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>
49 #include <linux/fsnotify.h>
50 #include <linux/fs_context.h>
51 #include <linux/fs_parser.h>
54 #include "ext4_extents.h" /* Needed for trace points definition */
55 #include "ext4_jbd2.h"
61 #define CREATE_TRACE_POINTS
62 #include <trace/events/ext4.h>
64 static struct ext4_lazy_init *ext4_li_info;
65 static DEFINE_MUTEX(ext4_li_mtx);
66 static struct ratelimit_state ext4_mount_msg_ratelimit;
68 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
69 unsigned long journal_devnum);
70 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
71 static void ext4_update_super(struct super_block *sb);
72 static int ext4_commit_super(struct super_block *sb);
73 static int ext4_mark_recovery_complete(struct super_block *sb,
74 struct ext4_super_block *es);
75 static int ext4_clear_journal_err(struct super_block *sb,
76 struct ext4_super_block *es);
77 static int ext4_sync_fs(struct super_block *sb, int wait);
78 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
79 static int ext4_unfreeze(struct super_block *sb);
80 static int ext4_freeze(struct super_block *sb);
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);
88 static int ext4_validate_options(struct fs_context *fc);
89 static int ext4_check_opt_consistency(struct fs_context *fc,
90 struct super_block *sb);
91 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb);
92 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
93 static int ext4_get_tree(struct fs_context *fc);
94 static int ext4_reconfigure(struct fs_context *fc);
95 static void ext4_fc_free(struct fs_context *fc);
96 static int ext4_init_fs_context(struct fs_context *fc);
97 static const struct fs_parameter_spec ext4_param_specs[];
103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
104 * -> page lock -> i_data_sem (rw)
106 * buffered write path:
107 * sb_start_write -> i_mutex -> mmap_lock
108 * sb_start_write -> i_mutex -> transaction start -> page lock ->
112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
118 * sb_start_write -> i_mutex -> mmap_lock
119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
122 * transaction start -> page lock(s) -> i_data_sem (rw)
125 static const struct fs_context_operations ext4_context_ops = {
126 .parse_param = ext4_parse_param,
127 .get_tree = ext4_get_tree,
128 .reconfigure = ext4_reconfigure,
129 .free = ext4_fc_free,
133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
134 static struct file_system_type ext2_fs_type = {
135 .owner = THIS_MODULE,
137 .init_fs_context = ext4_init_fs_context,
138 .parameters = ext4_param_specs,
139 .kill_sb = kill_block_super,
140 .fs_flags = FS_REQUIRES_DEV,
142 MODULE_ALIAS_FS("ext2");
143 MODULE_ALIAS("ext2");
144 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
146 #define IS_EXT2_SB(sb) (0)
150 static struct file_system_type ext3_fs_type = {
151 .owner = THIS_MODULE,
153 .init_fs_context = ext4_init_fs_context,
154 .parameters = ext4_param_specs,
155 .kill_sb = kill_block_super,
156 .fs_flags = FS_REQUIRES_DEV,
158 MODULE_ALIAS_FS("ext3");
159 MODULE_ALIAS("ext3");
160 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
163 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
167 * buffer's verified bit is no longer valid after reading from
168 * disk again due to write out error, clear it to make sure we
169 * recheck the buffer contents.
171 clear_buffer_verified(bh);
173 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
175 submit_bh(REQ_OP_READ, op_flags, bh);
178 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
181 BUG_ON(!buffer_locked(bh));
183 if (ext4_buffer_uptodate(bh)) {
187 __ext4_read_bh(bh, op_flags, end_io);
190 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
192 BUG_ON(!buffer_locked(bh));
194 if (ext4_buffer_uptodate(bh)) {
199 __ext4_read_bh(bh, op_flags, end_io);
202 if (buffer_uptodate(bh))
207 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
209 if (trylock_buffer(bh)) {
211 return ext4_read_bh(bh, op_flags, NULL);
212 ext4_read_bh_nowait(bh, op_flags, NULL);
217 if (buffer_uptodate(bh))
225 * This works like __bread_gfp() except it uses ERR_PTR for error
226 * returns. Currently with sb_bread it's impossible to distinguish
227 * between ENOMEM and EIO situations (since both result in a NULL
230 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
231 sector_t block, int op_flags,
234 struct buffer_head *bh;
237 bh = sb_getblk_gfp(sb, block, gfp);
239 return ERR_PTR(-ENOMEM);
240 if (ext4_buffer_uptodate(bh))
243 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
251 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
254 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
257 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
260 return __ext4_sb_bread_gfp(sb, block, 0, 0);
263 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
265 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
268 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
273 static int ext4_verify_csum_type(struct super_block *sb,
274 struct ext4_super_block *es)
276 if (!ext4_has_feature_metadata_csum(sb))
279 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
282 static __le32 ext4_superblock_csum(struct super_block *sb,
283 struct ext4_super_block *es)
285 struct ext4_sb_info *sbi = EXT4_SB(sb);
286 int offset = offsetof(struct ext4_super_block, s_checksum);
289 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
291 return cpu_to_le32(csum);
294 static int ext4_superblock_csum_verify(struct super_block *sb,
295 struct ext4_super_block *es)
297 if (!ext4_has_metadata_csum(sb))
300 return es->s_checksum == ext4_superblock_csum(sb, es);
303 void ext4_superblock_csum_set(struct super_block *sb)
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307 if (!ext4_has_metadata_csum(sb))
310 es->s_checksum = ext4_superblock_csum(sb, es);
313 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
314 struct ext4_group_desc *bg)
316 return le32_to_cpu(bg->bg_block_bitmap_lo) |
317 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
318 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
321 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
322 struct ext4_group_desc *bg)
324 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
325 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
326 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
329 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
330 struct ext4_group_desc *bg)
332 return le32_to_cpu(bg->bg_inode_table_lo) |
333 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
334 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
337 __u32 ext4_free_group_clusters(struct super_block *sb,
338 struct ext4_group_desc *bg)
340 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
341 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
342 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
345 __u32 ext4_free_inodes_count(struct super_block *sb,
346 struct ext4_group_desc *bg)
348 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
349 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
350 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
353 __u32 ext4_used_dirs_count(struct super_block *sb,
354 struct ext4_group_desc *bg)
356 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
357 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
358 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
361 __u32 ext4_itable_unused_count(struct super_block *sb,
362 struct ext4_group_desc *bg)
364 return le16_to_cpu(bg->bg_itable_unused_lo) |
365 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
366 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
369 void ext4_block_bitmap_set(struct super_block *sb,
370 struct ext4_group_desc *bg, ext4_fsblk_t blk)
372 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
373 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
374 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
377 void ext4_inode_bitmap_set(struct super_block *sb,
378 struct ext4_group_desc *bg, ext4_fsblk_t blk)
380 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
381 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
382 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
385 void ext4_inode_table_set(struct super_block *sb,
386 struct ext4_group_desc *bg, ext4_fsblk_t blk)
388 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
389 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
390 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
393 void ext4_free_group_clusters_set(struct super_block *sb,
394 struct ext4_group_desc *bg, __u32 count)
396 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
397 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
398 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
401 void ext4_free_inodes_set(struct super_block *sb,
402 struct ext4_group_desc *bg, __u32 count)
404 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
405 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
406 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
409 void ext4_used_dirs_set(struct super_block *sb,
410 struct ext4_group_desc *bg, __u32 count)
412 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
413 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
414 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
417 void ext4_itable_unused_set(struct super_block *sb,
418 struct ext4_group_desc *bg, __u32 count)
420 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
421 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
422 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
425 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
427 now = clamp_val(now, 0, (1ull << 40) - 1);
429 *lo = cpu_to_le32(lower_32_bits(now));
430 *hi = upper_32_bits(now);
433 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
435 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
437 #define ext4_update_tstamp(es, tstamp) \
438 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
439 ktime_get_real_seconds())
440 #define ext4_get_tstamp(es, tstamp) \
441 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
444 * The del_gendisk() function uninitializes the disk-specific data
445 * structures, including the bdi structure, without telling anyone
446 * else. Once this happens, any attempt to call mark_buffer_dirty()
447 * (for example, by ext4_commit_super), will cause a kernel OOPS.
448 * This is a kludge to prevent these oops until we can put in a proper
449 * hook in del_gendisk() to inform the VFS and file system layers.
451 static int block_device_ejected(struct super_block *sb)
453 struct inode *bd_inode = sb->s_bdev->bd_inode;
454 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
456 return bdi->dev == NULL;
459 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
461 struct super_block *sb = journal->j_private;
462 struct ext4_sb_info *sbi = EXT4_SB(sb);
463 int error = is_journal_aborted(journal);
464 struct ext4_journal_cb_entry *jce;
466 BUG_ON(txn->t_state == T_FINISHED);
468 ext4_process_freed_data(sb, txn->t_tid);
470 spin_lock(&sbi->s_md_lock);
471 while (!list_empty(&txn->t_private_list)) {
472 jce = list_entry(txn->t_private_list.next,
473 struct ext4_journal_cb_entry, jce_list);
474 list_del_init(&jce->jce_list);
475 spin_unlock(&sbi->s_md_lock);
476 jce->jce_func(sb, jce, error);
477 spin_lock(&sbi->s_md_lock);
479 spin_unlock(&sbi->s_md_lock);
483 * This writepage callback for write_cache_pages()
484 * takes care of a few cases after page cleaning.
486 * write_cache_pages() already checks for dirty pages
487 * and calls clear_page_dirty_for_io(), which we want,
488 * to write protect the pages.
490 * However, we may have to redirty a page (see below.)
492 static int ext4_journalled_writepage_callback(struct page *page,
493 struct writeback_control *wbc,
496 transaction_t *transaction = (transaction_t *) data;
497 struct buffer_head *bh, *head;
498 struct journal_head *jh;
500 bh = head = page_buffers(page);
503 * We have to redirty a page in these cases:
504 * 1) If buffer is dirty, it means the page was dirty because it
505 * contains a buffer that needs checkpointing. So the dirty bit
506 * needs to be preserved so that checkpointing writes the buffer
508 * 2) If buffer is not part of the committing transaction
509 * (we may have just accidentally come across this buffer because
510 * inode range tracking is not exact) or if the currently running
511 * transaction already contains this buffer as well, dirty bit
512 * needs to be preserved so that the buffer gets writeprotected
513 * properly on running transaction's commit.
516 if (buffer_dirty(bh) ||
517 (jh && (jh->b_transaction != transaction ||
518 jh->b_next_transaction))) {
519 redirty_page_for_writepage(wbc, page);
522 } while ((bh = bh->b_this_page) != head);
525 return AOP_WRITEPAGE_ACTIVATE;
528 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
530 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
531 struct writeback_control wbc = {
532 .sync_mode = WB_SYNC_ALL,
533 .nr_to_write = LONG_MAX,
534 .range_start = jinode->i_dirty_start,
535 .range_end = jinode->i_dirty_end,
538 return write_cache_pages(mapping, &wbc,
539 ext4_journalled_writepage_callback,
540 jinode->i_transaction);
543 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
547 if (ext4_should_journal_data(jinode->i_vfs_inode))
548 ret = ext4_journalled_submit_inode_data_buffers(jinode);
550 ret = jbd2_journal_submit_inode_data_buffers(jinode);
555 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
559 if (!ext4_should_journal_data(jinode->i_vfs_inode))
560 ret = jbd2_journal_finish_inode_data_buffers(jinode);
565 static bool system_going_down(void)
567 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
568 || system_state == SYSTEM_RESTART;
571 struct ext4_err_translation {
576 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
578 static struct ext4_err_translation err_translation[] = {
579 EXT4_ERR_TRANSLATE(EIO),
580 EXT4_ERR_TRANSLATE(ENOMEM),
581 EXT4_ERR_TRANSLATE(EFSBADCRC),
582 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
583 EXT4_ERR_TRANSLATE(ENOSPC),
584 EXT4_ERR_TRANSLATE(ENOKEY),
585 EXT4_ERR_TRANSLATE(EROFS),
586 EXT4_ERR_TRANSLATE(EFBIG),
587 EXT4_ERR_TRANSLATE(EEXIST),
588 EXT4_ERR_TRANSLATE(ERANGE),
589 EXT4_ERR_TRANSLATE(EOVERFLOW),
590 EXT4_ERR_TRANSLATE(EBUSY),
591 EXT4_ERR_TRANSLATE(ENOTDIR),
592 EXT4_ERR_TRANSLATE(ENOTEMPTY),
593 EXT4_ERR_TRANSLATE(ESHUTDOWN),
594 EXT4_ERR_TRANSLATE(EFAULT),
597 static int ext4_errno_to_code(int errno)
601 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
602 if (err_translation[i].errno == errno)
603 return err_translation[i].code;
604 return EXT4_ERR_UNKNOWN;
607 static void save_error_info(struct super_block *sb, int error,
608 __u32 ino, __u64 block,
609 const char *func, unsigned int line)
611 struct ext4_sb_info *sbi = EXT4_SB(sb);
613 /* We default to EFSCORRUPTED error... */
615 error = EFSCORRUPTED;
617 spin_lock(&sbi->s_error_lock);
618 sbi->s_add_error_count++;
619 sbi->s_last_error_code = error;
620 sbi->s_last_error_line = line;
621 sbi->s_last_error_ino = ino;
622 sbi->s_last_error_block = block;
623 sbi->s_last_error_func = func;
624 sbi->s_last_error_time = ktime_get_real_seconds();
625 if (!sbi->s_first_error_time) {
626 sbi->s_first_error_code = error;
627 sbi->s_first_error_line = line;
628 sbi->s_first_error_ino = ino;
629 sbi->s_first_error_block = block;
630 sbi->s_first_error_func = func;
631 sbi->s_first_error_time = sbi->s_last_error_time;
633 spin_unlock(&sbi->s_error_lock);
636 /* Deal with the reporting of failure conditions on a filesystem such as
637 * inconsistencies detected or read IO failures.
639 * On ext2, we can store the error state of the filesystem in the
640 * superblock. That is not possible on ext4, because we may have other
641 * write ordering constraints on the superblock which prevent us from
642 * writing it out straight away; and given that the journal is about to
643 * be aborted, we can't rely on the current, or future, transactions to
644 * write out the superblock safely.
646 * We'll just use the jbd2_journal_abort() error code to record an error in
647 * the journal instead. On recovery, the journal will complain about
648 * that error until we've noted it down and cleared it.
650 * If force_ro is set, we unconditionally force the filesystem into an
651 * ABORT|READONLY state, unless the error response on the fs has been set to
652 * panic in which case we take the easy way out and panic immediately. This is
653 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
654 * at a critical moment in log management.
656 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
657 __u32 ino, __u64 block,
658 const char *func, unsigned int line)
660 journal_t *journal = EXT4_SB(sb)->s_journal;
661 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
663 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
664 if (test_opt(sb, WARN_ON_ERROR))
667 if (!continue_fs && !sb_rdonly(sb)) {
668 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
670 jbd2_journal_abort(journal, -EIO);
673 if (!bdev_read_only(sb->s_bdev)) {
674 save_error_info(sb, error, ino, block, func, line);
676 * In case the fs should keep running, we need to writeout
677 * superblock through the journal. Due to lock ordering
678 * constraints, it may not be safe to do it right here so we
679 * defer superblock flushing to a workqueue.
681 if (continue_fs && journal)
682 schedule_work(&EXT4_SB(sb)->s_error_work);
684 ext4_commit_super(sb);
688 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
689 * could panic during 'reboot -f' as the underlying device got already
692 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
693 panic("EXT4-fs (device %s): panic forced after error\n",
697 if (sb_rdonly(sb) || continue_fs)
700 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
702 * Make sure updated value of ->s_mount_flags will be visible before
706 sb->s_flags |= SB_RDONLY;
709 static void flush_stashed_error_work(struct work_struct *work)
711 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
713 journal_t *journal = sbi->s_journal;
717 * If the journal is still running, we have to write out superblock
718 * through the journal to avoid collisions of other journalled sb
721 * We use directly jbd2 functions here to avoid recursing back into
722 * ext4 error handling code during handling of previous errors.
724 if (!sb_rdonly(sbi->s_sb) && journal) {
725 struct buffer_head *sbh = sbi->s_sbh;
726 handle = jbd2_journal_start(journal, 1);
729 if (jbd2_journal_get_write_access(handle, sbh)) {
730 jbd2_journal_stop(handle);
733 ext4_update_super(sbi->s_sb);
734 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
735 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
736 "superblock detected");
737 clear_buffer_write_io_error(sbh);
738 set_buffer_uptodate(sbh);
741 if (jbd2_journal_dirty_metadata(handle, sbh)) {
742 jbd2_journal_stop(handle);
745 jbd2_journal_stop(handle);
746 ext4_notify_error_sysfs(sbi);
751 * Write through journal failed. Write sb directly to get error info
752 * out and hope for the best.
754 ext4_commit_super(sbi->s_sb);
755 ext4_notify_error_sysfs(sbi);
758 #define ext4_error_ratelimit(sb) \
759 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
762 void __ext4_error(struct super_block *sb, const char *function,
763 unsigned int line, bool force_ro, int error, __u64 block,
764 const char *fmt, ...)
766 struct va_format vaf;
769 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
772 trace_ext4_error(sb, function, line);
773 if (ext4_error_ratelimit(sb)) {
778 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
779 sb->s_id, function, line, current->comm, &vaf);
782 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
784 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
787 void __ext4_error_inode(struct inode *inode, const char *function,
788 unsigned int line, ext4_fsblk_t block, int error,
789 const char *fmt, ...)
792 struct va_format vaf;
794 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
797 trace_ext4_error(inode->i_sb, function, line);
798 if (ext4_error_ratelimit(inode->i_sb)) {
803 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
804 "inode #%lu: block %llu: comm %s: %pV\n",
805 inode->i_sb->s_id, function, line, inode->i_ino,
806 block, current->comm, &vaf);
808 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
809 "inode #%lu: comm %s: %pV\n",
810 inode->i_sb->s_id, function, line, inode->i_ino,
811 current->comm, &vaf);
814 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
816 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
820 void __ext4_error_file(struct file *file, const char *function,
821 unsigned int line, ext4_fsblk_t block,
822 const char *fmt, ...)
825 struct va_format vaf;
826 struct inode *inode = file_inode(file);
827 char pathname[80], *path;
829 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
832 trace_ext4_error(inode->i_sb, function, line);
833 if (ext4_error_ratelimit(inode->i_sb)) {
834 path = file_path(file, pathname, sizeof(pathname));
842 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
843 "block %llu: comm %s: path %s: %pV\n",
844 inode->i_sb->s_id, function, line, inode->i_ino,
845 block, current->comm, path, &vaf);
848 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
849 "comm %s: path %s: %pV\n",
850 inode->i_sb->s_id, function, line, inode->i_ino,
851 current->comm, path, &vaf);
854 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
856 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
860 const char *ext4_decode_error(struct super_block *sb, int errno,
867 errstr = "Corrupt filesystem";
870 errstr = "Filesystem failed CRC";
873 errstr = "IO failure";
876 errstr = "Out of memory";
879 if (!sb || (EXT4_SB(sb)->s_journal &&
880 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
881 errstr = "Journal has aborted";
883 errstr = "Readonly filesystem";
886 /* If the caller passed in an extra buffer for unknown
887 * errors, textualise them now. Else we just return
890 /* Check for truncated error codes... */
891 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
900 /* __ext4_std_error decodes expected errors from journaling functions
901 * automatically and invokes the appropriate error response. */
903 void __ext4_std_error(struct super_block *sb, const char *function,
904 unsigned int line, int errno)
909 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
912 /* Special case: if the error is EROFS, and we're not already
913 * inside a transaction, then there's really no point in logging
915 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
918 if (ext4_error_ratelimit(sb)) {
919 errstr = ext4_decode_error(sb, errno, nbuf);
920 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
921 sb->s_id, function, line, errstr);
923 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
925 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
928 void __ext4_msg(struct super_block *sb,
929 const char *prefix, const char *fmt, ...)
931 struct va_format vaf;
935 atomic_inc(&EXT4_SB(sb)->s_msg_count);
936 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
945 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
947 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
951 static int ext4_warning_ratelimit(struct super_block *sb)
953 atomic_inc(&EXT4_SB(sb)->s_warning_count);
954 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
958 void __ext4_warning(struct super_block *sb, const char *function,
959 unsigned int line, const char *fmt, ...)
961 struct va_format vaf;
964 if (!ext4_warning_ratelimit(sb))
970 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
971 sb->s_id, function, line, &vaf);
975 void __ext4_warning_inode(const struct inode *inode, const char *function,
976 unsigned int line, const char *fmt, ...)
978 struct va_format vaf;
981 if (!ext4_warning_ratelimit(inode->i_sb))
987 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
988 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
989 function, line, inode->i_ino, current->comm, &vaf);
993 void __ext4_grp_locked_error(const char *function, unsigned int line,
994 struct super_block *sb, ext4_group_t grp,
995 unsigned long ino, ext4_fsblk_t block,
996 const char *fmt, ...)
