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/uaccess.h>
43 #include <linux/iversion.h>
44 #include <linux/unicode.h>
45 #include <linux/part_stat.h>
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
48 #include <linux/fsnotify.h>
49 #include <linux/fs_context.h>
50 #include <linux/fs_parser.h>
53 #include "ext4_extents.h" /* Needed for trace points definition */
54 #include "ext4_jbd2.h"
60 #define CREATE_TRACE_POINTS
61 #include <trace/events/ext4.h>
63 static struct ext4_lazy_init *ext4_li_info;
64 static DEFINE_MUTEX(ext4_li_mtx);
65 static struct ratelimit_state ext4_mount_msg_ratelimit;
67 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
68 unsigned long journal_devnum);
69 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
70 static void ext4_update_super(struct super_block *sb);
71 static int ext4_commit_super(struct super_block *sb);
72 static int ext4_mark_recovery_complete(struct super_block *sb,
73 struct ext4_super_block *es);
74 static int ext4_clear_journal_err(struct super_block *sb,
75 struct ext4_super_block *es);
76 static int ext4_sync_fs(struct super_block *sb, int wait);
77 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
78 static int ext4_unfreeze(struct super_block *sb);
79 static int ext4_freeze(struct super_block *sb);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
87 static int ext4_validate_options(struct fs_context *fc);
88 static int ext4_check_opt_consistency(struct fs_context *fc,
89 struct super_block *sb);
90 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
91 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
92 static int ext4_get_tree(struct fs_context *fc);
93 static int ext4_reconfigure(struct fs_context *fc);
94 static void ext4_fc_free(struct fs_context *fc);
95 static int ext4_init_fs_context(struct fs_context *fc);
96 static void ext4_kill_sb(struct super_block *sb);
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 = ext4_kill_sb,
140 .fs_flags = FS_REQUIRES_DEV,
142 MODULE_ALIAS_FS("ext2");
143 MODULE_ALIAS("ext2");
144 #define IS_EXT2_SB(sb) ((sb)->s_type == &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 = ext4_kill_sb,
156 .fs_flags = FS_REQUIRES_DEV,
158 MODULE_ALIAS_FS("ext3");
159 MODULE_ALIAS("ext3");
160 #define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
163 static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t 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, blk_opf_t 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, blk_opf_t 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, blk_opf_t op_flags, bool wait)
211 ext4_read_bh_nowait(bh, op_flags, NULL);
214 return ext4_read_bh(bh, op_flags, NULL);
218 * This works like __bread_gfp() except it uses ERR_PTR for error
219 * returns. Currently with sb_bread it's impossible to distinguish
220 * between ENOMEM and EIO situations (since both result in a NULL
223 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
225 blk_opf_t op_flags, gfp_t gfp)
227 struct buffer_head *bh;
230 bh = sb_getblk_gfp(sb, block, gfp);
232 return ERR_PTR(-ENOMEM);
233 if (ext4_buffer_uptodate(bh))
236 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
244 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
247 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
250 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
253 return __ext4_sb_bread_gfp(sb, block, 0, 0);
256 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
258 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
261 if (trylock_buffer(bh))
262 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
267 static int ext4_verify_csum_type(struct super_block *sb,
268 struct ext4_super_block *es)
270 if (!ext4_has_feature_metadata_csum(sb))
273 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
276 __le32 ext4_superblock_csum(struct super_block *sb,
277 struct ext4_super_block *es)
279 struct ext4_sb_info *sbi = EXT4_SB(sb);
280 int offset = offsetof(struct ext4_super_block, s_checksum);
283 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
285 return cpu_to_le32(csum);
288 static int ext4_superblock_csum_verify(struct super_block *sb,
289 struct ext4_super_block *es)
291 if (!ext4_has_metadata_csum(sb))
294 return es->s_checksum == ext4_superblock_csum(sb, es);
297 void ext4_superblock_csum_set(struct super_block *sb)
299 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
301 if (!ext4_has_metadata_csum(sb))
304 es->s_checksum = ext4_superblock_csum(sb, es);
307 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
308 struct ext4_group_desc *bg)
310 return le32_to_cpu(bg->bg_block_bitmap_lo) |
311 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
312 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
315 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
316 struct ext4_group_desc *bg)
318 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
319 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
320 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
323 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
324 struct ext4_group_desc *bg)
326 return le32_to_cpu(bg->bg_inode_table_lo) |
327 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
328 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
331 __u32 ext4_free_group_clusters(struct super_block *sb,
332 struct ext4_group_desc *bg)
334 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
335 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
336 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
339 __u32 ext4_free_inodes_count(struct super_block *sb,
340 struct ext4_group_desc *bg)
342 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
343 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
344 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
347 __u32 ext4_used_dirs_count(struct super_block *sb,
348 struct ext4_group_desc *bg)
350 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
351 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
352 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
355 __u32 ext4_itable_unused_count(struct super_block *sb,
356 struct ext4_group_desc *bg)
358 return le16_to_cpu(bg->bg_itable_unused_lo) |
359 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
360 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
363 void ext4_block_bitmap_set(struct super_block *sb,
364 struct ext4_group_desc *bg, ext4_fsblk_t blk)
366 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
367 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
368 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
371 void ext4_inode_bitmap_set(struct super_block *sb,
372 struct ext4_group_desc *bg, ext4_fsblk_t blk)
374 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
375 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
376 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
379 void ext4_inode_table_set(struct super_block *sb,
380 struct ext4_group_desc *bg, ext4_fsblk_t blk)
382 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
383 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
384 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
387 void ext4_free_group_clusters_set(struct super_block *sb,
388 struct ext4_group_desc *bg, __u32 count)
390 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
391 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
392 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
395 void ext4_free_inodes_set(struct super_block *sb,
396 struct ext4_group_desc *bg, __u32 count)
398 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
399 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
400 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
403 void ext4_used_dirs_set(struct super_block *sb,
404 struct ext4_group_desc *bg, __u32 count)
406 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
407 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
408 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
411 void ext4_itable_unused_set(struct super_block *sb,
412 struct ext4_group_desc *bg, __u32 count)
414 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
415 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
416 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
419 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
421 now = clamp_val(now, 0, (1ull << 40) - 1);
423 *lo = cpu_to_le32(lower_32_bits(now));
424 *hi = upper_32_bits(now);
427 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
429 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
431 #define ext4_update_tstamp(es, tstamp) \
432 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
433 ktime_get_real_seconds())
434 #define ext4_get_tstamp(es, tstamp) \
435 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
438 * The del_gendisk() function uninitializes the disk-specific data
439 * structures, including the bdi structure, without telling anyone
440 * else. Once this happens, any attempt to call mark_buffer_dirty()
441 * (for example, by ext4_commit_super), will cause a kernel OOPS.
442 * This is a kludge to prevent these oops until we can put in a proper
443 * hook in del_gendisk() to inform the VFS and file system layers.
445 static int block_device_ejected(struct super_block *sb)
447 struct inode *bd_inode = sb->s_bdev->bd_inode;
448 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
450 return bdi->dev == NULL;
453 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
455 struct super_block *sb = journal->j_private;
456 struct ext4_sb_info *sbi = EXT4_SB(sb);
457 int error = is_journal_aborted(journal);
458 struct ext4_journal_cb_entry *jce;
460 BUG_ON(txn->t_state == T_FINISHED);
462 ext4_process_freed_data(sb, txn->t_tid);
464 spin_lock(&sbi->s_md_lock);
465 while (!list_empty(&txn->t_private_list)) {
466 jce = list_entry(txn->t_private_list.next,
467 struct ext4_journal_cb_entry, jce_list);
468 list_del_init(&jce->jce_list);
469 spin_unlock(&sbi->s_md_lock);
470 jce->jce_func(sb, jce, error);
471 spin_lock(&sbi->s_md_lock);
473 spin_unlock(&sbi->s_md_lock);
477 * This writepage callback for write_cache_pages()
478 * takes care of a few cases after page cleaning.
480 * write_cache_pages() already checks for dirty pages
481 * and calls clear_page_dirty_for_io(), which we want,
482 * to write protect the pages.
484 * However, we may have to redirty a page (see below.)
486 static int ext4_journalled_writepage_callback(struct folio *folio,
487 struct writeback_control *wbc,
490 transaction_t *transaction = (transaction_t *) data;
491 struct buffer_head *bh, *head;
492 struct journal_head *jh;
494 bh = head = folio_buffers(folio);
497 * We have to redirty a page in these cases:
498 * 1) If buffer is dirty, it means the page was dirty because it
499 * contains a buffer that needs checkpointing. So the dirty bit
500 * needs to be preserved so that checkpointing writes the buffer
502 * 2) If buffer is not part of the committing transaction
503 * (we may have just accidentally come across this buffer because
504 * inode range tracking is not exact) or if the currently running
505 * transaction already contains this buffer as well, dirty bit
506 * needs to be preserved so that the buffer gets writeprotected
507 * properly on running transaction's commit.
510 if (buffer_dirty(bh) ||
511 (jh && (jh->b_transaction != transaction ||
512 jh->b_next_transaction))) {
513 folio_redirty_for_writepage(wbc, folio);
516 } while ((bh = bh->b_this_page) != head);
519 return AOP_WRITEPAGE_ACTIVATE;
522 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
524 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
525 struct writeback_control wbc = {
526 .sync_mode = WB_SYNC_ALL,
527 .nr_to_write = LONG_MAX,
528 .range_start = jinode->i_dirty_start,
529 .range_end = jinode->i_dirty_end,
532 return write_cache_pages(mapping, &wbc,
533 ext4_journalled_writepage_callback,
534 jinode->i_transaction);
537 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
541 if (ext4_should_journal_data(jinode->i_vfs_inode))
542 ret = ext4_journalled_submit_inode_data_buffers(jinode);
544 ret = ext4_normal_submit_inode_data_buffers(jinode);
548 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
552 if (!ext4_should_journal_data(jinode->i_vfs_inode))
553 ret = jbd2_journal_finish_inode_data_buffers(jinode);
558 static bool system_going_down(void)
560 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
561 || system_state == SYSTEM_RESTART;
564 struct ext4_err_translation {
569 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
571 static struct ext4_err_translation err_translation[] = {
572 EXT4_ERR_TRANSLATE(EIO),
573 EXT4_ERR_TRANSLATE(ENOMEM),
574 EXT4_ERR_TRANSLATE(EFSBADCRC),
575 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
576 EXT4_ERR_TRANSLATE(ENOSPC),
577 EXT4_ERR_TRANSLATE(ENOKEY),
578 EXT4_ERR_TRANSLATE(EROFS),
579 EXT4_ERR_TRANSLATE(EFBIG),
580 EXT4_ERR_TRANSLATE(EEXIST),
581 EXT4_ERR_TRANSLATE(ERANGE),
582 EXT4_ERR_TRANSLATE(EOVERFLOW),
583 EXT4_ERR_TRANSLATE(EBUSY),
584 EXT4_ERR_TRANSLATE(ENOTDIR),
585 EXT4_ERR_TRANSLATE(ENOTEMPTY),
586 EXT4_ERR_TRANSLATE(ESHUTDOWN),
587 EXT4_ERR_TRANSLATE(EFAULT),
590 static int ext4_errno_to_code(int errno)
594 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
595 if (err_translation[i].errno == errno)
596 return err_translation[i].code;
597 return EXT4_ERR_UNKNOWN;
600 static void save_error_info(struct super_block *sb, int error,
601 __u32 ino, __u64 block,
602 const char *func, unsigned int line)
604 struct ext4_sb_info *sbi = EXT4_SB(sb);
606 /* We default to EFSCORRUPTED error... */
608 error = EFSCORRUPTED;
610 spin_lock(&sbi->s_error_lock);
611 sbi->s_add_error_count++;
612 sbi->s_last_error_code = error;
613 sbi->s_last_error_line = line;
614 sbi->s_last_error_ino = ino;
615 sbi->s_last_error_block = block;
616 sbi->s_last_error_func = func;
617 sbi->s_last_error_time = ktime_get_real_seconds();
618 if (!sbi->s_first_error_time) {
619 sbi->s_first_error_code = error;
620 sbi->s_first_error_line = line;
621 sbi->s_first_error_ino = ino;
622 sbi->s_first_error_block = block;
623 sbi->s_first_error_func = func;
624 sbi->s_first_error_time = sbi->s_last_error_time;
626 spin_unlock(&sbi->s_error_lock);
629 /* Deal with the reporting of failure conditions on a filesystem such as
630 * inconsistencies detected or read IO failures.
632 * On ext2, we can store the error state of the filesystem in the
633 * superblock. That is not possible on ext4, because we may have other
634 * write ordering constraints on the superblock which prevent us from
635 * writing it out straight away; and given that the journal is about to
636 * be aborted, we can't rely on the current, or future, transactions to
637 * write out the superblock safely.
639 * We'll just use the jbd2_journal_abort() error code to record an error in
640 * the journal instead. On recovery, the journal will complain about
641 * that error until we've noted it down and cleared it.
643 * If force_ro is set, we unconditionally force the filesystem into an
644 * ABORT|READONLY state, unless the error response on the fs has been set to
645 * panic in which case we take the easy way out and panic immediately. This is
646 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
647 * at a critical moment in log management.
649 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
650 __u32 ino, __u64 block,
651 const char *func, unsigned int line)
653 journal_t *journal = EXT4_SB(sb)->s_journal;
654 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
656 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
657 if (test_opt(sb, WARN_ON_ERROR))
660 if (!continue_fs && !sb_rdonly(sb)) {
661 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
663 jbd2_journal_abort(journal, -EIO);
666 if (!bdev_read_only(sb->s_bdev)) {
667 save_error_info(sb, error, ino, block, func, line);
669 * In case the fs should keep running, we need to writeout
670 * superblock through the journal. Due to lock ordering
671 * constraints, it may not be safe to do it right here so we
672 * defer superblock flushing to a workqueue.
674 if (continue_fs && journal)
675 schedule_work(&EXT4_SB(sb)->s_error_work);
677 ext4_commit_super(sb);
681 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
682 * could panic during 'reboot -f' as the underlying device got already
685 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
686 panic("EXT4-fs (device %s): panic forced after error\n",
690 if (sb_rdonly(sb) || continue_fs)
693 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
695 * Make sure updated value of ->s_mount_flags will be visible before
699 sb->s_flags |= SB_RDONLY;
702 static void flush_stashed_error_work(struct work_struct *work)
704 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
706 journal_t *journal = sbi->s_journal;
710 * If the journal is still running, we have to write out superblock
711 * through the journal to avoid collisions of other journalled sb
714 * We use directly jbd2 functions here to avoid recursing back into
715 * ext4 error handling code during handling of previous errors.
717 if (!sb_rdonly(sbi->s_sb) && journal) {
718 struct buffer_head *sbh = sbi->s_sbh;
719 handle = jbd2_journal_start(journal, 1);
722 if (jbd2_journal_get_write_access(handle, sbh)) {
723 jbd2_journal_stop(handle);
726 ext4_update_super(sbi->s_sb);
727 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
728 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
729 "superblock detected");
730 clear_buffer_write_io_error(sbh);
731 set_buffer_uptodate(sbh);
734 if (jbd2_journal_dirty_metadata(handle, sbh)) {
735 jbd2_journal_stop(handle);
738 jbd2_journal_stop(handle);
739 ext4_notify_error_sysfs(sbi);
744 * Write through journal failed. Write sb directly to get error info
745 * out and hope for the best.
747 ext4_commit_super(sbi->s_sb);
748 ext4_notify_error_sysfs(sbi);
751 #define ext4_error_ratelimit(sb) \
752 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
755 void __ext4_error(struct super_block *sb, const char *function,
756 unsigned int line, bool force_ro, int error, __u64 block,
757 const char *fmt, ...)
759 struct va_format vaf;
762 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
765 trace_ext4_error(sb, function, line);
766 if (ext4_error_ratelimit(sb)) {
771 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
772 sb->s_id, function, line, current->comm, &vaf);
775 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
777 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
780 void __ext4_error_inode(struct inode *inode, const char *function,
781 unsigned int line, ext4_fsblk_t block, int error,
782 const char *fmt, ...)
785 struct va_format vaf;
787 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
790 trace_ext4_error(inode->i_sb, function, line);
791 if (ext4_error_ratelimit(inode->i_sb)) {
796 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
797 "inode #%lu: block %llu: comm %s: %pV\n",
798 inode->i_sb->s_id, function, line, inode->i_ino,
799 block, current->comm, &vaf);
801 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
802 "inode #%lu: comm %s: %pV\n",
803 inode->i_sb->s_id, function, line, inode->i_ino,
804 current->comm, &vaf);
807 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
809 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
813 void __ext4_error_file(struct file *file, const char *function,
814 unsigned int line, ext4_fsblk_t block,
815 const char *fmt, ...)
818 struct va_format vaf;
819 struct inode *inode = file_inode(file);
820 char pathname[80], *path;
822 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
825 trace_ext4_error(inode->i_sb, function, line);
826 if (ext4_error_ratelimit(inode->i_sb)) {
827 path = file_path(file, pathname, sizeof(pathname));
835 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
836 "block %llu: comm %s: path %s: %pV\n",
837 inode->i_sb->s_id, function, line, inode->i_ino,
838 block, current->comm, path, &vaf);
841 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
842 "comm %s: path %s: %pV\n",
843 inode->i_sb->s_id, function, line, inode->i_ino,
844 current->comm, path, &vaf);
847 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
849 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
853 const char *ext4_decode_error(struct super_block *sb, int errno,
860 errstr = "Corrupt filesystem";
863 errstr = "Filesystem failed CRC";
866 errstr = "IO failure";
869 errstr = "Out of memory";
872 if (!sb || (EXT4_SB(sb)->s_journal &&
873 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
874 errstr = "Journal has aborted";
876 errstr = "Readonly filesystem";
879 /* If the caller passed in an extra buffer for unknown
880 * errors, textualise them now. Else we just return
883 /* Check for truncated error codes... */
884 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
893 /* __ext4_std_error decodes expected errors from journaling functions
894 * automatically and invokes the appropriate error response. */
896 void __ext4_std_error(struct super_block *sb, const char *function,
897 unsigned int line, int errno)
902 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
905 /* Special case: if the error is EROFS, and we're not already
906 * inside a transaction, then there's really no point in logging
908 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
911 if (ext4_error_ratelimit(sb)) {
912 errstr = ext4_decode_error(sb, errno, nbuf);
913 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
914 sb->s_id, function, line, errstr);
916 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
918 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
921 void __ext4_msg(struct super_block *sb,
922 const char *prefix, const char *fmt, ...)
924 struct va_format vaf;
928 atomic_inc(&EXT4_SB(sb)->s_msg_count);
929 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
938 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
940 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
944 static int ext4_warning_ratelimit(struct super_block *sb)
946 atomic_inc(&EXT4_SB(sb)->s_warning_count);
947 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
951 void __ext4_warning(struct super_block *sb, const char *function,
952 unsigned int line, const char *fmt, ...)
954 struct va_format vaf;
957 if (!ext4_warning_ratelimit(sb))
963 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
964 sb->s_id, function, line, &vaf);
968 void __ext4_warning_inode(const struct inode *inode, const char *function,
969 unsigned int line, const char *fmt, ...)
971 struct va_format vaf;
974 if (!ext4_warning_ratelimit(inode->i_sb))
980 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
981 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
982 function, line, inode->i_ino, current->comm, &vaf);
986 void __ext4_grp_locked_error(const char *function, unsigned int line,
987 struct super_block *sb, ext4_group_t grp,
988 unsigned long ino, ext4_fsblk_t block,
989 const char *fmt, ...)
993 struct va_format vaf;
996 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
999 trace_ext4_error(sb, function, line);
1000 if (ext4_error_ratelimit(sb)) {
1001 va_start(args, fmt);
1004 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1005 sb->s_id, function, line, grp);
1007 printk(KERN_CONT "inode %lu: ", ino);
1009 printk(KERN_CONT "block %llu:",
1010 (unsigned long long) block);
1011 printk(KERN_CONT "%pV\n", &vaf);
1015 if (test_opt(sb, ERRORS_CONT)) {
1016 if (test_opt(sb, WARN_ON_ERROR))
1018 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1019 if (!bdev_read_only(sb->s_bdev)) {
1020 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1022 schedule_work(&EXT4_SB(sb)->s_error_work);
1026 ext4_unlock_group(sb, grp);
1027 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1029 * We only get here in the ERRORS_RO case; relocking the group
1030 * may be dangerous, but nothing bad will happen since the
1031 * filesystem will have already been marked read/only and the
1032 * journal has been aborted. We return 1 as a hint to callers
1033 * who might what to use the return value from
1034 * ext4_grp_locked_error() to distinguish between the
1035 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1036 * aggressively from the ext4 function in question, with a
1037 * more appropriate error code.
