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)
437 #define EXT4_SB_REFRESH_INTERVAL_SEC (3600) /* seconds (1 hour) */
438 #define EXT4_SB_REFRESH_INTERVAL_KB (16384) /* kilobytes (16MB) */
441 * The ext4_maybe_update_superblock() function checks and updates the
442 * superblock if needed.
444 * This function is designed to update the on-disk superblock only under
445 * certain conditions to prevent excessive disk writes and unnecessary
446 * waking of the disk from sleep. The superblock will be updated if:
447 * 1. More than an hour has passed since the last superblock update, and
448 * 2. More than 16MB have been written since the last superblock update.
450 * @sb: The superblock
452 static void ext4_maybe_update_superblock(struct super_block *sb)
454 struct ext4_sb_info *sbi = EXT4_SB(sb);
455 struct ext4_super_block *es = sbi->s_es;
456 journal_t *journal = sbi->s_journal;
459 __u64 lifetime_write_kbytes;
462 if (sb_rdonly(sb) || !(sb->s_flags & SB_ACTIVE) ||
463 !journal || (journal->j_flags & JBD2_UNMOUNT))
466 now = ktime_get_real_seconds();
467 last_update = ext4_get_tstamp(es, s_wtime);
469 if (likely(now - last_update < EXT4_SB_REFRESH_INTERVAL_SEC))
472 lifetime_write_kbytes = sbi->s_kbytes_written +
473 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
474 sbi->s_sectors_written_start) >> 1);
476 /* Get the number of kilobytes not written to disk to account
477 * for statistics and compare with a multiple of 16 MB. This
478 * is used to determine when the next superblock commit should
479 * occur (i.e. not more often than once per 16MB if there was
480 * less written in an hour).
482 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written);
484 if (diff_size > EXT4_SB_REFRESH_INTERVAL_KB)
485 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
489 * The del_gendisk() function uninitializes the disk-specific data
490 * structures, including the bdi structure, without telling anyone
491 * else. Once this happens, any attempt to call mark_buffer_dirty()
492 * (for example, by ext4_commit_super), will cause a kernel OOPS.
493 * This is a kludge to prevent these oops until we can put in a proper
494 * hook in del_gendisk() to inform the VFS and file system layers.
496 static int block_device_ejected(struct super_block *sb)
498 struct inode *bd_inode = sb->s_bdev->bd_inode;
499 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
501 return bdi->dev == NULL;
504 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
506 struct super_block *sb = journal->j_private;
507 struct ext4_sb_info *sbi = EXT4_SB(sb);
508 int error = is_journal_aborted(journal);
509 struct ext4_journal_cb_entry *jce;
511 BUG_ON(txn->t_state == T_FINISHED);
513 ext4_process_freed_data(sb, txn->t_tid);
514 ext4_maybe_update_superblock(sb);
516 spin_lock(&sbi->s_md_lock);
517 while (!list_empty(&txn->t_private_list)) {
518 jce = list_entry(txn->t_private_list.next,
519 struct ext4_journal_cb_entry, jce_list);
520 list_del_init(&jce->jce_list);
521 spin_unlock(&sbi->s_md_lock);
522 jce->jce_func(sb, jce, error);
523 spin_lock(&sbi->s_md_lock);
525 spin_unlock(&sbi->s_md_lock);
529 * This writepage callback for write_cache_pages()
530 * takes care of a few cases after page cleaning.
532 * write_cache_pages() already checks for dirty pages
533 * and calls clear_page_dirty_for_io(), which we want,
534 * to write protect the pages.
536 * However, we may have to redirty a page (see below.)