1000 struct va_format vaf;
1003 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
1006 trace_ext4_error(sb, function, line);
1007 if (ext4_error_ratelimit(sb)) {
1008 va_start(args, fmt);
1011 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1012 sb->s_id, function, line, grp);
1014 printk(KERN_CONT "inode %lu: ", ino);
1016 printk(KERN_CONT "block %llu:",
1017 (unsigned long long) block);
1018 printk(KERN_CONT "%pV\n", &vaf);
1022 if (test_opt(sb, ERRORS_CONT)) {
1023 if (test_opt(sb, WARN_ON_ERROR))
1025 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1026 if (!bdev_read_only(sb->s_bdev)) {
1027 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1029 schedule_work(&EXT4_SB(sb)->s_error_work);
1033 ext4_unlock_group(sb, grp);
1034 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1036 * We only get here in the ERRORS_RO case; relocking the group
1037 * may be dangerous, but nothing bad will happen since the
1038 * filesystem will have already been marked read/only and the
1039 * journal has been aborted. We return 1 as a hint to callers
1040 * who might what to use the return value from
1041 * ext4_grp_locked_error() to distinguish between the
1042 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1043 * aggressively from the ext4 function in question, with a
1044 * more appropriate error code.
1046 ext4_lock_group(sb, grp);
1050 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1054 struct ext4_sb_info *sbi = EXT4_SB(sb);
1055 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1056 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1059 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1060 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1063 percpu_counter_sub(&sbi->s_freeclusters_counter,
1067 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1068 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1073 count = ext4_free_inodes_count(sb, gdp);
1074 percpu_counter_sub(&sbi->s_freeinodes_counter,
1080 void ext4_update_dynamic_rev(struct super_block *sb)
1082 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1084 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1088 "updating to rev %d because of new feature flag, "
1089 "running e2fsck is recommended",
1092 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1093 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1094 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1095 /* leave es->s_feature_*compat flags alone */
1096 /* es->s_uuid will be set by e2fsck if empty */
1099 * The rest of the superblock fields should be zero, and if not it
1100 * means they are likely already in use, so leave them alone. We
1101 * can leave it up to e2fsck to clean up any inconsistencies there.
1106 * Open the external journal device
1108 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1110 struct block_device *bdev;
1112 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1118 ext4_msg(sb, KERN_ERR,
1119 "failed to open journal device unknown-block(%u,%u) %ld",
1120 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1125 * Release the journal device
1127 static void ext4_blkdev_put(struct block_device *bdev)
1129 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1132 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1134 struct block_device *bdev;
1135 bdev = sbi->s_journal_bdev;
1137 ext4_blkdev_put(bdev);
1138 sbi->s_journal_bdev = NULL;
1142 static inline struct inode *orphan_list_entry(struct list_head *l)
1144 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1147 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1149 struct list_head *l;
1151 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1152 le32_to_cpu(sbi->s_es->s_last_orphan));
1154 printk(KERN_ERR "sb_info orphan list:\n");
1155 list_for_each(l, &sbi->s_orphan) {
1156 struct inode *inode = orphan_list_entry(l);
1158 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1159 inode->i_sb->s_id, inode->i_ino, inode,
1160 inode->i_mode, inode->i_nlink,
1161 NEXT_ORPHAN(inode));
1166 static int ext4_quota_off(struct super_block *sb, int type);
1168 static inline void ext4_quota_off_umount(struct super_block *sb)
1172 /* Use our quota_off function to clear inode flags etc. */
1173 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1174 ext4_quota_off(sb, type);
1178 * This is a helper function which is used in the mount/remount
1179 * codepaths (which holds s_umount) to fetch the quota file name.
1181 static inline char *get_qf_name(struct super_block *sb,
1182 struct ext4_sb_info *sbi,
1185 return rcu_dereference_protected(sbi->s_qf_names[type],
1186 lockdep_is_held(&sb->s_umount));
1189 static inline void ext4_quota_off_umount(struct super_block *sb)
1194 static void ext4_put_super(struct super_block *sb)
1196 struct ext4_sb_info *sbi = EXT4_SB(sb);
1197 struct ext4_super_block *es = sbi->s_es;
1198 struct buffer_head **group_desc;
1199 struct flex_groups **flex_groups;
1203 ext4_unregister_li_request(sb);
1204 ext4_quota_off_umount(sb);
1206 flush_work(&sbi->s_error_work);
1207 destroy_workqueue(sbi->rsv_conversion_wq);
1208 ext4_release_orphan_info(sb);
1211 * Unregister sysfs before destroying jbd2 journal.
1212 * Since we could still access attr_journal_task attribute via sysfs
1213 * path which could have sbi->s_journal->j_task as NULL
1215 ext4_unregister_sysfs(sb);
1217 if (sbi->s_journal) {
1218 aborted = is_journal_aborted(sbi->s_journal);
1219 err = jbd2_journal_destroy(sbi->s_journal);
1220 sbi->s_journal = NULL;
1221 if ((err < 0) && !aborted) {
1222 ext4_abort(sb, -err, "Couldn't clean up the journal");
1226 ext4_es_unregister_shrinker(sbi);
1227 del_timer_sync(&sbi->s_err_report);
1228 ext4_release_system_zone(sb);
1229 ext4_mb_release(sb);
1230 ext4_ext_release(sb);
1232 if (!sb_rdonly(sb) && !aborted) {
1233 ext4_clear_feature_journal_needs_recovery(sb);
1234 ext4_clear_feature_orphan_present(sb);
1235 es->s_state = cpu_to_le16(sbi->s_mount_state);
1238 ext4_commit_super(sb);
1241 group_desc = rcu_dereference(sbi->s_group_desc);
1242 for (i = 0; i < sbi->s_gdb_count; i++)
1243 brelse(group_desc[i]);
1245 flex_groups = rcu_dereference(sbi->s_flex_groups);
1247 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1248 kvfree(flex_groups[i]);
1249 kvfree(flex_groups);
1252 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1253 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1254 percpu_counter_destroy(&sbi->s_dirs_counter);
1255 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1256 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1257 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1259 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1260 kfree(get_qf_name(sb, sbi, i));
1263 /* Debugging code just in case the in-memory inode orphan list
1264 * isn't empty. The on-disk one can be non-empty if we've
1265 * detected an error and taken the fs readonly, but the
1266 * in-memory list had better be clean by this point. */
1267 if (!list_empty(&sbi->s_orphan))
1268 dump_orphan_list(sb, sbi);
1269 ASSERT(list_empty(&sbi->s_orphan));
1271 sync_blockdev(sb->s_bdev);
1272 invalidate_bdev(sb->s_bdev);
1273 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1275 * Invalidate the journal device's buffers. We don't want them
1276 * floating about in memory - the physical journal device may
1277 * hotswapped, and it breaks the `ro-after' testing code.
1279 sync_blockdev(sbi->s_journal_bdev);
1280 invalidate_bdev(sbi->s_journal_bdev);
1281 ext4_blkdev_remove(sbi);
1284 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1285 sbi->s_ea_inode_cache = NULL;
1287 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1288 sbi->s_ea_block_cache = NULL;
1290 ext4_stop_mmpd(sbi);
1293 sb->s_fs_info = NULL;
1295 * Now that we are completely done shutting down the
1296 * superblock, we need to actually destroy the kobject.
1298 kobject_put(&sbi->s_kobj);
1299 wait_for_completion(&sbi->s_kobj_unregister);
1300 if (sbi->s_chksum_driver)
1301 crypto_free_shash(sbi->s_chksum_driver);
1302 kfree(sbi->s_blockgroup_lock);
1303 fs_put_dax(sbi->s_daxdev);
1304 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1305 #ifdef CONFIG_UNICODE
1306 utf8_unload(sb->s_encoding);
1311 static struct kmem_cache *ext4_inode_cachep;
1314 * Called inside transaction, so use GFP_NOFS
1316 static struct inode *ext4_alloc_inode(struct super_block *sb)
1318 struct ext4_inode_info *ei;
1320 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1324 inode_set_iversion(&ei->vfs_inode, 1);
1325 spin_lock_init(&ei->i_raw_lock);
1326 INIT_LIST_HEAD(&ei->i_prealloc_list);
1327 atomic_set(&ei->i_prealloc_active, 0);
1328 spin_lock_init(&ei->i_prealloc_lock);
1329 ext4_es_init_tree(&ei->i_es_tree);
1330 rwlock_init(&ei->i_es_lock);
1331 INIT_LIST_HEAD(&ei->i_es_list);
1332 ei->i_es_all_nr = 0;
1333 ei->i_es_shk_nr = 0;
1334 ei->i_es_shrink_lblk = 0;
1335 ei->i_reserved_data_blocks = 0;
1336 spin_lock_init(&(ei->i_block_reservation_lock));
1337 ext4_init_pending_tree(&ei->i_pending_tree);
1339 ei->i_reserved_quota = 0;
1340 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1343 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1344 spin_lock_init(&ei->i_completed_io_lock);
1346 ei->i_datasync_tid = 0;
1347 atomic_set(&ei->i_unwritten, 0);
1348 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1349 ext4_fc_init_inode(&ei->vfs_inode);
1350 mutex_init(&ei->i_fc_lock);
1351 return &ei->vfs_inode;
1354 static int ext4_drop_inode(struct inode *inode)
1356 int drop = generic_drop_inode(inode);
1359 drop = fscrypt_drop_inode(inode);
1361 trace_ext4_drop_inode(inode, drop);
1365 static void ext4_free_in_core_inode(struct inode *inode)
1367 fscrypt_free_inode(inode);
1368 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1369 pr_warn("%s: inode %ld still in fc list",
1370 __func__, inode->i_ino);
1372 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1375 static void ext4_destroy_inode(struct inode *inode)
1377 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1378 ext4_msg(inode->i_sb, KERN_ERR,
1379 "Inode %lu (%p): orphan list check failed!",
1380 inode->i_ino, EXT4_I(inode));
1381 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1382 EXT4_I(inode), sizeof(struct ext4_inode_info),
1387 if (EXT4_I(inode)->i_reserved_data_blocks)
1388 ext4_msg(inode->i_sb, KERN_ERR,
1389 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1390 inode->i_ino, EXT4_I(inode),
1391 EXT4_I(inode)->i_reserved_data_blocks);
1394 static void init_once(void *foo)
1396 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1398 INIT_LIST_HEAD(&ei->i_orphan);
1399 init_rwsem(&ei->xattr_sem);
1400 init_rwsem(&ei->i_data_sem);
1401 inode_init_once(&ei->vfs_inode);
1402 ext4_fc_init_inode(&ei->vfs_inode);
1405 static int __init init_inodecache(void)
1407 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1408 sizeof(struct ext4_inode_info), 0,
1409 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1411 offsetof(struct ext4_inode_info, i_data),
1412 sizeof_field(struct ext4_inode_info, i_data),
1414 if (ext4_inode_cachep == NULL)
1419 static void destroy_inodecache(void)
1422 * Make sure all delayed rcu free inodes are flushed before we
1426 kmem_cache_destroy(ext4_inode_cachep);
1429 void ext4_clear_inode(struct inode *inode)
1432 invalidate_inode_buffers(inode);
1434 ext4_discard_preallocations(inode, 0);
1435 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1437 if (EXT4_I(inode)->jinode) {
1438 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1439 EXT4_I(inode)->jinode);
1440 jbd2_free_inode(EXT4_I(inode)->jinode);
1441 EXT4_I(inode)->jinode = NULL;
1443 fscrypt_put_encryption_info(inode);
1444 fsverity_cleanup_inode(inode);
1447 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1448 u64 ino, u32 generation)
1450 struct inode *inode;
1453 * Currently we don't know the generation for parent directory, so
1454 * a generation of 0 means "accept any"
1456 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1458 return ERR_CAST(inode);
1459 if (generation && inode->i_generation != generation) {
1461 return ERR_PTR(-ESTALE);
1467 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1468 int fh_len, int fh_type)
1470 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1471 ext4_nfs_get_inode);
1474 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1475 int fh_len, int fh_type)
1477 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1478 ext4_nfs_get_inode);
1481 static int ext4_nfs_commit_metadata(struct inode *inode)
1483 struct writeback_control wbc = {
1484 .sync_mode = WB_SYNC_ALL
1487 trace_ext4_nfs_commit_metadata(inode);
1488 return ext4_write_inode(inode, &wbc);
1491 #ifdef CONFIG_FS_ENCRYPTION
1492 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1494 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1498 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1501 handle_t *handle = fs_data;
1502 int res, res2, credits, retries = 0;
1505 * Encrypting the root directory is not allowed because e2fsck expects
1506 * lost+found to exist and be unencrypted, and encrypting the root
1507 * directory would imply encrypting the lost+found directory as well as
1508 * the filename "lost+found" itself.
1510 if (inode->i_ino == EXT4_ROOT_INO)
1513 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1516 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1519 res = ext4_convert_inline_data(inode);
1524 * If a journal handle was specified, then the encryption context is
1525 * being set on a new inode via inheritance and is part of a larger
1526 * transaction to create the inode. Otherwise the encryption context is
1527 * being set on an existing inode in its own transaction. Only in the
1528 * latter case should the "retry on ENOSPC" logic be used.
1532 res = ext4_xattr_set_handle(handle, inode,
1533 EXT4_XATTR_INDEX_ENCRYPTION,
1534 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1537 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1538 ext4_clear_inode_state(inode,
1539 EXT4_STATE_MAY_INLINE_DATA);
1541 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1542 * S_DAX may be disabled
1544 ext4_set_inode_flags(inode, false);
1549 res = dquot_initialize(inode);
1553 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1558 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1560 return PTR_ERR(handle);
1562 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1563 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1566 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1568 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1569 * S_DAX may be disabled
1571 ext4_set_inode_flags(inode, false);
1572 res = ext4_mark_inode_dirty(handle, inode);
1574 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1576 res2 = ext4_journal_stop(handle);
1578 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1585 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1587 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1590 static bool ext4_has_stable_inodes(struct super_block *sb)
1592 return ext4_has_feature_stable_inodes(sb);
1595 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1596 int *ino_bits_ret, int *lblk_bits_ret)
1598 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1599 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1602 static const struct fscrypt_operations ext4_cryptops = {
1603 .key_prefix = "ext4:",
1604 .get_context = ext4_get_context,
1605 .set_context = ext4_set_context,
1606 .get_dummy_policy = ext4_get_dummy_policy,
1607 .empty_dir = ext4_empty_dir,
1608 .has_stable_inodes = ext4_has_stable_inodes,
1609 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1614 static const char * const quotatypes[] = INITQFNAMES;
1615 #define QTYPE2NAME(t) (quotatypes[t])
1617 static int ext4_write_dquot(struct dquot *dquot);
1618 static int ext4_acquire_dquot(struct dquot *dquot);
1619 static int ext4_release_dquot(struct dquot *dquot);
1620 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1621 static int ext4_write_info(struct super_block *sb, int type);
1622 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1623 const struct path *path);
1624 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1625 size_t len, loff_t off);
1626 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1627 const char *data, size_t len, loff_t off);
1628 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1629 unsigned int flags);
1631 static struct dquot **ext4_get_dquots(struct inode *inode)
1633 return EXT4_I(inode)->i_dquot;
1636 static const struct dquot_operations ext4_quota_operations = {
1637 .get_reserved_space = ext4_get_reserved_space,
1638 .write_dquot = ext4_write_dquot,
1639 .acquire_dquot = ext4_acquire_dquot,
1640 .release_dquot = ext4_release_dquot,
1641 .mark_dirty = ext4_mark_dquot_dirty,
1642 .write_info = ext4_write_info,
1643 .alloc_dquot = dquot_alloc,
1644 .destroy_dquot = dquot_destroy,
1645 .get_projid = ext4_get_projid,
1646 .get_inode_usage = ext4_get_inode_usage,
1647 .get_next_id = dquot_get_next_id,
1650 static const struct quotactl_ops ext4_qctl_operations = {
1651 .quota_on = ext4_quota_on,
1652 .quota_off = ext4_quota_off,
1653 .quota_sync = dquot_quota_sync,
1654 .get_state = dquot_get_state,
1655 .set_info = dquot_set_dqinfo,
1656 .get_dqblk = dquot_get_dqblk,
1657 .set_dqblk = dquot_set_dqblk,
1658 .get_nextdqblk = dquot_get_next_dqblk,
1662 static const struct super_operations ext4_sops = {
1663 .alloc_inode = ext4_alloc_inode,
1664 .free_inode = ext4_free_in_core_inode,
1665 .destroy_inode = ext4_destroy_inode,
1666 .write_inode = ext4_write_inode,
1667 .dirty_inode = ext4_dirty_inode,
1668 .drop_inode = ext4_drop_inode,
1669 .evict_inode = ext4_evict_inode,
1670 .put_super = ext4_put_super,
1671 .sync_fs = ext4_sync_fs,
1672 .freeze_fs = ext4_freeze,
1673 .unfreeze_fs = ext4_unfreeze,
1674 .statfs = ext4_statfs,
1675 .show_options = ext4_show_options,
1677 .quota_read = ext4_quota_read,
1678 .quota_write = ext4_quota_write,
1679 .get_dquots = ext4_get_dquots,
1683 static const struct export_operations ext4_export_ops = {
1684 .fh_to_dentry = ext4_fh_to_dentry,
1685 .fh_to_parent = ext4_fh_to_parent,
1686 .get_parent = ext4_get_parent,
1687 .commit_metadata = ext4_nfs_commit_metadata,
1691 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1692 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1693 Opt_nouid32, Opt_debug, Opt_removed,
1694 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1695 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1696 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1697 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1698 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1699 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1701 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1702 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1703 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1704 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1705 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1706 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1707 Opt_nowarn_on_error, Opt_mblk_io_submit,
1708 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1709 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1710 Opt_inode_readahead_blks, Opt_journal_ioprio,
1711 Opt_dioread_nolock, Opt_dioread_lock,
1712 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1713 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1714 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1715 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1716 #ifdef CONFIG_EXT4_DEBUG
1717 Opt_fc_debug_max_replay, Opt_fc_debug_force
1721 static const struct constant_table ext4_param_errors[] = {
1722 {"continue", Opt_err_cont},
1723 {"panic", Opt_err_panic},
1724 {"remount-ro", Opt_err_ro},
1728 static const struct constant_table ext4_param_data[] = {
1729 {"journal", Opt_data_journal},
1730 {"ordered", Opt_data_ordered},
1731 {"writeback", Opt_data_writeback},
1735 static const struct constant_table ext4_param_data_err[] = {
1736 {"abort", Opt_data_err_abort},
1737 {"ignore", Opt_data_err_ignore},
1741 static const struct constant_table ext4_param_jqfmt[] = {
1742 {"vfsold", Opt_jqfmt_vfsold},
1743 {"vfsv0", Opt_jqfmt_vfsv0},
1744 {"vfsv1", Opt_jqfmt_vfsv1},
1748 static const struct constant_table ext4_param_dax[] = {
1749 {"always", Opt_dax_always},
1750 {"inode", Opt_dax_inode},
1751 {"never", Opt_dax_never},
1755 /* String parameter that allows empty argument */
1756 #define fsparam_string_empty(NAME, OPT) \
1757 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1760 * Mount option specification
1761 * We don't use fsparam_flag_no because of the way we set the
1762 * options and the way we show them in _ext4_show_options(). To
1763 * keep the changes to a minimum, let's keep the negative options
1766 static const struct fs_parameter_spec ext4_param_specs[] = {
1767 fsparam_flag ("bsddf", Opt_bsd_df),
1768 fsparam_flag ("minixdf", Opt_minix_df),
1769 fsparam_flag ("grpid", Opt_grpid),
1770 fsparam_flag ("bsdgroups", Opt_grpid),
1771 fsparam_flag ("nogrpid", Opt_nogrpid),
1772 fsparam_flag ("sysvgroups", Opt_nogrpid),
1773 fsparam_u32 ("resgid", Opt_resgid),
1774 fsparam_u32 ("resuid", Opt_resuid),
1775 fsparam_u32 ("sb", Opt_sb),
1776 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1777 fsparam_flag ("nouid32", Opt_nouid32),
1778 fsparam_flag ("debug", Opt_debug),
1779 fsparam_flag ("oldalloc", Opt_removed),
1780 fsparam_flag ("orlov", Opt_removed),
1781 fsparam_flag ("user_xattr", Opt_user_xattr),
1782 fsparam_flag ("nouser_xattr", Opt_nouser_xattr),
1783 fsparam_flag ("acl", Opt_acl),
1784 fsparam_flag ("noacl", Opt_noacl),
1785 fsparam_flag ("norecovery", Opt_noload),
1786 fsparam_flag ("noload", Opt_noload),
1787 fsparam_flag ("bh", Opt_removed),
1788 fsparam_flag ("nobh", Opt_removed),
1789 fsparam_u32 ("commit", Opt_commit),
1790 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1791 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1792 fsparam_u32 ("journal_dev", Opt_journal_dev),
1793 fsparam_bdev ("journal_path", Opt_journal_path),
1794 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1795 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1796 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1797 fsparam_flag ("abort", Opt_abort),
1798 fsparam_enum ("data", Opt_data, ext4_param_data),
1799 fsparam_enum ("data_err", Opt_data_err,
1800 ext4_param_data_err),
1801 fsparam_string_empty
1802 ("usrjquota", Opt_usrjquota),
1803 fsparam_string_empty
1804 ("grpjquota", Opt_grpjquota),
1805 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1806 fsparam_flag ("grpquota", Opt_grpquota),
1807 fsparam_flag ("quota", Opt_quota),
1808 fsparam_flag ("noquota", Opt_noquota),
1809 fsparam_flag ("usrquota", Opt_usrquota),
1810 fsparam_flag ("prjquota", Opt_prjquota),
1811 fsparam_flag ("barrier", Opt_barrier),
1812 fsparam_u32 ("barrier", Opt_barrier),
1813 fsparam_flag ("nobarrier", Opt_nobarrier),
1814 fsparam_flag ("i_version", Opt_i_version),
1815 fsparam_flag ("dax", Opt_dax),
1816 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1817 fsparam_u32 ("stripe", Opt_stripe),
1818 fsparam_flag ("delalloc", Opt_delalloc),
1819 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1820 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1821 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1822 fsparam_flag ("lazytime", Opt_lazytime),
1823 fsparam_flag ("nolazytime", Opt_nolazytime),
1824 fsparam_u32 ("debug_want_extra_isize",
1825 Opt_debug_want_extra_isize),
1826 fsparam_flag ("mblk_io_submit", Opt_removed),
1827 fsparam_flag ("nomblk_io_submit", Opt_removed),
1828 fsparam_flag ("block_validity", Opt_block_validity),
1829 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1830 fsparam_u32 ("inode_readahead_blks",
1831 Opt_inode_readahead_blks),
1832 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1833 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1834 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1835 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1836 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1837 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1838 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1839 fsparam_flag ("discard", Opt_discard),
1840 fsparam_flag ("nodiscard", Opt_nodiscard),
1841 fsparam_u32 ("init_itable", Opt_init_itable),
1842 fsparam_flag ("init_itable", Opt_init_itable),
1843 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1844 #ifdef CONFIG_EXT4_DEBUG
1845 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1846 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1848 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1849 fsparam_flag ("test_dummy_encryption",
1850 Opt_test_dummy_encryption),
1851 fsparam_string ("test_dummy_encryption",
1852 Opt_test_dummy_encryption),
1853 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1854 fsparam_flag ("nombcache", Opt_nombcache),
1855 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1856 fsparam_flag ("prefetch_block_bitmaps",
1858 fsparam_flag ("no_prefetch_block_bitmaps",
1859 Opt_no_prefetch_block_bitmaps),
1860 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1861 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1862 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1863 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1864 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1865 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1869 static const match_table_t tokens = {