1039 ext4_lock_group(sb, grp);
1043 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1047 struct ext4_sb_info *sbi = EXT4_SB(sb);
1048 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1049 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1054 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1055 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1058 percpu_counter_sub(&sbi->s_freeclusters_counter,
1062 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1063 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1068 count = ext4_free_inodes_count(sb, gdp);
1069 percpu_counter_sub(&sbi->s_freeinodes_counter,
1075 void ext4_update_dynamic_rev(struct super_block *sb)
1077 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1079 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1083 "updating to rev %d because of new feature flag, "
1084 "running e2fsck is recommended",
1087 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1088 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1089 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1090 /* leave es->s_feature_*compat flags alone */
1091 /* es->s_uuid will be set by e2fsck if empty */
1094 * The rest of the superblock fields should be zero, and if not it
1095 * means they are likely already in use, so leave them alone. We
1096 * can leave it up to e2fsck to clean up any inconsistencies there.
1100 static void ext4_bdev_mark_dead(struct block_device *bdev)
1102 ext4_force_shutdown(bdev->bd_holder, EXT4_GOING_FLAGS_NOLOGFLUSH);
1105 static const struct blk_holder_ops ext4_holder_ops = {
1106 .mark_dead = ext4_bdev_mark_dead,
1110 * Open the external journal device
1112 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1114 struct block_device *bdev;
1116 bdev = blkdev_get_by_dev(dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb,
1123 ext4_msg(sb, KERN_ERR,
1124 "failed to open journal device unknown-block(%u,%u) %ld",
1125 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1129 static inline struct inode *orphan_list_entry(struct list_head *l)
1131 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1134 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1136 struct list_head *l;
1138 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1139 le32_to_cpu(sbi->s_es->s_last_orphan));
1141 printk(KERN_ERR "sb_info orphan list:\n");
1142 list_for_each(l, &sbi->s_orphan) {
1143 struct inode *inode = orphan_list_entry(l);
1145 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1146 inode->i_sb->s_id, inode->i_ino, inode,
1147 inode->i_mode, inode->i_nlink,
1148 NEXT_ORPHAN(inode));
1153 static int ext4_quota_off(struct super_block *sb, int type);
1155 static inline void ext4_quotas_off(struct super_block *sb, int type)
1157 BUG_ON(type > EXT4_MAXQUOTAS);
1159 /* Use our quota_off function to clear inode flags etc. */
1160 for (type--; type >= 0; type--)
1161 ext4_quota_off(sb, type);
1165 * This is a helper function which is used in the mount/remount
1166 * codepaths (which holds s_umount) to fetch the quota file name.
1168 static inline char *get_qf_name(struct super_block *sb,
1169 struct ext4_sb_info *sbi,
1172 return rcu_dereference_protected(sbi->s_qf_names[type],
1173 lockdep_is_held(&sb->s_umount));
1176 static inline void ext4_quotas_off(struct super_block *sb, int type)
1181 static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1186 block = ext4_count_free_clusters(sbi->s_sb);
1187 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1188 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1191 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1192 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1193 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1197 err = percpu_counter_init(&sbi->s_dirs_counter,
1198 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1200 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1203 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1206 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1209 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1214 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1216 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1217 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1218 percpu_counter_destroy(&sbi->s_dirs_counter);
1219 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1220 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1221 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1224 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1226 struct buffer_head **group_desc;
1230 group_desc = rcu_dereference(sbi->s_group_desc);
1231 for (i = 0; i < sbi->s_gdb_count; i++)
1232 brelse(group_desc[i]);
1237 static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1239 struct flex_groups **flex_groups;
1243 flex_groups = rcu_dereference(sbi->s_flex_groups);
1245 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1246 kvfree(flex_groups[i]);
1247 kvfree(flex_groups);
1252 static void ext4_put_super(struct super_block *sb)
1254 struct ext4_sb_info *sbi = EXT4_SB(sb);
1255 struct ext4_super_block *es = sbi->s_es;
1260 * Unregister sysfs before destroying jbd2 journal.
1261 * Since we could still access attr_journal_task attribute via sysfs
1262 * path which could have sbi->s_journal->j_task as NULL
1263 * Unregister sysfs before flush sbi->s_error_work.
1264 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1265 * read metadata verify failed then will queue error work.
1266 * flush_stashed_error_work will call start_this_handle may trigger
1269 ext4_unregister_sysfs(sb);
1271 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1272 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1275 ext4_unregister_li_request(sb);
1276 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1278 flush_work(&sbi->s_error_work);
1279 destroy_workqueue(sbi->rsv_conversion_wq);
1280 ext4_release_orphan_info(sb);
1282 if (sbi->s_journal) {
1283 aborted = is_journal_aborted(sbi->s_journal);
1284 err = jbd2_journal_destroy(sbi->s_journal);
1285 sbi->s_journal = NULL;
1286 if ((err < 0) && !aborted) {
1287 ext4_abort(sb, -err, "Couldn't clean up the journal");
1291 ext4_es_unregister_shrinker(sbi);
1292 timer_shutdown_sync(&sbi->s_err_report);
1293 ext4_release_system_zone(sb);
1294 ext4_mb_release(sb);
1295 ext4_ext_release(sb);
1297 if (!sb_rdonly(sb) && !aborted) {
1298 ext4_clear_feature_journal_needs_recovery(sb);
1299 ext4_clear_feature_orphan_present(sb);
1300 es->s_state = cpu_to_le16(sbi->s_mount_state);
1303 ext4_commit_super(sb);
1305 ext4_group_desc_free(sbi);
1306 ext4_flex_groups_free(sbi);
1307 ext4_percpu_param_destroy(sbi);
1309 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1310 kfree(get_qf_name(sb, sbi, i));
1313 /* Debugging code just in case the in-memory inode orphan list
1314 * isn't empty. The on-disk one can be non-empty if we've
1315 * detected an error and taken the fs readonly, but the
1316 * in-memory list had better be clean by this point. */
1317 if (!list_empty(&sbi->s_orphan))
1318 dump_orphan_list(sb, sbi);
1319 ASSERT(list_empty(&sbi->s_orphan));
1321 sync_blockdev(sb->s_bdev);
1322 invalidate_bdev(sb->s_bdev);
1323 if (sbi->s_journal_bdev) {
1325 * Invalidate the journal device's buffers. We don't want them
1326 * floating about in memory - the physical journal device may
1327 * hotswapped, and it breaks the `ro-after' testing code.
1329 sync_blockdev(sbi->s_journal_bdev);
1330 invalidate_bdev(sbi->s_journal_bdev);
1333 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1334 sbi->s_ea_inode_cache = NULL;
1336 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1337 sbi->s_ea_block_cache = NULL;
1339 ext4_stop_mmpd(sbi);
1342 sb->s_fs_info = NULL;
1344 * Now that we are completely done shutting down the
1345 * superblock, we need to actually destroy the kobject.
1347 kobject_put(&sbi->s_kobj);
1348 wait_for_completion(&sbi->s_kobj_unregister);
1349 if (sbi->s_chksum_driver)
1350 crypto_free_shash(sbi->s_chksum_driver);
1351 kfree(sbi->s_blockgroup_lock);
1352 fs_put_dax(sbi->s_daxdev, NULL);
1353 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1354 #if IS_ENABLED(CONFIG_UNICODE)
1355 utf8_unload(sb->s_encoding);
1360 static struct kmem_cache *ext4_inode_cachep;
1363 * Called inside transaction, so use GFP_NOFS
1365 static struct inode *ext4_alloc_inode(struct super_block *sb)
1367 struct ext4_inode_info *ei;
1369 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1373 inode_set_iversion(&ei->vfs_inode, 1);
1375 spin_lock_init(&ei->i_raw_lock);
1376 ei->i_prealloc_node = RB_ROOT;
1377 atomic_set(&ei->i_prealloc_active, 0);
1378 rwlock_init(&ei->i_prealloc_lock);
1379 ext4_es_init_tree(&ei->i_es_tree);
1380 rwlock_init(&ei->i_es_lock);
1381 INIT_LIST_HEAD(&ei->i_es_list);
1382 ei->i_es_all_nr = 0;
1383 ei->i_es_shk_nr = 0;
1384 ei->i_es_shrink_lblk = 0;
1385 ei->i_reserved_data_blocks = 0;
1386 spin_lock_init(&(ei->i_block_reservation_lock));
1387 ext4_init_pending_tree(&ei->i_pending_tree);
1389 ei->i_reserved_quota = 0;
1390 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1393 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1394 spin_lock_init(&ei->i_completed_io_lock);
1396 ei->i_datasync_tid = 0;
1397 atomic_set(&ei->i_unwritten, 0);
1398 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1399 ext4_fc_init_inode(&ei->vfs_inode);
1400 mutex_init(&ei->i_fc_lock);
1401 return &ei->vfs_inode;
1404 static int ext4_drop_inode(struct inode *inode)
1406 int drop = generic_drop_inode(inode);
1409 drop = fscrypt_drop_inode(inode);
1411 trace_ext4_drop_inode(inode, drop);
1415 static void ext4_free_in_core_inode(struct inode *inode)
1417 fscrypt_free_inode(inode);
1418 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1419 pr_warn("%s: inode %ld still in fc list",
1420 __func__, inode->i_ino);
1422 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1425 static void ext4_destroy_inode(struct inode *inode)
1427 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1428 ext4_msg(inode->i_sb, KERN_ERR,
1429 "Inode %lu (%p): orphan list check failed!",
1430 inode->i_ino, EXT4_I(inode));
1431 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1432 EXT4_I(inode), sizeof(struct ext4_inode_info),
1437 if (EXT4_I(inode)->i_reserved_data_blocks)
1438 ext4_msg(inode->i_sb, KERN_ERR,
1439 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1440 inode->i_ino, EXT4_I(inode),
1441 EXT4_I(inode)->i_reserved_data_blocks);
1444 static void ext4_shutdown(struct super_block *sb)
1446 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1449 static void init_once(void *foo)
1451 struct ext4_inode_info *ei = foo;
1453 INIT_LIST_HEAD(&ei->i_orphan);
1454 init_rwsem(&ei->xattr_sem);
1455 init_rwsem(&ei->i_data_sem);
1456 inode_init_once(&ei->vfs_inode);
1457 ext4_fc_init_inode(&ei->vfs_inode);
1460 static int __init init_inodecache(void)
1462 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1463 sizeof(struct ext4_inode_info), 0,
1464 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1466 offsetof(struct ext4_inode_info, i_data),
1467 sizeof_field(struct ext4_inode_info, i_data),
1469 if (ext4_inode_cachep == NULL)
1474 static void destroy_inodecache(void)
1477 * Make sure all delayed rcu free inodes are flushed before we
1481 kmem_cache_destroy(ext4_inode_cachep);
1484 void ext4_clear_inode(struct inode *inode)
1487 invalidate_inode_buffers(inode);
1489 ext4_discard_preallocations(inode, 0);
1490 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1492 if (EXT4_I(inode)->jinode) {
1493 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1494 EXT4_I(inode)->jinode);
1495 jbd2_free_inode(EXT4_I(inode)->jinode);
1496 EXT4_I(inode)->jinode = NULL;
1498 fscrypt_put_encryption_info(inode);
1499 fsverity_cleanup_inode(inode);
1502 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1503 u64 ino, u32 generation)
1505 struct inode *inode;
1508 * Currently we don't know the generation for parent directory, so
1509 * a generation of 0 means "accept any"
1511 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1513 return ERR_CAST(inode);
1514 if (generation && inode->i_generation != generation) {
1516 return ERR_PTR(-ESTALE);
1522 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1523 int fh_len, int fh_type)
1525 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1526 ext4_nfs_get_inode);
1529 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1530 int fh_len, int fh_type)
1532 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1533 ext4_nfs_get_inode);
1536 static int ext4_nfs_commit_metadata(struct inode *inode)
1538 struct writeback_control wbc = {
1539 .sync_mode = WB_SYNC_ALL
1542 trace_ext4_nfs_commit_metadata(inode);
1543 return ext4_write_inode(inode, &wbc);
1547 static const char * const quotatypes[] = INITQFNAMES;
1548 #define QTYPE2NAME(t) (quotatypes[t])
1550 static int ext4_write_dquot(struct dquot *dquot);
1551 static int ext4_acquire_dquot(struct dquot *dquot);
1552 static int ext4_release_dquot(struct dquot *dquot);
1553 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1554 static int ext4_write_info(struct super_block *sb, int type);
1555 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1556 const struct path *path);
1557 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1558 size_t len, loff_t off);
1559 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1560 const char *data, size_t len, loff_t off);
1561 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1562 unsigned int flags);
1564 static struct dquot **ext4_get_dquots(struct inode *inode)
1566 return EXT4_I(inode)->i_dquot;
1569 static const struct dquot_operations ext4_quota_operations = {
1570 .get_reserved_space = ext4_get_reserved_space,
1571 .write_dquot = ext4_write_dquot,
1572 .acquire_dquot = ext4_acquire_dquot,
1573 .release_dquot = ext4_release_dquot,
1574 .mark_dirty = ext4_mark_dquot_dirty,
1575 .write_info = ext4_write_info,
1576 .alloc_dquot = dquot_alloc,
1577 .destroy_dquot = dquot_destroy,
1578 .get_projid = ext4_get_projid,
1579 .get_inode_usage = ext4_get_inode_usage,
1580 .get_next_id = dquot_get_next_id,
1583 static const struct quotactl_ops ext4_qctl_operations = {
1584 .quota_on = ext4_quota_on,
1585 .quota_off = ext4_quota_off,
1586 .quota_sync = dquot_quota_sync,
1587 .get_state = dquot_get_state,
1588 .set_info = dquot_set_dqinfo,
1589 .get_dqblk = dquot_get_dqblk,
1590 .set_dqblk = dquot_set_dqblk,
1591 .get_nextdqblk = dquot_get_next_dqblk,
1595 static const struct super_operations ext4_sops = {
1596 .alloc_inode = ext4_alloc_inode,
1597 .free_inode = ext4_free_in_core_inode,
1598 .destroy_inode = ext4_destroy_inode,
1599 .write_inode = ext4_write_inode,
1600 .dirty_inode = ext4_dirty_inode,
1601 .drop_inode = ext4_drop_inode,
1602 .evict_inode = ext4_evict_inode,
1603 .put_super = ext4_put_super,
1604 .sync_fs = ext4_sync_fs,
1605 .freeze_fs = ext4_freeze,
1606 .unfreeze_fs = ext4_unfreeze,
1607 .statfs = ext4_statfs,
1608 .show_options = ext4_show_options,
1609 .shutdown = ext4_shutdown,
1611 .quota_read = ext4_quota_read,
1612 .quota_write = ext4_quota_write,
1613 .get_dquots = ext4_get_dquots,
1617 static const struct export_operations ext4_export_ops = {
1618 .fh_to_dentry = ext4_fh_to_dentry,
1619 .fh_to_parent = ext4_fh_to_parent,
1620 .get_parent = ext4_get_parent,
1621 .commit_metadata = ext4_nfs_commit_metadata,
1625 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1626 Opt_resgid, Opt_resuid, Opt_sb,
1627 Opt_nouid32, Opt_debug, Opt_removed,
1628 Opt_user_xattr, Opt_acl,
1629 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1630 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1631 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1632 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1633 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1635 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1636 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1637 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1638 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1639 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1640 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1641 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1642 Opt_inode_readahead_blks, Opt_journal_ioprio,
1643 Opt_dioread_nolock, Opt_dioread_lock,
1644 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1645 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1646 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1647 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1648 #ifdef CONFIG_EXT4_DEBUG
1649 Opt_fc_debug_max_replay, Opt_fc_debug_force
1653 static const struct constant_table ext4_param_errors[] = {
1654 {"continue", EXT4_MOUNT_ERRORS_CONT},
1655 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1656 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1660 static const struct constant_table ext4_param_data[] = {
1661 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1662 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1663 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1667 static const struct constant_table ext4_param_data_err[] = {
1668 {"abort", Opt_data_err_abort},
1669 {"ignore", Opt_data_err_ignore},
1673 static const struct constant_table ext4_param_jqfmt[] = {
1674 {"vfsold", QFMT_VFS_OLD},
1675 {"vfsv0", QFMT_VFS_V0},
1676 {"vfsv1", QFMT_VFS_V1},
1680 static const struct constant_table ext4_param_dax[] = {
1681 {"always", Opt_dax_always},
1682 {"inode", Opt_dax_inode},
1683 {"never", Opt_dax_never},
1687 /* String parameter that allows empty argument */
1688 #define fsparam_string_empty(NAME, OPT) \
1689 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1692 * Mount option specification
1693 * We don't use fsparam_flag_no because of the way we set the
1694 * options and the way we show them in _ext4_show_options(). To
1695 * keep the changes to a minimum, let's keep the negative options
1698 static const struct fs_parameter_spec ext4_param_specs[] = {
1699 fsparam_flag ("bsddf", Opt_bsd_df),
1700 fsparam_flag ("minixdf", Opt_minix_df),
1701 fsparam_flag ("grpid", Opt_grpid),
1702 fsparam_flag ("bsdgroups", Opt_grpid),
1703 fsparam_flag ("nogrpid", Opt_nogrpid),
1704 fsparam_flag ("sysvgroups", Opt_nogrpid),
1705 fsparam_u32 ("resgid", Opt_resgid),
1706 fsparam_u32 ("resuid", Opt_resuid),
1707 fsparam_u32 ("sb", Opt_sb),
1708 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1709 fsparam_flag ("nouid32", Opt_nouid32),
1710 fsparam_flag ("debug", Opt_debug),
1711 fsparam_flag ("oldalloc", Opt_removed),
1712 fsparam_flag ("orlov", Opt_removed),
1713 fsparam_flag ("user_xattr", Opt_user_xattr),
1714 fsparam_flag ("acl", Opt_acl),
1715 fsparam_flag ("norecovery", Opt_noload),
1716 fsparam_flag ("noload", Opt_noload),
1717 fsparam_flag ("bh", Opt_removed),
1718 fsparam_flag ("nobh", Opt_removed),
1719 fsparam_u32 ("commit", Opt_commit),
1720 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1721 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1722 fsparam_u32 ("journal_dev", Opt_journal_dev),
1723 fsparam_bdev ("journal_path", Opt_journal_path),
1724 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1725 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1726 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1727 fsparam_flag ("abort", Opt_abort),
1728 fsparam_enum ("data", Opt_data, ext4_param_data),
1729 fsparam_enum ("data_err", Opt_data_err,
1730 ext4_param_data_err),
1731 fsparam_string_empty
1732 ("usrjquota", Opt_usrjquota),
1733 fsparam_string_empty
1734 ("grpjquota", Opt_grpjquota),
1735 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1736 fsparam_flag ("grpquota", Opt_grpquota),
1737 fsparam_flag ("quota", Opt_quota),
1738 fsparam_flag ("noquota", Opt_noquota),
1739 fsparam_flag ("usrquota", Opt_usrquota),
1740 fsparam_flag ("prjquota", Opt_prjquota),
1741 fsparam_flag ("barrier", Opt_barrier),
1742 fsparam_u32 ("barrier", Opt_barrier),
1743 fsparam_flag ("nobarrier", Opt_nobarrier),
1744 fsparam_flag ("i_version", Opt_removed),
1745 fsparam_flag ("dax", Opt_dax),
1746 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1747 fsparam_u32 ("stripe", Opt_stripe),
1748 fsparam_flag ("delalloc", Opt_delalloc),
1749 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1750 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1751 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1752 fsparam_u32 ("debug_want_extra_isize",
1753 Opt_debug_want_extra_isize),
1754 fsparam_flag ("mblk_io_submit", Opt_removed),
1755 fsparam_flag ("nomblk_io_submit", Opt_removed),
1756 fsparam_flag ("block_validity", Opt_block_validity),
1757 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1758 fsparam_u32 ("inode_readahead_blks",
1759 Opt_inode_readahead_blks),
1760 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1761 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1762 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1763 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1764 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1765 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1766 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1767 fsparam_flag ("discard", Opt_discard),
1768 fsparam_flag ("nodiscard", Opt_nodiscard),
1769 fsparam_u32 ("init_itable", Opt_init_itable),
1770 fsparam_flag ("init_itable", Opt_init_itable),
1771 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1772 #ifdef CONFIG_EXT4_DEBUG
1773 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1774 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1776 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1777 fsparam_flag ("test_dummy_encryption",
1778 Opt_test_dummy_encryption),
1779 fsparam_string ("test_dummy_encryption",
1780 Opt_test_dummy_encryption),
1781 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1782 fsparam_flag ("nombcache", Opt_nombcache),
1783 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1784 fsparam_flag ("prefetch_block_bitmaps",
1786 fsparam_flag ("no_prefetch_block_bitmaps",
1787 Opt_no_prefetch_block_bitmaps),
1788 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1789 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1790 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1791 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1792 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1793 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1797 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1799 #define MOPT_SET 0x0001
1800 #define MOPT_CLEAR 0x0002
1801 #define MOPT_NOSUPPORT 0x0004
1802 #define MOPT_EXPLICIT 0x0008
1805 #define MOPT_QFMT 0x0010
1807 #define MOPT_Q MOPT_NOSUPPORT
1808 #define MOPT_QFMT MOPT_NOSUPPORT
1810 #define MOPT_NO_EXT2 0x0020
1811 #define MOPT_NO_EXT3 0x0040
1812 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1813 #define MOPT_SKIP 0x0080
1814 #define MOPT_2 0x0100
1816 static const struct mount_opts {
1820 } ext4_mount_opts[] = {
1821 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1822 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1823 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1824 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1825 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1826 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1827 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1828 MOPT_EXT4_ONLY | MOPT_SET},
1829 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1830 MOPT_EXT4_ONLY | MOPT_CLEAR},
1831 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1832 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1833 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1834 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1835 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1836 MOPT_EXT4_ONLY | MOPT_CLEAR},
1837 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1838 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1839 {Opt_commit, 0, MOPT_NO_EXT2},
1840 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1841 MOPT_EXT4_ONLY | MOPT_CLEAR},
1842 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1843 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1844 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1845 EXT4_MOUNT_JOURNAL_CHECKSUM),
1846 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1847 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1848 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1849 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1850 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1851 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1852 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1853 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1854 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1855 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1856 {Opt_journal_path, 0, MOPT_NO_EXT2},
1857 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1858 {Opt_data, 0, MOPT_NO_EXT2},
1859 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1860 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1861 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1863 {Opt_acl, 0, MOPT_NOSUPPORT},
1865 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1866 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1867 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1868 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1870 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1872 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1874 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1875 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1876 MOPT_CLEAR | MOPT_Q},
1877 {Opt_usrjquota, 0, MOPT_Q},
1878 {Opt_grpjquota, 0, MOPT_Q},
1879 {Opt_jqfmt, 0, MOPT_QFMT},
1880 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1881 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1883 #ifdef CONFIG_EXT4_DEBUG
1884 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1885 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1890 #if IS_ENABLED(CONFIG_UNICODE)
1891 static const struct ext4_sb_encodings {
1894 unsigned int version;
1895 } ext4_sb_encoding_map[] = {
1896 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1899 static const struct ext4_sb_encodings *
1900 ext4_sb_read_encoding(const struct ext4_super_block *es)
1902 __u16 magic = le16_to_cpu(es->s_encoding);
1905 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1906 if (magic == ext4_sb_encoding_map[i].magic)
1907 return &ext4_sb_encoding_map[i];
1913 #define EXT4_SPEC_JQUOTA (1 << 0)
1914 #define EXT4_SPEC_JQFMT (1 << 1)
1915 #define EXT4_SPEC_DATAJ (1 << 2)
1916 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1917 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1918 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1919 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1920 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1921 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1922 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1923 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1924 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1925 #define EXT4_SPEC_s_stripe (1 << 13)
1926 #define EXT4_SPEC_s_resuid (1 << 14)
1927 #define EXT4_SPEC_s_resgid (1 << 15)
1928 #define EXT4_SPEC_s_commit_interval (1 << 16)
1929 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1930 #define EXT4_SPEC_s_sb_block (1 << 18)
1931 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1933 struct ext4_fs_context {
1934 char *s_qf_names[EXT4_MAXQUOTAS];
1935 struct fscrypt_dummy_policy dummy_enc_policy;
1936 int s_jquota_fmt; /* Format of quota to use */
1937 #ifdef CONFIG_EXT4_DEBUG
1938 int s_fc_debug_max_replay;
1940 unsigned short qname_spec;
1941 unsigned long vals_s_flags; /* Bits to set in s_flags */
1942 unsigned long mask_s_flags; /* Bits changed in s_flags */
1943 unsigned long journal_devnum;
1944 unsigned long s_commit_interval;
1945 unsigned long s_stripe;
1946 unsigned int s_inode_readahead_blks;
1947 unsigned int s_want_extra_isize;
1948 unsigned int s_li_wait_mult;
1949 unsigned int s_max_dir_size_kb;
1950 unsigned int journal_ioprio;
1951 unsigned int vals_s_mount_opt;
1952 unsigned int mask_s_mount_opt;
1953 unsigned int vals_s_mount_opt2;
1954 unsigned int mask_s_mount_opt2;
1955 unsigned long vals_s_mount_flags;
1956 unsigned long mask_s_mount_flags;
1957 unsigned int opt_flags; /* MOPT flags */
1959 u32 s_max_batch_time;
1960 u32 s_min_batch_time;
1963 ext4_fsblk_t s_sb_block;
1966 static void ext4_fc_free(struct fs_context *fc)
1968 struct ext4_fs_context *ctx = fc->fs_private;
1974 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1975 kfree(ctx->s_qf_names[i]);
1977 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1981 int ext4_init_fs_context(struct fs_context *fc)
1983 struct ext4_fs_context *ctx;
1985 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1989 fc->fs_private = ctx;
1990 fc->ops = &ext4_context_ops;
1997 * Note the name of the specified quota file.
