool quota_mopt[EXT4_MAXQUOTAS] = {
6331 test_opt(sb, USRQUOTA),
6332 test_opt(sb, GRPQUOTA),
6333 test_opt(sb, PRJQUOTA),
6336 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6337 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6338 if (qf_inums[type]) {
6339 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6340 DQUOT_USAGE_ENABLED |
6341 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6344 "Failed to enable quota tracking "
6345 "(type=%d, err=%d). Please run "
6346 "e2fsck to fix.", type, err);
6347 for (type--; type >= 0; type--)
6348 dquot_quota_off(sb, type);
6357 static int ext4_quota_off(struct super_block *sb, int type)
6359 struct inode *inode = sb_dqopt(sb)->files[type];
6363 /* Force all delayed allocation blocks to be allocated.
6364 * Caller already holds s_umount sem */
6365 if (test_opt(sb, DELALLOC))
6366 sync_filesystem(sb);
6368 if (!inode || !igrab(inode))
6371 err = dquot_quota_off(sb, type);
6372 if (err || ext4_has_feature_quota(sb))
6377 * Update modification times of quota files when userspace can
6378 * start looking at them. If we fail, we return success anyway since
6379 * this is not a hard failure and quotas are already disabled.
6381 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6382 if (IS_ERR(handle)) {
6383 err = PTR_ERR(handle);
6386 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6387 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6388 inode->i_mtime = inode->i_ctime = current_time(inode);
6389 err = ext4_mark_inode_dirty(handle, inode);
6390 ext4_journal_stop(handle);
6392 inode_unlock(inode);
6394 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6398 return dquot_quota_off(sb, type);
6401 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6402 * acquiring the locks... As quota files are never truncated and quota code
6403 * itself serializes the operations (and no one else should touch the files)
6404 * we don't have to be afraid of races */
6405 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6406 size_t len, loff_t off)
6408 struct inode *inode = sb_dqopt(sb)->files[type];
6409 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6410 int offset = off & (sb->s_blocksize - 1);
6413 struct buffer_head *bh;
6414 loff_t i_size = i_size_read(inode);
6418 if (off+len > i_size)
6421 while (toread > 0) {
6422 tocopy = sb->s_blocksize - offset < toread ?
6423 sb->s_blocksize - offset : toread;
6424 bh = ext4_bread(NULL, inode, blk, 0);
6427 if (!bh) /* A hole? */
6428 memset(data, 0, tocopy);
6430 memcpy(data, bh->b_data+offset, tocopy);
6440 /* Write to quotafile (we know the transaction is already started and has
6441 * enough credits) */
6442 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6443 const char *data, size_t len, loff_t off)
6445 struct inode *inode = sb_dqopt(sb)->files[type];
6446 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6447 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6449 struct buffer_head *bh;
6450 handle_t *handle = journal_current_handle();
6452 if (EXT4_SB(sb)->s_journal && !handle) {
6453 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6454 " cancelled because transaction is not started",
6455 (unsigned long long)off, (unsigned long long)len);
6459 * Since we account only one data block in transaction credits,
6460 * then it is impossible to cross a block boundary.
6462 if (sb->s_blocksize - offset < len) {
6463 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6464 " cancelled because not block aligned",
6465 (unsigned long long)off, (unsigned long long)len);
6470 bh = ext4_bread(handle, inode, blk,
6471 EXT4_GET_BLOCKS_CREATE |
6472 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6473 } while (PTR_ERR(bh) == -ENOSPC &&
6474 ext4_should_retry_alloc(inode->i_sb, &retries));
6479 BUFFER_TRACE(bh, "get write access");
6480 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6486 memcpy(bh->b_data+offset, data, len);
6487 flush_dcache_page(bh->b_page);
6489 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6492 if (inode->i_size < off + len) {
6493 i_size_write(inode, off + len);
6494 EXT4_I(inode)->i_disksize = inode->i_size;
6495 err2 = ext4_mark_inode_dirty(handle, inode);
6496 if (unlikely(err2 && !err))
6499 return err ? err : len;
6503 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6504 const char *dev_name, void *data)
6506 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6509 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6510 static inline void register_as_ext2(void)
6512 int err = register_filesystem(&ext2_fs_type);
6515 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6518 static inline void unregister_as_ext2(void)
6520 unregister_filesystem(&ext2_fs_type);
6523 static inline int ext2_feature_set_ok(struct super_block *sb)
6525 if (ext4_has_unknown_ext2_incompat_features(sb))
6529 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6534 static inline void register_as_ext2(void) { }
6535 static inline void unregister_as_ext2(void) { }
6536 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6539 static inline void register_as_ext3(void)
6541 int err = register_filesystem(&ext3_fs_type);
6544 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6547 static inline void unregister_as_ext3(void)
6549 unregister_filesystem(&ext3_fs_type);
6552 static inline int ext3_feature_set_ok(struct super_block *sb)
6554 if (ext4_has_unknown_ext3_incompat_features(sb))
6556 if (!ext4_has_feature_journal(sb))
6560 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6565 static struct file_system_type ext4_fs_type = {
6566 .owner = THIS_MODULE,
6568 .mount = ext4_mount,
6569 .kill_sb = kill_block_super,
6570 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6572 MODULE_ALIAS_FS("ext4");
6574 /* Shared across all ext4 file systems */
6575 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6577 static int __init ext4_init_fs(void)
6581 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6582 ext4_li_info = NULL;
6584 /* Build-time check for flags consistency */
6585 ext4_check_flag_values();
6587 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6588 init_waitqueue_head(&ext4__ioend_wq[i]);
6590 err = ext4_init_es();
6594 err = ext4_init_pending();
6598 err = ext4_init_post_read_processing();
6602 err = ext4_init_pageio();
6606 err = ext4_init_system_zone();
6610 err = ext4_init_sysfs();
6614 err = ext4_init_mballoc();
6617 err = init_inodecache();
6621 err = ext4_fc_init_dentry_cache();
6627 err = register_filesystem(&ext4_fs_type);
6633 unregister_as_ext2();
6634 unregister_as_ext3();
6636 destroy_inodecache();
6638 ext4_exit_mballoc();
6642 ext4_exit_system_zone();
6646 ext4_exit_post_read_processing();
6648 ext4_exit_pending();
6655 static void __exit ext4_exit_fs(void)
6657 ext4_destroy_lazyinit_thread();
6658 unregister_as_ext2();
6659 unregister_as_ext3();
6660 unregister_filesystem(&ext4_fs_type);
6661 destroy_inodecache();
6662 ext4_exit_mballoc();
6664 ext4_exit_system_zone();
6666 ext4_exit_post_read_processing();
6668 ext4_exit_pending();
6671 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6672 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6673 MODULE_LICENSE("GPL");
6674 MODULE_SOFTDEP("pre: crc32c");
6675 module_init(ext4_init_fs)
6676 module_exit(ext4_exit_fs)