1870 {Opt_bsd_df, "bsddf"},
1871 {Opt_minix_df, "minixdf"},
1872 {Opt_grpid, "grpid"},
1873 {Opt_grpid, "bsdgroups"},
1874 {Opt_nogrpid, "nogrpid"},
1875 {Opt_nogrpid, "sysvgroups"},
1876 {Opt_resgid, "resgid=%u"},
1877 {Opt_resuid, "resuid=%u"},
1879 {Opt_err_cont, "errors=continue"},
1880 {Opt_err_panic, "errors=panic"},
1881 {Opt_err_ro, "errors=remount-ro"},
1882 {Opt_nouid32, "nouid32"},
1883 {Opt_debug, "debug"},
1884 {Opt_removed, "oldalloc"},
1885 {Opt_removed, "orlov"},
1886 {Opt_user_xattr, "user_xattr"},
1887 {Opt_nouser_xattr, "nouser_xattr"},
1889 {Opt_noacl, "noacl"},
1890 {Opt_noload, "norecovery"},
1891 {Opt_noload, "noload"},
1892 {Opt_removed, "nobh"},
1893 {Opt_removed, "bh"},
1894 {Opt_commit, "commit=%u"},
1895 {Opt_min_batch_time, "min_batch_time=%u"},
1896 {Opt_max_batch_time, "max_batch_time=%u"},
1897 {Opt_journal_dev, "journal_dev=%u"},
1898 {Opt_journal_path, "journal_path=%s"},
1899 {Opt_journal_checksum, "journal_checksum"},
1900 {Opt_nojournal_checksum, "nojournal_checksum"},
1901 {Opt_journal_async_commit, "journal_async_commit"},
1902 {Opt_abort, "abort"},
1903 {Opt_data_journal, "data=journal"},
1904 {Opt_data_ordered, "data=ordered"},
1905 {Opt_data_writeback, "data=writeback"},
1906 {Opt_data_err_abort, "data_err=abort"},
1907 {Opt_data_err_ignore, "data_err=ignore"},
1908 {Opt_offusrjquota, "usrjquota="},
1909 {Opt_usrjquota, "usrjquota=%s"},
1910 {Opt_offgrpjquota, "grpjquota="},
1911 {Opt_grpjquota, "grpjquota=%s"},
1912 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1913 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1914 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1915 {Opt_grpquota, "grpquota"},
1916 {Opt_noquota, "noquota"},
1917 {Opt_quota, "quota"},
1918 {Opt_usrquota, "usrquota"},
1919 {Opt_prjquota, "prjquota"},
1920 {Opt_barrier, "barrier=%u"},
1921 {Opt_barrier, "barrier"},
1922 {Opt_nobarrier, "nobarrier"},
1923 {Opt_i_version, "i_version"},
1925 {Opt_dax_always, "dax=always"},
1926 {Opt_dax_inode, "dax=inode"},
1927 {Opt_dax_never, "dax=never"},
1928 {Opt_stripe, "stripe=%u"},
1929 {Opt_delalloc, "delalloc"},
1930 {Opt_warn_on_error, "warn_on_error"},
1931 {Opt_nowarn_on_error, "nowarn_on_error"},
1932 {Opt_lazytime, "lazytime"},
1933 {Opt_nolazytime, "nolazytime"},
1934 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1935 {Opt_nodelalloc, "nodelalloc"},
1936 {Opt_removed, "mblk_io_submit"},
1937 {Opt_removed, "nomblk_io_submit"},
1938 {Opt_block_validity, "block_validity"},
1939 {Opt_noblock_validity, "noblock_validity"},
1940 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1941 {Opt_journal_ioprio, "journal_ioprio=%u"},
1942 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1943 {Opt_auto_da_alloc, "auto_da_alloc"},
1944 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1945 {Opt_dioread_nolock, "dioread_nolock"},
1946 {Opt_dioread_lock, "nodioread_nolock"},
1947 {Opt_dioread_lock, "dioread_lock"},
1948 {Opt_discard, "discard"},
1949 {Opt_nodiscard, "nodiscard"},
1950 {Opt_init_itable, "init_itable=%u"},
1951 {Opt_init_itable, "init_itable"},
1952 {Opt_noinit_itable, "noinit_itable"},
1953 #ifdef CONFIG_EXT4_DEBUG
1954 {Opt_fc_debug_force, "fc_debug_force"},
1955 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1957 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1958 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1959 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1960 {Opt_inlinecrypt, "inlinecrypt"},
1961 {Opt_nombcache, "nombcache"},
1962 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1963 {Opt_removed, "prefetch_block_bitmaps"},
1964 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1965 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1966 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1967 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1968 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1969 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1970 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1974 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1975 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1977 static const char deprecated_msg[] =
1978 "Mount option \"%s\" will be removed by %s\n"
1979 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1981 #define MOPT_SET 0x0001
1982 #define MOPT_CLEAR 0x0002
1983 #define MOPT_NOSUPPORT 0x0004
1984 #define MOPT_EXPLICIT 0x0008
1985 #define MOPT_CLEAR_ERR 0x0010
1986 #define MOPT_GTE0 0x0020
1989 #define MOPT_QFMT 0x0040
1991 #define MOPT_Q MOPT_NOSUPPORT
1992 #define MOPT_QFMT MOPT_NOSUPPORT
1994 #define MOPT_DATAJ 0x0080
1995 #define MOPT_NO_EXT2 0x0100
1996 #define MOPT_NO_EXT3 0x0200
1997 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1998 #define MOPT_STRING 0x0400
1999 #define MOPT_SKIP 0x0800
2000 #define MOPT_2 0x1000
2002 static const struct mount_opts {
2006 } ext4_mount_opts[] = {
2007 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
2008 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
2009 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
2010 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
2011 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
2012 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
2013 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
2014 MOPT_EXT4_ONLY | MOPT_SET},
2015 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
2016 MOPT_EXT4_ONLY | MOPT_CLEAR},
2017 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
2018 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
2019 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
2020 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
2021 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
2022 MOPT_EXT4_ONLY | MOPT_CLEAR},
2023 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
2024 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
2025 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
2026 MOPT_EXT4_ONLY | MOPT_CLEAR},
2027 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
2028 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
2029 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
2030 EXT4_MOUNT_JOURNAL_CHECKSUM),
2031 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
2032 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
2033 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
2034 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
2035 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
2036 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
2038 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
2040 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
2041 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
2042 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
2043 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
2044 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
2045 {Opt_commit, 0, MOPT_GTE0},
2046 {Opt_max_batch_time, 0, MOPT_GTE0},
2047 {Opt_min_batch_time, 0, MOPT_GTE0},
2048 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
2049 {Opt_init_itable, 0, MOPT_GTE0},
2050 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
2051 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
2052 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
2053 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
2054 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
2055 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
2056 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
2057 {Opt_stripe, 0, MOPT_GTE0},
2058 {Opt_resuid, 0, MOPT_GTE0},
2059 {Opt_resgid, 0, MOPT_GTE0},
2060 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
2061 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
2062 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
2063 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
2064 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
2065 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
2066 MOPT_NO_EXT2 | MOPT_DATAJ},
2067 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
2068 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
2069 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2070 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
2071 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
2073 {Opt_acl, 0, MOPT_NOSUPPORT},
2074 {Opt_noacl, 0, MOPT_NOSUPPORT},
2076 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
2077 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
2078 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
2079 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2080 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2082 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2084 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2086 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2087 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2088 MOPT_CLEAR | MOPT_Q},
2089 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2090 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
2091 {Opt_offusrjquota, 0, MOPT_Q},
2092 {Opt_offgrpjquota, 0, MOPT_Q},
2093 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2094 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2095 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2096 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
2097 {Opt_test_dummy_encryption, 0, MOPT_STRING},
2098 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2099 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2101 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2102 #ifdef CONFIG_EXT4_DEBUG
2103 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2104 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2105 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2110 #ifdef CONFIG_UNICODE
2111 static const struct ext4_sb_encodings {
2115 } ext4_sb_encoding_map[] = {
2116 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2119 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2120 const struct ext4_sb_encodings **encoding,
2123 __u16 magic = le16_to_cpu(es->s_encoding);
2126 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2127 if (magic == ext4_sb_encoding_map[i].magic)
2130 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2133 *encoding = &ext4_sb_encoding_map[i];
2134 *flags = le16_to_cpu(es->s_encoding_flags);
2140 static int ext4_set_test_dummy_encryption(struct super_block *sb, char *arg)
2142 #ifdef CONFIG_FS_ENCRYPTION
2143 struct ext4_sb_info *sbi = EXT4_SB(sb);
2146 err = fscrypt_set_test_dummy_encryption(sb, arg,
2147 &sbi->s_dummy_enc_policy);
2149 ext4_msg(sb, KERN_WARNING,
2150 "Error while setting test dummy encryption [%d]", err);
2153 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2158 #define EXT4_SPEC_JQUOTA (1 << 0)
2159 #define EXT4_SPEC_JQFMT (1 << 1)
2160 #define EXT4_SPEC_DATAJ (1 << 2)
2161 #define EXT4_SPEC_SB_BLOCK (1 << 3)
2162 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
2163 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
2164 #define EXT4_SPEC_DUMMY_ENCRYPTION (1 << 6)
2165 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
2166 #define EXT4_SPEC_s_max_batch_time (1 << 8)
2167 #define EXT4_SPEC_s_min_batch_time (1 << 9)
2168 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
2169 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
2170 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
2171 #define EXT4_SPEC_s_stripe (1 << 13)
2172 #define EXT4_SPEC_s_resuid (1 << 14)
2173 #define EXT4_SPEC_s_resgid (1 << 15)
2174 #define EXT4_SPEC_s_commit_interval (1 << 16)
2175 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
2176 #define EXT4_SPEC_s_sb_block (1 << 18)
2178 struct ext4_fs_context {
2179 char *s_qf_names[EXT4_MAXQUOTAS];
2180 char *test_dummy_enc_arg;
2181 int s_jquota_fmt; /* Format of quota to use */
2182 int mb_optimize_scan;
2183 #ifdef CONFIG_EXT4_DEBUG
2184 int s_fc_debug_max_replay;
2186 unsigned short qname_spec;
2187 unsigned long vals_s_flags; /* Bits to set in s_flags */
2188 unsigned long mask_s_flags; /* Bits changed in s_flags */
2189 unsigned long journal_devnum;
2190 unsigned long s_commit_interval;
2191 unsigned long s_stripe;
2192 unsigned int s_inode_readahead_blks;
2193 unsigned int s_want_extra_isize;
2194 unsigned int s_li_wait_mult;
2195 unsigned int s_max_dir_size_kb;
2196 unsigned int journal_ioprio;
2197 unsigned int vals_s_mount_opt;
2198 unsigned int mask_s_mount_opt;
2199 unsigned int vals_s_mount_opt2;
2200 unsigned int mask_s_mount_opt2;
2201 unsigned int vals_s_mount_flags;
2202 unsigned int mask_s_mount_flags;
2203 unsigned int opt_flags; /* MOPT flags */
2205 u32 s_max_batch_time;
2206 u32 s_min_batch_time;
2209 ext4_fsblk_t s_sb_block;
2212 static void ext4_fc_free(struct fs_context *fc)
2214 struct ext4_fs_context *ctx = fc->fs_private;
2220 for (i = 0; i < EXT4_MAXQUOTAS; i++)
2221 kfree(ctx->s_qf_names[i]);
2223 kfree(ctx->test_dummy_enc_arg);
2227 int ext4_init_fs_context(struct fs_context *fc)
2229 struct xfs_fs_context *ctx;
2231 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2235 fc->fs_private = ctx;
2236 fc->ops = &ext4_context_ops;
2243 * Note the name of the specified quota file.
2245 static int note_qf_name(struct fs_context *fc, int qtype,
2246 struct fs_parameter *param)
2248 struct ext4_fs_context *ctx = fc->fs_private;
2251 if (param->size < 1) {
2252 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2255 if (strchr(param->string, '/')) {
2256 ext4_msg(NULL, KERN_ERR,
2257 "quotafile must be on filesystem root");
2260 if (ctx->s_qf_names[qtype]) {
2261 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2262 ext4_msg(NULL, KERN_ERR,
2263 "%s quota file already specified",
2270 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2272 ext4_msg(NULL, KERN_ERR,
2273 "Not enough memory for storing quotafile name");
2276 ctx->s_qf_names[qtype] = qname;
2277 ctx->qname_spec |= 1 << qtype;
2278 ctx->spec |= EXT4_SPEC_JQUOTA;
2283 * Clear the name of the specified quota file.
2285 static int unnote_qf_name(struct fs_context *fc, int qtype)
2287 struct ext4_fs_context *ctx = fc->fs_private;
2289 if (ctx->s_qf_names[qtype])
2290 kfree(ctx->s_qf_names[qtype]);
2292 ctx->s_qf_names[qtype] = NULL;
2293 ctx->qname_spec |= 1 << qtype;
2294 ctx->spec |= EXT4_SPEC_JQUOTA;
2299 #define EXT4_SET_CTX(name) \
2300 static inline void ctx_set_##name(struct ext4_fs_context *ctx, int flag)\
2302 ctx->mask_s_##name |= flag; \
2303 ctx->vals_s_##name |= flag; \
2305 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, int flag)\
2307 ctx->mask_s_##name |= flag; \
2308 ctx->vals_s_##name &= ~flag; \
2310 static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\
2312 return ((ctx->vals_s_##name & flag) != 0); \
2315 EXT4_SET_CTX(flags);
2316 EXT4_SET_CTX(mount_opt);
2317 EXT4_SET_CTX(mount_opt2);
2318 EXT4_SET_CTX(mount_flags);
2320 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2322 struct ext4_fs_context *ctx = fc->fs_private;
2323 struct fs_parse_result result;
2324 const struct mount_opts *m;
2330 token = fs_parse(fc, ext4_param_specs, param, &result);
2333 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2336 if (token == Opt_usrjquota) {
2337 if (!*param->string)
2338 return unnote_qf_name(fc, USRQUOTA);
2340 return note_qf_name(fc, USRQUOTA, param);
2341 } else if (token == Opt_grpjquota) {
2342 if (!*param->string)
2343 return unnote_qf_name(fc, GRPQUOTA);
2345 return note_qf_name(fc, GRPQUOTA, param);
2350 case Opt_nouser_xattr:
2351 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5");
2354 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2355 ext4_msg(NULL, KERN_WARNING,
2356 "Ignoring %s option on remount", param->key);
2358 ctx->s_sb_block = result.uint_32;
2359 ctx->spec |= EXT4_SPEC_s_sb_block;
2363 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2367 ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
2370 ctx_set_flags(ctx, SB_I_VERSION);
2373 ctx_set_flags(ctx, SB_LAZYTIME);
2375 case Opt_nolazytime:
2376 ctx_clear_flags(ctx, SB_LAZYTIME);
2378 case Opt_inlinecrypt:
2379 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2380 ctx_set_flags(ctx, SB_INLINECRYPT);
2382 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2390 token = result.uint_32;
2393 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2394 if (token == m->token)
2397 ctx->opt_flags |= m->flags;
2399 if (m->token == Opt_err) {
2400 ext4_msg(NULL, KERN_ERR, "Unrecognized mount option \"%s\" "
2401 "or missing value", param->key);
2405 if (m->flags & MOPT_EXPLICIT) {
2406 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2407 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2408 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2409 ctx_set_mount_opt2(ctx,
2410 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2414 if (m->flags & MOPT_CLEAR_ERR)
2415 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2417 if (m->flags & MOPT_NOSUPPORT) {
2418 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2420 } else if (token == Opt_commit) {
2421 if (result.uint_32 == 0)
2422 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
2423 else if (result.uint_32 > INT_MAX / HZ) {
2424 ext4_msg(NULL, KERN_ERR,
2425 "Invalid commit interval %d, "
2426 "must be smaller than %d",
2427 result.uint_32, INT_MAX / HZ);
2430 ctx->s_commit_interval = HZ * result.uint_32;
2431 ctx->spec |= EXT4_SPEC_s_commit_interval;
2432 } else if (token == Opt_debug_want_extra_isize) {
2433 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2434 ext4_msg(NULL, KERN_ERR,
2435 "Invalid want_extra_isize %d", result.uint_32);
2438 ctx->s_want_extra_isize = result.uint_32;
2439 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2440 } else if (token == Opt_max_batch_time) {
2441 ctx->s_max_batch_time = result.uint_32;
2442 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2443 } else if (token == Opt_min_batch_time) {
2444 ctx->s_min_batch_time = result.uint_32;
2445 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2446 } else if (token == Opt_inode_readahead_blks) {
2447 if (result.uint_32 &&
2448 (result.uint_32 > (1 << 30) ||
2449 !is_power_of_2(result.uint_32))) {
2450 ext4_msg(NULL, KERN_ERR,
2451 "EXT4-fs: inode_readahead_blks must be "
2452 "0 or a power of 2 smaller than 2^31");
2455 ctx->s_inode_readahead_blks = result.uint_32;
2456 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2457 } else if (token == Opt_init_itable) {
2458 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2459 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2460 if (param->type == fs_value_is_string)
2461 ctx->s_li_wait_mult = result.uint_32;
2462 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2463 } else if (token == Opt_max_dir_size_kb) {
2464 ctx->s_max_dir_size_kb = result.uint_32;
2465 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2466 #ifdef CONFIG_EXT4_DEBUG
2467 } else if (token == Opt_fc_debug_max_replay) {
2468 ctx->s_fc_debug_max_replay = result.uint_32;
2469 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2471 } else if (token == Opt_stripe) {
2472 ctx->s_stripe = result.uint_32;
2473 ctx->spec |= EXT4_SPEC_s_stripe;
2474 } else if (token == Opt_resuid) {
2475 uid = make_kuid(current_user_ns(), result.uint_32);
2476 if (!uid_valid(uid)) {
2477 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2481 ctx->s_resuid = uid;
2482 ctx->spec |= EXT4_SPEC_s_resuid;
2483 } else if (token == Opt_resgid) {
2484 gid = make_kgid(current_user_ns(), result.uint_32);
2485 if (!gid_valid(gid)) {
2486 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2490 ctx->s_resgid = gid;
2491 ctx->spec |= EXT4_SPEC_s_resgid;
2492 } else if (token == Opt_journal_dev) {
2494 ext4_msg(NULL, KERN_ERR,
2495 "Cannot specify journal on remount");
2498 ctx->journal_devnum = result.uint_32;
2499 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2500 } else if (token == Opt_journal_path) {
2501 struct inode *journal_inode;
2506 ext4_msg(NULL, KERN_ERR,
2507 "Cannot specify journal on remount");
2511 error = fs_lookup_param(fc, param, 1, &path);
2513 ext4_msg(NULL, KERN_ERR, "error: could not find "
2514 "journal device path");
2518 journal_inode = d_inode(path.dentry);
2519 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2520 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2522 } else if (token == Opt_journal_ioprio) {
2523 if (result.uint_32 > 7) {
2524 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2528 ctx->journal_ioprio =
2529 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2530 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2531 } else if (token == Opt_test_dummy_encryption) {
2532 #ifdef CONFIG_FS_ENCRYPTION
2533 if (param->type == fs_value_is_flag) {
2534 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION;
2535 ctx->test_dummy_enc_arg = NULL;
2538 if (*param->string &&
2539 !(!strcmp(param->string, "v1") ||
2540 !strcmp(param->string, "v2"))) {
2541 ext4_msg(NULL, KERN_WARNING,
2542 "Value of option \"%s\" is unrecognized",
2546 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION;
2547 ctx->test_dummy_enc_arg = kmemdup_nul(param->string, param->size,
2550 ext4_msg(NULL, KERN_WARNING,
2551 "Test dummy encryption mount option ignored");
2553 } else if (m->flags & MOPT_DATAJ) {
2554 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2555 ctx_set_mount_opt(ctx, m->mount_opt);
2556 ctx->spec |= EXT4_SPEC_DATAJ;
2558 } else if (m->flags & MOPT_QFMT) {
2559 ctx->s_jquota_fmt = m->mount_opt;
2560 ctx->spec |= EXT4_SPEC_JQFMT;
2562 } else if (token == Opt_dax || token == Opt_dax_always ||
2563 token == Opt_dax_inode || token == Opt_dax_never) {
2564 #ifdef CONFIG_FS_DAX
2567 case Opt_dax_always:
2568 ctx_set_mount_opt(ctx, m->mount_opt);
2569 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2572 ctx_set_mount_opt2(ctx, m->mount_opt);
2573 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2576 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2577 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2578 /* Strictly for printing options */
2579 ctx_set_mount_opt2(ctx, m->mount_opt);
2583 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2584 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2585 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2588 } else if (token == Opt_data_err_abort) {
2589 ctx_set_mount_opt(ctx, m->mount_opt);
2590 } else if (token == Opt_data_err_ignore) {
2591 ctx_clear_mount_opt(ctx, m->mount_opt);
2592 } else if (token == Opt_mb_optimize_scan) {
2593 if (result.int_32 != 0 && result.int_32 != 1) {
2594 ext4_msg(NULL, KERN_WARNING,
2595 "mb_optimize_scan should be set to 0 or 1.");
2598 ctx->mb_optimize_scan = result.int_32;
2600 unsigned int set = 0;
2602 if ((param->type == fs_value_is_flag) ||
2606 if (m->flags & MOPT_CLEAR)
2608 else if (unlikely(!(m->flags & MOPT_SET))) {
2609 ext4_msg(NULL, KERN_WARNING,
2610 "buggy handling of option %s",
2615 if (m->flags & MOPT_2) {
2617 ctx_set_mount_opt2(ctx, m->mount_opt);
2619 ctx_clear_mount_opt2(ctx, m->mount_opt);
2622 ctx_set_mount_opt(ctx, m->mount_opt);
2624 ctx_clear_mount_opt(ctx, m->mount_opt);
2630 static int parse_options(struct fs_context *fc, char *options)
2632 struct fs_parameter param;
2639 while ((key = strsep(&options, ",")) != NULL) {
2642 char *value = strchr(key, '=');
2644 param.type = fs_value_is_flag;
2645 param.string = NULL;
2652 v_len = strlen(value);
2653 param.string = kmemdup_nul(value, v_len,
2657 param.type = fs_value_is_string;
2663 ret = ext4_parse_param(fc, ¶m);
2665 kfree(param.string);
2671 ret = ext4_validate_options(fc);
2678 static int parse_apply_sb_mount_options(struct super_block *sb,
2679 struct ext4_fs_context *m_ctx)
2681 struct ext4_sb_info *sbi = EXT4_SB(sb);
2682 char *s_mount_opts = NULL;
2683 struct ext4_fs_context *s_ctx = NULL;
2684 struct fs_context *fc = NULL;
2687 if (!sbi->s_es->s_mount_opts[0])
2690 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2691 sizeof(sbi->s_es->s_mount_opts),
2696 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2700 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2704 fc->fs_private = s_ctx;
2705 fc->s_fs_info = sbi;
2707 ret = parse_options(fc, s_mount_opts);
2711 ret = ext4_check_opt_consistency(fc, sb);
2714 ext4_msg(sb, KERN_WARNING,
2715 "failed to parse options in superblock: %s",
2721 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2722 m_ctx->journal_devnum = s_ctx->journal_devnum;
2723 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2724 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2726 ret = ext4_apply_options(fc, sb);
2731 kfree(s_mount_opts);
2735 static void ext4_apply_quota_options(struct fs_context *fc,
2736 struct super_block *sb)
2739 bool quota_feature = ext4_has_feature_quota(sb);
2740 struct ext4_fs_context *ctx = fc->fs_private;
2741 struct ext4_sb_info *sbi = EXT4_SB(sb);
2748 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2749 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2750 if (!(ctx->qname_spec & (1 << i)))
2753 qname = ctx->s_qf_names[i]; /* May be NULL */
2754 ctx->s_qf_names[i] = NULL;
2755 kfree(sbi->s_qf_names[i]);
2756 rcu_assign_pointer(sbi->s_qf_names[i], qname);
2761 if (ctx->spec & EXT4_SPEC_JQFMT)
2762 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2767 * Check quota settings consistency.