1999 static int note_qf_name(struct fs_context *fc, int qtype,
2000 struct fs_parameter *param)
2002 struct ext4_fs_context *ctx = fc->fs_private;
2005 if (param->size < 1) {
2006 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2009 if (strchr(param->string, '/')) {
2010 ext4_msg(NULL, KERN_ERR,
2011 "quotafile must be on filesystem root");
2014 if (ctx->s_qf_names[qtype]) {
2015 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2016 ext4_msg(NULL, KERN_ERR,
2017 "%s quota file already specified",
2024 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2026 ext4_msg(NULL, KERN_ERR,
2027 "Not enough memory for storing quotafile name");
2030 ctx->s_qf_names[qtype] = qname;
2031 ctx->qname_spec |= 1 << qtype;
2032 ctx->spec |= EXT4_SPEC_JQUOTA;
2037 * Clear the name of the specified quota file.
2039 static int unnote_qf_name(struct fs_context *fc, int qtype)
2041 struct ext4_fs_context *ctx = fc->fs_private;
2043 if (ctx->s_qf_names[qtype])
2044 kfree(ctx->s_qf_names[qtype]);
2046 ctx->s_qf_names[qtype] = NULL;
2047 ctx->qname_spec |= 1 << qtype;
2048 ctx->spec |= EXT4_SPEC_JQUOTA;
2053 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2054 struct ext4_fs_context *ctx)
2058 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2059 ext4_msg(NULL, KERN_WARNING,
2060 "test_dummy_encryption option not supported");
2063 err = fscrypt_parse_test_dummy_encryption(param,
2064 &ctx->dummy_enc_policy);
2065 if (err == -EINVAL) {
2066 ext4_msg(NULL, KERN_WARNING,
2067 "Value of option \"%s\" is unrecognized", param->key);
2068 } else if (err == -EEXIST) {
2069 ext4_msg(NULL, KERN_WARNING,
2070 "Conflicting test_dummy_encryption options");
2076 #define EXT4_SET_CTX(name) \
2077 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2078 unsigned long flag) \
2080 ctx->mask_s_##name |= flag; \
2081 ctx->vals_s_##name |= flag; \
2084 #define EXT4_CLEAR_CTX(name) \
2085 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2086 unsigned long flag) \
2088 ctx->mask_s_##name |= flag; \
2089 ctx->vals_s_##name &= ~flag; \
2092 #define EXT4_TEST_CTX(name) \
2093 static inline unsigned long \
2094 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2096 return (ctx->vals_s_##name & flag); \
2099 EXT4_SET_CTX(flags); /* set only */
2100 EXT4_SET_CTX(mount_opt);
2101 EXT4_CLEAR_CTX(mount_opt);
2102 EXT4_TEST_CTX(mount_opt);
2103 EXT4_SET_CTX(mount_opt2);
2104 EXT4_CLEAR_CTX(mount_opt2);
2105 EXT4_TEST_CTX(mount_opt2);
2107 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2109 set_bit(bit, &ctx->mask_s_mount_flags);
2110 set_bit(bit, &ctx->vals_s_mount_flags);
2113 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2115 struct ext4_fs_context *ctx = fc->fs_private;
2116 struct fs_parse_result result;
2117 const struct mount_opts *m;
2123 token = fs_parse(fc, ext4_param_specs, param, &result);
2126 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2128 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2129 if (token == m->token)
2132 ctx->opt_flags |= m->flags;
2134 if (m->flags & MOPT_EXPLICIT) {
2135 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2136 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2137 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2138 ctx_set_mount_opt2(ctx,
2139 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2144 if (m->flags & MOPT_NOSUPPORT) {
2145 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2153 if (!*param->string)
2154 return unnote_qf_name(fc, USRQUOTA);
2156 return note_qf_name(fc, USRQUOTA, param);
2158 if (!*param->string)
2159 return unnote_qf_name(fc, GRPQUOTA);
2161 return note_qf_name(fc, GRPQUOTA, param);
2164 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2165 ext4_msg(NULL, KERN_WARNING,
2166 "Ignoring %s option on remount", param->key);
2168 ctx->s_sb_block = result.uint_32;
2169 ctx->spec |= EXT4_SPEC_s_sb_block;
2173 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2177 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2179 case Opt_inlinecrypt:
2180 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2181 ctx_set_flags(ctx, SB_INLINECRYPT);
2183 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2187 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2188 ctx_set_mount_opt(ctx, result.uint_32);
2192 ctx->s_jquota_fmt = result.uint_32;
2193 ctx->spec |= EXT4_SPEC_JQFMT;
2197 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2198 ctx_set_mount_opt(ctx, result.uint_32);
2199 ctx->spec |= EXT4_SPEC_DATAJ;
2202 if (result.uint_32 == 0)
2203 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2204 else if (result.uint_32 > INT_MAX / HZ) {
2205 ext4_msg(NULL, KERN_ERR,
2206 "Invalid commit interval %d, "
2207 "must be smaller than %d",
2208 result.uint_32, INT_MAX / HZ);
2211 ctx->s_commit_interval = HZ * result.uint_32;
2212 ctx->spec |= EXT4_SPEC_s_commit_interval;
2214 case Opt_debug_want_extra_isize:
2215 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2216 ext4_msg(NULL, KERN_ERR,
2217 "Invalid want_extra_isize %d", result.uint_32);
2220 ctx->s_want_extra_isize = result.uint_32;
2221 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2223 case Opt_max_batch_time:
2224 ctx->s_max_batch_time = result.uint_32;
2225 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2227 case Opt_min_batch_time:
2228 ctx->s_min_batch_time = result.uint_32;
2229 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2231 case Opt_inode_readahead_blks:
2232 if (result.uint_32 &&
2233 (result.uint_32 > (1 << 30) ||
2234 !is_power_of_2(result.uint_32))) {
2235 ext4_msg(NULL, KERN_ERR,
2236 "EXT4-fs: inode_readahead_blks must be "
2237 "0 or a power of 2 smaller than 2^31");
2240 ctx->s_inode_readahead_blks = result.uint_32;
2241 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2243 case Opt_init_itable:
2244 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2245 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2246 if (param->type == fs_value_is_string)
2247 ctx->s_li_wait_mult = result.uint_32;
2248 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2250 case Opt_max_dir_size_kb:
2251 ctx->s_max_dir_size_kb = result.uint_32;
2252 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2254 #ifdef CONFIG_EXT4_DEBUG
2255 case Opt_fc_debug_max_replay:
2256 ctx->s_fc_debug_max_replay = result.uint_32;
2257 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2261 ctx->s_stripe = result.uint_32;
2262 ctx->spec |= EXT4_SPEC_s_stripe;
2265 uid = make_kuid(current_user_ns(), result.uint_32);
2266 if (!uid_valid(uid)) {
2267 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2271 ctx->s_resuid = uid;
2272 ctx->spec |= EXT4_SPEC_s_resuid;
2275 gid = make_kgid(current_user_ns(), result.uint_32);
2276 if (!gid_valid(gid)) {
2277 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2281 ctx->s_resgid = gid;
2282 ctx->spec |= EXT4_SPEC_s_resgid;
2284 case Opt_journal_dev:
2286 ext4_msg(NULL, KERN_ERR,
2287 "Cannot specify journal on remount");
2290 ctx->journal_devnum = result.uint_32;
2291 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2293 case Opt_journal_path:
2295 struct inode *journal_inode;
2300 ext4_msg(NULL, KERN_ERR,
2301 "Cannot specify journal on remount");
2305 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2307 ext4_msg(NULL, KERN_ERR, "error: could not find "
2308 "journal device path");
2312 journal_inode = d_inode(path.dentry);
2313 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2314 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2318 case Opt_journal_ioprio:
2319 if (result.uint_32 > 7) {
2320 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2324 ctx->journal_ioprio =
2325 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2326 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2328 case Opt_test_dummy_encryption:
2329 return ext4_parse_test_dummy_encryption(param, ctx);
2332 #ifdef CONFIG_FS_DAX
2334 int type = (token == Opt_dax) ?
2335 Opt_dax : result.uint_32;
2339 case Opt_dax_always:
2340 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2341 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2344 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2345 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2348 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2349 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2350 /* Strictly for printing options */
2351 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2357 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2361 if (result.uint_32 == Opt_data_err_abort)
2362 ctx_set_mount_opt(ctx, m->mount_opt);
2363 else if (result.uint_32 == Opt_data_err_ignore)
2364 ctx_clear_mount_opt(ctx, m->mount_opt);
2366 case Opt_mb_optimize_scan:
2367 if (result.int_32 == 1) {
2368 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2369 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2370 } else if (result.int_32 == 0) {
2371 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2372 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2374 ext4_msg(NULL, KERN_WARNING,
2375 "mb_optimize_scan should be set to 0 or 1.");
2382 * At this point we should only be getting options requiring MOPT_SET,
2383 * or MOPT_CLEAR. Anything else is a bug
2385 if (m->token == Opt_err) {
2386 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2393 unsigned int set = 0;
2395 if ((param->type == fs_value_is_flag) ||
2399 if (m->flags & MOPT_CLEAR)
2401 else if (unlikely(!(m->flags & MOPT_SET))) {
2402 ext4_msg(NULL, KERN_WARNING,
2403 "buggy handling of option %s",
2408 if (m->flags & MOPT_2) {
2410 ctx_set_mount_opt2(ctx, m->mount_opt);
2412 ctx_clear_mount_opt2(ctx, m->mount_opt);
2415 ctx_set_mount_opt(ctx, m->mount_opt);
2417 ctx_clear_mount_opt(ctx, m->mount_opt);
2424 static int parse_options(struct fs_context *fc, char *options)
2426 struct fs_parameter param;
2433 while ((key = strsep(&options, ",")) != NULL) {
2436 char *value = strchr(key, '=');
2438 param.type = fs_value_is_flag;
2439 param.string = NULL;
2446 v_len = strlen(value);
2447 param.string = kmemdup_nul(value, v_len,
2451 param.type = fs_value_is_string;
2457 ret = ext4_parse_param(fc, ¶m);
2459 kfree(param.string);
2465 ret = ext4_validate_options(fc);
2472 static int parse_apply_sb_mount_options(struct super_block *sb,
2473 struct ext4_fs_context *m_ctx)
2475 struct ext4_sb_info *sbi = EXT4_SB(sb);
2476 char *s_mount_opts = NULL;
2477 struct ext4_fs_context *s_ctx = NULL;
2478 struct fs_context *fc = NULL;
2481 if (!sbi->s_es->s_mount_opts[0])
2484 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2485 sizeof(sbi->s_es->s_mount_opts),
2490 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2494 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2498 fc->fs_private = s_ctx;
2499 fc->s_fs_info = sbi;
2501 ret = parse_options(fc, s_mount_opts);
2505 ret = ext4_check_opt_consistency(fc, sb);
2508 ext4_msg(sb, KERN_WARNING,
2509 "failed to parse options in superblock: %s",
2515 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2516 m_ctx->journal_devnum = s_ctx->journal_devnum;
2517 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2518 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2520 ext4_apply_options(fc, sb);
2528 kfree(s_mount_opts);
2532 static void ext4_apply_quota_options(struct fs_context *fc,
2533 struct super_block *sb)
2536 bool quota_feature = ext4_has_feature_quota(sb);
2537 struct ext4_fs_context *ctx = fc->fs_private;
2538 struct ext4_sb_info *sbi = EXT4_SB(sb);
2545 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2546 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2547 if (!(ctx->qname_spec & (1 << i)))
2550 qname = ctx->s_qf_names[i]; /* May be NULL */
2553 ctx->s_qf_names[i] = NULL;
2554 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2555 lockdep_is_held(&sb->s_umount));
2557 kfree_rcu_mightsleep(qname);
2561 if (ctx->spec & EXT4_SPEC_JQFMT)
2562 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2567 * Check quota settings consistency.
2569 static int ext4_check_quota_consistency(struct fs_context *fc,
2570 struct super_block *sb)
2573 struct ext4_fs_context *ctx = fc->fs_private;
2574 struct ext4_sb_info *sbi = EXT4_SB(sb);
2575 bool quota_feature = ext4_has_feature_quota(sb);
2576 bool quota_loaded = sb_any_quota_loaded(sb);
2577 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2581 * We do the test below only for project quotas. 'usrquota' and
2582 * 'grpquota' mount options are allowed even without quota feature
2583 * to support legacy quotas in quota files.
2585 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2586 !ext4_has_feature_project(sb)) {
2587 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2588 "Cannot enable project quota enforcement.");
2592 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2593 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2595 ctx->mask_s_mount_opt & quota_flags &&
2596 !ctx_test_mount_opt(ctx, quota_flags))
2597 goto err_quota_change;
2599 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2601 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2602 if (!(ctx->qname_spec & (1 << i)))
2606 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2607 goto err_jquota_change;
2609 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2610 strcmp(get_qf_name(sb, sbi, i),
2611 ctx->s_qf_names[i]) != 0)
2612 goto err_jquota_specified;
2615 if (quota_feature) {
2616 ext4_msg(NULL, KERN_INFO,
2617 "Journaled quota options ignored when "
2618 "QUOTA feature is enabled");
2623 if (ctx->spec & EXT4_SPEC_JQFMT) {
2624 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2625 goto err_jquota_change;
2626 if (quota_feature) {
2627 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2628 "ignored when QUOTA feature is enabled");
2633 /* Make sure we don't mix old and new quota format */
2634 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2635 ctx->s_qf_names[USRQUOTA]);
2636 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2637 ctx->s_qf_names[GRPQUOTA]);
2639 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2640 test_opt(sb, USRQUOTA));
2642 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2643 test_opt(sb, GRPQUOTA));
2646 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2650 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2654 if (usr_qf_name || grp_qf_name) {
2655 if (usrquota || grpquota) {
2656 ext4_msg(NULL, KERN_ERR, "old and new quota "
2661 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2662 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2671 ext4_msg(NULL, KERN_ERR,
2672 "Cannot change quota options when quota turned on");
2675 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2676 "options when quota turned on");
2678 err_jquota_specified:
2679 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2687 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2688 struct super_block *sb)
2690 const struct ext4_fs_context *ctx = fc->fs_private;
2691 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2693 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2696 if (!ext4_has_feature_encrypt(sb)) {
2697 ext4_msg(NULL, KERN_WARNING,
2698 "test_dummy_encryption requires encrypt feature");
2702 * This mount option is just for testing, and it's not worthwhile to
2703 * implement the extra complexity (e.g. RCU protection) that would be
2704 * needed to allow it to be set or changed during remount. We do allow
2705 * it to be specified during remount, but only if there is no change.