2769 static int ext4_check_quota_consistency(struct fs_context *fc,
2770 struct super_block *sb)
2773 struct ext4_fs_context *ctx = fc->fs_private;
2774 struct ext4_sb_info *sbi = EXT4_SB(sb);
2775 bool quota_feature = ext4_has_feature_quota(sb);
2776 bool quota_loaded = sb_any_quota_loaded(sb);
2777 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2781 * We do the test below only for project quotas. 'usrquota' and
2782 * 'grpquota' mount options are allowed even without quota feature
2783 * to support legacy quotas in quota files.
2785 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2786 !ext4_has_feature_project(sb)) {
2787 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2788 "Cannot enable project quota enforcement.");
2792 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2793 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2795 ctx->mask_s_mount_opt & quota_flags &&
2796 !ctx_test_mount_opt(ctx, quota_flags))
2797 goto err_quota_change;
2799 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2801 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2802 if (!(ctx->qname_spec & (1 << i)))
2806 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2807 goto err_jquota_change;
2809 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2810 strcmp(sbi->s_qf_names[i],
2811 ctx->s_qf_names[i]) != 0)
2812 goto err_jquota_specified;
2815 if (quota_feature) {
2816 ext4_msg(NULL, KERN_INFO,
2817 "Journaled quota options ignored when "
2818 "QUOTA feature is enabled");
2823 if (ctx->spec & EXT4_SPEC_JQFMT) {
2824 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2825 goto err_jquota_change;
2826 if (quota_feature) {
2827 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2828 "ignored when QUOTA feature is enabled");
2833 /* Make sure we don't mix old and new quota format */
2834 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2835 ctx->s_qf_names[USRQUOTA]);
2836 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2837 ctx->s_qf_names[GRPQUOTA]);
2839 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2840 test_opt(sb, USRQUOTA));
2842 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2843 test_opt(sb, GRPQUOTA));
2846 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2850 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2854 if (usr_qf_name || grp_qf_name) {
2855 if (usrquota || grpquota) {
2856 ext4_msg(NULL, KERN_ERR, "old and new quota "
2861 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2862 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2871 ext4_msg(NULL, KERN_ERR,
2872 "Cannot change quota options when quota turned on");
2875 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2876 "options when quota turned on");
2878 err_jquota_specified:
2879 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2887 static int ext4_check_opt_consistency(struct fs_context *fc,
2888 struct super_block *sb)
2890 struct ext4_fs_context *ctx = fc->fs_private;
2891 struct ext4_sb_info *sbi = fc->s_fs_info;
2892 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2894 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2895 ext4_msg(NULL, KERN_ERR,
2896 "Mount option(s) incompatible with ext2");
2899 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2900 ext4_msg(NULL, KERN_ERR,
2901 "Mount option(s) incompatible with ext3");
2905 if (ctx->s_want_extra_isize >
2906 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2907 ext4_msg(NULL, KERN_ERR,
2908 "Invalid want_extra_isize %d",
2909 ctx->s_want_extra_isize);
2913 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2915 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2916 if (blocksize < PAGE_SIZE)
2917 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2918 "experimental mount option 'dioread_nolock' "
2919 "for blocksize < PAGE_SIZE");
2922 #ifdef CONFIG_FS_ENCRYPTION
2924 * This mount option is just for testing, and it's not worthwhile to
2925 * implement the extra complexity (e.g. RCU protection) that would be
2926 * needed to allow it to be set or changed during remount. We do allow
2927 * it to be specified during remount, but only if there is no change.
2929 if ((ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) &&
2930 is_remount && !sbi->s_dummy_enc_policy.policy) {
2931 ext4_msg(NULL, KERN_WARNING,
2932 "Can't set test_dummy_encryption on remount");
2937 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2938 if (!sbi->s_journal) {
2939 ext4_msg(NULL, KERN_WARNING,
2940 "Remounting file system with no journal "
2941 "so ignoring journalled data option");
2942 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2943 } else if (ctx->mask_s_mount_opt & EXT4_MOUNT_DATA_FLAGS) {
2944 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2951 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2952 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2953 ext4_msg(NULL, KERN_ERR, "can't mount with "
2954 "both data=journal and dax");
2958 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2959 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2960 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2961 fail_dax_change_remount:
2962 ext4_msg(NULL, KERN_ERR, "can't change "
2963 "dax mount option while remounting");
2965 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2966 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2967 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2968 goto fail_dax_change_remount;
2969 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2970 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2971 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2972 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2973 goto fail_dax_change_remount;
2977 return ext4_check_quota_consistency(fc, sb);
2980 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2982 struct ext4_fs_context *ctx = fc->fs_private;
2983 struct ext4_sb_info *sbi = fc->s_fs_info;
2986 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2987 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2988 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2989 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2990 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2991 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2992 sb->s_flags &= ~ctx->mask_s_flags;
2993 sb->s_flags |= ctx->vals_s_flags;
2995 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2996 APPLY(s_commit_interval);
2998 APPLY(s_max_batch_time);
2999 APPLY(s_min_batch_time);
3000 APPLY(s_want_extra_isize);
3001 APPLY(s_inode_readahead_blks);
3002 APPLY(s_max_dir_size_kb);
3003 APPLY(s_li_wait_mult);
3007 #ifdef CONFIG_EXT4_DEBUG
3008 APPLY(s_fc_debug_max_replay);
3011 ext4_apply_quota_options(fc, sb);
3013 if (ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION)
3014 ret = ext4_set_test_dummy_encryption(sb, ctx->test_dummy_enc_arg);
3020 static int ext4_validate_options(struct fs_context *fc)
3023 struct ext4_fs_context *ctx = fc->fs_private;
3024 char *usr_qf_name, *grp_qf_name;
3026 usr_qf_name = ctx->s_qf_names[USRQUOTA];
3027 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
3029 if (usr_qf_name || grp_qf_name) {
3030 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
3031 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
3033 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
3034 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
3036 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
3037 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
3038 ext4_msg(NULL, KERN_ERR, "old and new quota "
3047 static inline void ext4_show_quota_options(struct seq_file *seq,
3048 struct super_block *sb)
3050 #if defined(CONFIG_QUOTA)
3051 struct ext4_sb_info *sbi = EXT4_SB(sb);
3052 char *usr_qf_name, *grp_qf_name;
3054 if (sbi->s_jquota_fmt) {
3057 switch (sbi->s_jquota_fmt) {
3068 seq_printf(seq, ",jqfmt=%s", fmtname);
3072 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
3073 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
3075 seq_show_option(seq, "usrjquota", usr_qf_name);
3077 seq_show_option(seq, "grpjquota", grp_qf_name);
3082 static const char *token2str(int token)
3084 const struct fs_parameter_spec *spec;
3086 for (spec = ext4_param_specs; spec->name != NULL; spec++)
3087 if (spec->opt == token && !spec->type)
3094 * - it's set to a non-default value OR
3095 * - if the per-sb default is different from the global default
3097 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101 struct ext4_super_block *es = sbi->s_es;
3102 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
3103 const struct mount_opts *m;
3104 char sep = nodefs ? '\n' : ',';
3106 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
3107 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
3109 if (sbi->s_sb_block != 1)
3110 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
3112 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
3113 int want_set = m->flags & MOPT_SET;
3114 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
3115 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
3117 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
3118 continue; /* skip if same as the default */
3120 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
3121 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
3122 continue; /* select Opt_noFoo vs Opt_Foo */
3123 SEQ_OPTS_PRINT("%s", token2str(m->token));
3126 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
3127 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
3128 SEQ_OPTS_PRINT("resuid=%u",
3129 from_kuid_munged(&init_user_ns, sbi->s_resuid));
3130 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
3131 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
3132 SEQ_OPTS_PRINT("resgid=%u",
3133 from_kgid_munged(&init_user_ns, sbi->s_resgid));
3134 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
3135 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3136 SEQ_OPTS_PUTS("errors=remount-ro");
3137 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
3138 SEQ_OPTS_PUTS("errors=continue");
3139 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
3140 SEQ_OPTS_PUTS("errors=panic");
3141 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
3142 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
3143 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
3144 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
3145 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
3146 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3147 if (sb->s_flags & SB_I_VERSION)
3148 SEQ_OPTS_PUTS("i_version");
3149 if (nodefs || sbi->s_stripe)
3150 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3151 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3152 (sbi->s_mount_opt ^ def_mount_opt)) {
3153 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3154 SEQ_OPTS_PUTS("data=journal");
3155 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3156 SEQ_OPTS_PUTS("data=ordered");
3157 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3158 SEQ_OPTS_PUTS("data=writeback");
3161 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3162 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3163 sbi->s_inode_readahead_blks);
3165 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3166 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3167 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3168 if (nodefs || sbi->s_max_dir_size_kb)
3169 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3170 if (test_opt(sb, DATA_ERR_ABORT))
3171 SEQ_OPTS_PUTS("data_err=abort");
3173 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3175 if (sb->s_flags & SB_INLINECRYPT)
3176 SEQ_OPTS_PUTS("inlinecrypt");
3178 if (test_opt(sb, DAX_ALWAYS)) {
3180 SEQ_OPTS_PUTS("dax");
3182 SEQ_OPTS_PUTS("dax=always");
3183 } else if (test_opt2(sb, DAX_NEVER)) {
3184 SEQ_OPTS_PUTS("dax=never");
3185 } else if (test_opt2(sb, DAX_INODE)) {
3186 SEQ_OPTS_PUTS("dax=inode");
3188 ext4_show_quota_options(seq, sb);
3192 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3194 return _ext4_show_options(seq, root->d_sb, 0);
3197 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3199 struct super_block *sb = seq->private;
3202 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3203 rc = _ext4_show_options(seq, sb, 1);
3204 seq_puts(seq, "\n");
3208 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3215 ext4_msg(sb, KERN_ERR, "revision level too high, "
3216 "forcing read-only mode");
3222 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3223 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3224 "running e2fsck is recommended");
3225 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3226 ext4_msg(sb, KERN_WARNING,
3227 "warning: mounting fs with errors, "
3228 "running e2fsck is recommended");
3229 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3230 le16_to_cpu(es->s_mnt_count) >=
3231 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3232 ext4_msg(sb, KERN_WARNING,
3233 "warning: maximal mount count reached, "
3234 "running e2fsck is recommended");
3235 else if (le32_to_cpu(es->s_checkinterval) &&
3236 (ext4_get_tstamp(es, s_lastcheck) +
3237 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3238 ext4_msg(sb, KERN_WARNING,
3239 "warning: checktime reached, "
3240 "running e2fsck is recommended");
3241 if (!sbi->s_journal)
3242 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3243 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3244 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3245 le16_add_cpu(&es->s_mnt_count, 1);
3246 ext4_update_tstamp(es, s_mtime);
3247 if (sbi->s_journal) {
3248 ext4_set_feature_journal_needs_recovery(sb);
3249 if (ext4_has_feature_orphan_file(sb))
3250 ext4_set_feature_orphan_present(sb);
3253 err = ext4_commit_super(sb);
3255 if (test_opt(sb, DEBUG))
3256 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3257 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3259 sbi->s_groups_count,
3260 EXT4_BLOCKS_PER_GROUP(sb),
3261 EXT4_INODES_PER_GROUP(sb),
3262 sbi->s_mount_opt, sbi->s_mount_opt2);
3264 cleancache_init_fs(sb);
3268 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3270 struct ext4_sb_info *sbi = EXT4_SB(sb);
3271 struct flex_groups **old_groups, **new_groups;
3274 if (!sbi->s_log_groups_per_flex)
3277 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3278 if (size <= sbi->s_flex_groups_allocated)
3281 new_groups = kvzalloc(roundup_pow_of_two(size *
3282 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3284 ext4_msg(sb, KERN_ERR,
3285 "not enough memory for %d flex group pointers", size);
3288 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3289 new_groups[i] = kvzalloc(roundup_pow_of_two(
3290 sizeof(struct flex_groups)),
3292 if (!new_groups[i]) {
3293 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3294 kvfree(new_groups[j]);
3296 ext4_msg(sb, KERN_ERR,
3297 "not enough memory for %d flex groups", size);
3302 old_groups = rcu_dereference(sbi->s_flex_groups);
3304 memcpy(new_groups, old_groups,
3305 (sbi->s_flex_groups_allocated *
3306 sizeof(struct flex_groups *)));
3308 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3309 sbi->s_flex_groups_allocated = size;
3311 ext4_kvfree_array_rcu(old_groups);
3315 static int ext4_fill_flex_info(struct super_block *sb)
3317 struct ext4_sb_info *sbi = EXT4_SB(sb);
3318 struct ext4_group_desc *gdp = NULL;
3319 struct flex_groups *fg;
3320 ext4_group_t flex_group;
3323 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3324 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3325 sbi->s_log_groups_per_flex = 0;
3329 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3333 for (i = 0; i < sbi->s_groups_count; i++) {
3334 gdp = ext4_get_group_desc(sb, i, NULL);
3336 flex_group = ext4_flex_group(sbi, i);
3337 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3338 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3339 atomic64_add(ext4_free_group_clusters(sb, gdp),
3340 &fg->free_clusters);
3341 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3349 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3350 struct ext4_group_desc *gdp)
3352 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3354 __le32 le_group = cpu_to_le32(block_group);
3355 struct ext4_sb_info *sbi = EXT4_SB(sb);
3357 if (ext4_has_metadata_csum(sbi->s_sb)) {
3358 /* Use new metadata_csum algorithm */
3360 __u16 dummy_csum = 0;
3362 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3364 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3365 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3366 sizeof(dummy_csum));
3367 offset += sizeof(dummy_csum);
3368 if (offset < sbi->s_desc_size)
3369 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3370 sbi->s_desc_size - offset);
3372 crc = csum32 & 0xFFFF;
3376 /* old crc16 code */
3377 if (!ext4_has_feature_gdt_csum(sb))
3380 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3381 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3382 crc = crc16(crc, (__u8 *)gdp, offset);
3383 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3384 /* for checksum of struct ext4_group_desc do the rest...*/
3385 if (ext4_has_feature_64bit(sb) &&
3386 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3387 crc = crc16(crc, (__u8 *)gdp + offset,
3388 le16_to_cpu(sbi->s_es->s_desc_size) -
3392 return cpu_to_le16(crc);
3395 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3396 struct ext4_group_desc *gdp)
3398 if (ext4_has_group_desc_csum(sb) &&
3399 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3405 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3406 struct ext4_group_desc *gdp)
3408 if (!ext4_has_group_desc_csum(sb))
3410 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3413 /* Called at mount-time, super-block is locked */
3414 static int ext4_check_descriptors(struct super_block *sb,
3415 ext4_fsblk_t sb_block,
3416 ext4_group_t *first_not_zeroed)
3418 struct ext4_sb_info *sbi = EXT4_SB(sb);
3419 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3420 ext4_fsblk_t last_block;
3421 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3422 ext4_fsblk_t block_bitmap;
3423 ext4_fsblk_t inode_bitmap;
3424 ext4_fsblk_t inode_table;
3425 int flexbg_flag = 0;
3426 ext4_group_t i, grp = sbi->s_groups_count;
3428 if (ext4_has_feature_flex_bg(sb))
3431 ext4_debug("Checking group descriptors");
3433 for (i = 0; i < sbi->s_groups_count; i++) {
3434 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3436 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3437 last_block = ext4_blocks_count(sbi->s_es) - 1;
3439 last_block = first_block +
3440 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3442 if ((grp == sbi->s_groups_count) &&
3443 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3446 block_bitmap = ext4_block_bitmap(sb, gdp);
3447 if (block_bitmap == sb_block) {
3448 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3449 "Block bitmap for group %u overlaps "
3454 if (block_bitmap >= sb_block + 1 &&
3455 block_bitmap <= last_bg_block) {
3456 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3457 "Block bitmap for group %u overlaps "
3458 "block group descriptors", i);
3462 if (block_bitmap < first_block || block_bitmap > last_block) {
3463 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3464 "Block bitmap for group %u not in group "
3465 "(block %llu)!", i, block_bitmap);
3468 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3469 if (inode_bitmap == sb_block) {
3470 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3471 "Inode bitmap for group %u overlaps "
3476 if (inode_bitmap >= sb_block + 1 &&
3477 inode_bitmap <= last_bg_block) {
3478 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3479 "Inode bitmap for group %u overlaps "
3480 "block group descriptors", i);
3484 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3485 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3486 "Inode bitmap for group %u not in group "
3487 "(block %llu)!", i, inode_bitmap);
3490 inode_table = ext4_inode_table(sb, gdp);
3491 if (inode_table == sb_block) {
3492 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3493 "Inode table for group %u overlaps "
3498 if (inode_table >= sb_block + 1 &&
3499 inode_table <= last_bg_block) {
3500 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3501 "Inode table for group %u overlaps "
3502 "block group descriptors", i);
3506 if (inode_table < first_block ||
3507 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3508 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3509 "Inode table for group %u not in group "
3510 "(block %llu)!", i, inode_table);
3513 ext4_lock_group(sb, i);
3514 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3515 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3516 "Checksum for group %u failed (%u!=%u)",
3517 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3518 gdp)), le16_to_cpu(gdp->bg_checksum));
3519 if (!sb_rdonly(sb)) {
3520 ext4_unlock_group(sb, i);
3524 ext4_unlock_group(sb, i);
3526 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3528 if (NULL != first_not_zeroed)
3529 *first_not_zeroed = grp;
3534 * Maximal extent format file size.