2707 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2708 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2709 &ctx->dummy_enc_policy))
2711 ext4_msg(NULL, KERN_WARNING,
2712 "Can't set or change test_dummy_encryption on remount");
2715 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2716 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2717 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2718 &ctx->dummy_enc_policy))
2720 ext4_msg(NULL, KERN_WARNING,
2721 "Conflicting test_dummy_encryption options");
2727 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2728 struct super_block *sb)
2730 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2731 /* if already set, it was already verified to be the same */
2732 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2734 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2735 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2736 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2739 static int ext4_check_opt_consistency(struct fs_context *fc,
2740 struct super_block *sb)
2742 struct ext4_fs_context *ctx = fc->fs_private;
2743 struct ext4_sb_info *sbi = fc->s_fs_info;
2744 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2747 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2748 ext4_msg(NULL, KERN_ERR,
2749 "Mount option(s) incompatible with ext2");
2752 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2753 ext4_msg(NULL, KERN_ERR,
2754 "Mount option(s) incompatible with ext3");
2758 if (ctx->s_want_extra_isize >
2759 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2760 ext4_msg(NULL, KERN_ERR,
2761 "Invalid want_extra_isize %d",
2762 ctx->s_want_extra_isize);
2766 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2768 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2769 if (blocksize < PAGE_SIZE)
2770 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2771 "experimental mount option 'dioread_nolock' "
2772 "for blocksize < PAGE_SIZE");
2775 err = ext4_check_test_dummy_encryption(fc, sb);
2779 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2780 if (!sbi->s_journal) {
2781 ext4_msg(NULL, KERN_WARNING,
2782 "Remounting file system with no journal "
2783 "so ignoring journalled data option");
2784 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2785 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2786 test_opt(sb, DATA_FLAGS)) {
2787 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2794 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2795 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2796 ext4_msg(NULL, KERN_ERR, "can't mount with "
2797 "both data=journal and dax");
2801 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2802 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2803 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2804 fail_dax_change_remount:
2805 ext4_msg(NULL, KERN_ERR, "can't change "
2806 "dax mount option while remounting");
2808 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2809 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2810 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2811 goto fail_dax_change_remount;
2812 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2813 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2814 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2815 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2816 goto fail_dax_change_remount;
2820 return ext4_check_quota_consistency(fc, sb);
2823 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2825 struct ext4_fs_context *ctx = fc->fs_private;
2826 struct ext4_sb_info *sbi = fc->s_fs_info;
2828 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2829 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2830 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2831 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2832 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2833 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2834 sb->s_flags &= ~ctx->mask_s_flags;
2835 sb->s_flags |= ctx->vals_s_flags;
2837 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2838 APPLY(s_commit_interval);
2840 APPLY(s_max_batch_time);
2841 APPLY(s_min_batch_time);
2842 APPLY(s_want_extra_isize);
2843 APPLY(s_inode_readahead_blks);
2844 APPLY(s_max_dir_size_kb);
2845 APPLY(s_li_wait_mult);
2849 #ifdef CONFIG_EXT4_DEBUG
2850 APPLY(s_fc_debug_max_replay);
2853 ext4_apply_quota_options(fc, sb);
2854 ext4_apply_test_dummy_encryption(ctx, sb);
2858 static int ext4_validate_options(struct fs_context *fc)
2861 struct ext4_fs_context *ctx = fc->fs_private;
2862 char *usr_qf_name, *grp_qf_name;
2864 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2865 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2867 if (usr_qf_name || grp_qf_name) {
2868 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2869 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2871 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2872 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2874 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2875 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2876 ext4_msg(NULL, KERN_ERR, "old and new quota "
2885 static inline void ext4_show_quota_options(struct seq_file *seq,
2886 struct super_block *sb)
2888 #if defined(CONFIG_QUOTA)
2889 struct ext4_sb_info *sbi = EXT4_SB(sb);
2890 char *usr_qf_name, *grp_qf_name;
2892 if (sbi->s_jquota_fmt) {
2895 switch (sbi->s_jquota_fmt) {
2906 seq_printf(seq, ",jqfmt=%s", fmtname);
2910 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2911 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2913 seq_show_option(seq, "usrjquota", usr_qf_name);
2915 seq_show_option(seq, "grpjquota", grp_qf_name);
2920 static const char *token2str(int token)
2922 const struct fs_parameter_spec *spec;
2924 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2925 if (spec->opt == token && !spec->type)
2932 * - it's set to a non-default value OR
2933 * - if the per-sb default is different from the global default
2935 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2938 struct ext4_sb_info *sbi = EXT4_SB(sb);
2939 struct ext4_super_block *es = sbi->s_es;
2941 const struct mount_opts *m;
2942 char sep = nodefs ? '\n' : ',';
2944 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2945 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2947 if (sbi->s_sb_block != 1)
2948 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2950 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2951 int want_set = m->flags & MOPT_SET;
2952 int opt_2 = m->flags & MOPT_2;
2953 unsigned int mount_opt, def_mount_opt;
2955 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2956 m->flags & MOPT_SKIP)
2960 mount_opt = sbi->s_mount_opt2;
2961 def_mount_opt = sbi->s_def_mount_opt2;
2963 mount_opt = sbi->s_mount_opt;
2964 def_mount_opt = sbi->s_def_mount_opt;
2966 /* skip if same as the default */
2967 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2969 /* select Opt_noFoo vs Opt_Foo */
2971 (mount_opt & m->mount_opt) != m->mount_opt) ||
2972 (!want_set && (mount_opt & m->mount_opt)))
2974 SEQ_OPTS_PRINT("%s", token2str(m->token));
2977 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2978 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2979 SEQ_OPTS_PRINT("resuid=%u",
2980 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2981 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2982 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2983 SEQ_OPTS_PRINT("resgid=%u",
2984 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2985 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2986 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2987 SEQ_OPTS_PUTS("errors=remount-ro");
2988 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2989 SEQ_OPTS_PUTS("errors=continue");
2990 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2991 SEQ_OPTS_PUTS("errors=panic");
2992 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2993 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2994 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2995 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2996 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2997 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2998 if (nodefs || sbi->s_stripe)
2999 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3000 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3001 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3002 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3003 SEQ_OPTS_PUTS("data=journal");
3004 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3005 SEQ_OPTS_PUTS("data=ordered");
3006 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3007 SEQ_OPTS_PUTS("data=writeback");
3010 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3011 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3012 sbi->s_inode_readahead_blks);
3014 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3015 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3016 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3017 if (nodefs || sbi->s_max_dir_size_kb)
3018 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3019 if (test_opt(sb, DATA_ERR_ABORT))
3020 SEQ_OPTS_PUTS("data_err=abort");
3022 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3024 if (sb->s_flags & SB_INLINECRYPT)
3025 SEQ_OPTS_PUTS("inlinecrypt");
3027 if (test_opt(sb, DAX_ALWAYS)) {
3029 SEQ_OPTS_PUTS("dax");
3031 SEQ_OPTS_PUTS("dax=always");
3032 } else if (test_opt2(sb, DAX_NEVER)) {
3033 SEQ_OPTS_PUTS("dax=never");
3034 } else if (test_opt2(sb, DAX_INODE)) {
3035 SEQ_OPTS_PUTS("dax=inode");
3038 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3039 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3040 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3041 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3042 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3043 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3046 ext4_show_quota_options(seq, sb);
3050 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3052 return _ext4_show_options(seq, root->d_sb, 0);
3055 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3057 struct super_block *sb = seq->private;
3060 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3061 rc = _ext4_show_options(seq, sb, 1);
3062 seq_puts(seq, "\n");
3066 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3069 struct ext4_sb_info *sbi = EXT4_SB(sb);
3072 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3073 ext4_msg(sb, KERN_ERR, "revision level too high, "
3074 "forcing read-only mode");
3080 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3081 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3082 "running e2fsck is recommended");
3083 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3084 ext4_msg(sb, KERN_WARNING,
3085 "warning: mounting fs with errors, "
3086 "running e2fsck is recommended");
3087 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3088 le16_to_cpu(es->s_mnt_count) >=
3089 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3090 ext4_msg(sb, KERN_WARNING,
3091 "warning: maximal mount count reached, "
3092 "running e2fsck is recommended");
3093 else if (le32_to_cpu(es->s_checkinterval) &&
3094 (ext4_get_tstamp(es, s_lastcheck) +
3095 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3096 ext4_msg(sb, KERN_WARNING,
3097 "warning: checktime reached, "
3098 "running e2fsck is recommended");
3099 if (!sbi->s_journal)
3100 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3101 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3102 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3103 le16_add_cpu(&es->s_mnt_count, 1);
3104 ext4_update_tstamp(es, s_mtime);
3105 if (sbi->s_journal) {
3106 ext4_set_feature_journal_needs_recovery(sb);
3107 if (ext4_has_feature_orphan_file(sb))
3108 ext4_set_feature_orphan_present(sb);
3111 err = ext4_commit_super(sb);
3113 if (test_opt(sb, DEBUG))
3114 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3115 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3117 sbi->s_groups_count,
3118 EXT4_BLOCKS_PER_GROUP(sb),
3119 EXT4_INODES_PER_GROUP(sb),
3120 sbi->s_mount_opt, sbi->s_mount_opt2);
3124 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3126 struct ext4_sb_info *sbi = EXT4_SB(sb);
3127 struct flex_groups **old_groups, **new_groups;
3130 if (!sbi->s_log_groups_per_flex)
3133 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3134 if (size <= sbi->s_flex_groups_allocated)
3137 new_groups = kvzalloc(roundup_pow_of_two(size *
3138 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3140 ext4_msg(sb, KERN_ERR,
3141 "not enough memory for %d flex group pointers", size);
3144 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3145 new_groups[i] = kvzalloc(roundup_pow_of_two(
3146 sizeof(struct flex_groups)),
3148 if (!new_groups[i]) {
3149 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3150 kvfree(new_groups[j]);
3152 ext4_msg(sb, KERN_ERR,
3153 "not enough memory for %d flex groups", size);
3158 old_groups = rcu_dereference(sbi->s_flex_groups);
3160 memcpy(new_groups, old_groups,
3161 (sbi->s_flex_groups_allocated *
3162 sizeof(struct flex_groups *)));
3164 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3165 sbi->s_flex_groups_allocated = size;
3167 ext4_kvfree_array_rcu(old_groups);
3171 static int ext4_fill_flex_info(struct super_block *sb)
3173 struct ext4_sb_info *sbi = EXT4_SB(sb);
3174 struct ext4_group_desc *gdp = NULL;
3175 struct flex_groups *fg;
3176 ext4_group_t flex_group;
3179 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3180 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3181 sbi->s_log_groups_per_flex = 0;
3185 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3189 for (i = 0; i < sbi->s_groups_count; i++) {
3190 gdp = ext4_get_group_desc(sb, i, NULL);
3192 flex_group = ext4_flex_group(sbi, i);
3193 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3194 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3195 atomic64_add(ext4_free_group_clusters(sb, gdp),
3196 &fg->free_clusters);
3197 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3205 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3206 struct ext4_group_desc *gdp)
3208 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3210 __le32 le_group = cpu_to_le32(block_group);
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3213 if (ext4_has_metadata_csum(sbi->s_sb)) {
3214 /* Use new metadata_csum algorithm */
3216 __u16 dummy_csum = 0;
3218 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3220 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3221 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3222 sizeof(dummy_csum));
3223 offset += sizeof(dummy_csum);
3224 if (offset < sbi->s_desc_size)
3225 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3226 sbi->s_desc_size - offset);
3228 crc = csum32 & 0xFFFF;
3232 /* old crc16 code */
3233 if (!ext4_has_feature_gdt_csum(sb))
3236 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3237 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3238 crc = crc16(crc, (__u8 *)gdp, offset);
3239 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3240 /* for checksum of struct ext4_group_desc do the rest...*/
3241 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3242 crc = crc16(crc, (__u8 *)gdp + offset,
3243 sbi->s_desc_size - offset);
3246 return cpu_to_le16(crc);
3249 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3250 struct ext4_group_desc *gdp)
3252 if (ext4_has_group_desc_csum(sb) &&
3253 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3259 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3260 struct ext4_group_desc *gdp)
3262 if (!ext4_has_group_desc_csum(sb))
3264 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3267 /* Called at mount-time, super-block is locked */
3268 static int ext4_check_descriptors(struct super_block *sb,
3269 ext4_fsblk_t sb_block,
3270 ext4_group_t *first_not_zeroed)
3272 struct ext4_sb_info *sbi = EXT4_SB(sb);
3273 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3274 ext4_fsblk_t last_block;
3275 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3276 ext4_fsblk_t block_bitmap;
3277 ext4_fsblk_t inode_bitmap;
3278 ext4_fsblk_t inode_table;
3279 int flexbg_flag = 0;
3280 ext4_group_t i, grp = sbi->s_groups_count;
3282 if (ext4_has_feature_flex_bg(sb))
3285 ext4_debug("Checking group descriptors");
3287 for (i = 0; i < sbi->s_groups_count; i++) {
3288 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3290 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3291 last_block = ext4_blocks_count(sbi->s_es) - 1;
3293 last_block = first_block +
3294 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3296 if ((grp == sbi->s_groups_count) &&
3297 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3300 block_bitmap = ext4_block_bitmap(sb, gdp);
3301 if (block_bitmap == sb_block) {
3302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3303 "Block bitmap for group %u overlaps "
3308 if (block_bitmap >= sb_block + 1 &&
3309 block_bitmap <= last_bg_block) {
3310 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3311 "Block bitmap for group %u overlaps "
3312 "block group descriptors", i);
3316 if (block_bitmap < first_block || block_bitmap > last_block) {
3317 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3318 "Block bitmap for group %u not in group "
3319 "(block %llu)!", i, block_bitmap);
3322 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3323 if (inode_bitmap == sb_block) {
3324 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3325 "Inode bitmap for group %u overlaps "
3330 if (inode_bitmap >= sb_block + 1 &&
3331 inode_bitmap <= last_bg_block) {
3332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3333 "Inode bitmap for group %u overlaps "
3334 "block group descriptors", i);
3338 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3339 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3340 "Inode bitmap for group %u not in group "
3341 "(block %llu)!", i, inode_bitmap);
3344 inode_table = ext4_inode_table(sb, gdp);
3345 if (inode_table == sb_block) {
3346 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3347 "Inode table for group %u overlaps "
3352 if (inode_table >= sb_block + 1 &&
3353 inode_table <= last_bg_block) {
3354 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3355 "Inode table for group %u overlaps "
3356 "block group descriptors", i);
3360 if (inode_table < first_block ||
3361 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3362 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3363 "Inode table for group %u not in group "
3364 "(block %llu)!", i, inode_table);
3367 ext4_lock_group(sb, i);
3368 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3369 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3370 "Checksum for group %u failed (%u!=%u)",
3371 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3372 gdp)), le16_to_cpu(gdp->bg_checksum));
3373 if (!sb_rdonly(sb)) {
3374 ext4_unlock_group(sb, i);
3378 ext4_unlock_group(sb, i);
3380 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3382 if (NULL != first_not_zeroed)
3383 *first_not_zeroed = grp;
3388 * Maximal extent format file size.
3389 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3390 * extent format containers, within a sector_t, and within i_blocks
3391 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3392 * so that won't be a limiting factor.
3394 * However there is other limiting factor. We do store extents in the form
3395 * of starting block and length, hence the resulting length of the extent
3396 * covering maximum file size must fit into on-disk format containers as
3397 * well. Given that length is always by 1 unit bigger than max unit (because
3398 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3400 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3402 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3405 loff_t upper_limit = MAX_LFS_FILESIZE;
3407 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3409 if (!has_huge_files) {
3410 upper_limit = (1LL << 32) - 1;
3412 /* total blocks in file system block size */
3413 upper_limit >>= (blkbits - 9);
3414 upper_limit <<= blkbits;
3418 * 32-bit extent-start container, ee_block. We lower the maxbytes
3419 * by one fs block, so ee_len can cover the extent of maximum file
3422 res = (1LL << 32) - 1;
3425 /* Sanity check against vm- & vfs- imposed limits */
3426 if (res > upper_limit)
3433 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3434 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3435 * We need to be 1 filesystem block less than the 2^48 sector limit.
3437 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3439 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3441 unsigned int ppb = 1 << (bits - 2);
3444 * This is calculated to be the largest file size for a dense, block
3445 * mapped file such that the file's total number of 512-byte sectors,
3446 * including data and all indirect blocks, does not exceed (2^48 - 1).
3448 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3449 * number of 512-byte sectors of the file.
3451 if (!has_huge_files) {
3453 * !has_huge_files or implies that the inode i_block field
3454 * represents total file blocks in 2^32 512-byte sectors ==
3455 * size of vfs inode i_blocks * 8
3457 upper_limit = (1LL << 32) - 1;
3459 /* total blocks in file system block size */
3460 upper_limit >>= (bits - 9);
3464 * We use 48 bit ext4_inode i_blocks
3465 * With EXT4_HUGE_FILE_FL set the i_blocks
3466 * represent total number of blocks in
3467 * file system block size
3469 upper_limit = (1LL << 48) - 1;
3473 /* Compute how many blocks we can address by block tree */
3476 res += ((loff_t)ppb) * ppb * ppb;
3477 /* Compute how many metadata blocks are needed */
3479 meta_blocks += 1 + ppb;
3480 meta_blocks += 1 + ppb + ppb * ppb;
3481 /* Does block tree limit file size? */
3482 if (res + meta_blocks <= upper_limit)
3486 /* How many metadata blocks are needed for addressing upper_limit? */
3487 upper_limit -= EXT4_NDIR_BLOCKS;
3488 /* indirect blocks */
3491 /* double indirect blocks */
3492 if (upper_limit < ppb * ppb) {
3493 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3497 meta_blocks += 1 + ppb;
3498 upper_limit -= ppb * ppb;
3499 /* tripple indirect blocks for the rest */
3500 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3501 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3505 if (res > MAX_LFS_FILESIZE)
3506 res = MAX_LFS_FILESIZE;
3511 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3512 ext4_fsblk_t logical_sb_block, int nr)
3514 struct ext4_sb_info *sbi = EXT4_SB(sb);
3515 ext4_group_t bg, first_meta_bg;
3518 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3520 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3521 return logical_sb_block + nr + 1;
3522 bg = sbi->s_desc_per_block * nr;
3523 if (ext4_bg_has_super(sb, bg))
3527 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3528 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3529 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3532 if (sb->s_blocksize == 1024 && nr == 0 &&
3533 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3536 return (has_super + ext4_group_first_block_no(sb, bg));
3540 * ext4_get_stripe_size: Get the stripe size.
3541 * @sbi: In memory super block info
3543 * If we have specified it via mount option, then
3544 * use the mount option value. If the value specified at mount time is
3545 * greater than the blocks per group use the super block value.
3546 * If the super block value is greater than blocks per group return 0.
3547 * Allocator needs it be less than blocks per group.
3550 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3552 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3553 unsigned long stripe_width =
3554 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3557 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3558 ret = sbi->s_stripe;
3559 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3561 else if (stride && stride <= sbi->s_blocks_per_group)
3567 * If the stripe width is 1, this makes no sense and
3568 * we set it to 0 to turn off stripe handling code.
3577 * Check whether this filesystem can be mounted based on
3578 * the features present and the RDONLY/RDWR mount requested.
3579 * Returns 1 if this filesystem can be mounted as requested,
3580 * 0 if it cannot be.
3582 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3584 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3585 ext4_msg(sb, KERN_ERR,
3586 "Couldn't mount because of "
3587 "unsupported optional features (%x)",
3588 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3589 ~EXT4_FEATURE_INCOMPAT_SUPP));
3593 #if !IS_ENABLED(CONFIG_UNICODE)
3594 if (ext4_has_feature_casefold(sb)) {
3595 ext4_msg(sb, KERN_ERR,
3596 "Filesystem with casefold feature cannot be "
3597 "mounted without CONFIG_UNICODE");
3605 if (ext4_has_feature_readonly(sb)) {
3606 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3607 sb->s_flags |= SB_RDONLY;
3611 /* Check that feature set is OK for a read-write mount */
3612 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3613 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3614 "unsupported optional features (%x)",
3615 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3616 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3619 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3620 ext4_msg(sb, KERN_ERR,
3621 "Can't support bigalloc feature without "
3622 "extents feature\n");
3626 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3627 if (!readonly && (ext4_has_feature_quota(sb) ||
3628 ext4_has_feature_project(sb))) {
3629 ext4_msg(sb, KERN_ERR,
3630 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3633 #endif /* CONFIG_QUOTA */
3638 * This function is called once a day if we have errors logged
3639 * on the file system
3641 static void print_daily_error_info(struct timer_list *t)
3643 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3644 struct super_block *sb = sbi->s_sb;
3645 struct ext4_super_block *es = sbi->s_es;
3647 if (es->s_error_count)
3648 /* fsck newer than v1.41.13 is needed to clean this condition. */
3649 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3650 le32_to_cpu(es->s_error_count));
3651 if (es->s_first_error_time) {
3652 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3654 ext4_get_tstamp(es, s_first_error_time),
3655 (int) sizeof(es->s_first_error_func),
3656 es->s_first_error_func,
3657 le32_to_cpu(es->s_first_error_line));
3658 if (es->s_first_error_ino)
3659 printk(KERN_CONT ": inode %u",
3660 le32_to_cpu(es->s_first_error_ino));
3661 if (es->s_first_error_block)
3662 printk(KERN_CONT ": block %llu", (unsigned long long)
3663 le64_to_cpu(es->s_first_error_block));
3664 printk(KERN_CONT "\n");
3666 if (es->s_last_error_time) {
3667 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3669 ext4_get_tstamp(es, s_last_error_time),
3670 (int) sizeof(es->s_last_error_func),
3671 es->s_last_error_func,
3672 le32_to_cpu(es->s_last_error_line));
3673 if (es->s_last_error_ino)
3674 printk(KERN_CONT ": inode %u",
3675 le32_to_cpu(es->s_last_error_ino));
3676 if (es->s_last_error_block)
3677 printk(KERN_CONT ": block %llu", (unsigned long long)
3678 le64_to_cpu(es->s_last_error_block));
3679 printk(KERN_CONT "\n");
3681 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3684 /* Find next suitable group and run ext4_init_inode_table */
3685 static int ext4_run_li_request(struct ext4_li_request *elr)
3687 struct ext4_group_desc *gdp = NULL;
3688 struct super_block *sb = elr->lr_super;
3689 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3690 ext4_group_t group = elr->lr_next_group;
3691 unsigned int prefetch_ios = 0;
3693 int nr = EXT4_SB(sb)->s_mb_prefetch;
3696 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3697 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3698 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3699 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3700 if (group >= elr->lr_next_group) {
3702 if (elr->lr_first_not_zeroed != ngroups &&
3703 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3704 elr->lr_next_group = elr->lr_first_not_zeroed;
3705 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3712 for (; group < ngroups; group++) {
3713 gdp = ext4_get_group_desc(sb, group, NULL);
3719 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3723 if (group >= ngroups)
3727 start_time = ktime_get_real_ns();
3728 ret = ext4_init_inode_table(sb, group,
3729 elr->lr_timeout ? 0 : 1);
3730 trace_ext4_lazy_itable_init(sb, group);
3731 if (elr->lr_timeout == 0) {
3732 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3733 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3735 elr->lr_next_sched = jiffies + elr->lr_timeout;
3736 elr->lr_next_group = group + 1;
3742 * Remove lr_request from the list_request and free the
3743 * request structure. Should be called with li_list_mtx held
3745 static void ext4_remove_li_request(struct ext4_li_request *elr)
3750 list_del(&elr->lr_request);
3751 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3755 static void ext4_unregister_li_request(struct super_block *sb)
3757 mutex_lock(&ext4_li_mtx);
3758 if (!ext4_li_info) {
3759 mutex_unlock(&ext4_li_mtx);
3763 mutex_lock(&ext4_li_info->li_list_mtx);
3764 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3765 mutex_unlock(&ext4_li_info->li_list_mtx);
3766 mutex_unlock(&ext4_li_mtx);
3769 static struct task_struct *ext4_lazyinit_task;
3772 * This is the function where ext4lazyinit thread lives. It walks
3773 * through the request list searching for next scheduled filesystem.