3535 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3536 * extent format containers, within a sector_t, and within i_blocks
3537 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3538 * so that won't be a limiting factor.
3540 * However there is other limiting factor. We do store extents in the form
3541 * of starting block and length, hence the resulting length of the extent
3542 * covering maximum file size must fit into on-disk format containers as
3543 * well. Given that length is always by 1 unit bigger than max unit (because
3544 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3546 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3548 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3551 loff_t upper_limit = MAX_LFS_FILESIZE;
3553 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3555 if (!has_huge_files) {
3556 upper_limit = (1LL << 32) - 1;
3558 /* total blocks in file system block size */
3559 upper_limit >>= (blkbits - 9);
3560 upper_limit <<= blkbits;
3564 * 32-bit extent-start container, ee_block. We lower the maxbytes
3565 * by one fs block, so ee_len can cover the extent of maximum file
3568 res = (1LL << 32) - 1;
3571 /* Sanity check against vm- & vfs- imposed limits */
3572 if (res > upper_limit)
3579 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3580 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3581 * We need to be 1 filesystem block less than the 2^48 sector limit.
3583 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3585 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS;
3589 * This is calculated to be the largest file size for a dense, block
3590 * mapped file such that the file's total number of 512-byte sectors,
3591 * including data and all indirect blocks, does not exceed (2^48 - 1).
3593 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3594 * number of 512-byte sectors of the file.
3596 if (!has_huge_files) {
3598 * !has_huge_files or implies that the inode i_block field
3599 * represents total file blocks in 2^32 512-byte sectors ==
3600 * size of vfs inode i_blocks * 8
3602 upper_limit = (1LL << 32) - 1;
3604 /* total blocks in file system block size */
3605 upper_limit >>= (bits - 9);
3609 * We use 48 bit ext4_inode i_blocks
3610 * With EXT4_HUGE_FILE_FL set the i_blocks
3611 * represent total number of blocks in
3612 * file system block size
3614 upper_limit = (1LL << 48) - 1;
3618 /* indirect blocks */
3620 /* double indirect blocks */
3621 meta_blocks += 1 + (1LL << (bits-2));
3622 /* tripple indirect blocks */
3623 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3625 upper_limit -= meta_blocks;
3626 upper_limit <<= bits;
3628 res += 1LL << (bits-2);
3629 res += 1LL << (2*(bits-2));
3630 res += 1LL << (3*(bits-2));
3632 if (res > upper_limit)
3635 if (res > MAX_LFS_FILESIZE)
3636 res = MAX_LFS_FILESIZE;
3641 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3642 ext4_fsblk_t logical_sb_block, int nr)
3644 struct ext4_sb_info *sbi = EXT4_SB(sb);
3645 ext4_group_t bg, first_meta_bg;
3648 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3650 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3651 return logical_sb_block + nr + 1;
3652 bg = sbi->s_desc_per_block * nr;
3653 if (ext4_bg_has_super(sb, bg))
3657 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3658 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3659 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3662 if (sb->s_blocksize == 1024 && nr == 0 &&
3663 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3666 return (has_super + ext4_group_first_block_no(sb, bg));
3670 * ext4_get_stripe_size: Get the stripe size.
3671 * @sbi: In memory super block info
3673 * If we have specified it via mount option, then
3674 * use the mount option value. If the value specified at mount time is
3675 * greater than the blocks per group use the super block value.
3676 * If the super block value is greater than blocks per group return 0.
3677 * Allocator needs it be less than blocks per group.
3680 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3682 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3683 unsigned long stripe_width =
3684 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3687 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3688 ret = sbi->s_stripe;
3689 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3691 else if (stride && stride <= sbi->s_blocks_per_group)
3697 * If the stripe width is 1, this makes no sense and
3698 * we set it to 0 to turn off stripe handling code.
3707 * Check whether this filesystem can be mounted based on
3708 * the features present and the RDONLY/RDWR mount requested.
3709 * Returns 1 if this filesystem can be mounted as requested,
3710 * 0 if it cannot be.
3712 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3714 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3715 ext4_msg(sb, KERN_ERR,
3716 "Couldn't mount because of "
3717 "unsupported optional features (%x)",
3718 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3719 ~EXT4_FEATURE_INCOMPAT_SUPP));
3723 #ifndef CONFIG_UNICODE
3724 if (ext4_has_feature_casefold(sb)) {
3725 ext4_msg(sb, KERN_ERR,
3726 "Filesystem with casefold feature cannot be "
3727 "mounted without CONFIG_UNICODE");
3735 if (ext4_has_feature_readonly(sb)) {
3736 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3737 sb->s_flags |= SB_RDONLY;
3741 /* Check that feature set is OK for a read-write mount */
3742 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3743 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3744 "unsupported optional features (%x)",
3745 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3746 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3749 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3750 ext4_msg(sb, KERN_ERR,
3751 "Can't support bigalloc feature without "
3752 "extents feature\n");
3756 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3757 if (!readonly && (ext4_has_feature_quota(sb) ||
3758 ext4_has_feature_project(sb))) {
3759 ext4_msg(sb, KERN_ERR,
3760 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3763 #endif /* CONFIG_QUOTA */
3768 * This function is called once a day if we have errors logged
3769 * on the file system
3771 static void print_daily_error_info(struct timer_list *t)
3773 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3774 struct super_block *sb = sbi->s_sb;
3775 struct ext4_super_block *es = sbi->s_es;
3777 if (es->s_error_count)
3778 /* fsck newer than v1.41.13 is needed to clean this condition. */
3779 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3780 le32_to_cpu(es->s_error_count));
3781 if (es->s_first_error_time) {
3782 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3784 ext4_get_tstamp(es, s_first_error_time),
3785 (int) sizeof(es->s_first_error_func),
3786 es->s_first_error_func,
3787 le32_to_cpu(es->s_first_error_line));
3788 if (es->s_first_error_ino)
3789 printk(KERN_CONT ": inode %u",
3790 le32_to_cpu(es->s_first_error_ino));
3791 if (es->s_first_error_block)
3792 printk(KERN_CONT ": block %llu", (unsigned long long)
3793 le64_to_cpu(es->s_first_error_block));
3794 printk(KERN_CONT "\n");
3796 if (es->s_last_error_time) {
3797 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3799 ext4_get_tstamp(es, s_last_error_time),
3800 (int) sizeof(es->s_last_error_func),
3801 es->s_last_error_func,
3802 le32_to_cpu(es->s_last_error_line));
3803 if (es->s_last_error_ino)
3804 printk(KERN_CONT ": inode %u",
3805 le32_to_cpu(es->s_last_error_ino));
3806 if (es->s_last_error_block)
3807 printk(KERN_CONT ": block %llu", (unsigned long long)
3808 le64_to_cpu(es->s_last_error_block));
3809 printk(KERN_CONT "\n");
3811 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3814 /* Find next suitable group and run ext4_init_inode_table */
3815 static int ext4_run_li_request(struct ext4_li_request *elr)
3817 struct ext4_group_desc *gdp = NULL;
3818 struct super_block *sb = elr->lr_super;
3819 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3820 ext4_group_t group = elr->lr_next_group;
3821 unsigned int prefetch_ios = 0;
3825 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3826 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3827 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3829 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3831 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3833 if (group >= elr->lr_next_group) {
3835 if (elr->lr_first_not_zeroed != ngroups &&
3836 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3837 elr->lr_next_group = elr->lr_first_not_zeroed;
3838 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3845 for (; group < ngroups; group++) {
3846 gdp = ext4_get_group_desc(sb, group, NULL);
3852 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3856 if (group >= ngroups)
3860 start_time = ktime_get_real_ns();
3861 ret = ext4_init_inode_table(sb, group,
3862 elr->lr_timeout ? 0 : 1);
3863 trace_ext4_lazy_itable_init(sb, group);
3864 if (elr->lr_timeout == 0) {
3865 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3866 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3868 elr->lr_next_sched = jiffies + elr->lr_timeout;
3869 elr->lr_next_group = group + 1;
3875 * Remove lr_request from the list_request and free the
3876 * request structure. Should be called with li_list_mtx held
3878 static void ext4_remove_li_request(struct ext4_li_request *elr)
3883 list_del(&elr->lr_request);
3884 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3888 static void ext4_unregister_li_request(struct super_block *sb)
3890 mutex_lock(&ext4_li_mtx);
3891 if (!ext4_li_info) {
3892 mutex_unlock(&ext4_li_mtx);
3896 mutex_lock(&ext4_li_info->li_list_mtx);
3897 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3898 mutex_unlock(&ext4_li_info->li_list_mtx);
3899 mutex_unlock(&ext4_li_mtx);
3902 static struct task_struct *ext4_lazyinit_task;
3905 * This is the function where ext4lazyinit thread lives. It walks
3906 * through the request list searching for next scheduled filesystem.
3907 * When such a fs is found, run the lazy initialization request
3908 * (ext4_rn_li_request) and keep track of the time spend in this
3909 * function. Based on that time we compute next schedule time of
3910 * the request. When walking through the list is complete, compute
3911 * next waking time and put itself into sleep.
3913 static int ext4_lazyinit_thread(void *arg)
3915 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3916 struct list_head *pos, *n;
3917 struct ext4_li_request *elr;
3918 unsigned long next_wakeup, cur;
3920 BUG_ON(NULL == eli);
3924 next_wakeup = MAX_JIFFY_OFFSET;
3926 mutex_lock(&eli->li_list_mtx);
3927 if (list_empty(&eli->li_request_list)) {
3928 mutex_unlock(&eli->li_list_mtx);
3931 list_for_each_safe(pos, n, &eli->li_request_list) {
3934 elr = list_entry(pos, struct ext4_li_request,
3937 if (time_before(jiffies, elr->lr_next_sched)) {
3938 if (time_before(elr->lr_next_sched, next_wakeup))
3939 next_wakeup = elr->lr_next_sched;
3942 if (down_read_trylock(&elr->lr_super->s_umount)) {
3943 if (sb_start_write_trylock(elr->lr_super)) {
3946 * We hold sb->s_umount, sb can not
3947 * be removed from the list, it is
3948 * now safe to drop li_list_mtx
3950 mutex_unlock(&eli->li_list_mtx);
3951 err = ext4_run_li_request(elr);
3952 sb_end_write(elr->lr_super);
3953 mutex_lock(&eli->li_list_mtx);
3956 up_read((&elr->lr_super->s_umount));
3958 /* error, remove the lazy_init job */
3960 ext4_remove_li_request(elr);
3964 elr->lr_next_sched = jiffies +
3966 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3968 if (time_before(elr->lr_next_sched, next_wakeup))
3969 next_wakeup = elr->lr_next_sched;
3971 mutex_unlock(&eli->li_list_mtx);
3976 if ((time_after_eq(cur, next_wakeup)) ||
3977 (MAX_JIFFY_OFFSET == next_wakeup)) {
3982 schedule_timeout_interruptible(next_wakeup - cur);
3984 if (kthread_should_stop()) {
3985 ext4_clear_request_list();
3992 * It looks like the request list is empty, but we need
3993 * to check it under the li_list_mtx lock, to prevent any
3994 * additions into it, and of course we should lock ext4_li_mtx
3995 * to atomically free the list and ext4_li_info, because at
3996 * this point another ext4 filesystem could be registering
3999 mutex_lock(&ext4_li_mtx);
4000 mutex_lock(&eli->li_list_mtx);
4001 if (!list_empty(&eli->li_request_list)) {
4002 mutex_unlock(&eli->li_list_mtx);
4003 mutex_unlock(&ext4_li_mtx);
4006 mutex_unlock(&eli->li_list_mtx);
4007 kfree(ext4_li_info);
4008 ext4_li_info = NULL;
4009 mutex_unlock(&ext4_li_mtx);
4014 static void ext4_clear_request_list(void)
4016 struct list_head *pos, *n;
4017 struct ext4_li_request *elr;
4019 mutex_lock(&ext4_li_info->li_list_mtx);
4020 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
4021 elr = list_entry(pos, struct ext4_li_request,
4023 ext4_remove_li_request(elr);
4025 mutex_unlock(&ext4_li_info->li_list_mtx);
4028 static int ext4_run_lazyinit_thread(void)
4030 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
4031 ext4_li_info, "ext4lazyinit");
4032 if (IS_ERR(ext4_lazyinit_task)) {
4033 int err = PTR_ERR(ext4_lazyinit_task);
4034 ext4_clear_request_list();
4035 kfree(ext4_li_info);
4036 ext4_li_info = NULL;
4037 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
4038 "initialization thread\n",
4042 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
4047 * Check whether it make sense to run itable init. thread or not.
4048 * If there is at least one uninitialized inode table, return
4049 * corresponding group number, else the loop goes through all
4050 * groups and return total number of groups.
4052 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
4054 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
4055 struct ext4_group_desc *gdp = NULL;
4057 if (!ext4_has_group_desc_csum(sb))
4060 for (group = 0; group < ngroups; group++) {
4061 gdp = ext4_get_group_desc(sb, group, NULL);
4065 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
4072 static int ext4_li_info_new(void)
4074 struct ext4_lazy_init *eli = NULL;
4076 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
4080 INIT_LIST_HEAD(&eli->li_request_list);
4081 mutex_init(&eli->li_list_mtx);
4083 eli->li_state |= EXT4_LAZYINIT_QUIT;
4090 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
4093 struct ext4_li_request *elr;
4095 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
4100 elr->lr_first_not_zeroed = start;
4101 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
4102 elr->lr_mode = EXT4_LI_MODE_ITABLE;
4103 elr->lr_next_group = start;
4105 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
4109 * Randomize first schedule time of the request to
4110 * spread the inode table initialization requests
4113 elr->lr_next_sched = jiffies + (prandom_u32() %
4114 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
4118 int ext4_register_li_request(struct super_block *sb,
4119 ext4_group_t first_not_zeroed)
4121 struct ext4_sb_info *sbi = EXT4_SB(sb);
4122 struct ext4_li_request *elr = NULL;
4123 ext4_group_t ngroups = sbi->s_groups_count;
4126 mutex_lock(&ext4_li_mtx);
4127 if (sbi->s_li_request != NULL) {
4129 * Reset timeout so it can be computed again, because
4130 * s_li_wait_mult might have changed.
4132 sbi->s_li_request->lr_timeout = 0;
4136 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4137 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
4138 !test_opt(sb, INIT_INODE_TABLE)))
4141 elr = ext4_li_request_new(sb, first_not_zeroed);
4147 if (NULL == ext4_li_info) {
4148 ret = ext4_li_info_new();
4153 mutex_lock(&ext4_li_info->li_list_mtx);
4154 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4155 mutex_unlock(&ext4_li_info->li_list_mtx);
4157 sbi->s_li_request = elr;
4159 * set elr to NULL here since it has been inserted to
4160 * the request_list and the removal and free of it is
4161 * handled by ext4_clear_request_list from now on.
4165 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4166 ret = ext4_run_lazyinit_thread();
4171 mutex_unlock(&ext4_li_mtx);
4178 * We do not need to lock anything since this is called on
4181 static void ext4_destroy_lazyinit_thread(void)
4184 * If thread exited earlier
4185 * there's nothing to be done.
4187 if (!ext4_li_info || !ext4_lazyinit_task)
4190 kthread_stop(ext4_lazyinit_task);
4193 static int set_journal_csum_feature_set(struct super_block *sb)
4196 int compat, incompat;
4197 struct ext4_sb_info *sbi = EXT4_SB(sb);
4199 if (ext4_has_metadata_csum(sb)) {
4200 /* journal checksum v3 */
4202 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4204 /* journal checksum v1 */
4205 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4209 jbd2_journal_clear_features(sbi->s_journal,
4210 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4211 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4212 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4213 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4214 ret = jbd2_journal_set_features(sbi->s_journal,
4216 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4218 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4219 ret = jbd2_journal_set_features(sbi->s_journal,
4222 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4223 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4225 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4226 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4233 * Note: calculating the overhead so we can be compatible with
4234 * historical BSD practice is quite difficult in the face of
4235 * clusters/bigalloc. This is because multiple metadata blocks from
4236 * different block group can end up in the same allocation cluster.
4237 * Calculating the exact overhead in the face of clustered allocation
4238 * requires either O(all block bitmaps) in memory or O(number of block
4239 * groups**2) in time. We will still calculate the superblock for
4240 * older file systems --- and if we come across with a bigalloc file
4241 * system with zero in s_overhead_clusters the estimate will be close to
4242 * correct especially for very large cluster sizes --- but for newer
4243 * file systems, it's better to calculate this figure once at mkfs
4244 * time, and store it in the superblock. If the superblock value is
4245 * present (even for non-bigalloc file systems), we will use it.
4247 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4250 struct ext4_sb_info *sbi = EXT4_SB(sb);
4251 struct ext4_group_desc *gdp;
4252 ext4_fsblk_t first_block, last_block, b;
4253 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4254 int s, j, count = 0;
4256 if (!ext4_has_feature_bigalloc(sb))
4257 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
4258 sbi->s_itb_per_group + 2);
4260 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4261 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4262 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4263 for (i = 0; i < ngroups; i++) {
4264 gdp = ext4_get_group_desc(sb, i, NULL);
4265 b = ext4_block_bitmap(sb, gdp);
4266 if (b >= first_block && b <= last_block) {
4267 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4270 b = ext4_inode_bitmap(sb, gdp);
4271 if (b >= first_block && b <= last_block) {
4272 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4275 b = ext4_inode_table(sb, gdp);
4276 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4277 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4278 int c = EXT4_B2C(sbi, b - first_block);
4279 ext4_set_bit(c, buf);
4285 if (ext4_bg_has_super(sb, grp)) {
4286 ext4_set_bit(s++, buf);
4289 j = ext4_bg_num_gdb(sb, grp);
4290 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4291 ext4_error(sb, "Invalid number of block group "
4292 "descriptor blocks: %d", j);
4293 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4297 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4301 return EXT4_CLUSTERS_PER_GROUP(sb) -
4302 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4306 * Compute the overhead and stash it in sbi->s_overhead
4308 int ext4_calculate_overhead(struct super_block *sb)
4310 struct ext4_sb_info *sbi = EXT4_SB(sb);
4311 struct ext4_super_block *es = sbi->s_es;
4312 struct inode *j_inode;
4313 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4314 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4315 ext4_fsblk_t overhead = 0;
4316 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4322 * Compute the overhead (FS structures). This is constant
4323 * for a given filesystem unless the number of block groups
4324 * changes so we cache the previous value until it does.
4328 * All of the blocks before first_data_block are overhead
4330 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4333 * Add the overhead found in each block group
4335 for (i = 0; i < ngroups; i++) {
4338 blks = count_overhead(sb, i, buf);
4341 memset(buf, 0, PAGE_SIZE);
4346 * Add the internal journal blocks whether the journal has been
4349 if (sbi->s_journal && !sbi->s_journal_bdev)
4350 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4351 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4352 /* j_inum for internal journal is non-zero */
4353 j_inode = ext4_get_journal_inode(sb, j_inum);
4355 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4356 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4359 ext4_msg(sb, KERN_ERR, "can't get journal size");
4362 sbi->s_overhead = overhead;
4364 free_page((unsigned long) buf);
4368 static void ext4_set_resv_clusters(struct super_block *sb)
4370 ext4_fsblk_t resv_clusters;
4371 struct ext4_sb_info *sbi = EXT4_SB(sb);
4374 * There's no need to reserve anything when we aren't using extents.
4375 * The space estimates are exact, there are no unwritten extents,
4376 * hole punching doesn't need new metadata... This is needed especially
4377 * to keep ext2/3 backward compatibility.
4379 if (!ext4_has_feature_extents(sb))
4382 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4383 * This should cover the situations where we can not afford to run
4384 * out of space like for example punch hole, or converting
4385 * unwritten extents in delalloc path. In most cases such
4386 * allocation would require 1, or 2 blocks, higher numbers are
4389 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4390 sbi->s_cluster_bits);
4392 do_div(resv_clusters, 50);
4393 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4395 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4398 static const char *ext4_quota_mode(struct super_block *sb)
4401 if (!ext4_quota_capable(sb))
4404 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4405 return "journalled";
4413 static void ext4_setup_csum_trigger(struct super_block *sb,
4414 enum ext4_journal_trigger_type type,
4416 struct jbd2_buffer_trigger_type *type,
4417 struct buffer_head *bh,
4421 struct ext4_sb_info *sbi = EXT4_SB(sb);
4423 sbi->s_journal_triggers[type].sb = sb;
4424 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4427 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4432 kfree(sbi->s_blockgroup_lock);
4433 fs_put_dax(sbi->s_daxdev);
4437 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4439 struct ext4_sb_info *sbi;
4441 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4445 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev);
4447 sbi->s_blockgroup_lock =
4448 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4450 if (!sbi->s_blockgroup_lock)
4453 sb->s_fs_info = sbi;
4457 fs_put_dax(sbi->s_daxdev);
4462 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb,
4465 struct buffer_head *bh, **group_desc;
4466 struct ext4_super_block *es = NULL;
4467 struct ext4_sb_info *sbi = EXT4_SB(sb);
4468 struct flex_groups **flex_groups;
4470 ext4_fsblk_t logical_sb_block;
4471 unsigned long offset = 0;
4472 unsigned long def_mount_opts;
4475 int blocksize, clustersize;
4476 unsigned int db_count;
4478 int needs_recovery, has_huge_files;
4481 ext4_group_t first_not_zeroed;
4482 struct ext4_fs_context *ctx = fc->fs_private;
4484 /* Set defaults for the variables that will be set during parsing */
4485 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4486 ctx->mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
4488 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4489 sbi->s_sectors_written_start =
4490 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4492 /* -EINVAL is default */
4494 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4496 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4501 * The ext4 superblock will not be buffer aligned for other than 1kB
4502 * block sizes. We need to calculate the offset from buffer start.