3774 * When such a fs is found, run the lazy initialization request
3775 * (ext4_rn_li_request) and keep track of the time spend in this
3776 * function. Based on that time we compute next schedule time of
3777 * the request. When walking through the list is complete, compute
3778 * next waking time and put itself into sleep.
3780 static int ext4_lazyinit_thread(void *arg)
3782 struct ext4_lazy_init *eli = arg;
3783 struct list_head *pos, *n;
3784 struct ext4_li_request *elr;
3785 unsigned long next_wakeup, cur;
3787 BUG_ON(NULL == eli);
3792 next_wakeup = MAX_JIFFY_OFFSET;
3794 mutex_lock(&eli->li_list_mtx);
3795 if (list_empty(&eli->li_request_list)) {
3796 mutex_unlock(&eli->li_list_mtx);
3799 list_for_each_safe(pos, n, &eli->li_request_list) {
3802 elr = list_entry(pos, struct ext4_li_request,
3805 if (time_before(jiffies, elr->lr_next_sched)) {
3806 if (time_before(elr->lr_next_sched, next_wakeup))
3807 next_wakeup = elr->lr_next_sched;
3810 if (down_read_trylock(&elr->lr_super->s_umount)) {
3811 if (sb_start_write_trylock(elr->lr_super)) {
3814 * We hold sb->s_umount, sb can not
3815 * be removed from the list, it is
3816 * now safe to drop li_list_mtx
3818 mutex_unlock(&eli->li_list_mtx);
3819 err = ext4_run_li_request(elr);
3820 sb_end_write(elr->lr_super);
3821 mutex_lock(&eli->li_list_mtx);
3824 up_read((&elr->lr_super->s_umount));
3826 /* error, remove the lazy_init job */
3828 ext4_remove_li_request(elr);
3832 elr->lr_next_sched = jiffies +
3833 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3835 if (time_before(elr->lr_next_sched, next_wakeup))
3836 next_wakeup = elr->lr_next_sched;
3838 mutex_unlock(&eli->li_list_mtx);
3843 if ((time_after_eq(cur, next_wakeup)) ||
3844 (MAX_JIFFY_OFFSET == next_wakeup)) {
3849 schedule_timeout_interruptible(next_wakeup - cur);
3851 if (kthread_should_stop()) {
3852 ext4_clear_request_list();
3859 * It looks like the request list is empty, but we need
3860 * to check it under the li_list_mtx lock, to prevent any
3861 * additions into it, and of course we should lock ext4_li_mtx
3862 * to atomically free the list and ext4_li_info, because at
3863 * this point another ext4 filesystem could be registering
3866 mutex_lock(&ext4_li_mtx);
3867 mutex_lock(&eli->li_list_mtx);
3868 if (!list_empty(&eli->li_request_list)) {
3869 mutex_unlock(&eli->li_list_mtx);
3870 mutex_unlock(&ext4_li_mtx);
3873 mutex_unlock(&eli->li_list_mtx);
3874 kfree(ext4_li_info);
3875 ext4_li_info = NULL;
3876 mutex_unlock(&ext4_li_mtx);
3881 static void ext4_clear_request_list(void)
3883 struct list_head *pos, *n;
3884 struct ext4_li_request *elr;
3886 mutex_lock(&ext4_li_info->li_list_mtx);
3887 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3888 elr = list_entry(pos, struct ext4_li_request,
3890 ext4_remove_li_request(elr);
3892 mutex_unlock(&ext4_li_info->li_list_mtx);
3895 static int ext4_run_lazyinit_thread(void)
3897 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3898 ext4_li_info, "ext4lazyinit");
3899 if (IS_ERR(ext4_lazyinit_task)) {
3900 int err = PTR_ERR(ext4_lazyinit_task);
3901 ext4_clear_request_list();
3902 kfree(ext4_li_info);
3903 ext4_li_info = NULL;
3904 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3905 "initialization thread\n",
3909 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3914 * Check whether it make sense to run itable init. thread or not.
3915 * If there is at least one uninitialized inode table, return
3916 * corresponding group number, else the loop goes through all
3917 * groups and return total number of groups.
3919 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3921 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3922 struct ext4_group_desc *gdp = NULL;
3924 if (!ext4_has_group_desc_csum(sb))
3927 for (group = 0; group < ngroups; group++) {
3928 gdp = ext4_get_group_desc(sb, group, NULL);
3932 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3939 static int ext4_li_info_new(void)
3941 struct ext4_lazy_init *eli = NULL;
3943 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3947 INIT_LIST_HEAD(&eli->li_request_list);
3948 mutex_init(&eli->li_list_mtx);
3950 eli->li_state |= EXT4_LAZYINIT_QUIT;
3957 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3960 struct ext4_li_request *elr;
3962 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3967 elr->lr_first_not_zeroed = start;
3968 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3969 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3970 elr->lr_next_group = start;
3972 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3976 * Randomize first schedule time of the request to
3977 * spread the inode table initialization requests
3980 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3984 int ext4_register_li_request(struct super_block *sb,
3985 ext4_group_t first_not_zeroed)
3987 struct ext4_sb_info *sbi = EXT4_SB(sb);
3988 struct ext4_li_request *elr = NULL;
3989 ext4_group_t ngroups = sbi->s_groups_count;
3992 mutex_lock(&ext4_li_mtx);
3993 if (sbi->s_li_request != NULL) {
3995 * Reset timeout so it can be computed again, because
3996 * s_li_wait_mult might have changed.
3998 sbi->s_li_request->lr_timeout = 0;
4002 if (sb_rdonly(sb) ||
4003 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4004 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4007 elr = ext4_li_request_new(sb, first_not_zeroed);
4013 if (NULL == ext4_li_info) {
4014 ret = ext4_li_info_new();
4019 mutex_lock(&ext4_li_info->li_list_mtx);
4020 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4021 mutex_unlock(&ext4_li_info->li_list_mtx);
4023 sbi->s_li_request = elr;
4025 * set elr to NULL here since it has been inserted to
4026 * the request_list and the removal and free of it is
4027 * handled by ext4_clear_request_list from now on.
4031 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4032 ret = ext4_run_lazyinit_thread();
4037 mutex_unlock(&ext4_li_mtx);
4044 * We do not need to lock anything since this is called on
4047 static void ext4_destroy_lazyinit_thread(void)
4050 * If thread exited earlier
4051 * there's nothing to be done.
4053 if (!ext4_li_info || !ext4_lazyinit_task)
4056 kthread_stop(ext4_lazyinit_task);
4059 static int set_journal_csum_feature_set(struct super_block *sb)
4062 int compat, incompat;
4063 struct ext4_sb_info *sbi = EXT4_SB(sb);
4065 if (ext4_has_metadata_csum(sb)) {
4066 /* journal checksum v3 */
4068 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4070 /* journal checksum v1 */
4071 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4075 jbd2_journal_clear_features(sbi->s_journal,
4076 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4077 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4078 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4079 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4080 ret = jbd2_journal_set_features(sbi->s_journal,
4082 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4084 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4085 ret = jbd2_journal_set_features(sbi->s_journal,
4088 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4089 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4091 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4092 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4099 * Note: calculating the overhead so we can be compatible with
4100 * historical BSD practice is quite difficult in the face of
4101 * clusters/bigalloc. This is because multiple metadata blocks from
4102 * different block group can end up in the same allocation cluster.
4103 * Calculating the exact overhead in the face of clustered allocation
4104 * requires either O(all block bitmaps) in memory or O(number of block
4105 * groups**2) in time. We will still calculate the superblock for
4106 * older file systems --- and if we come across with a bigalloc file
4107 * system with zero in s_overhead_clusters the estimate will be close to
4108 * correct especially for very large cluster sizes --- but for newer
4109 * file systems, it's better to calculate this figure once at mkfs
4110 * time, and store it in the superblock. If the superblock value is
4111 * present (even for non-bigalloc file systems), we will use it.
4113 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4116 struct ext4_sb_info *sbi = EXT4_SB(sb);
4117 struct ext4_group_desc *gdp;
4118 ext4_fsblk_t first_block, last_block, b;
4119 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4120 int s, j, count = 0;
4121 int has_super = ext4_bg_has_super(sb, grp);
4123 if (!ext4_has_feature_bigalloc(sb))
4124 return (has_super + ext4_bg_num_gdb(sb, grp) +
4125 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4126 sbi->s_itb_per_group + 2);
4128 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4129 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4130 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4131 for (i = 0; i < ngroups; i++) {
4132 gdp = ext4_get_group_desc(sb, i, NULL);
4133 b = ext4_block_bitmap(sb, gdp);
4134 if (b >= first_block && b <= last_block) {
4135 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4138 b = ext4_inode_bitmap(sb, gdp);
4139 if (b >= first_block && b <= last_block) {
4140 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4143 b = ext4_inode_table(sb, gdp);
4144 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4145 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4146 int c = EXT4_B2C(sbi, b - first_block);
4147 ext4_set_bit(c, buf);
4153 if (ext4_bg_has_super(sb, grp)) {
4154 ext4_set_bit(s++, buf);
4157 j = ext4_bg_num_gdb(sb, grp);
4158 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4159 ext4_error(sb, "Invalid number of block group "
4160 "descriptor blocks: %d", j);
4161 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4165 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4169 return EXT4_CLUSTERS_PER_GROUP(sb) -
4170 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4174 * Compute the overhead and stash it in sbi->s_overhead
4176 int ext4_calculate_overhead(struct super_block *sb)
4178 struct ext4_sb_info *sbi = EXT4_SB(sb);
4179 struct ext4_super_block *es = sbi->s_es;
4180 struct inode *j_inode;
4181 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4182 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4183 ext4_fsblk_t overhead = 0;
4184 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4190 * Compute the overhead (FS structures). This is constant
4191 * for a given filesystem unless the number of block groups
4192 * changes so we cache the previous value until it does.
4196 * All of the blocks before first_data_block are overhead
4198 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4201 * Add the overhead found in each block group
4203 for (i = 0; i < ngroups; i++) {
4206 blks = count_overhead(sb, i, buf);
4209 memset(buf, 0, PAGE_SIZE);
4214 * Add the internal journal blocks whether the journal has been
4217 if (sbi->s_journal && !sbi->s_journal_bdev)
4218 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4219 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4220 /* j_inum for internal journal is non-zero */
4221 j_inode = ext4_get_journal_inode(sb, j_inum);
4223 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4224 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4227 ext4_msg(sb, KERN_ERR, "can't get journal size");
4230 sbi->s_overhead = overhead;
4232 free_page((unsigned long) buf);
4236 static void ext4_set_resv_clusters(struct super_block *sb)
4238 ext4_fsblk_t resv_clusters;
4239 struct ext4_sb_info *sbi = EXT4_SB(sb);
4242 * There's no need to reserve anything when we aren't using extents.
4243 * The space estimates are exact, there are no unwritten extents,
4244 * hole punching doesn't need new metadata... This is needed especially
4245 * to keep ext2/3 backward compatibility.
4247 if (!ext4_has_feature_extents(sb))
4250 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4251 * This should cover the situations where we can not afford to run
4252 * out of space like for example punch hole, or converting
4253 * unwritten extents in delalloc path. In most cases such
4254 * allocation would require 1, or 2 blocks, higher numbers are
4257 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4258 sbi->s_cluster_bits);
4260 do_div(resv_clusters, 50);
4261 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4263 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4266 static const char *ext4_quota_mode(struct super_block *sb)
4269 if (!ext4_quota_capable(sb))
4272 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4273 return "journalled";
4281 static void ext4_setup_csum_trigger(struct super_block *sb,
4282 enum ext4_journal_trigger_type type,
4284 struct jbd2_buffer_trigger_type *type,
4285 struct buffer_head *bh,
4289 struct ext4_sb_info *sbi = EXT4_SB(sb);
4291 sbi->s_journal_triggers[type].sb = sb;
4292 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4295 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4300 kfree(sbi->s_blockgroup_lock);
4301 fs_put_dax(sbi->s_daxdev, NULL);
4305 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4307 struct ext4_sb_info *sbi;
4309 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4313 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4316 sbi->s_blockgroup_lock =
4317 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4319 if (!sbi->s_blockgroup_lock)
4322 sb->s_fs_info = sbi;
4326 fs_put_dax(sbi->s_daxdev, NULL);
4331 static void ext4_set_def_opts(struct super_block *sb,
4332 struct ext4_super_block *es)
4334 unsigned long def_mount_opts;
4336 /* Set defaults before we parse the mount options */
4337 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4338 set_opt(sb, INIT_INODE_TABLE);
4339 if (def_mount_opts & EXT4_DEFM_DEBUG)
4341 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4343 if (def_mount_opts & EXT4_DEFM_UID16)
4344 set_opt(sb, NO_UID32);
4345 /* xattr user namespace & acls are now defaulted on */
4346 set_opt(sb, XATTR_USER);
4347 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4348 set_opt(sb, POSIX_ACL);
4350 if (ext4_has_feature_fast_commit(sb))
4351 set_opt2(sb, JOURNAL_FAST_COMMIT);
4352 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4353 if (ext4_has_metadata_csum(sb))
4354 set_opt(sb, JOURNAL_CHECKSUM);
4356 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4357 set_opt(sb, JOURNAL_DATA);
4358 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4359 set_opt(sb, ORDERED_DATA);
4360 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4361 set_opt(sb, WRITEBACK_DATA);
4363 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4364 set_opt(sb, ERRORS_PANIC);
4365 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4366 set_opt(sb, ERRORS_CONT);
4368 set_opt(sb, ERRORS_RO);
4369 /* block_validity enabled by default; disable with noblock_validity */
4370 set_opt(sb, BLOCK_VALIDITY);
4371 if (def_mount_opts & EXT4_DEFM_DISCARD)
4372 set_opt(sb, DISCARD);
4374 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4375 set_opt(sb, BARRIER);
4378 * enable delayed allocation by default
4379 * Use -o nodelalloc to turn it off
4381 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4382 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4383 set_opt(sb, DELALLOC);
4385 if (sb->s_blocksize == PAGE_SIZE)
4386 set_opt(sb, DIOREAD_NOLOCK);
4389 static int ext4_handle_clustersize(struct super_block *sb)
4391 struct ext4_sb_info *sbi = EXT4_SB(sb);
4392 struct ext4_super_block *es = sbi->s_es;
4395 /* Handle clustersize */
4396 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4397 if (ext4_has_feature_bigalloc(sb)) {
4398 if (clustersize < sb->s_blocksize) {
4399 ext4_msg(sb, KERN_ERR,
4400 "cluster size (%d) smaller than "
4401 "block size (%lu)", clustersize, sb->s_blocksize);
4404 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4405 le32_to_cpu(es->s_log_block_size);
4406 sbi->s_clusters_per_group =
4407 le32_to_cpu(es->s_clusters_per_group);
4408 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4409 ext4_msg(sb, KERN_ERR,
4410 "#clusters per group too big: %lu",
4411 sbi->s_clusters_per_group);
4414 if (sbi->s_blocks_per_group !=
4415 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4416 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4417 "clusters per group (%lu) inconsistent",
4418 sbi->s_blocks_per_group,
4419 sbi->s_clusters_per_group);
4423 if (clustersize != sb->s_blocksize) {
4424 ext4_msg(sb, KERN_ERR,
4425 "fragment/cluster size (%d) != "
4426 "block size (%lu)", clustersize, sb->s_blocksize);
4429 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4430 ext4_msg(sb, KERN_ERR,
4431 "#blocks per group too big: %lu",
4432 sbi->s_blocks_per_group);
4435 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4436 sbi->s_cluster_bits = 0;
4438 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4440 /* Do we have standard group size of clustersize * 8 blocks ? */
4441 if (sbi->s_blocks_per_group == clustersize << 3)
4442 set_opt2(sb, STD_GROUP_SIZE);
4447 static void ext4_fast_commit_init(struct super_block *sb)
4449 struct ext4_sb_info *sbi = EXT4_SB(sb);
4451 /* Initialize fast commit stuff */
4452 atomic_set(&sbi->s_fc_subtid, 0);
4453 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4454 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4455 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4456 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4457 sbi->s_fc_bytes = 0;
4458 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4459 sbi->s_fc_ineligible_tid = 0;
4460 spin_lock_init(&sbi->s_fc_lock);
4461 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4462 sbi->s_fc_replay_state.fc_regions = NULL;
4463 sbi->s_fc_replay_state.fc_regions_size = 0;
4464 sbi->s_fc_replay_state.fc_regions_used = 0;
4465 sbi->s_fc_replay_state.fc_regions_valid = 0;
4466 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4467 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4468 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4471 static int ext4_inode_info_init(struct super_block *sb,
4472 struct ext4_super_block *es)
4474 struct ext4_sb_info *sbi = EXT4_SB(sb);
4476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4477 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4478 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4480 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4481 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4482 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4483 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4487 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4488 (!is_power_of_2(sbi->s_inode_size)) ||
4489 (sbi->s_inode_size > sb->s_blocksize)) {
4490 ext4_msg(sb, KERN_ERR,
4491 "unsupported inode size: %d",
4493 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4497 * i_atime_extra is the last extra field available for
4498 * [acm]times in struct ext4_inode. Checking for that
4499 * field should suffice to ensure we have extra space
4502 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4503 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4504 sb->s_time_gran = 1;
4505 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4507 sb->s_time_gran = NSEC_PER_SEC;
4508 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4510 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4513 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4514 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4515 EXT4_GOOD_OLD_INODE_SIZE;
4516 if (ext4_has_feature_extra_isize(sb)) {
4517 unsigned v, max = (sbi->s_inode_size -
4518 EXT4_GOOD_OLD_INODE_SIZE);
4520 v = le16_to_cpu(es->s_want_extra_isize);
4522 ext4_msg(sb, KERN_ERR,
4523 "bad s_want_extra_isize: %d", v);
4526 if (sbi->s_want_extra_isize < v)
4527 sbi->s_want_extra_isize = v;
4529 v = le16_to_cpu(es->s_min_extra_isize);
4531 ext4_msg(sb, KERN_ERR,
4532 "bad s_min_extra_isize: %d", v);
4535 if (sbi->s_want_extra_isize < v)
4536 sbi->s_want_extra_isize = v;
4543 #if IS_ENABLED(CONFIG_UNICODE)
4544 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4546 const struct ext4_sb_encodings *encoding_info;
4547 struct unicode_map *encoding;
4548 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4550 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4553 encoding_info = ext4_sb_read_encoding(es);
4554 if (!encoding_info) {
4555 ext4_msg(sb, KERN_ERR,
4556 "Encoding requested by superblock is unknown");
4560 encoding = utf8_load(encoding_info->version);
4561 if (IS_ERR(encoding)) {
4562 ext4_msg(sb, KERN_ERR,
4563 "can't mount with superblock charset: %s-%u.%u.%u "
4564 "not supported by the kernel. flags: 0x%x.",
4565 encoding_info->name,
4566 unicode_major(encoding_info->version),
4567 unicode_minor(encoding_info->version),
4568 unicode_rev(encoding_info->version),
4572 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4573 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4574 unicode_major(encoding_info->version),
4575 unicode_minor(encoding_info->version),
4576 unicode_rev(encoding_info->version),
4579 sb->s_encoding = encoding;
4580 sb->s_encoding_flags = encoding_flags;
4585 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4591 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4593 struct ext4_sb_info *sbi = EXT4_SB(sb);
4595 /* Warn if metadata_csum and gdt_csum are both set. */
4596 if (ext4_has_feature_metadata_csum(sb) &&
4597 ext4_has_feature_gdt_csum(sb))
4598 ext4_warning(sb, "metadata_csum and uninit_bg are "
4599 "redundant flags; please run fsck.");
4601 /* Check for a known checksum algorithm */
4602 if (!ext4_verify_csum_type(sb, es)) {
4603 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4604 "unknown checksum algorithm.");
4607 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4608 ext4_orphan_file_block_trigger);
4610 /* Load the checksum driver */
4611 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4612 if (IS_ERR(sbi->s_chksum_driver)) {
4613 int ret = PTR_ERR(sbi->s_chksum_driver);
4614 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4615 sbi->s_chksum_driver = NULL;
4619 /* Check superblock checksum */
4620 if (!ext4_superblock_csum_verify(sb, es)) {
4621 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4622 "invalid superblock checksum. Run e2fsck?");
4626 /* Precompute checksum seed for all metadata */
4627 if (ext4_has_feature_csum_seed(sb))
4628 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4629 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4630 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4631 sizeof(es->s_uuid));
4635 static int ext4_check_feature_compatibility(struct super_block *sb,
4636 struct ext4_super_block *es,
4639 struct ext4_sb_info *sbi = EXT4_SB(sb);
4641 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4642 (ext4_has_compat_features(sb) ||
4643 ext4_has_ro_compat_features(sb) ||
4644 ext4_has_incompat_features(sb)))
4645 ext4_msg(sb, KERN_WARNING,
4646 "feature flags set on rev 0 fs, "
4647 "running e2fsck is recommended");
4649 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4650 set_opt2(sb, HURD_COMPAT);
4651 if (ext4_has_feature_64bit(sb)) {
4652 ext4_msg(sb, KERN_ERR,
4653 "The Hurd can't support 64-bit file systems");
4658 * ea_inode feature uses l_i_version field which is not
4659 * available in HURD_COMPAT mode.