4504 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4505 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4506 offset = do_div(logical_sb_block, blocksize);
4508 logical_sb_block = sbi->s_sb_block;
4511 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4513 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4518 * Note: s_es must be initialized as soon as possible because
4519 * some ext4 macro-instructions depend on its value
4521 es = (struct ext4_super_block *) (bh->b_data + offset);
4523 sb->s_magic = le16_to_cpu(es->s_magic);
4524 if (sb->s_magic != EXT4_SUPER_MAGIC)
4526 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4528 /* Warn if metadata_csum and gdt_csum are both set. */
4529 if (ext4_has_feature_metadata_csum(sb) &&
4530 ext4_has_feature_gdt_csum(sb))
4531 ext4_warning(sb, "metadata_csum and uninit_bg are "
4532 "redundant flags; please run fsck.");
4534 /* Check for a known checksum algorithm */
4535 if (!ext4_verify_csum_type(sb, es)) {
4536 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4537 "unknown checksum algorithm.");
4541 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4542 ext4_orphan_file_block_trigger);
4544 /* Load the checksum driver */
4545 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4546 if (IS_ERR(sbi->s_chksum_driver)) {
4547 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4548 ret = PTR_ERR(sbi->s_chksum_driver);
4549 sbi->s_chksum_driver = NULL;
4553 /* Check superblock checksum */
4554 if (!ext4_superblock_csum_verify(sb, es)) {
4555 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4556 "invalid superblock checksum. Run e2fsck?");
4562 /* Precompute checksum seed for all metadata */
4563 if (ext4_has_feature_csum_seed(sb))
4564 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4565 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4566 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4567 sizeof(es->s_uuid));
4569 /* Set defaults before we parse the mount options */
4570 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4571 set_opt(sb, INIT_INODE_TABLE);
4572 if (def_mount_opts & EXT4_DEFM_DEBUG)
4574 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4576 if (def_mount_opts & EXT4_DEFM_UID16)
4577 set_opt(sb, NO_UID32);
4578 /* xattr user namespace & acls are now defaulted on */
4579 set_opt(sb, XATTR_USER);
4580 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4581 set_opt(sb, POSIX_ACL);
4583 if (ext4_has_feature_fast_commit(sb))
4584 set_opt2(sb, JOURNAL_FAST_COMMIT);
4585 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4586 if (ext4_has_metadata_csum(sb))
4587 set_opt(sb, JOURNAL_CHECKSUM);
4589 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4590 set_opt(sb, JOURNAL_DATA);
4591 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4592 set_opt(sb, ORDERED_DATA);
4593 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4594 set_opt(sb, WRITEBACK_DATA);
4596 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4597 set_opt(sb, ERRORS_PANIC);
4598 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4599 set_opt(sb, ERRORS_CONT);
4601 set_opt(sb, ERRORS_RO);
4602 /* block_validity enabled by default; disable with noblock_validity */
4603 set_opt(sb, BLOCK_VALIDITY);
4604 if (def_mount_opts & EXT4_DEFM_DISCARD)
4605 set_opt(sb, DISCARD);
4607 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4608 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4609 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4610 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4611 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4613 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4614 set_opt(sb, BARRIER);
4617 * enable delayed allocation by default
4618 * Use -o nodelalloc to turn it off
4620 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4621 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4622 set_opt(sb, DELALLOC);
4625 * set default s_li_wait_mult for lazyinit, for the case there is
4626 * no mount option specified.
4628 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4630 if (le32_to_cpu(es->s_log_block_size) >
4631 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4632 ext4_msg(sb, KERN_ERR,
4633 "Invalid log block size: %u",
4634 le32_to_cpu(es->s_log_block_size));
4637 if (le32_to_cpu(es->s_log_cluster_size) >
4638 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4639 ext4_msg(sb, KERN_ERR,
4640 "Invalid log cluster size: %u",
4641 le32_to_cpu(es->s_log_cluster_size));
4645 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4647 if (blocksize == PAGE_SIZE)
4648 set_opt(sb, DIOREAD_NOLOCK);
4650 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4651 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4652 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4654 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4655 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4656 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4657 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4661 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4662 (!is_power_of_2(sbi->s_inode_size)) ||
4663 (sbi->s_inode_size > blocksize)) {
4664 ext4_msg(sb, KERN_ERR,
4665 "unsupported inode size: %d",
4667 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4671 * i_atime_extra is the last extra field available for
4672 * [acm]times in struct ext4_inode. Checking for that
4673 * field should suffice to ensure we have extra space
4676 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4677 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4678 sb->s_time_gran = 1;
4679 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4681 sb->s_time_gran = NSEC_PER_SEC;
4682 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4684 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4686 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4687 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4688 EXT4_GOOD_OLD_INODE_SIZE;
4689 if (ext4_has_feature_extra_isize(sb)) {
4690 unsigned v, max = (sbi->s_inode_size -
4691 EXT4_GOOD_OLD_INODE_SIZE);
4693 v = le16_to_cpu(es->s_want_extra_isize);
4695 ext4_msg(sb, KERN_ERR,
4696 "bad s_want_extra_isize: %d", v);
4699 if (sbi->s_want_extra_isize < v)
4700 sbi->s_want_extra_isize = v;
4702 v = le16_to_cpu(es->s_min_extra_isize);
4704 ext4_msg(sb, KERN_ERR,
4705 "bad s_min_extra_isize: %d", v);
4708 if (sbi->s_want_extra_isize < v)
4709 sbi->s_want_extra_isize = v;
4713 err = parse_apply_sb_mount_options(sb, ctx);
4717 sbi->s_def_mount_opt = sbi->s_mount_opt;
4719 err = ext4_check_opt_consistency(fc, sb);
4723 err = ext4_apply_options(fc, sb);
4727 #ifdef CONFIG_UNICODE
4728 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4729 const struct ext4_sb_encodings *encoding_info;
4730 struct unicode_map *encoding;
4731 __u16 encoding_flags;
4733 if (ext4_sb_read_encoding(es, &encoding_info,
4735 ext4_msg(sb, KERN_ERR,
4736 "Encoding requested by superblock is unknown");
4740 encoding = utf8_load(encoding_info->version);
4741 if (IS_ERR(encoding)) {
4742 ext4_msg(sb, KERN_ERR,
4743 "can't mount with superblock charset: %s-%s "
4744 "not supported by the kernel. flags: 0x%x.",
4745 encoding_info->name, encoding_info->version,
4749 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4750 "%s-%s with flags 0x%hx", encoding_info->name,
4751 encoding_info->version?:"\b", encoding_flags);
4753 sb->s_encoding = encoding;
4754 sb->s_encoding_flags = encoding_flags;
4758 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4759 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4760 /* can't mount with both data=journal and dioread_nolock. */
4761 clear_opt(sb, DIOREAD_NOLOCK);
4762 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4763 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4764 ext4_msg(sb, KERN_ERR, "can't mount with "
4765 "both data=journal and delalloc");
4768 if (test_opt(sb, DAX_ALWAYS)) {
4769 ext4_msg(sb, KERN_ERR, "can't mount with "
4770 "both data=journal and dax");
4773 if (ext4_has_feature_encrypt(sb)) {
4774 ext4_msg(sb, KERN_WARNING,
4775 "encrypted files will use data=ordered "
4776 "instead of data journaling mode");
4778 if (test_opt(sb, DELALLOC))
4779 clear_opt(sb, DELALLOC);
4781 sb->s_iflags |= SB_I_CGROUPWB;
4784 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4785 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4787 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4788 (ext4_has_compat_features(sb) ||
4789 ext4_has_ro_compat_features(sb) ||
4790 ext4_has_incompat_features(sb)))
4791 ext4_msg(sb, KERN_WARNING,
4792 "feature flags set on rev 0 fs, "
4793 "running e2fsck is recommended");
4795 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4796 set_opt2(sb, HURD_COMPAT);
4797 if (ext4_has_feature_64bit(sb)) {
4798 ext4_msg(sb, KERN_ERR,
4799 "The Hurd can't support 64-bit file systems");
4804 * ea_inode feature uses l_i_version field which is not
4805 * available in HURD_COMPAT mode.
4807 if (ext4_has_feature_ea_inode(sb)) {
4808 ext4_msg(sb, KERN_ERR,
4809 "ea_inode feature is not supported for Hurd");
4814 if (IS_EXT2_SB(sb)) {
4815 if (ext2_feature_set_ok(sb))
4816 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4817 "using the ext4 subsystem");
4820 * If we're probing be silent, if this looks like
4821 * it's actually an ext[34] filesystem.
4823 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4825 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4826 "to feature incompatibilities");
4831 if (IS_EXT3_SB(sb)) {
4832 if (ext3_feature_set_ok(sb))
4833 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4834 "using the ext4 subsystem");
4837 * If we're probing be silent, if this looks like
4838 * it's actually an ext4 filesystem.
4840 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4842 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4843 "to feature incompatibilities");
4849 * Check feature flags regardless of the revision level, since we
4850 * previously didn't change the revision level when setting the flags,
4851 * so there is a chance incompat flags are set on a rev 0 filesystem.
4853 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4856 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4857 ext4_msg(sb, KERN_ERR,
4858 "Number of reserved GDT blocks insanely large: %d",
4859 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4863 if (dax_supported(sbi->s_daxdev, sb->s_bdev, blocksize, 0,
4864 bdev_nr_sectors(sb->s_bdev)))
4865 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4867 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4868 if (ext4_has_feature_inline_data(sb)) {
4869 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4870 " that may contain inline data");
4873 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4874 ext4_msg(sb, KERN_ERR,
4875 "DAX unsupported by block device.");
4880 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4881 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4882 es->s_encryption_level);
4886 if (sb->s_blocksize != blocksize) {
4888 * bh must be released before kill_bdev(), otherwise
4889 * it won't be freed and its page also. kill_bdev()
4890 * is called by sb_set_blocksize().
4893 /* Validate the filesystem blocksize */
4894 if (!sb_set_blocksize(sb, blocksize)) {
4895 ext4_msg(sb, KERN_ERR, "bad block size %d",
4901 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4902 offset = do_div(logical_sb_block, blocksize);
4903 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4905 ext4_msg(sb, KERN_ERR,
4906 "Can't read superblock on 2nd try");
4911 es = (struct ext4_super_block *)(bh->b_data + offset);
4913 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4914 ext4_msg(sb, KERN_ERR,
4915 "Magic mismatch, very weird!");
4920 has_huge_files = ext4_has_feature_huge_file(sb);
4921 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4923 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4925 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4926 if (ext4_has_feature_64bit(sb)) {
4927 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4928 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4929 !is_power_of_2(sbi->s_desc_size)) {
4930 ext4_msg(sb, KERN_ERR,
4931 "unsupported descriptor size %lu",
4936 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4938 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4939 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4941 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4942 if (sbi->s_inodes_per_block == 0)
4944 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4945 sbi->s_inodes_per_group > blocksize * 8) {
4946 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4947 sbi->s_inodes_per_group);
4950 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4951 sbi->s_inodes_per_block;
4952 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4954 sbi->s_mount_state = le16_to_cpu(es->s_state);
4955 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4956 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4958 for (i = 0; i < 4; i++)
4959 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4960 sbi->s_def_hash_version = es->s_def_hash_version;
4961 if (ext4_has_feature_dir_index(sb)) {
4962 i = le32_to_cpu(es->s_flags);
4963 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4964 sbi->s_hash_unsigned = 3;
4965 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4966 #ifdef __CHAR_UNSIGNED__
4969 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4970 sbi->s_hash_unsigned = 3;
4974 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4979 /* Handle clustersize */
4980 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4981 if (ext4_has_feature_bigalloc(sb)) {
4982 if (clustersize < blocksize) {
4983 ext4_msg(sb, KERN_ERR,
4984 "cluster size (%d) smaller than "
4985 "block size (%d)", clustersize, blocksize);
4988 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4989 le32_to_cpu(es->s_log_block_size);
4990 sbi->s_clusters_per_group =
4991 le32_to_cpu(es->s_clusters_per_group);
4992 if (sbi->s_clusters_per_group > blocksize * 8) {
4993 ext4_msg(sb, KERN_ERR,
4994 "#clusters per group too big: %lu",
4995 sbi->s_clusters_per_group);
4998 if (sbi->s_blocks_per_group !=
4999 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
5000 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
5001 "clusters per group (%lu) inconsistent",
5002 sbi->s_blocks_per_group,
5003 sbi->s_clusters_per_group);
5007 if (clustersize != blocksize) {
5008 ext4_msg(sb, KERN_ERR,
5009 "fragment/cluster size (%d) != "
5010 "block size (%d)", clustersize, blocksize);
5013 if (sbi->s_blocks_per_group > blocksize * 8) {
5014 ext4_msg(sb, KERN_ERR,
5015 "#blocks per group too big: %lu",
5016 sbi->s_blocks_per_group);
5019 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
5020 sbi->s_cluster_bits = 0;
5022 sbi->s_cluster_ratio = clustersize / blocksize;
5024 /* Do we have standard group size of clustersize * 8 blocks ? */
5025 if (sbi->s_blocks_per_group == clustersize << 3)
5026 set_opt2(sb, STD_GROUP_SIZE);
5029 * Test whether we have more sectors than will fit in sector_t,
5030 * and whether the max offset is addressable by the page cache.
5032 err = generic_check_addressable(sb->s_blocksize_bits,
5033 ext4_blocks_count(es));
5035 ext4_msg(sb, KERN_ERR, "filesystem"
5036 " too large to mount safely on this system");
5040 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
5043 /* check blocks count against device size */
5044 blocks_count = sb_bdev_nr_blocks(sb);
5045 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
5046 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
5047 "exceeds size of device (%llu blocks)",
5048 ext4_blocks_count(es), blocks_count);
5053 * It makes no sense for the first data block to be beyond the end
5054 * of the filesystem.
5056 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5057 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
5058 "block %u is beyond end of filesystem (%llu)",
5059 le32_to_cpu(es->s_first_data_block),
5060 ext4_blocks_count(es));
5063 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
5064 (sbi->s_cluster_ratio == 1)) {
5065 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
5066 "block is 0 with a 1k block and cluster size");
5070 blocks_count = (ext4_blocks_count(es) -
5071 le32_to_cpu(es->s_first_data_block) +
5072 EXT4_BLOCKS_PER_GROUP(sb) - 1);
5073 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
5074 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
5075 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
5076 "(block count %llu, first data block %u, "
5077 "blocks per group %lu)", blocks_count,
5078 ext4_blocks_count(es),
5079 le32_to_cpu(es->s_first_data_block),
5080 EXT4_BLOCKS_PER_GROUP(sb));
5083 sbi->s_groups_count = blocks_count;
5084 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
5085 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
5086 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
5087 le32_to_cpu(es->s_inodes_count)) {
5088 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
5089 le32_to_cpu(es->s_inodes_count),
5090 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
5094 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
5095 EXT4_DESC_PER_BLOCK(sb);
5096 if (ext4_has_feature_meta_bg(sb)) {
5097 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
5098 ext4_msg(sb, KERN_WARNING,
5099 "first meta block group too large: %u "
5100 "(group descriptor block count %u)",
5101 le32_to_cpu(es->s_first_meta_bg), db_count);
5105 rcu_assign_pointer(sbi->s_group_desc,
5106 kvmalloc_array(db_count,
5107 sizeof(struct buffer_head *),
5109 if (sbi->s_group_desc == NULL) {
5110 ext4_msg(sb, KERN_ERR, "not enough memory");
5115 bgl_lock_init(sbi->s_blockgroup_lock);
5117 /* Pre-read the descriptors into the buffer cache */
5118 for (i = 0; i < db_count; i++) {
5119 block = descriptor_loc(sb, logical_sb_block, i);
5120 ext4_sb_breadahead_unmovable(sb, block);
5123 for (i = 0; i < db_count; i++) {
5124 struct buffer_head *bh;
5126 block = descriptor_loc(sb, logical_sb_block, i);
5127 bh = ext4_sb_bread_unmovable(sb, block);
5129 ext4_msg(sb, KERN_ERR,
5130 "can't read group descriptor %d", i);
5136 rcu_dereference(sbi->s_group_desc)[i] = bh;
5139 sbi->s_gdb_count = db_count;
5140 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
5141 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
5142 ret = -EFSCORRUPTED;
5146 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5147 spin_lock_init(&sbi->s_error_lock);
5148 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5150 /* Register extent status tree shrinker */
5151 if (ext4_es_register_shrinker(sbi))
5154 sbi->s_stripe = ext4_get_stripe_size(sbi);
5155 sbi->s_extent_max_zeroout_kb = 32;
5158 * set up enough so that it can read an inode
5160 sb->s_op = &ext4_sops;
5161 sb->s_export_op = &ext4_export_ops;
5162 sb->s_xattr = ext4_xattr_handlers;
5163 #ifdef CONFIG_FS_ENCRYPTION
5164 sb->s_cop = &ext4_cryptops;
5166 #ifdef CONFIG_FS_VERITY
5167 sb->s_vop = &ext4_verityops;
5170 sb->dq_op = &ext4_quota_operations;
5171 if (ext4_has_feature_quota(sb))
5172 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5174 sb->s_qcop = &ext4_qctl_operations;
5175 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5177 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5179 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5180 mutex_init(&sbi->s_orphan_lock);
5182 /* Initialize fast commit stuff */
5183 atomic_set(&sbi->s_fc_subtid, 0);
5184 atomic_set(&sbi->s_fc_ineligible_updates, 0);
5185 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
5186 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
5187 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
5188 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
5189 sbi->s_fc_bytes = 0;
5190 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
5191 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
5192 spin_lock_init(&sbi->s_fc_lock);
5193 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
5194 sbi->s_fc_replay_state.fc_regions = NULL;
5195 sbi->s_fc_replay_state.fc_regions_size = 0;
5196 sbi->s_fc_replay_state.fc_regions_used = 0;
5197 sbi->s_fc_replay_state.fc_regions_valid = 0;
5198 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
5199 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
5200 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
5204 needs_recovery = (es->s_last_orphan != 0 ||
5205 ext4_has_feature_orphan_present(sb) ||
5206 ext4_has_feature_journal_needs_recovery(sb));
5208 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5209 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5210 goto failed_mount3a;
5213 * The first inode we look at is the journal inode. Don't try
5214 * root first: it may be modified in the journal!
5216 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5217 err = ext4_load_journal(sb, es, ctx->journal_devnum);
5219 goto failed_mount3a;
5220 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5221 ext4_has_feature_journal_needs_recovery(sb)) {
5222 ext4_msg(sb, KERN_ERR, "required journal recovery "
5223 "suppressed and not mounted read-only");
5224 goto failed_mount_wq;
5226 /* Nojournal mode, all journal mount options are illegal */
5227 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5228 ext4_msg(sb, KERN_ERR, "can't mount with "
5229 "journal_checksum, fs mounted w/o journal");
5230 goto failed_mount_wq;
5232 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5233 ext4_msg(sb, KERN_ERR, "can't mount with "
5234 "journal_async_commit, fs mounted w/o journal");
5235 goto failed_mount_wq;
5237 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5238 ext4_msg(sb, KERN_ERR, "can't mount with "
5239 "commit=%lu, fs mounted w/o journal",
5240 sbi->s_commit_interval / HZ);
5241 goto failed_mount_wq;
5243 if (EXT4_MOUNT_DATA_FLAGS &
5244 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5245 ext4_msg(sb, KERN_ERR, "can't mount with "
5246 "data=, fs mounted w/o journal");
5247 goto failed_mount_wq;
5249 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5250 clear_opt(sb, JOURNAL_CHECKSUM);
5251 clear_opt(sb, DATA_FLAGS);
5252 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5253 sbi->s_journal = NULL;
5258 if (ext4_has_feature_64bit(sb) &&
5259 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5260 JBD2_FEATURE_INCOMPAT_64BIT)) {
5261 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
5262 goto failed_mount_wq;
5265 if (!set_journal_csum_feature_set(sb)) {
5266 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
5268 goto failed_mount_wq;
5271 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
5272 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5273 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
5274 ext4_msg(sb, KERN_ERR,
5275 "Failed to set fast commit journal feature");
5276 goto failed_mount_wq;
5279 /* We have now updated the journal if required, so we can
5280 * validate the data journaling mode. */
5281 switch (test_opt(sb, DATA_FLAGS)) {
5283 /* No mode set, assume a default based on the journal
5284 * capabilities: ORDERED_DATA if the journal can
5285 * cope, else JOURNAL_DATA
5287 if (jbd2_journal_check_available_features
5288 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5289 set_opt(sb, ORDERED_DATA);
5290 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
5292 set_opt(sb, JOURNAL_DATA);
5293 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
5297 case EXT4_MOUNT_ORDERED_DATA:
5298 case EXT4_MOUNT_WRITEBACK_DATA:
5299 if (!jbd2_journal_check_available_features
5300 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5301 ext4_msg(sb, KERN_ERR, "Journal does not support "
5302 "requested data journaling mode");
5303 goto failed_mount_wq;
5310 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
5311 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5312 ext4_msg(sb, KERN_ERR, "can't mount with "
5313 "journal_async_commit in data=ordered mode");
5314 goto failed_mount_wq;
5317 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
5319 sbi->s_journal->j_submit_inode_data_buffers =
5320 ext4_journal_submit_inode_data_buffers;
5321 sbi->s_journal->j_finish_inode_data_buffers =
5322 ext4_journal_finish_inode_data_buffers;
5325 if (!test_opt(sb, NO_MBCACHE)) {
5326 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5327 if (!sbi->s_ea_block_cache) {
5328 ext4_msg(sb, KERN_ERR,
5329 "Failed to create ea_block_cache");
5330 goto failed_mount_wq;
5333 if (ext4_has_feature_ea_inode(sb)) {
5334 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5335 if (!sbi->s_ea_inode_cache) {
5336 ext4_msg(sb, KERN_ERR,
5337 "Failed to create ea_inode_cache");
5338 goto failed_mount_wq;
5343 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
5344 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5345 goto failed_mount_wq;
5348 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
5349 !ext4_has_feature_encrypt(sb)) {
5350 ext4_set_feature_encrypt(sb);
5351 ext4_commit_super(sb);
5355 * Get the # of file system overhead blocks from the
5356 * superblock if present.