4661 if (ext4_has_feature_ea_inode(sb)) {
4662 ext4_msg(sb, KERN_ERR,
4663 "ea_inode feature is not supported for Hurd");
4668 if (IS_EXT2_SB(sb)) {
4669 if (ext2_feature_set_ok(sb))
4670 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4671 "using the ext4 subsystem");
4674 * If we're probing be silent, if this looks like
4675 * it's actually an ext[34] filesystem.
4677 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4679 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4680 "to feature incompatibilities");
4685 if (IS_EXT3_SB(sb)) {
4686 if (ext3_feature_set_ok(sb))
4687 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4688 "using the ext4 subsystem");
4691 * If we're probing be silent, if this looks like
4692 * it's actually an ext4 filesystem.
4694 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4696 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4697 "to feature incompatibilities");
4703 * Check feature flags regardless of the revision level, since we
4704 * previously didn't change the revision level when setting the flags,
4705 * so there is a chance incompat flags are set on a rev 0 filesystem.
4707 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4710 if (sbi->s_daxdev) {
4711 if (sb->s_blocksize == PAGE_SIZE)
4712 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4714 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4717 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4718 if (ext4_has_feature_inline_data(sb)) {
4719 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4720 " that may contain inline data");
4723 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4724 ext4_msg(sb, KERN_ERR,
4725 "DAX unsupported by block device.");
4730 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4731 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4732 es->s_encryption_level);
4739 static int ext4_check_geometry(struct super_block *sb,
4740 struct ext4_super_block *es)
4742 struct ext4_sb_info *sbi = EXT4_SB(sb);
4746 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4747 ext4_msg(sb, KERN_ERR,
4748 "Number of reserved GDT blocks insanely large: %d",
4749 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4753 * Test whether we have more sectors than will fit in sector_t,
4754 * and whether the max offset is addressable by the page cache.
4756 err = generic_check_addressable(sb->s_blocksize_bits,
4757 ext4_blocks_count(es));
4759 ext4_msg(sb, KERN_ERR, "filesystem"
4760 " too large to mount safely on this system");
4764 /* check blocks count against device size */
4765 blocks_count = sb_bdev_nr_blocks(sb);
4766 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4767 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4768 "exceeds size of device (%llu blocks)",
4769 ext4_blocks_count(es), blocks_count);
4774 * It makes no sense for the first data block to be beyond the end
4775 * of the filesystem.
4777 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4778 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4779 "block %u is beyond end of filesystem (%llu)",
4780 le32_to_cpu(es->s_first_data_block),
4781 ext4_blocks_count(es));
4784 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4785 (sbi->s_cluster_ratio == 1)) {
4786 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4787 "block is 0 with a 1k block and cluster size");
4791 blocks_count = (ext4_blocks_count(es) -
4792 le32_to_cpu(es->s_first_data_block) +
4793 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4794 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4795 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4796 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4797 "(block count %llu, first data block %u, "
4798 "blocks per group %lu)", blocks_count,
4799 ext4_blocks_count(es),
4800 le32_to_cpu(es->s_first_data_block),
4801 EXT4_BLOCKS_PER_GROUP(sb));
4804 sbi->s_groups_count = blocks_count;
4805 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4806 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4807 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4808 le32_to_cpu(es->s_inodes_count)) {
4809 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4810 le32_to_cpu(es->s_inodes_count),
4811 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4818 static int ext4_group_desc_init(struct super_block *sb,
4819 struct ext4_super_block *es,
4820 ext4_fsblk_t logical_sb_block,
4821 ext4_group_t *first_not_zeroed)
4823 struct ext4_sb_info *sbi = EXT4_SB(sb);
4824 unsigned int db_count;
4828 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4829 EXT4_DESC_PER_BLOCK(sb);
4830 if (ext4_has_feature_meta_bg(sb)) {
4831 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4832 ext4_msg(sb, KERN_WARNING,
4833 "first meta block group too large: %u "
4834 "(group descriptor block count %u)",
4835 le32_to_cpu(es->s_first_meta_bg), db_count);
4839 rcu_assign_pointer(sbi->s_group_desc,
4840 kvmalloc_array(db_count,
4841 sizeof(struct buffer_head *),
4843 if (sbi->s_group_desc == NULL) {
4844 ext4_msg(sb, KERN_ERR, "not enough memory");
4848 bgl_lock_init(sbi->s_blockgroup_lock);
4850 /* Pre-read the descriptors into the buffer cache */
4851 for (i = 0; i < db_count; i++) {
4852 block = descriptor_loc(sb, logical_sb_block, i);
4853 ext4_sb_breadahead_unmovable(sb, block);
4856 for (i = 0; i < db_count; i++) {
4857 struct buffer_head *bh;
4859 block = descriptor_loc(sb, logical_sb_block, i);
4860 bh = ext4_sb_bread_unmovable(sb, block);
4862 ext4_msg(sb, KERN_ERR,
4863 "can't read group descriptor %d", i);
4864 sbi->s_gdb_count = i;
4868 rcu_dereference(sbi->s_group_desc)[i] = bh;
4871 sbi->s_gdb_count = db_count;
4872 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4873 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4874 return -EFSCORRUPTED;
4880 static int ext4_load_and_init_journal(struct super_block *sb,
4881 struct ext4_super_block *es,
4882 struct ext4_fs_context *ctx)
4884 struct ext4_sb_info *sbi = EXT4_SB(sb);
4887 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4891 if (ext4_has_feature_64bit(sb) &&
4892 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4893 JBD2_FEATURE_INCOMPAT_64BIT)) {
4894 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4898 if (!set_journal_csum_feature_set(sb)) {
4899 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4904 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4905 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4906 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4907 ext4_msg(sb, KERN_ERR,
4908 "Failed to set fast commit journal feature");
4912 /* We have now updated the journal if required, so we can
4913 * validate the data journaling mode. */
4914 switch (test_opt(sb, DATA_FLAGS)) {
4916 /* No mode set, assume a default based on the journal
4917 * capabilities: ORDERED_DATA if the journal can
4918 * cope, else JOURNAL_DATA
4920 if (jbd2_journal_check_available_features
4921 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4922 set_opt(sb, ORDERED_DATA);
4923 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4925 set_opt(sb, JOURNAL_DATA);
4926 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4930 case EXT4_MOUNT_ORDERED_DATA:
4931 case EXT4_MOUNT_WRITEBACK_DATA:
4932 if (!jbd2_journal_check_available_features
4933 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4934 ext4_msg(sb, KERN_ERR, "Journal does not support "
4935 "requested data journaling mode");
4943 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4944 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4945 ext4_msg(sb, KERN_ERR, "can't mount with "
4946 "journal_async_commit in data=ordered mode");
4950 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4952 sbi->s_journal->j_submit_inode_data_buffers =
4953 ext4_journal_submit_inode_data_buffers;
4954 sbi->s_journal->j_finish_inode_data_buffers =
4955 ext4_journal_finish_inode_data_buffers;
4960 /* flush s_error_work before journal destroy. */
4961 flush_work(&sbi->s_error_work);
4962 jbd2_journal_destroy(sbi->s_journal);
4963 sbi->s_journal = NULL;
4967 static int ext4_check_journal_data_mode(struct super_block *sb)
4969 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4970 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4971 "data=journal disables delayed allocation, "
4972 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4973 /* can't mount with both data=journal and dioread_nolock. */
4974 clear_opt(sb, DIOREAD_NOLOCK);
4975 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4976 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4977 ext4_msg(sb, KERN_ERR, "can't mount with "
4978 "both data=journal and delalloc");
4981 if (test_opt(sb, DAX_ALWAYS)) {
4982 ext4_msg(sb, KERN_ERR, "can't mount with "
4983 "both data=journal and dax");
4986 if (ext4_has_feature_encrypt(sb)) {
4987 ext4_msg(sb, KERN_WARNING,
4988 "encrypted files will use data=ordered "
4989 "instead of data journaling mode");
4991 if (test_opt(sb, DELALLOC))
4992 clear_opt(sb, DELALLOC);
4994 sb->s_iflags |= SB_I_CGROUPWB;
5000 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5003 struct ext4_sb_info *sbi = EXT4_SB(sb);
5004 struct ext4_super_block *es;
5005 ext4_fsblk_t logical_sb_block;
5006 unsigned long offset = 0;
5007 struct buffer_head *bh;
5011 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5013 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5018 * The ext4 superblock will not be buffer aligned for other than 1kB
5019 * block sizes. We need to calculate the offset from buffer start.
5021 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5022 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5023 offset = do_div(logical_sb_block, blocksize);
5025 logical_sb_block = sbi->s_sb_block;
5028 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5030 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5034 * Note: s_es must be initialized as soon as possible because
5035 * some ext4 macro-instructions depend on its value
5037 es = (struct ext4_super_block *) (bh->b_data + offset);
5039 sb->s_magic = le16_to_cpu(es->s_magic);
5040 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5042 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5046 if (le32_to_cpu(es->s_log_block_size) >
5047 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5048 ext4_msg(sb, KERN_ERR,
5049 "Invalid log block size: %u",
5050 le32_to_cpu(es->s_log_block_size));
5053 if (le32_to_cpu(es->s_log_cluster_size) >
5054 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5055 ext4_msg(sb, KERN_ERR,
5056 "Invalid log cluster size: %u",
5057 le32_to_cpu(es->s_log_cluster_size));
5061 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5064 * If the default block size is not the same as the real block size,
5065 * we need to reload it.
5067 if (sb->s_blocksize == blocksize) {
5068 *lsb = logical_sb_block;
5074 * bh must be released before kill_bdev(), otherwise
5075 * it won't be freed and its page also. kill_bdev()
5076 * is called by sb_set_blocksize().
5079 /* Validate the filesystem blocksize */
5080 if (!sb_set_blocksize(sb, blocksize)) {
5081 ext4_msg(sb, KERN_ERR, "bad block size %d",
5087 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5088 offset = do_div(logical_sb_block, blocksize);
5089 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5091 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5096 es = (struct ext4_super_block *)(bh->b_data + offset);
5098 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5099 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5102 *lsb = logical_sb_block;
5110 static void ext4_hash_info_init(struct super_block *sb)
5112 struct ext4_sb_info *sbi = EXT4_SB(sb);
5113 struct ext4_super_block *es = sbi->s_es;
5116 for (i = 0; i < 4; i++)
5117 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5119 sbi->s_def_hash_version = es->s_def_hash_version;
5120 if (ext4_has_feature_dir_index(sb)) {
5121 i = le32_to_cpu(es->s_flags);
5122 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5123 sbi->s_hash_unsigned = 3;
5124 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5125 #ifdef __CHAR_UNSIGNED__
5128 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5129 sbi->s_hash_unsigned = 3;
5133 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5139 static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5141 struct ext4_sb_info *sbi = EXT4_SB(sb);
5142 struct ext4_super_block *es = sbi->s_es;
5145 has_huge_files = ext4_has_feature_huge_file(sb);
5146 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5148 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5150 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5151 if (ext4_has_feature_64bit(sb)) {
5152 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5153 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5154 !is_power_of_2(sbi->s_desc_size)) {
5155 ext4_msg(sb, KERN_ERR,
5156 "unsupported descriptor size %lu",
5161 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5163 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5164 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5166 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5167 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5169 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5172 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5173 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5174 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5175 sbi->s_inodes_per_group);
5178 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5179 sbi->s_inodes_per_block;
5180 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5181 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5182 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5183 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5188 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5190 struct ext4_super_block *es = NULL;
5191 struct ext4_sb_info *sbi = EXT4_SB(sb);
5192 ext4_fsblk_t logical_sb_block;
5196 ext4_group_t first_not_zeroed;
5197 struct ext4_fs_context *ctx = fc->fs_private;
5198 int silent = fc->sb_flags & SB_SILENT;
5200 /* Set defaults for the variables that will be set during parsing */
5201 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5202 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5204 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5205 sbi->s_sectors_written_start =
5206 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5208 err = ext4_load_super(sb, &logical_sb_block, silent);
5213 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5215 err = ext4_init_metadata_csum(sb, es);
5219 ext4_set_def_opts(sb, es);
5221 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5222 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5223 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5224 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5225 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5228 * set default s_li_wait_mult for lazyinit, for the case there is
5229 * no mount option specified.
5231 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5233 err = ext4_inode_info_init(sb, es);
5237 err = parse_apply_sb_mount_options(sb, ctx);
5241 sbi->s_def_mount_opt = sbi->s_mount_opt;
5242 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5244 err = ext4_check_opt_consistency(fc, sb);
5248 ext4_apply_options(fc, sb);
5250 err = ext4_encoding_init(sb, es);
5254 err = ext4_check_journal_data_mode(sb);
5258 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5259 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5261 /* i_version is always enabled now */
5262 sb->s_flags |= SB_I_VERSION;
5264 err = ext4_check_feature_compatibility(sb, es, silent);
5268 err = ext4_block_group_meta_init(sb, silent);
5272 ext4_hash_info_init(sb);
5274 err = ext4_handle_clustersize(sb);
5278 err = ext4_check_geometry(sb, es);
5282 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5283 spin_lock_init(&sbi->s_error_lock);
5284 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5286 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5290 err = ext4_es_register_shrinker(sbi);
5294 sbi->s_stripe = ext4_get_stripe_size(sbi);
5296 * It's hard to get stripe aligned blocks if stripe is not aligned with
5297 * cluster, just disable stripe and alert user to simpfy code and avoid
5298 * stripe aligned allocation which will rarely successes.
5300 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5301 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5302 ext4_msg(sb, KERN_WARNING,
5303 "stripe (%lu) is not aligned with cluster size (%u), "
5304 "stripe is disabled",
5305 sbi->s_stripe, sbi->s_cluster_ratio);
5308 sbi->s_extent_max_zeroout_kb = 32;
5311 * set up enough so that it can read an inode
5313 sb->s_op = &ext4_sops;
5314 sb->s_export_op = &ext4_export_ops;
5315 sb->s_xattr = ext4_xattr_handlers;
5316 #ifdef CONFIG_FS_ENCRYPTION
5317 sb->s_cop = &ext4_cryptops;
5319 #ifdef CONFIG_FS_VERITY
5320 sb->s_vop = &ext4_verityops;
5323 sb->dq_op = &ext4_quota_operations;
5324 if (ext4_has_feature_quota(sb))
5325 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5327 sb->s_qcop = &ext4_qctl_operations;
5328 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5330 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5332 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5333 mutex_init(&sbi->s_orphan_lock);
5335 ext4_fast_commit_init(sb);
5339 needs_recovery = (es->s_last_orphan != 0 ||
5340 ext4_has_feature_orphan_present(sb) ||
5341 ext4_has_feature_journal_needs_recovery(sb));
5343 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5344 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5346 goto failed_mount3a;
5351 * The first inode we look at is the journal inode. Don't try
5352 * root first: it may be modified in the journal!
5354 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5355 err = ext4_load_and_init_journal(sb, es, ctx);
5357 goto failed_mount3a;
5358 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5359 ext4_has_feature_journal_needs_recovery(sb)) {
5360 ext4_msg(sb, KERN_ERR, "required journal recovery "
5361 "suppressed and not mounted read-only");
5362 goto failed_mount3a;
5364 /* Nojournal mode, all journal mount options are illegal */
5365 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5366 ext4_msg(sb, KERN_ERR, "can't mount with "
5367 "journal_async_commit, fs mounted w/o journal");
5368 goto failed_mount3a;
5371 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5372 ext4_msg(sb, KERN_ERR, "can't mount with "
5373 "journal_checksum, fs mounted w/o journal");
5374 goto failed_mount3a;
5376 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5377 ext4_msg(sb, KERN_ERR, "can't mount with "
5378 "commit=%lu, fs mounted w/o journal",
5379 sbi->s_commit_interval / HZ);
5380 goto failed_mount3a;
5382 if (EXT4_MOUNT_DATA_FLAGS &
5383 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5384 ext4_msg(sb, KERN_ERR, "can't mount with "
5385 "data=, fs mounted w/o journal");
5386 goto failed_mount3a;
5388 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5389 clear_opt(sb, JOURNAL_CHECKSUM);
5390 clear_opt(sb, DATA_FLAGS);
5391 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5392 sbi->s_journal = NULL;
5396 if (!test_opt(sb, NO_MBCACHE)) {
5397 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5398 if (!sbi->s_ea_block_cache) {
5399 ext4_msg(sb, KERN_ERR,
5400 "Failed to create ea_block_cache");
5402 goto failed_mount_wq;
5405 if (ext4_has_feature_ea_inode(sb)) {
5406 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5407 if (!sbi->s_ea_inode_cache) {
5408 ext4_msg(sb, KERN_ERR,
5409 "Failed to create ea_inode_cache");
5411 goto failed_mount_wq;
5417 * Get the # of file system overhead blocks from the
5418 * superblock if present.
5420 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5421 /* ignore the precalculated value if it is ridiculous */
5422 if (sbi->s_overhead > ext4_blocks_count(es))
5423 sbi->s_overhead = 0;
5425 * If the bigalloc feature is not enabled recalculating the
5426 * overhead doesn't take long, so we might as well just redo
5427 * it to make sure we are using the correct value.
5429 if (!ext4_has_feature_bigalloc(sb))
5430 sbi->s_overhead = 0;
5431 if (sbi->s_overhead == 0) {
5432 err = ext4_calculate_overhead(sb);
5434 goto failed_mount_wq;
5438 * The maximum number of concurrent works can be high and
5439 * concurrency isn't really necessary. Limit it to 1.
5441 EXT4_SB(sb)->rsv_conversion_wq =
5442 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5443 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5444 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5450 * The jbd2_journal_load will have done any necessary log recovery,
5451 * so we can safely mount the rest of the filesystem now.
5454 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5456 ext4_msg(sb, KERN_ERR, "get root inode failed");
5457 err = PTR_ERR(root);
5461 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5462 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5464 err = -EFSCORRUPTED;
5468 sb->s_root = d_make_root(root);
5470 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5475 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5476 if (err == -EROFS) {
5477 sb->s_flags |= SB_RDONLY;
5479 goto failed_mount4a;
5481 ext4_set_resv_clusters(sb);
5483 if (test_opt(sb, BLOCK_VALIDITY)) {
5484 err = ext4_setup_system_zone(sb);
5486 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5488 goto failed_mount4a;
5491 ext4_fc_replay_cleanup(sb);
5496 * Enable optimize_scan if number of groups is > threshold. This can be
5497 * turned off by passing "mb_optimize_scan=0". This can also be
5498 * turned on forcefully by passing "mb_optimize_scan=1".