5358 if (es->s_overhead_clusters)
5359 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5361 err = ext4_calculate_overhead(sb);
5363 goto failed_mount_wq;
5367 * The maximum number of concurrent works can be high and
5368 * concurrency isn't really necessary. Limit it to 1.
5370 EXT4_SB(sb)->rsv_conversion_wq =
5371 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5372 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5373 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5379 * The jbd2_journal_load will have done any necessary log recovery,
5380 * so we can safely mount the rest of the filesystem now.
5383 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5385 ext4_msg(sb, KERN_ERR, "get root inode failed");
5386 ret = PTR_ERR(root);
5390 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5391 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5396 sb->s_root = d_make_root(root);
5398 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5403 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5404 if (ret == -EROFS) {
5405 sb->s_flags |= SB_RDONLY;
5408 goto failed_mount4a;
5410 ext4_set_resv_clusters(sb);
5412 if (test_opt(sb, BLOCK_VALIDITY)) {
5413 err = ext4_setup_system_zone(sb);
5415 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5417 goto failed_mount4a;
5420 ext4_fc_replay_cleanup(sb);
5425 * Enable optimize_scan if number of groups is > threshold. This can be
5426 * turned off by passing "mb_optimize_scan=0". This can also be
5427 * turned on forcefully by passing "mb_optimize_scan=1".
5429 if (ctx->mb_optimize_scan == 1)
5430 set_opt2(sb, MB_OPTIMIZE_SCAN);
5431 else if (ctx->mb_optimize_scan == 0)
5432 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5433 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5434 set_opt2(sb, MB_OPTIMIZE_SCAN);
5436 err = ext4_mb_init(sb);
5438 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5444 * We can only set up the journal commit callback once
5445 * mballoc is initialized
5448 sbi->s_journal->j_commit_callback =
5449 ext4_journal_commit_callback;
5451 block = ext4_count_free_clusters(sb);
5452 ext4_free_blocks_count_set(sbi->s_es,
5453 EXT4_C2B(sbi, block));
5454 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5457 unsigned long freei = ext4_count_free_inodes(sb);
5458 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5459 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5463 * Update the checksum after updating free space/inode
5464 * counters. Otherwise the superblock can have an incorrect
5465 * checksum in the buffer cache until it is written out and
5466 * e2fsprogs programs trying to open a file system immediately
5467 * after it is mounted can fail.
5469 ext4_superblock_csum_set(sb);
5471 err = percpu_counter_init(&sbi->s_dirs_counter,
5472 ext4_count_dirs(sb), GFP_KERNEL);
5474 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5477 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5480 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5483 ext4_msg(sb, KERN_ERR, "insufficient memory");
5487 if (ext4_has_feature_flex_bg(sb))
5488 if (!ext4_fill_flex_info(sb)) {
5489 ext4_msg(sb, KERN_ERR,
5490 "unable to initialize "
5491 "flex_bg meta info!");
5496 err = ext4_register_li_request(sb, first_not_zeroed);
5500 err = ext4_register_sysfs(sb);
5504 err = ext4_init_orphan_info(sb);
5508 /* Enable quota usage during mount. */
5509 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5510 err = ext4_enable_quotas(sb);
5514 #endif /* CONFIG_QUOTA */
5517 * Save the original bdev mapping's wb_err value which could be
5518 * used to detect the metadata async write error.
5520 spin_lock_init(&sbi->s_bdev_wb_lock);
5521 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5522 &sbi->s_bdev_wb_err);
5523 sb->s_bdev->bd_super = sb;
5524 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5525 ext4_orphan_cleanup(sb, es);
5526 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5527 if (needs_recovery) {
5528 ext4_msg(sb, KERN_INFO, "recovery complete");
5529 err = ext4_mark_recovery_complete(sb, es);
5534 if (test_opt(sb, DISCARD)) {
5535 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5536 if (!blk_queue_discard(q))
5537 ext4_msg(sb, KERN_WARNING,
5538 "mounting with \"discard\" option, but "
5539 "the device does not support discard");
5542 if (es->s_error_count)
5543 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5545 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5546 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5547 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5548 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5549 atomic_set(&sbi->s_warning_count, 0);
5550 atomic_set(&sbi->s_msg_count, 0);
5556 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5560 ext4_release_orphan_info(sb);
5562 ext4_unregister_sysfs(sb);
5563 kobject_put(&sbi->s_kobj);
5565 ext4_unregister_li_request(sb);
5567 ext4_mb_release(sb);
5569 flex_groups = rcu_dereference(sbi->s_flex_groups);
5571 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5572 kvfree(flex_groups[i]);
5573 kvfree(flex_groups);
5576 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5577 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5578 percpu_counter_destroy(&sbi->s_dirs_counter);
5579 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5580 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5581 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5583 ext4_ext_release(sb);
5584 ext4_release_system_zone(sb);
5589 ext4_msg(sb, KERN_ERR, "mount failed");
5590 if (EXT4_SB(sb)->rsv_conversion_wq)
5591 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5593 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5594 sbi->s_ea_inode_cache = NULL;
5596 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5597 sbi->s_ea_block_cache = NULL;
5599 if (sbi->s_journal) {
5600 /* flush s_error_work before journal destroy. */
5601 flush_work(&sbi->s_error_work);
5602 jbd2_journal_destroy(sbi->s_journal);
5603 sbi->s_journal = NULL;
5606 ext4_es_unregister_shrinker(sbi);
5608 /* flush s_error_work before sbi destroy */
5609 flush_work(&sbi->s_error_work);
5610 del_timer_sync(&sbi->s_err_report);
5611 ext4_stop_mmpd(sbi);
5614 group_desc = rcu_dereference(sbi->s_group_desc);
5615 for (i = 0; i < db_count; i++)
5616 brelse(group_desc[i]);
5620 if (sbi->s_chksum_driver)
5621 crypto_free_shash(sbi->s_chksum_driver);
5623 #ifdef CONFIG_UNICODE
5624 utf8_unload(sb->s_encoding);
5628 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5629 kfree(get_qf_name(sb, sbi, i));
5631 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5632 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5634 ext4_blkdev_remove(sbi);
5636 sb->s_fs_info = NULL;
5637 return err ? err : ret;
5640 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5642 struct ext4_fs_context *ctx = fc->fs_private;
5643 struct ext4_sb_info *sbi;
5647 sbi = ext4_alloc_sbi(sb);
5651 fc->s_fs_info = sbi;
5653 /* Cleanup superblock name */
5654 strreplace(sb->s_id, '/', '!');
5656 sbi->s_sb_block = 1; /* Default super block location */
5657 if (ctx->spec & EXT4_SPEC_s_sb_block)
5658 sbi->s_sb_block = ctx->s_sb_block;
5660 ret = __ext4_fill_super(fc, sb, fc->sb_flags & SB_SILENT);
5664 if (sbi->s_journal) {
5665 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5666 descr = " journalled data mode";
5667 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5668 descr = " ordered data mode";
5670 descr = " writeback data mode";
5672 descr = "out journal";
5674 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5675 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5676 "Quota mode: %s.", descr, ext4_quota_mode(sb));
5682 fc->s_fs_info = NULL;
5686 static int ext4_get_tree(struct fs_context *fc)
5688 return get_tree_bdev(fc, ext4_fill_super);
5692 * Setup any per-fs journal parameters now. We'll do this both on
5693 * initial mount, once the journal has been initialised but before we've
5694 * done any recovery; and again on any subsequent remount.
5696 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5698 struct ext4_sb_info *sbi = EXT4_SB(sb);
5700 journal->j_commit_interval = sbi->s_commit_interval;
5701 journal->j_min_batch_time = sbi->s_min_batch_time;
5702 journal->j_max_batch_time = sbi->s_max_batch_time;
5703 ext4_fc_init(sb, journal);
5705 write_lock(&journal->j_state_lock);
5706 if (test_opt(sb, BARRIER))
5707 journal->j_flags |= JBD2_BARRIER;
5709 journal->j_flags &= ~JBD2_BARRIER;
5710 if (test_opt(sb, DATA_ERR_ABORT))
5711 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5713 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5714 write_unlock(&journal->j_state_lock);
5717 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5718 unsigned int journal_inum)
5720 struct inode *journal_inode;
5723 * Test for the existence of a valid inode on disk. Bad things
5724 * happen if we iget() an unused inode, as the subsequent iput()
5725 * will try to delete it.
5727 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5728 if (IS_ERR(journal_inode)) {
5729 ext4_msg(sb, KERN_ERR, "no journal found");
5732 if (!journal_inode->i_nlink) {
5733 make_bad_inode(journal_inode);
5734 iput(journal_inode);
5735 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5739 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5740 journal_inode, journal_inode->i_size);
5741 if (!S_ISREG(journal_inode->i_mode)) {
5742 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5743 iput(journal_inode);
5746 return journal_inode;
5749 static journal_t *ext4_get_journal(struct super_block *sb,
5750 unsigned int journal_inum)
5752 struct inode *journal_inode;
5755 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5758 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5762 journal = jbd2_journal_init_inode(journal_inode);
5764 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5765 iput(journal_inode);
5768 journal->j_private = sb;
5769 ext4_init_journal_params(sb, journal);
5773 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5776 struct buffer_head *bh;
5780 int hblock, blocksize;
5781 ext4_fsblk_t sb_block;
5782 unsigned long offset;
5783 struct ext4_super_block *es;
5784 struct block_device *bdev;
5786 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5789 bdev = ext4_blkdev_get(j_dev, sb);
5793 blocksize = sb->s_blocksize;
5794 hblock = bdev_logical_block_size(bdev);
5795 if (blocksize < hblock) {
5796 ext4_msg(sb, KERN_ERR,
5797 "blocksize too small for journal device");
5801 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5802 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5803 set_blocksize(bdev, blocksize);
5804 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5805 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5806 "external journal");
5810 es = (struct ext4_super_block *) (bh->b_data + offset);
5811 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5812 !(le32_to_cpu(es->s_feature_incompat) &
5813 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5814 ext4_msg(sb, KERN_ERR, "external journal has "
5820 if ((le32_to_cpu(es->s_feature_ro_compat) &
5821 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5822 es->s_checksum != ext4_superblock_csum(sb, es)) {
5823 ext4_msg(sb, KERN_ERR, "external journal has "
5824 "corrupt superblock");
5829 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5830 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5835 len = ext4_blocks_count(es);
5836 start = sb_block + 1;
5837 brelse(bh); /* we're done with the superblock */
5839 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5840 start, len, blocksize);
5842 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5845 journal->j_private = sb;
5846 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5847 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5850 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5851 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5852 "user (unsupported) - %d",
5853 be32_to_cpu(journal->j_superblock->s_nr_users));
5856 EXT4_SB(sb)->s_journal_bdev = bdev;
5857 ext4_init_journal_params(sb, journal);
5861 jbd2_journal_destroy(journal);
5863 ext4_blkdev_put(bdev);
5867 static int ext4_load_journal(struct super_block *sb,
5868 struct ext4_super_block *es,
5869 unsigned long journal_devnum)
5872 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5875 int really_read_only;
5878 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5879 return -EFSCORRUPTED;
5881 if (journal_devnum &&
5882 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5883 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5884 "numbers have changed");
5885 journal_dev = new_decode_dev(journal_devnum);
5887 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5889 if (journal_inum && journal_dev) {
5890 ext4_msg(sb, KERN_ERR,
5891 "filesystem has both journal inode and journal device!");
5896 journal = ext4_get_journal(sb, journal_inum);
5900 journal = ext4_get_dev_journal(sb, journal_dev);
5905 journal_dev_ro = bdev_read_only(journal->j_dev);
5906 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5908 if (journal_dev_ro && !sb_rdonly(sb)) {
5909 ext4_msg(sb, KERN_ERR,
5910 "journal device read-only, try mounting with '-o ro'");
5916 * Are we loading a blank journal or performing recovery after a
5917 * crash? For recovery, we need to check in advance whether we
5918 * can get read-write access to the device.
5920 if (ext4_has_feature_journal_needs_recovery(sb)) {
5921 if (sb_rdonly(sb)) {
5922 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5923 "required on readonly filesystem");
5924 if (really_read_only) {
5925 ext4_msg(sb, KERN_ERR, "write access "
5926 "unavailable, cannot proceed "
5927 "(try mounting with noload)");
5931 ext4_msg(sb, KERN_INFO, "write access will "
5932 "be enabled during recovery");
5936 if (!(journal->j_flags & JBD2_BARRIER))
5937 ext4_msg(sb, KERN_INFO, "barriers disabled");
5939 if (!ext4_has_feature_journal_needs_recovery(sb))
5940 err = jbd2_journal_wipe(journal, !really_read_only);
5942 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5944 memcpy(save, ((char *) es) +
5945 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5946 err = jbd2_journal_load(journal);
5948 memcpy(((char *) es) + EXT4_S_ERR_START,
5949 save, EXT4_S_ERR_LEN);
5954 ext4_msg(sb, KERN_ERR, "error loading journal");
5958 EXT4_SB(sb)->s_journal = journal;
5959 err = ext4_clear_journal_err(sb, es);
5961 EXT4_SB(sb)->s_journal = NULL;
5962 jbd2_journal_destroy(journal);
5966 if (!really_read_only && journal_devnum &&
5967 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5968 es->s_journal_dev = cpu_to_le32(journal_devnum);
5970 /* Make sure we flush the recovery flag to disk. */
5971 ext4_commit_super(sb);
5977 jbd2_journal_destroy(journal);
5981 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5982 static void ext4_update_super(struct super_block *sb)
5984 struct ext4_sb_info *sbi = EXT4_SB(sb);
5985 struct ext4_super_block *es = sbi->s_es;
5986 struct buffer_head *sbh = sbi->s_sbh;
5990 * If the file system is mounted read-only, don't update the
5991 * superblock write time. This avoids updating the superblock
5992 * write time when we are mounting the root file system
5993 * read/only but we need to replay the journal; at that point,
5994 * for people who are east of GMT and who make their clock
5995 * tick in localtime for Windows bug-for-bug compatibility,
5996 * the clock is set in the future, and this will cause e2fsck
5997 * to complain and force a full file system check.
5999 if (!(sb->s_flags & SB_RDONLY))
6000 ext4_update_tstamp(es, s_wtime);
6001 es->s_kbytes_written =
6002 cpu_to_le64(sbi->s_kbytes_written +
6003 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6004 sbi->s_sectors_written_start) >> 1));
6005 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6006 ext4_free_blocks_count_set(es,
6007 EXT4_C2B(sbi, percpu_counter_sum_positive(
6008 &sbi->s_freeclusters_counter)));
6009 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6010 es->s_free_inodes_count =
6011 cpu_to_le32(percpu_counter_sum_positive(
6012 &sbi->s_freeinodes_counter));
6013 /* Copy error information to the on-disk superblock */
6014 spin_lock(&sbi->s_error_lock);
6015 if (sbi->s_add_error_count > 0) {
6016 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6017 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6018 __ext4_update_tstamp(&es->s_first_error_time,
6019 &es->s_first_error_time_hi,
6020 sbi->s_first_error_time);
6021 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6022 sizeof(es->s_first_error_func));
6023 es->s_first_error_line =
6024 cpu_to_le32(sbi->s_first_error_line);
6025 es->s_first_error_ino =
6026 cpu_to_le32(sbi->s_first_error_ino);
6027 es->s_first_error_block =
6028 cpu_to_le64(sbi->s_first_error_block);
6029 es->s_first_error_errcode =
6030 ext4_errno_to_code(sbi->s_first_error_code);
6032 __ext4_update_tstamp(&es->s_last_error_time,
6033 &es->s_last_error_time_hi,
6034 sbi->s_last_error_time);
6035 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6036 sizeof(es->s_last_error_func));
6037 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6038 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6039 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6040 es->s_last_error_errcode =
6041 ext4_errno_to_code(sbi->s_last_error_code);
6043 * Start the daily error reporting function if it hasn't been
6046 if (!es->s_error_count)
6047 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6048 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6049 sbi->s_add_error_count = 0;
6051 spin_unlock(&sbi->s_error_lock);
6053 ext4_superblock_csum_set(sb);
6057 static int ext4_commit_super(struct super_block *sb)
6059 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6064 if (block_device_ejected(sb))
6067 ext4_update_super(sb);
6069 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6071 * Oh, dear. A previous attempt to write the
6072 * superblock failed. This could happen because the
6073 * USB device was yanked out. Or it could happen to
6074 * be a transient write error and maybe the block will
6075 * be remapped. Nothing we can do but to retry the
6076 * write and hope for the best.
6078 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6079 "superblock detected");
6080 clear_buffer_write_io_error(sbh);
6081 set_buffer_uptodate(sbh);
6083 BUFFER_TRACE(sbh, "marking dirty");
6084 mark_buffer_dirty(sbh);
6085 error = __sync_dirty_buffer(sbh,
6086 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
6087 if (buffer_write_io_error(sbh)) {
6088 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6090 clear_buffer_write_io_error(sbh);
6091 set_buffer_uptodate(sbh);
6097 * Have we just finished recovery? If so, and if we are mounting (or
6098 * remounting) the filesystem readonly, then we will end up with a
6099 * consistent fs on disk. Record that fact.
6101 static int ext4_mark_recovery_complete(struct super_block *sb,
6102 struct ext4_super_block *es)
6105 journal_t *journal = EXT4_SB(sb)->s_journal;
6107 if (!ext4_has_feature_journal(sb)) {
6108 if (journal != NULL) {
6109 ext4_error(sb, "Journal got removed while the fs was "
6111 return -EFSCORRUPTED;
6115 jbd2_journal_lock_updates(journal);
6116 err = jbd2_journal_flush(journal, 0);
6120 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6121 ext4_has_feature_orphan_present(sb))) {
6122 if (!ext4_orphan_file_empty(sb)) {
6123 ext4_error(sb, "Orphan file not empty on read-only fs.");
6124 err = -EFSCORRUPTED;
6127 ext4_clear_feature_journal_needs_recovery(sb);
6128 ext4_clear_feature_orphan_present(sb);
6129 ext4_commit_super(sb);
6132 jbd2_journal_unlock_updates(journal);
6137 * If we are mounting (or read-write remounting) a filesystem whose journal
6138 * has recorded an error from a previous lifetime, move that error to the
6139 * main filesystem now.
6141 static int ext4_clear_journal_err(struct super_block *sb,
6142 struct ext4_super_block *es)
6148 if (!ext4_has_feature_journal(sb)) {
6149 ext4_error(sb, "Journal got removed while the fs was mounted!");
6150 return -EFSCORRUPTED;
6153 journal = EXT4_SB(sb)->s_journal;
6156 * Now check for any error status which may have been recorded in the
6157 * journal by a prior ext4_error() or ext4_abort()
6160 j_errno = jbd2_journal_errno(journal);
6164 errstr = ext4_decode_error(sb, j_errno, nbuf);
6165 ext4_warning(sb, "Filesystem error recorded "
6166 "from previous mount: %s", errstr);
6167 ext4_warning(sb, "Marking fs in need of filesystem check.");
6169 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6170 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6171 ext4_commit_super(sb);
6173 jbd2_journal_clear_err(journal);
6174 jbd2_journal_update_sb_errno(journal);
6180 * Force the running and committing transactions to commit,
6181 * and wait on the commit.
6183 int ext4_force_commit(struct super_block *sb)
6190 journal = EXT4_SB(sb)->s_journal;
6191 return ext4_journal_force_commit(journal);
6194 static int ext4_sync_fs(struct super_block *sb, int wait)
6198 bool needs_barrier = false;
6199 struct ext4_sb_info *sbi = EXT4_SB(sb);
6201 if (unlikely(ext4_forced_shutdown(sbi)))
6204 trace_ext4_sync_fs(sb, wait);
6205 flush_workqueue(sbi->rsv_conversion_wq);
6207 * Writeback quota in non-journalled quota case - journalled quota has
6210 dquot_writeback_dquots(sb, -1);
6212 * Data writeback is possible w/o journal transaction, so barrier must
6213 * being sent at the end of the function. But we can skip it if
6214 * transaction_commit will do it for us.
6216 if (sbi->s_journal) {
6217 target = jbd2_get_latest_transaction(sbi->s_journal);
6218 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6219 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6220 needs_barrier = true;
6222 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6224 ret = jbd2_log_wait_commit(sbi->s_journal,
6227 } else if (wait && test_opt(sb, BARRIER))
6228 needs_barrier = true;
6229 if (needs_barrier) {
6231 err = blkdev_issue_flush(sb->s_bdev);
6240 * LVM calls this function before a (read-only) snapshot is created. This
6241 * gives us a chance to flush the journal completely and mark the fs clean.