5500 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5501 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5502 set_opt2(sb, MB_OPTIMIZE_SCAN);
5504 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5507 err = ext4_mb_init(sb);
5509 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5515 * We can only set up the journal commit callback once
5516 * mballoc is initialized
5519 sbi->s_journal->j_commit_callback =
5520 ext4_journal_commit_callback;
5522 err = ext4_percpu_param_init(sbi);
5526 if (ext4_has_feature_flex_bg(sb))
5527 if (!ext4_fill_flex_info(sb)) {
5528 ext4_msg(sb, KERN_ERR,
5529 "unable to initialize "
5530 "flex_bg meta info!");
5535 err = ext4_register_li_request(sb, first_not_zeroed);
5539 err = ext4_register_sysfs(sb);
5543 err = ext4_init_orphan_info(sb);
5547 /* Enable quota usage during mount. */
5548 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5549 err = ext4_enable_quotas(sb);
5553 #endif /* CONFIG_QUOTA */
5556 * Save the original bdev mapping's wb_err value which could be
5557 * used to detect the metadata async write error.
5559 spin_lock_init(&sbi->s_bdev_wb_lock);
5560 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5561 &sbi->s_bdev_wb_err);
5562 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5563 ext4_orphan_cleanup(sb, es);
5564 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5566 * Update the checksum after updating free space/inode counters and
5567 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5568 * checksum in the buffer cache until it is written out and
5569 * e2fsprogs programs trying to open a file system immediately
5570 * after it is mounted can fail.
5572 ext4_superblock_csum_set(sb);
5573 if (needs_recovery) {
5574 ext4_msg(sb, KERN_INFO, "recovery complete");
5575 err = ext4_mark_recovery_complete(sb, es);
5577 goto failed_mount10;
5580 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5581 ext4_msg(sb, KERN_WARNING,
5582 "mounting with \"discard\" option, but the device does not support discard");
5584 if (es->s_error_count)
5585 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5587 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5588 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5589 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5590 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5591 atomic_set(&sbi->s_warning_count, 0);
5592 atomic_set(&sbi->s_msg_count, 0);
5597 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5598 failed_mount9: __maybe_unused
5599 ext4_release_orphan_info(sb);
5601 ext4_unregister_sysfs(sb);
5602 kobject_put(&sbi->s_kobj);
5604 ext4_unregister_li_request(sb);
5606 ext4_mb_release(sb);
5607 ext4_flex_groups_free(sbi);
5608 ext4_percpu_param_destroy(sbi);
5610 ext4_ext_release(sb);
5611 ext4_release_system_zone(sb);
5616 ext4_msg(sb, KERN_ERR, "mount failed");
5617 if (EXT4_SB(sb)->rsv_conversion_wq)
5618 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5620 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5621 sbi->s_ea_inode_cache = NULL;
5623 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5624 sbi->s_ea_block_cache = NULL;
5626 if (sbi->s_journal) {
5627 /* flush s_error_work before journal destroy. */
5628 flush_work(&sbi->s_error_work);
5629 jbd2_journal_destroy(sbi->s_journal);
5630 sbi->s_journal = NULL;
5633 ext4_es_unregister_shrinker(sbi);
5635 /* flush s_error_work before sbi destroy */
5636 flush_work(&sbi->s_error_work);
5637 del_timer_sync(&sbi->s_err_report);
5638 ext4_stop_mmpd(sbi);
5639 ext4_group_desc_free(sbi);
5641 if (sbi->s_chksum_driver)
5642 crypto_free_shash(sbi->s_chksum_driver);
5644 #if IS_ENABLED(CONFIG_UNICODE)
5645 utf8_unload(sb->s_encoding);
5649 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5650 kfree(get_qf_name(sb, sbi, i));
5652 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5654 if (sbi->s_journal_bdev) {
5655 invalidate_bdev(sbi->s_journal_bdev);
5656 blkdev_put(sbi->s_journal_bdev, sb);
5659 invalidate_bdev(sb->s_bdev);
5660 sb->s_fs_info = NULL;
5664 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5666 struct ext4_fs_context *ctx = fc->fs_private;
5667 struct ext4_sb_info *sbi;
5671 sbi = ext4_alloc_sbi(sb);
5675 fc->s_fs_info = sbi;
5677 /* Cleanup superblock name */
5678 strreplace(sb->s_id, '/', '!');
5680 sbi->s_sb_block = 1; /* Default super block location */
5681 if (ctx->spec & EXT4_SPEC_s_sb_block)
5682 sbi->s_sb_block = ctx->s_sb_block;
5684 ret = __ext4_fill_super(fc, sb);
5688 if (sbi->s_journal) {
5689 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5690 descr = " journalled data mode";
5691 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5692 descr = " ordered data mode";
5694 descr = " writeback data mode";
5696 descr = "out journal";
5698 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5699 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5700 "Quota mode: %s.", &sb->s_uuid,
5701 sb_rdonly(sb) ? "ro" : "r/w", descr,
5702 ext4_quota_mode(sb));
5704 /* Update the s_overhead_clusters if necessary */
5705 ext4_update_overhead(sb, false);
5710 fc->s_fs_info = NULL;
5714 static int ext4_get_tree(struct fs_context *fc)
5716 return get_tree_bdev(fc, ext4_fill_super);
5720 * Setup any per-fs journal parameters now. We'll do this both on
5721 * initial mount, once the journal has been initialised but before we've
5722 * done any recovery; and again on any subsequent remount.
5724 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5726 struct ext4_sb_info *sbi = EXT4_SB(sb);
5728 journal->j_commit_interval = sbi->s_commit_interval;
5729 journal->j_min_batch_time = sbi->s_min_batch_time;
5730 journal->j_max_batch_time = sbi->s_max_batch_time;
5731 ext4_fc_init(sb, journal);
5733 write_lock(&journal->j_state_lock);
5734 if (test_opt(sb, BARRIER))
5735 journal->j_flags |= JBD2_BARRIER;
5737 journal->j_flags &= ~JBD2_BARRIER;
5738 if (test_opt(sb, DATA_ERR_ABORT))
5739 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5741 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5743 * Always enable journal cycle record option, letting the journal
5744 * records log transactions continuously between each mount.
5746 journal->j_flags |= JBD2_CYCLE_RECORD;
5747 write_unlock(&journal->j_state_lock);
5750 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5751 unsigned int journal_inum)
5753 struct inode *journal_inode;
5756 * Test for the existence of a valid inode on disk. Bad things
5757 * happen if we iget() an unused inode, as the subsequent iput()
5758 * will try to delete it.
5760 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5761 if (IS_ERR(journal_inode)) {
5762 ext4_msg(sb, KERN_ERR, "no journal found");
5765 if (!journal_inode->i_nlink) {
5766 make_bad_inode(journal_inode);
5767 iput(journal_inode);
5768 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5772 ext4_debug("Journal inode found at %p: %lld bytes\n",
5773 journal_inode, journal_inode->i_size);
5774 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5775 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5776 iput(journal_inode);
5779 return journal_inode;
5782 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5784 struct ext4_map_blocks map;
5787 if (journal->j_inode == NULL)
5790 map.m_lblk = *block;
5792 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5794 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5795 "journal bmap failed: block %llu ret %d\n",
5797 jbd2_journal_abort(journal, ret ? ret : -EIO);
5800 *block = map.m_pblk;
5804 static journal_t *ext4_get_journal(struct super_block *sb,
5805 unsigned int journal_inum)
5807 struct inode *journal_inode;
5810 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5813 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5817 journal = jbd2_journal_init_inode(journal_inode);
5819 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5820 iput(journal_inode);
5823 journal->j_private = sb;
5824 journal->j_bmap = ext4_journal_bmap;
5825 ext4_init_journal_params(sb, journal);
5829 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5832 struct buffer_head *bh;
5836 int hblock, blocksize;
5837 ext4_fsblk_t sb_block;
5838 unsigned long offset;
5839 struct ext4_super_block *es;
5840 struct block_device *bdev;
5842 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5845 /* see get_tree_bdev why this is needed and safe */
5846 up_write(&sb->s_umount);
5847 bdev = ext4_blkdev_get(j_dev, sb);
5848 down_write(&sb->s_umount);
5852 blocksize = sb->s_blocksize;
5853 hblock = bdev_logical_block_size(bdev);
5854 if (blocksize < hblock) {
5855 ext4_msg(sb, KERN_ERR,
5856 "blocksize too small for journal device");
5860 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5861 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5862 set_blocksize(bdev, blocksize);
5863 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5864 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5865 "external journal");
5869 es = (struct ext4_super_block *) (bh->b_data + offset);
5870 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5871 !(le32_to_cpu(es->s_feature_incompat) &
5872 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5873 ext4_msg(sb, KERN_ERR, "external journal has "
5879 if ((le32_to_cpu(es->s_feature_ro_compat) &
5880 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5881 es->s_checksum != ext4_superblock_csum(sb, es)) {
5882 ext4_msg(sb, KERN_ERR, "external journal has "
5883 "corrupt superblock");
5888 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5889 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5894 len = ext4_blocks_count(es);
5895 start = sb_block + 1;
5896 brelse(bh); /* we're done with the superblock */
5898 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5899 start, len, blocksize);
5901 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5904 journal->j_private = sb;
5905 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5906 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5909 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5910 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5911 "user (unsupported) - %d",
5912 be32_to_cpu(journal->j_superblock->s_nr_users));
5915 EXT4_SB(sb)->s_journal_bdev = bdev;
5916 ext4_init_journal_params(sb, journal);
5920 jbd2_journal_destroy(journal);
5922 blkdev_put(bdev, sb);
5926 static int ext4_load_journal(struct super_block *sb,
5927 struct ext4_super_block *es,
5928 unsigned long journal_devnum)
5931 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5934 int really_read_only;
5937 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5938 return -EFSCORRUPTED;
5940 if (journal_devnum &&
5941 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5942 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5943 "numbers have changed");
5944 journal_dev = new_decode_dev(journal_devnum);
5946 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5948 if (journal_inum && journal_dev) {
5949 ext4_msg(sb, KERN_ERR,
5950 "filesystem has both journal inode and journal device!");
5955 journal = ext4_get_journal(sb, journal_inum);
5959 journal = ext4_get_dev_journal(sb, journal_dev);
5964 journal_dev_ro = bdev_read_only(journal->j_dev);
5965 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5967 if (journal_dev_ro && !sb_rdonly(sb)) {
5968 ext4_msg(sb, KERN_ERR,
5969 "journal device read-only, try mounting with '-o ro'");
5975 * Are we loading a blank journal or performing recovery after a
5976 * crash? For recovery, we need to check in advance whether we
5977 * can get read-write access to the device.
5979 if (ext4_has_feature_journal_needs_recovery(sb)) {
5980 if (sb_rdonly(sb)) {
5981 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5982 "required on readonly filesystem");
5983 if (really_read_only) {
5984 ext4_msg(sb, KERN_ERR, "write access "
5985 "unavailable, cannot proceed "
5986 "(try mounting with noload)");
5990 ext4_msg(sb, KERN_INFO, "write access will "
5991 "be enabled during recovery");
5995 if (!(journal->j_flags & JBD2_BARRIER))
5996 ext4_msg(sb, KERN_INFO, "barriers disabled");
5998 if (!ext4_has_feature_journal_needs_recovery(sb))
5999 err = jbd2_journal_wipe(journal, !really_read_only);
6001 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6003 bool changed = false;
6006 memcpy(save, ((char *) es) +
6007 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6008 err = jbd2_journal_load(journal);
6009 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6010 save, EXT4_S_ERR_LEN)) {
6011 memcpy(((char *) es) + EXT4_S_ERR_START,
6012 save, EXT4_S_ERR_LEN);
6016 orig_state = es->s_state;
6017 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6019 if (orig_state != es->s_state)
6021 /* Write out restored error information to the superblock */
6022 if (changed && !really_read_only) {
6024 err2 = ext4_commit_super(sb);
6030 ext4_msg(sb, KERN_ERR, "error loading journal");
6034 EXT4_SB(sb)->s_journal = journal;
6035 err = ext4_clear_journal_err(sb, es);
6037 EXT4_SB(sb)->s_journal = NULL;
6038 jbd2_journal_destroy(journal);
6042 if (!really_read_only && journal_devnum &&
6043 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6044 es->s_journal_dev = cpu_to_le32(journal_devnum);
6045 ext4_commit_super(sb);
6047 if (!really_read_only && journal_inum &&
6048 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6049 es->s_journal_inum = cpu_to_le32(journal_inum);
6050 ext4_commit_super(sb);
6056 jbd2_journal_destroy(journal);
6060 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6061 static void ext4_update_super(struct super_block *sb)
6063 struct ext4_sb_info *sbi = EXT4_SB(sb);
6064 struct ext4_super_block *es = sbi->s_es;
6065 struct buffer_head *sbh = sbi->s_sbh;
6069 * If the file system is mounted read-only, don't update the
6070 * superblock write time. This avoids updating the superblock
6071 * write time when we are mounting the root file system
6072 * read/only but we need to replay the journal; at that point,
6073 * for people who are east of GMT and who make their clock
6074 * tick in localtime for Windows bug-for-bug compatibility,
6075 * the clock is set in the future, and this will cause e2fsck
6076 * to complain and force a full file system check.
6078 if (!(sb->s_flags & SB_RDONLY))
6079 ext4_update_tstamp(es, s_wtime);
6080 es->s_kbytes_written =
6081 cpu_to_le64(sbi->s_kbytes_written +
6082 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6083 sbi->s_sectors_written_start) >> 1));
6084 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6085 ext4_free_blocks_count_set(es,
6086 EXT4_C2B(sbi, percpu_counter_sum_positive(
6087 &sbi->s_freeclusters_counter)));
6088 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6089 es->s_free_inodes_count =
6090 cpu_to_le32(percpu_counter_sum_positive(
6091 &sbi->s_freeinodes_counter));
6092 /* Copy error information to the on-disk superblock */
6093 spin_lock(&sbi->s_error_lock);
6094 if (sbi->s_add_error_count > 0) {
6095 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6096 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6097 __ext4_update_tstamp(&es->s_first_error_time,
6098 &es->s_first_error_time_hi,
6099 sbi->s_first_error_time);
6100 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6101 sizeof(es->s_first_error_func));
6102 es->s_first_error_line =
6103 cpu_to_le32(sbi->s_first_error_line);
6104 es->s_first_error_ino =
6105 cpu_to_le32(sbi->s_first_error_ino);
6106 es->s_first_error_block =
6107 cpu_to_le64(sbi->s_first_error_block);
6108 es->s_first_error_errcode =
6109 ext4_errno_to_code(sbi->s_first_error_code);
6111 __ext4_update_tstamp(&es->s_last_error_time,
6112 &es->s_last_error_time_hi,
6113 sbi->s_last_error_time);
6114 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6115 sizeof(es->s_last_error_func));
6116 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6117 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6118 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6119 es->s_last_error_errcode =
6120 ext4_errno_to_code(sbi->s_last_error_code);
6122 * Start the daily error reporting function if it hasn't been
6125 if (!es->s_error_count)
6126 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6127 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6128 sbi->s_add_error_count = 0;
6130 spin_unlock(&sbi->s_error_lock);
6132 ext4_superblock_csum_set(sb);
6136 static int ext4_commit_super(struct super_block *sb)
6138 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6142 if (block_device_ejected(sb))
6145 ext4_update_super(sb);
6148 /* Buffer got discarded which means block device got invalidated */
6149 if (!buffer_mapped(sbh)) {
6154 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6156 * Oh, dear. A previous attempt to write the
6157 * superblock failed. This could happen because the
6158 * USB device was yanked out. Or it could happen to
6159 * be a transient write error and maybe the block will
6160 * be remapped. Nothing we can do but to retry the
6161 * write and hope for the best.
6163 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6164 "superblock detected");
6165 clear_buffer_write_io_error(sbh);
6166 set_buffer_uptodate(sbh);
6169 /* Clear potential dirty bit if it was journalled update */
6170 clear_buffer_dirty(sbh);
6171 sbh->b_end_io = end_buffer_write_sync;
6172 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6173 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6174 wait_on_buffer(sbh);
6175 if (buffer_write_io_error(sbh)) {
6176 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6178 clear_buffer_write_io_error(sbh);
6179 set_buffer_uptodate(sbh);
6186 * Have we just finished recovery? If so, and if we are mounting (or
6187 * remounting) the filesystem readonly, then we will end up with a
6188 * consistent fs on disk. Record that fact.
6190 static int ext4_mark_recovery_complete(struct super_block *sb,
6191 struct ext4_super_block *es)
6194 journal_t *journal = EXT4_SB(sb)->s_journal;
6196 if (!ext4_has_feature_journal(sb)) {
6197 if (journal != NULL) {
6198 ext4_error(sb, "Journal got removed while the fs was "
6200 return -EFSCORRUPTED;
6204 jbd2_journal_lock_updates(journal);
6205 err = jbd2_journal_flush(journal, 0);
6209 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6210 ext4_has_feature_orphan_present(sb))) {
6211 if (!ext4_orphan_file_empty(sb)) {
6212 ext4_error(sb, "Orphan file not empty on read-only fs.");
6213 err = -EFSCORRUPTED;
6216 ext4_clear_feature_journal_needs_recovery(sb);
6217 ext4_clear_feature_orphan_present(sb);
6218 ext4_commit_super(sb);
6221 jbd2_journal_unlock_updates(journal);
6226 * If we are mounting (or read-write remounting) a filesystem whose journal
6227 * has recorded an error from a previous lifetime, move that error to the
6228 * main filesystem now.
6230 static int ext4_clear_journal_err(struct super_block *sb,
6231 struct ext4_super_block *es)
6237 if (!ext4_has_feature_journal(sb)) {
6238 ext4_error(sb, "Journal got removed while the fs was mounted!");
6239 return -EFSCORRUPTED;
6242 journal = EXT4_SB(sb)->s_journal;
6245 * Now check for any error status which may have been recorded in the
6246 * journal by a prior ext4_error() or ext4_abort()
6249 j_errno = jbd2_journal_errno(journal);
6253 errstr = ext4_decode_error(sb, j_errno, nbuf);
6254 ext4_warning(sb, "Filesystem error recorded "
6255 "from previous mount: %s", errstr);
6257 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6258 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6259 j_errno = ext4_commit_super(sb);
6262 ext4_warning(sb, "Marked fs in need of filesystem check.");
6264 jbd2_journal_clear_err(journal);
6265 jbd2_journal_update_sb_errno(journal);
6271 * Force the running and committing transactions to commit,
6272 * and wait on the commit.
6274 int ext4_force_commit(struct super_block *sb)
6281 journal = EXT4_SB(sb)->s_journal;
6282 return ext4_journal_force_commit(journal);
6285 static int ext4_sync_fs(struct super_block *sb, int wait)
6289 bool needs_barrier = false;
6290 struct ext4_sb_info *sbi = EXT4_SB(sb);
6292 if (unlikely(ext4_forced_shutdown(sbi)))
6295 trace_ext4_sync_fs(sb, wait);
6296 flush_workqueue(sbi->rsv_conversion_wq);
6298 * Writeback quota in non-journalled quota case - journalled quota has
6301 dquot_writeback_dquots(sb, -1);
6303 * Data writeback is possible w/o journal transaction, so barrier must
6304 * being sent at the end of the function. But we can skip it if
6305 * transaction_commit will do it for us.
6307 if (sbi->s_journal) {
6308 target = jbd2_get_latest_transaction(sbi->s_journal);
6309 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6310 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6311 needs_barrier = true;
6313 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6315 ret = jbd2_log_wait_commit(sbi->s_journal,
6318 } else if (wait && test_opt(sb, BARRIER))
6319 needs_barrier = true;
6320 if (needs_barrier) {
6322 err = blkdev_issue_flush(sb->s_bdev);
6331 * LVM calls this function before a (read-only) snapshot is created. This
6332 * gives us a chance to flush the journal completely and mark the fs clean.
6334 * Note that only this function cannot bring a filesystem to be in a clean
6335 * state independently. It relies on upper layer to stop all data & metadata
6338 static int ext4_freeze(struct super_block *sb)
6346 journal = EXT4_SB(sb)->s_journal;
6349 /* Now we set up the journal barrier. */
6350 jbd2_journal_lock_updates(journal);
6353 * Don't clear the needs_recovery flag if we failed to
6354 * flush the journal.
6356 error = jbd2_journal_flush(journal, 0);
6360 /* Journal blocked and flushed, clear needs_recovery flag. */
6361 ext4_clear_feature_journal_needs_recovery(sb);
6362 if (ext4_orphan_file_empty(sb))
6363 ext4_clear_feature_orphan_present(sb);
6366 error = ext4_commit_super(sb);
6369 /* we rely on upper layer to stop further updates */
6370 jbd2_journal_unlock_updates(journal);
6375 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6376 * flag here, even though the filesystem is not technically dirty yet.
6378 static int ext4_unfreeze(struct super_block *sb)
6380 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6383 if (EXT4_SB(sb)->s_journal) {
6384 /* Reset the needs_recovery flag before the fs is unlocked. */
6385 ext4_set_feature_journal_needs_recovery(sb);
6386 if (ext4_has_feature_orphan_file(sb))
6387 ext4_set_feature_orphan_present(sb);
6390 ext4_commit_super(sb);
6395 * Structure to save mount options for ext4_remount's benefit
6397 struct ext4_mount_options {
6398 unsigned long s_mount_opt;
6399 unsigned long s_mount_opt2;
6402 unsigned long s_commit_interval;
6403 u32 s_min_batch_time, s_max_batch_time;
6406 char *s_qf_names[EXT4_MAXQUOTAS];
6410 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6412 struct ext4_fs_context *ctx = fc->fs_private;
6413 struct ext4_super_block *es;
6414 struct ext4_sb_info *sbi = EXT4_SB(sb);
6415 unsigned long old_sb_flags;
6416 struct ext4_mount_options old_opts;
6420 int enable_quota = 0;
6422 char *to_free[EXT4_MAXQUOTAS];
6426 /* Store the original options */
6427 old_sb_flags = sb->s_flags;
6428 old_opts.s_mount_opt = sbi->s_mount_opt;
6429 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6430 old_opts.s_resuid = sbi->s_resuid;
6431 old_opts.s_resgid = sbi->s_resgid;
6432 old_opts.s_commit_interval = sbi->s_commit_interval;
6433 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6434 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6436 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6437 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6438 if (sbi->s_qf_names[i]) {
6439 char *qf_name = get_qf_name(sb, sbi, i);
6441 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6442 if (!old_opts.s_qf_names[i]) {
6443 for (j = 0; j < i; j++)
6444 kfree(old_opts.s_qf_names[j]);
6448 old_opts.s_qf_names[i] = NULL;
6450 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6451 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6452 ctx->journal_ioprio =
6453 sbi->s_journal->j_task->io_context->ioprio;
6455 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6459 ext4_apply_options(fc, sb);
6461 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6462 test_opt(sb, JOURNAL_CHECKSUM)) {
6463 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6464 "during remount not supported; ignoring");
6465 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6468 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6469 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6470 ext4_msg(sb, KERN_ERR, "can't mount with "
6471 "both data=journal and delalloc");
6475 if (test_opt(sb, DIOREAD_NOLOCK)) {
6476 ext4_msg(sb, KERN_ERR, "can't mount with "
6477 "both data=journal and dioread_nolock");
6481 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6482 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6483 ext4_msg(sb, KERN_ERR, "can't mount with "
6484 "journal_async_commit in data=ordered mode");
6490 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6491 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6496 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6497 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6499 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6500 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6504 if (sbi->s_journal) {
6505 ext4_init_journal_params(sb, sbi->s_journal);
6506 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6509 /* Flush outstanding errors before changing fs state */
6510 flush_work(&sbi->s_error_work);
6512 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6513 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6518 if (fc->sb_flags & SB_RDONLY) {
6519 err = sync_filesystem(sb);
6522 err = dquot_suspend(sb, -1);
6527 * First of all, the unconditional stuff we have to do
6528 * to disable replay of the journal when we next remount
6530 sb->s_flags |= SB_RDONLY;
6533 * OK, test if we are remounting a valid rw partition
6534 * readonly, and if so set the rdonly flag and then
6535 * mark the partition as valid again.
6537 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6538 (sbi->s_mount_state & EXT4_VALID_FS))
6539 es->s_state = cpu_to_le16(sbi->s_mount_state);
6541 if (sbi->s_journal) {
6543 * We let remount-ro finish even if marking fs
6544 * as clean failed...
6546 ext4_mark_recovery_complete(sb, es);
6549 /* Make sure we can mount this feature set readwrite */
6550 if (ext4_has_feature_readonly(sb) ||
6551 !ext4_feature_set_ok(sb, 0)) {
6556 * Make sure the group descriptor checksums
6557 * are sane. If they aren't, refuse to remount r/w.
6559 for (g = 0; g < sbi->s_groups_count; g++) {
6560 struct ext4_group_desc *gdp =
6561 ext4_get_group_desc(sb, g, NULL);
6563 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6564 ext4_msg(sb, KERN_ERR,
6565 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6566 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6567 le16_to_cpu(gdp->bg_checksum));
6574 * If we have an unprocessed orphan list hanging
6575 * around from a previously readonly bdev mount,
6576 * require a full umount/remount for now.
6578 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6579 ext4_msg(sb, KERN_WARNING, "Couldn't "
6580 "remount RDWR because of unprocessed "
6581 "orphan inode list. Please "
6582 "umount/remount instead");
6588 * Mounting a RDONLY partition read-write, so reread
6589 * and store the current valid flag. (It may have
6590 * been changed by e2fsck since we originally mounted
6593 if (sbi->s_journal) {
6594 err = ext4_clear_journal_err(sb, es);
6598 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6601 err = ext4_setup_super(sb, es, 0);
6605 sb->s_flags &= ~SB_RDONLY;
6606 if (ext4_has_feature_mmp(sb)) {
6607 err = ext4_multi_mount_protect(sb,
6608 le64_to_cpu(es->s_mmp_block));
6619 * Handle creation of system zone data early because it can fail.
6620 * Releasing of existing data is done when we are sure remount will
6623 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6624 err = ext4_setup_system_zone(sb);
6629 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6630 err = ext4_commit_super(sb);
6637 if (sb_any_quota_suspended(sb))
6638 dquot_resume(sb, -1);
6639 else if (ext4_has_feature_quota(sb)) {
6640 err = ext4_enable_quotas(sb);
6645 /* Release old quota file names */
6646 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6647 kfree(old_opts.s_qf_names[i]);
6649 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6650 ext4_release_system_zone(sb);
6653 * Reinitialize lazy itable initialization thread based on
6656 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6657 ext4_unregister_li_request(sb);
6659 ext4_group_t first_not_zeroed;
6660 first_not_zeroed = ext4_has_uninit_itable(sb);
6661 ext4_register_li_request(sb, first_not_zeroed);
6664 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6665 ext4_stop_mmpd(sbi);
6671 * If there was a failing r/w to ro transition, we may need to
6674 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6675 sb_any_quota_suspended(sb))
6676 dquot_resume(sb, -1);
6677 sb->s_flags = old_sb_flags;
6678 sbi->s_mount_opt = old_opts.s_mount_opt;
6679 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6680 sbi->s_resuid = old_opts.s_resuid;
6681 sbi->s_resgid = old_opts.s_resgid;
6682 sbi->s_commit_interval = old_opts.s_commit_interval;
6683 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6684 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6685 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6686 ext4_release_system_zone(sb);
6688 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6689 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6690 to_free[i] = get_qf_name(sb, sbi, i);
6691 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6694 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6697 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6698 ext4_stop_mmpd(sbi);
6702 static int ext4_reconfigure(struct fs_context *fc)
6704 struct super_block *sb = fc->root->d_sb;
6707 fc->s_fs_info = EXT4_SB(sb);
6709 ret = ext4_check_opt_consistency(fc, sb);
6713 ret = __ext4_remount(fc, sb);
6717 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6718 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6719 ext4_quota_mode(sb));
6725 static int ext4_statfs_project(struct super_block *sb,
6726 kprojid_t projid, struct kstatfs *buf)
6729 struct dquot *dquot;
6733 qid = make_kqid_projid(projid);
6734 dquot = dqget(sb, qid);
6736 return PTR_ERR(dquot);
6737 spin_lock(&dquot->dq_dqb_lock);
6739 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6740 dquot->dq_dqb.dqb_bhardlimit);
6741 limit >>= sb->s_blocksize_bits;
6743 if (limit && buf->f_blocks > limit) {
6744 curblock = (dquot->dq_dqb.dqb_curspace +
6745 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6746 buf->f_blocks = limit;
6747 buf->f_bfree = buf->f_bavail =
6748 (buf->f_blocks > curblock) ?
6749 (buf->f_blocks - curblock) : 0;
6752 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6753 dquot->dq_dqb.dqb_ihardlimit);
6754 if (limit && buf->f_files > limit) {
6755 buf->f_files = limit;
6757 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6758 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6761 spin_unlock(&dquot->dq_dqb_lock);
6767 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6769 struct super_block *sb = dentry->d_sb;
6770 struct ext4_sb_info *sbi = EXT4_SB(sb);
6771 struct ext4_super_block *es = sbi->s_es;
6772 ext4_fsblk_t overhead = 0, resv_blocks;
6774 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6776 if (!test_opt(sb, MINIX_DF))
6777 overhead = sbi->s_overhead;
6779 buf->f_type = EXT4_SUPER_MAGIC;
6780 buf->f_bsize = sb->s_blocksize;
6781 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6782 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6783 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6784 /* prevent underflow in case that few free space is available */
6785 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6786 buf->f_bavail = buf->f_bfree -
6787 (ext4_r_blocks_count(es) + resv_blocks);
6788 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6790 buf->f_files = le32_to_cpu(es->s_inodes_count);
6791 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6792 buf->f_namelen = EXT4_NAME_LEN;
6793 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6796 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6797 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6798 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6807 * Helper functions so that transaction is started before we acquire dqio_sem
6808 * to keep correct lock ordering of transaction > dqio_sem
6810 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6812 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6815 static int ext4_write_dquot(struct dquot *dquot)
6819 struct inode *inode;
6821 inode = dquot_to_inode(dquot);
6822 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6823 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6825 return PTR_ERR(handle);
6826 ret = dquot_commit(dquot);
6827 err = ext4_journal_stop(handle);
6833 static int ext4_acquire_dquot(struct dquot *dquot)
6838 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6839 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6841 return PTR_ERR(handle);
6842 ret = dquot_acquire(dquot);
6843 err = ext4_journal_stop(handle);
6849 static int ext4_release_dquot(struct dquot *dquot)
6854 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6855 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6856 if (IS_ERR(handle)) {
6857 /* Release dquot anyway to avoid endless cycle in dqput() */
6858 dquot_release(dquot);
6859 return PTR_ERR(handle);
6861 ret = dquot_release(dquot);
6862 err = ext4_journal_stop(handle);
6868 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6870 struct super_block *sb = dquot->dq_sb;
6872 if (ext4_is_quota_journalled(sb)) {
6873 dquot_mark_dquot_dirty(dquot);
6874 return ext4_write_dquot(dquot);
6876 return dquot_mark_dquot_dirty(dquot);
6880 static int ext4_write_info(struct super_block *sb, int type)
6885 /* Data block + inode block */
6886 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6888 return PTR_ERR(handle);
6889 ret = dquot_commit_info(sb, type);
6890 err = ext4_journal_stop(handle);
6896 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6898 struct ext4_inode_info *ei = EXT4_I(inode);
6900 /* The first argument of lockdep_set_subclass has to be
6901 * *exactly* the same as the argument to init_rwsem() --- in
6902 * this case, in init_once() --- or lockdep gets unhappy
6903 * because the name of the lock is set using the
6904 * stringification of the argument to init_rwsem().
6906 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6907 lockdep_set_subclass(&ei->i_data_sem, subclass);
6911 * Standard function to be called on quota_on
6913 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6914 const struct path *path)
6918 if (!test_opt(sb, QUOTA))
6921 /* Quotafile not on the same filesystem? */
6922 if (path->dentry->d_sb != sb)
6925 /* Quota already enabled for this file? */
6926 if (IS_NOQUOTA(d_inode(path->dentry)))
6929 /* Journaling quota? */
6930 if (EXT4_SB(sb)->s_qf_names[type]) {
6931 /* Quotafile not in fs root? */
6932 if (path->dentry->d_parent != sb->s_root)
6933 ext4_msg(sb, KERN_WARNING,
6934 "Quota file not on filesystem root. "
6935 "Journaled quota will not work");
6936 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6939 * Clear the flag just in case mount options changed since
6942 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6945 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6946 err = dquot_quota_on(sb, type, format_id, path);
6948 struct inode *inode = d_inode(path->dentry);
6952 * Set inode flags to prevent userspace from messing with quota
6953 * files. If this fails, we return success anyway since quotas
6954 * are already enabled and this is not a hard failure.
6957 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6960 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6961 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6962 S_NOATIME | S_IMMUTABLE);
6963 err = ext4_mark_inode_dirty(handle, inode);
6964 ext4_journal_stop(handle);
6966 inode_unlock(inode);
6968 dquot_quota_off(sb, type);
6971 lockdep_set_quota_inode(path->dentry->d_inode,
6976 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6980 return qf_inum == EXT4_USR_QUOTA_INO;
6982 return qf_inum == EXT4_GRP_QUOTA_INO;
6984 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6990 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6994 struct inode *qf_inode;
6995 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6996 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6997 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6998 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7001 BUG_ON(!ext4_has_feature_quota(sb));
7003 if (!qf_inums[type])
7006 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7007 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7008 qf_inums[type], type);
7012 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7013 if (IS_ERR(qf_inode)) {
7014 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7015 qf_inums[type], type);
7016 return PTR_ERR(qf_inode);
7019 /* Don't account quota for quota files to avoid recursion */
7020 qf_inode->i_flags |= S_NOQUOTA;
7021 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7022 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
7024 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7030 /* Enable usage tracking for all quota types. */
7031 int ext4_enable_quotas(struct super_block *sb)
7034 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7035 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7036 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7037 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7039 bool quota_mopt[EXT4_MAXQUOTAS] = {
7040 test_opt(sb, USRQUOTA),
7041 test_opt(sb, GRPQUOTA),
7042 test_opt(sb, PRJQUOTA),
7045 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7046 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7047 if (qf_inums[type]) {
7048 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7049 DQUOT_USAGE_ENABLED |
7050 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7053 "Failed to enable quota tracking "
7054 "(type=%d, err=%d, ino=%lu). "
7055 "Please run e2fsck to fix.", type,
7056 err, qf_inums[type]);
7058 ext4_quotas_off(sb, type);
7066 static int ext4_quota_off(struct super_block *sb, int type)
7068 struct inode *inode = sb_dqopt(sb)->files[type];
7072 /* Force all delayed allocation blocks to be allocated.
7073 * Caller already holds s_umount sem */
7074 if (test_opt(sb, DELALLOC))
7075 sync_filesystem(sb);
7077 if (!inode || !igrab(inode))
7080 err = dquot_quota_off(sb, type);
7081 if (err || ext4_has_feature_quota(sb))
7086 * Update modification times of quota files when userspace can
7087 * start looking at them. If we fail, we return success anyway since
7088 * this is not a hard failure and quotas are already disabled.
7090 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7091 if (IS_ERR(handle)) {
7092 err = PTR_ERR(handle);
7095 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7096 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7097 inode->i_mtime = inode->i_ctime = current_time(inode);
7098 err = ext4_mark_inode_dirty(handle, inode);
7099 ext4_journal_stop(handle);
7101 inode_unlock(inode);
7103 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7107 return dquot_quota_off(sb, type);
7110 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7111 * acquiring the locks... As quota files are never truncated and quota code
7112 * itself serializes the operations (and no one else should touch the files)
7113 * we don't have to be afraid of races */
7114 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7115 size_t len, loff_t off)
7117 struct inode *inode = sb_dqopt(sb)->files[type];
7118 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7119 int offset = off & (sb->s_blocksize - 1);
7122 struct buffer_head *bh;
7123 loff_t i_size = i_size_read(inode);
7127 if (off+len > i_size)
7130 while (toread > 0) {
7131 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7132 bh = ext4_bread(NULL, inode, blk, 0);
7135 if (!bh) /* A hole? */
7136 memset(data, 0, tocopy);
7138 memcpy(data, bh->b_data+offset, tocopy);
7148 /* Write to quotafile (we know the transaction is already started and has
7149 * enough credits) */
7150 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7151 const char *data, size_t len, loff_t off)
7153 struct inode *inode = sb_dqopt(sb)->files[type];
7154 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7155 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7157 struct buffer_head *bh;
7158 handle_t *handle = journal_current_handle();
7161 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7162 " cancelled because transaction is not started",
7163 (unsigned long long)off, (unsigned long long)len);
7167 * Since we account only one data block in transaction credits,
7168 * then it is impossible to cross a block boundary.
7170 if (sb->s_blocksize - offset < len) {
7171 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7172 " cancelled because not block aligned",
7173 (unsigned long long)off, (unsigned long long)len);
7178 bh = ext4_bread(handle, inode, blk,
7179 EXT4_GET_BLOCKS_CREATE |
7180 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7181 } while (PTR_ERR(bh) == -ENOSPC &&
7182 ext4_should_retry_alloc(inode->i_sb, &retries));
7187 BUFFER_TRACE(bh, "get write access");
7188 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7194 memcpy(bh->b_data+offset, data, len);
7195 flush_dcache_page(bh->b_page);
7197 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7200 if (inode->i_size < off + len) {
7201 i_size_write(inode, off + len);
7202 EXT4_I(inode)->i_disksize = inode->i_size;
7203 err2 = ext4_mark_inode_dirty(handle, inode);
7204 if (unlikely(err2 && !err))
7207 return err ? err : len;
7211 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7212 static inline void register_as_ext2(void)
7214 int err = register_filesystem(&ext2_fs_type);
7217 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7220 static inline void unregister_as_ext2(void)
7222 unregister_filesystem(&ext2_fs_type);
7225 static inline int ext2_feature_set_ok(struct super_block *sb)
7227 if (ext4_has_unknown_ext2_incompat_features(sb))
7231 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7236 static inline void register_as_ext2(void) { }
7237 static inline void unregister_as_ext2(void) { }
7238 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7241 static inline void register_as_ext3(void)
7243 int err = register_filesystem(&ext3_fs_type);
7246 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7249 static inline void unregister_as_ext3(void)
7251 unregister_filesystem(&ext3_fs_type);
7254 static inline int ext3_feature_set_ok(struct super_block *sb)
7256 if (ext4_has_unknown_ext3_incompat_features(sb))
7258 if (!ext4_has_feature_journal(sb))
7262 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7267 static void ext4_kill_sb(struct super_block *sb)
7269 struct ext4_sb_info *sbi = EXT4_SB(sb);
7270 struct block_device *journal_bdev = sbi ? sbi->s_journal_bdev : NULL;
7272 kill_block_super(sb);
7275 blkdev_put(journal_bdev, sb);
7278 static struct file_system_type ext4_fs_type = {
7279 .owner = THIS_MODULE,
7281 .init_fs_context = ext4_init_fs_context,
7282 .parameters = ext4_param_specs,
7283 .kill_sb = ext4_kill_sb,
7284 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7286 MODULE_ALIAS_FS("ext4");
7288 /* Shared across all ext4 file systems */
7289 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7291 static int __init ext4_init_fs(void)
7295 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7296 ext4_li_info = NULL;
7298 /* Build-time check for flags consistency */
7299 ext4_check_flag_values();
7301 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7302 init_waitqueue_head(&ext4__ioend_wq[i]);
7304 err = ext4_init_es();
7308 err = ext4_init_pending();
7312 err = ext4_init_post_read_processing();
7316 err = ext4_init_pageio();
7320 err = ext4_init_system_zone();
7324 err = ext4_init_sysfs();
7328 err = ext4_init_mballoc();
7331 err = init_inodecache();
7335 err = ext4_fc_init_dentry_cache();
7341 err = register_filesystem(&ext4_fs_type);
7347 unregister_as_ext2();
7348 unregister_as_ext3();
7349 ext4_fc_destroy_dentry_cache();
7351 destroy_inodecache();
7353 ext4_exit_mballoc();
7357 ext4_exit_system_zone();
7361 ext4_exit_post_read_processing();
7363 ext4_exit_pending();
7370 static void __exit ext4_exit_fs(void)
7372 ext4_destroy_lazyinit_thread();
7373 unregister_as_ext2();
7374 unregister_as_ext3();
7375 unregister_filesystem(&ext4_fs_type);
7376 ext4_fc_destroy_dentry_cache();
7377 destroy_inodecache();
7378 ext4_exit_mballoc();
7380 ext4_exit_system_zone();
7382 ext4_exit_post_read_processing();
7384 ext4_exit_pending();
7387 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7388 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7389 MODULE_LICENSE("GPL");
7390 MODULE_SOFTDEP("pre: crc32c");
7391 module_init(ext4_init_fs)
7392 module_exit(ext4_exit_fs)