6243 * Note that only this function cannot bring a filesystem to be in a clean
6244 * state independently. It relies on upper layer to stop all data & metadata
6247 static int ext4_freeze(struct super_block *sb)
6255 journal = EXT4_SB(sb)->s_journal;
6258 /* Now we set up the journal barrier. */
6259 jbd2_journal_lock_updates(journal);
6262 * Don't clear the needs_recovery flag if we failed to
6263 * flush the journal.
6265 error = jbd2_journal_flush(journal, 0);
6269 /* Journal blocked and flushed, clear needs_recovery flag. */
6270 ext4_clear_feature_journal_needs_recovery(sb);
6271 if (ext4_orphan_file_empty(sb))
6272 ext4_clear_feature_orphan_present(sb);
6275 error = ext4_commit_super(sb);
6278 /* we rely on upper layer to stop further updates */
6279 jbd2_journal_unlock_updates(journal);
6284 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6285 * flag here, even though the filesystem is not technically dirty yet.
6287 static int ext4_unfreeze(struct super_block *sb)
6289 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6292 if (EXT4_SB(sb)->s_journal) {
6293 /* Reset the needs_recovery flag before the fs is unlocked. */
6294 ext4_set_feature_journal_needs_recovery(sb);
6295 if (ext4_has_feature_orphan_file(sb))
6296 ext4_set_feature_orphan_present(sb);
6299 ext4_commit_super(sb);
6304 * Structure to save mount options for ext4_remount's benefit
6306 struct ext4_mount_options {
6307 unsigned long s_mount_opt;
6308 unsigned long s_mount_opt2;
6311 unsigned long s_commit_interval;
6312 u32 s_min_batch_time, s_max_batch_time;
6315 char *s_qf_names[EXT4_MAXQUOTAS];
6319 static int __ext4_remount(struct fs_context *fc, struct super_block *sb,
6322 struct ext4_fs_context *ctx = fc->fs_private;
6323 struct ext4_super_block *es;
6324 struct ext4_sb_info *sbi = EXT4_SB(sb);
6325 unsigned long old_sb_flags, vfs_flags;
6326 struct ext4_mount_options old_opts;
6330 int enable_quota = 0;
6332 char *to_free[EXT4_MAXQUOTAS];
6335 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6337 /* Store the original options */
6338 old_sb_flags = sb->s_flags;
6339 old_opts.s_mount_opt = sbi->s_mount_opt;
6340 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6341 old_opts.s_resuid = sbi->s_resuid;
6342 old_opts.s_resgid = sbi->s_resgid;
6343 old_opts.s_commit_interval = sbi->s_commit_interval;
6344 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6345 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6347 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6348 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6349 if (sbi->s_qf_names[i]) {
6350 char *qf_name = get_qf_name(sb, sbi, i);
6352 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6353 if (!old_opts.s_qf_names[i]) {
6354 for (j = 0; j < i; j++)
6355 kfree(old_opts.s_qf_names[j]);
6359 old_opts.s_qf_names[i] = NULL;
6361 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6362 ctx->journal_ioprio =
6363 sbi->s_journal->j_task->io_context->ioprio;
6366 * Some options can be enabled by ext4 and/or by VFS mount flag
6367 * either way we need to make sure it matches in both *flags and
6368 * s_flags. Copy those selected flags from *flags to s_flags
6370 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
6371 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
6373 ext4_apply_options(fc, sb);
6375 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6376 test_opt(sb, JOURNAL_CHECKSUM)) {
6377 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6378 "during remount not supported; ignoring");
6379 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6382 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6383 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6384 ext4_msg(sb, KERN_ERR, "can't mount with "
6385 "both data=journal and delalloc");
6389 if (test_opt(sb, DIOREAD_NOLOCK)) {
6390 ext4_msg(sb, KERN_ERR, "can't mount with "
6391 "both data=journal and dioread_nolock");
6395 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6396 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6397 ext4_msg(sb, KERN_ERR, "can't mount with "
6398 "journal_async_commit in data=ordered mode");
6404 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6405 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6410 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6411 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6413 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6414 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6418 if (sbi->s_journal) {
6419 ext4_init_journal_params(sb, sbi->s_journal);
6420 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6423 /* Flush outstanding errors before changing fs state */
6424 flush_work(&sbi->s_error_work);
6426 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
6427 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6432 if (*flags & SB_RDONLY) {
6433 err = sync_filesystem(sb);
6436 err = dquot_suspend(sb, -1);
6441 * First of all, the unconditional stuff we have to do
6442 * to disable replay of the journal when we next remount
6444 sb->s_flags |= SB_RDONLY;
6447 * OK, test if we are remounting a valid rw partition
6448 * readonly, and if so set the rdonly flag and then
6449 * mark the partition as valid again.
6451 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6452 (sbi->s_mount_state & EXT4_VALID_FS))
6453 es->s_state = cpu_to_le16(sbi->s_mount_state);
6455 if (sbi->s_journal) {
6457 * We let remount-ro finish even if marking fs
6458 * as clean failed...
6460 ext4_mark_recovery_complete(sb, es);
6463 /* Make sure we can mount this feature set readwrite */
6464 if (ext4_has_feature_readonly(sb) ||
6465 !ext4_feature_set_ok(sb, 0)) {
6470 * Make sure the group descriptor checksums
6471 * are sane. If they aren't, refuse to remount r/w.
6473 for (g = 0; g < sbi->s_groups_count; g++) {
6474 struct ext4_group_desc *gdp =
6475 ext4_get_group_desc(sb, g, NULL);
6477 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6478 ext4_msg(sb, KERN_ERR,
6479 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6480 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6481 le16_to_cpu(gdp->bg_checksum));
6488 * If we have an unprocessed orphan list hanging
6489 * around from a previously readonly bdev mount,
6490 * require a full umount/remount for now.
6492 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6493 ext4_msg(sb, KERN_WARNING, "Couldn't "
6494 "remount RDWR because of unprocessed "
6495 "orphan inode list. Please "
6496 "umount/remount instead");
6502 * Mounting a RDONLY partition read-write, so reread
6503 * and store the current valid flag. (It may have
6504 * been changed by e2fsck since we originally mounted
6507 if (sbi->s_journal) {
6508 err = ext4_clear_journal_err(sb, es);
6512 sbi->s_mount_state = le16_to_cpu(es->s_state);
6514 err = ext4_setup_super(sb, es, 0);
6518 sb->s_flags &= ~SB_RDONLY;
6519 if (ext4_has_feature_mmp(sb))
6520 if (ext4_multi_mount_protect(sb,
6521 le64_to_cpu(es->s_mmp_block))) {
6532 * Reinitialize lazy itable initialization thread based on
6535 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6536 ext4_unregister_li_request(sb);
6538 ext4_group_t first_not_zeroed;
6539 first_not_zeroed = ext4_has_uninit_itable(sb);
6540 ext4_register_li_request(sb, first_not_zeroed);
6544 * Handle creation of system zone data early because it can fail.
6545 * Releasing of existing data is done when we are sure remount will
6548 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6549 err = ext4_setup_system_zone(sb);
6554 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6555 err = ext4_commit_super(sb);
6561 /* Release old quota file names */
6562 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6563 kfree(old_opts.s_qf_names[i]);
6565 if (sb_any_quota_suspended(sb))
6566 dquot_resume(sb, -1);
6567 else if (ext4_has_feature_quota(sb)) {
6568 err = ext4_enable_quotas(sb);
6574 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6575 ext4_release_system_zone(sb);
6577 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6578 ext4_stop_mmpd(sbi);
6581 * Some options can be enabled by ext4 and/or by VFS mount flag
6582 * either way we need to make sure it matches in both *flags and
6583 * s_flags. Copy those selected flags from s_flags to *flags
6585 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6590 sb->s_flags = old_sb_flags;
6591 sbi->s_mount_opt = old_opts.s_mount_opt;
6592 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6593 sbi->s_resuid = old_opts.s_resuid;
6594 sbi->s_resgid = old_opts.s_resgid;
6595 sbi->s_commit_interval = old_opts.s_commit_interval;
6596 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6597 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6598 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6599 ext4_release_system_zone(sb);
6601 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6602 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6603 to_free[i] = get_qf_name(sb, sbi, i);
6604 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6607 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6610 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6611 ext4_stop_mmpd(sbi);
6615 static int ext4_reconfigure(struct fs_context *fc)
6617 struct super_block *sb = fc->root->d_sb;
6618 int flags = fc->sb_flags;
6621 fc->s_fs_info = EXT4_SB(sb);
6623 ret = ext4_check_opt_consistency(fc, sb);
6627 ret = __ext4_remount(fc, sb, &flags);
6631 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.",
6632 ext4_quota_mode(sb));
6638 static int ext4_statfs_project(struct super_block *sb,
6639 kprojid_t projid, struct kstatfs *buf)
6642 struct dquot *dquot;
6646 qid = make_kqid_projid(projid);
6647 dquot = dqget(sb, qid);
6649 return PTR_ERR(dquot);
6650 spin_lock(&dquot->dq_dqb_lock);
6652 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6653 dquot->dq_dqb.dqb_bhardlimit);
6654 limit >>= sb->s_blocksize_bits;
6656 if (limit && buf->f_blocks > limit) {
6657 curblock = (dquot->dq_dqb.dqb_curspace +
6658 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6659 buf->f_blocks = limit;
6660 buf->f_bfree = buf->f_bavail =
6661 (buf->f_blocks > curblock) ?
6662 (buf->f_blocks - curblock) : 0;
6665 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6666 dquot->dq_dqb.dqb_ihardlimit);
6667 if (limit && buf->f_files > limit) {
6668 buf->f_files = limit;
6670 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6671 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6674 spin_unlock(&dquot->dq_dqb_lock);
6680 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6682 struct super_block *sb = dentry->d_sb;
6683 struct ext4_sb_info *sbi = EXT4_SB(sb);
6684 struct ext4_super_block *es = sbi->s_es;
6685 ext4_fsblk_t overhead = 0, resv_blocks;
6687 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6689 if (!test_opt(sb, MINIX_DF))
6690 overhead = sbi->s_overhead;
6692 buf->f_type = EXT4_SUPER_MAGIC;
6693 buf->f_bsize = sb->s_blocksize;
6694 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6695 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6696 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6697 /* prevent underflow in case that few free space is available */
6698 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6699 buf->f_bavail = buf->f_bfree -
6700 (ext4_r_blocks_count(es) + resv_blocks);
6701 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6703 buf->f_files = le32_to_cpu(es->s_inodes_count);
6704 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6705 buf->f_namelen = EXT4_NAME_LEN;
6706 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6709 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6710 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6711 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6720 * Helper functions so that transaction is started before we acquire dqio_sem
6721 * to keep correct lock ordering of transaction > dqio_sem
6723 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6725 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6728 static int ext4_write_dquot(struct dquot *dquot)
6732 struct inode *inode;
6734 inode = dquot_to_inode(dquot);
6735 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6736 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6738 return PTR_ERR(handle);
6739 ret = dquot_commit(dquot);
6740 err = ext4_journal_stop(handle);
6746 static int ext4_acquire_dquot(struct dquot *dquot)
6751 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6752 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6754 return PTR_ERR(handle);
6755 ret = dquot_acquire(dquot);
6756 err = ext4_journal_stop(handle);
6762 static int ext4_release_dquot(struct dquot *dquot)
6767 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6768 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6769 if (IS_ERR(handle)) {
6770 /* Release dquot anyway to avoid endless cycle in dqput() */
6771 dquot_release(dquot);
6772 return PTR_ERR(handle);
6774 ret = dquot_release(dquot);
6775 err = ext4_journal_stop(handle);
6781 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6783 struct super_block *sb = dquot->dq_sb;
6785 if (ext4_is_quota_journalled(sb)) {
6786 dquot_mark_dquot_dirty(dquot);
6787 return ext4_write_dquot(dquot);
6789 return dquot_mark_dquot_dirty(dquot);
6793 static int ext4_write_info(struct super_block *sb, int type)
6798 /* Data block + inode block */
6799 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6801 return PTR_ERR(handle);
6802 ret = dquot_commit_info(sb, type);
6803 err = ext4_journal_stop(handle);
6809 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6811 struct ext4_inode_info *ei = EXT4_I(inode);
6813 /* The first argument of lockdep_set_subclass has to be
6814 * *exactly* the same as the argument to init_rwsem() --- in
6815 * this case, in init_once() --- or lockdep gets unhappy
6816 * because the name of the lock is set using the
6817 * stringification of the argument to init_rwsem().
6819 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6820 lockdep_set_subclass(&ei->i_data_sem, subclass);
6824 * Standard function to be called on quota_on
6826 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6827 const struct path *path)
6831 if (!test_opt(sb, QUOTA))
6834 /* Quotafile not on the same filesystem? */
6835 if (path->dentry->d_sb != sb)
6838 /* Quota already enabled for this file? */
6839 if (IS_NOQUOTA(d_inode(path->dentry)))
6842 /* Journaling quota? */
6843 if (EXT4_SB(sb)->s_qf_names[type]) {
6844 /* Quotafile not in fs root? */
6845 if (path->dentry->d_parent != sb->s_root)
6846 ext4_msg(sb, KERN_WARNING,
6847 "Quota file not on filesystem root. "
6848 "Journaled quota will not work");
6849 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6852 * Clear the flag just in case mount options changed since
6855 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6859 * When we journal data on quota file, we have to flush journal to see
6860 * all updates to the file when we bypass pagecache...
6862 if (EXT4_SB(sb)->s_journal &&
6863 ext4_should_journal_data(d_inode(path->dentry))) {
6865 * We don't need to lock updates but journal_flush() could
6866 * otherwise be livelocked...
6868 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6869 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6870 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6875 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6876 err = dquot_quota_on(sb, type, format_id, path);
6878 lockdep_set_quota_inode(path->dentry->d_inode,
6881 struct inode *inode = d_inode(path->dentry);
6885 * Set inode flags to prevent userspace from messing with quota
6886 * files. If this fails, we return success anyway since quotas
6887 * are already enabled and this is not a hard failure.
6890 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6893 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6894 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6895 S_NOATIME | S_IMMUTABLE);
6896 err = ext4_mark_inode_dirty(handle, inode);
6897 ext4_journal_stop(handle);
6899 inode_unlock(inode);
6904 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6908 struct inode *qf_inode;
6909 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6910 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6911 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6912 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6915 BUG_ON(!ext4_has_feature_quota(sb));
6917 if (!qf_inums[type])
6920 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6921 if (IS_ERR(qf_inode)) {
6922 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6923 return PTR_ERR(qf_inode);
6926 /* Don't account quota for quota files to avoid recursion */
6927 qf_inode->i_flags |= S_NOQUOTA;
6928 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6929 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6931 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6937 /* Enable usage tracking for all quota types. */
6938 int ext4_enable_quotas(struct super_block *sb)
6941 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6942 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6943 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6944 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6946 bool quota_mopt[EXT4_MAXQUOTAS] = {
6947 test_opt(sb, USRQUOTA),
6948 test_opt(sb, GRPQUOTA),
6949 test_opt(sb, PRJQUOTA),
6952 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6953 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6954 if (qf_inums[type]) {
6955 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6956 DQUOT_USAGE_ENABLED |
6957 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6960 "Failed to enable quota tracking "
6961 "(type=%d, err=%d). Please run "
6962 "e2fsck to fix.", type, err);
6963 for (type--; type >= 0; type--)
6964 dquot_quota_off(sb, type);
6973 static int ext4_quota_off(struct super_block *sb, int type)
6975 struct inode *inode = sb_dqopt(sb)->files[type];
6979 /* Force all delayed allocation blocks to be allocated.
6980 * Caller already holds s_umount sem */
6981 if (test_opt(sb, DELALLOC))
6982 sync_filesystem(sb);
6984 if (!inode || !igrab(inode))
6987 err = dquot_quota_off(sb, type);
6988 if (err || ext4_has_feature_quota(sb))
6993 * Update modification times of quota files when userspace can
6994 * start looking at them. If we fail, we return success anyway since
6995 * this is not a hard failure and quotas are already disabled.
6997 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6998 if (IS_ERR(handle)) {
6999 err = PTR_ERR(handle);
7002 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7003 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7004 inode->i_mtime = inode->i_ctime = current_time(inode);
7005 err = ext4_mark_inode_dirty(handle, inode);
7006 ext4_journal_stop(handle);
7008 inode_unlock(inode);
7010 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7014 return dquot_quota_off(sb, type);
7017 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7018 * acquiring the locks... As quota files are never truncated and quota code
7019 * itself serializes the operations (and no one else should touch the files)
7020 * we don't have to be afraid of races */
7021 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7022 size_t len, loff_t off)
7024 struct inode *inode = sb_dqopt(sb)->files[type];
7025 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7026 int offset = off & (sb->s_blocksize - 1);
7029 struct buffer_head *bh;
7030 loff_t i_size = i_size_read(inode);
7034 if (off+len > i_size)
7037 while (toread > 0) {
7038 tocopy = sb->s_blocksize - offset < toread ?
7039 sb->s_blocksize - offset : toread;
7040 bh = ext4_bread(NULL, inode, blk, 0);
7043 if (!bh) /* A hole? */
7044 memset(data, 0, tocopy);
7046 memcpy(data, bh->b_data+offset, tocopy);
7056 /* Write to quotafile (we know the transaction is already started and has
7057 * enough credits) */
7058 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7059 const char *data, size_t len, loff_t off)
7061 struct inode *inode = sb_dqopt(sb)->files[type];
7062 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7063 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7065 struct buffer_head *bh;
7066 handle_t *handle = journal_current_handle();
7068 if (EXT4_SB(sb)->s_journal && !handle) {
7069 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7070 " cancelled because transaction is not started",
7071 (unsigned long long)off, (unsigned long long)len);
7075 * Since we account only one data block in transaction credits,
7076 * then it is impossible to cross a block boundary.
7078 if (sb->s_blocksize - offset < len) {
7079 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7080 " cancelled because not block aligned",
7081 (unsigned long long)off, (unsigned long long)len);
7086 bh = ext4_bread(handle, inode, blk,
7087 EXT4_GET_BLOCKS_CREATE |
7088 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7089 } while (PTR_ERR(bh) == -ENOSPC &&
7090 ext4_should_retry_alloc(inode->i_sb, &retries));
7095 BUFFER_TRACE(bh, "get write access");
7096 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7102 memcpy(bh->b_data+offset, data, len);
7103 flush_dcache_page(bh->b_page);
7105 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7108 if (inode->i_size < off + len) {
7109 i_size_write(inode, off + len);
7110 EXT4_I(inode)->i_disksize = inode->i_size;
7111 err2 = ext4_mark_inode_dirty(handle, inode);
7112 if (unlikely(err2 && !err))
7115 return err ? err : len;
7119 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7120 static inline void register_as_ext2(void)
7122 int err = register_filesystem(&ext2_fs_type);
7125 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7128 static inline void unregister_as_ext2(void)
7130 unregister_filesystem(&ext2_fs_type);
7133 static inline int ext2_feature_set_ok(struct super_block *sb)
7135 if (ext4_has_unknown_ext2_incompat_features(sb))
7139 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7144 static inline void register_as_ext2(void) { }
7145 static inline void unregister_as_ext2(void) { }
7146 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7149 static inline void register_as_ext3(void)
7151 int err = register_filesystem(&ext3_fs_type);
7154 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7157 static inline void unregister_as_ext3(void)
7159 unregister_filesystem(&ext3_fs_type);
7162 static inline int ext3_feature_set_ok(struct super_block *sb)
7164 if (ext4_has_unknown_ext3_incompat_features(sb))
7166 if (!ext4_has_feature_journal(sb))
7170 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7175 static struct file_system_type ext4_fs_type = {
7176 .owner = THIS_MODULE,
7178 .init_fs_context = ext4_init_fs_context,
7179 .parameters = ext4_param_specs,
7180 .kill_sb = kill_block_super,
7181 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7183 MODULE_ALIAS_FS("ext4");
7185 /* Shared across all ext4 file systems */
7186 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7188 static int __init ext4_init_fs(void)
7192 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7193 ext4_li_info = NULL;
7195 /* Build-time check for flags consistency */
7196 ext4_check_flag_values();
7198 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7199 init_waitqueue_head(&ext4__ioend_wq[i]);
7201 err = ext4_init_es();
7205 err = ext4_init_pending();
7209 err = ext4_init_post_read_processing();
7213 err = ext4_init_pageio();
7217 err = ext4_init_system_zone();
7221 err = ext4_init_sysfs();
7225 err = ext4_init_mballoc();
7228 err = init_inodecache();
7232 err = ext4_fc_init_dentry_cache();
7238 err = register_filesystem(&ext4_fs_type);
7244 unregister_as_ext2();
7245 unregister_as_ext3();
7247 destroy_inodecache();
7249 ext4_exit_mballoc();
7253 ext4_exit_system_zone();
7257 ext4_exit_post_read_processing();
7259 ext4_exit_pending();
7266 static void __exit ext4_exit_fs(void)
7268 ext4_destroy_lazyinit_thread();
7269 unregister_as_ext2();
7270 unregister_as_ext3();
7271 unregister_filesystem(&ext4_fs_type);
7272 destroy_inodecache();
7273 ext4_exit_mballoc();
7275 ext4_exit_system_zone();
7277 ext4_exit_post_read_processing();
7279 ext4_exit_pending();
7282 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7283 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7284 MODULE_LICENSE("GPL");
7285 MODULE_SOFTDEP("pre: crc32c");
7286 module_init(ext4_init_fs)
7287 module_exit(ext4_exit_fs)