538 static int ext4_journalled_writepage_callback(struct folio *folio,
539 struct writeback_control *wbc,
542 transaction_t *transaction = (transaction_t *) data;
543 struct buffer_head *bh, *head;
544 struct journal_head *jh;
546 bh = head = folio_buffers(folio);
549 * We have to redirty a page in these cases:
550 * 1) If buffer is dirty, it means the page was dirty because it
551 * contains a buffer that needs checkpointing. So the dirty bit
552 * needs to be preserved so that checkpointing writes the buffer
554 * 2) If buffer is not part of the committing transaction
555 * (we may have just accidentally come across this buffer because
556 * inode range tracking is not exact) or if the currently running
557 * transaction already contains this buffer as well, dirty bit
558 * needs to be preserved so that the buffer gets writeprotected
559 * properly on running transaction's commit.
562 if (buffer_dirty(bh) ||
563 (jh && (jh->b_transaction != transaction ||
564 jh->b_next_transaction))) {
565 folio_redirty_for_writepage(wbc, folio);
568 } while ((bh = bh->b_this_page) != head);
571 return AOP_WRITEPAGE_ACTIVATE;
574 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
576 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
577 struct writeback_control wbc = {
578 .sync_mode = WB_SYNC_ALL,
579 .nr_to_write = LONG_MAX,
580 .range_start = jinode->i_dirty_start,
581 .range_end = jinode->i_dirty_end,
584 return write_cache_pages(mapping, &wbc,
585 ext4_journalled_writepage_callback,
586 jinode->i_transaction);
589 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
593 if (ext4_should_journal_data(jinode->i_vfs_inode))
594 ret = ext4_journalled_submit_inode_data_buffers(jinode);
596 ret = ext4_normal_submit_inode_data_buffers(jinode);
600 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
604 if (!ext4_should_journal_data(jinode->i_vfs_inode))
605 ret = jbd2_journal_finish_inode_data_buffers(jinode);
610 static bool system_going_down(void)
612 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
613 || system_state == SYSTEM_RESTART;
616 struct ext4_err_translation {
621 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
623 static struct ext4_err_translation err_translation[] = {
624 EXT4_ERR_TRANSLATE(EIO),
625 EXT4_ERR_TRANSLATE(ENOMEM),
626 EXT4_ERR_TRANSLATE(EFSBADCRC),
627 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
628 EXT4_ERR_TRANSLATE(ENOSPC),
629 EXT4_ERR_TRANSLATE(ENOKEY),
630 EXT4_ERR_TRANSLATE(EROFS),
631 EXT4_ERR_TRANSLATE(EFBIG),
632 EXT4_ERR_TRANSLATE(EEXIST),
633 EXT4_ERR_TRANSLATE(ERANGE),
634 EXT4_ERR_TRANSLATE(EOVERFLOW),
635 EXT4_ERR_TRANSLATE(EBUSY),
636 EXT4_ERR_TRANSLATE(ENOTDIR),
637 EXT4_ERR_TRANSLATE(ENOTEMPTY),
638 EXT4_ERR_TRANSLATE(ESHUTDOWN),
639 EXT4_ERR_TRANSLATE(EFAULT),
642 static int ext4_errno_to_code(int errno)
646 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
647 if (err_translation[i].errno == errno)
648 return err_translation[i].code;
649 return EXT4_ERR_UNKNOWN;
652 static void save_error_info(struct super_block *sb, int error,
653 __u32 ino, __u64 block,
654 const char *func, unsigned int line)
656 struct ext4_sb_info *sbi = EXT4_SB(sb);
658 /* We default to EFSCORRUPTED error... */
660 error = EFSCORRUPTED;
662 spin_lock(&sbi->s_error_lock);
663 sbi->s_add_error_count++;
664 sbi->s_last_error_code = error;
665 sbi->s_last_error_line = line;
666 sbi->s_last_error_ino = ino;
667 sbi->s_last_error_block = block;
668 sbi->s_last_error_func = func;
669 sbi->s_last_error_time = ktime_get_real_seconds();
670 if (!sbi->s_first_error_time) {
671 sbi->s_first_error_code = error;
672 sbi->s_first_error_line = line;
673 sbi->s_first_error_ino = ino;
674 sbi->s_first_error_block = block;
675 sbi->s_first_error_func = func;
676 sbi->s_first_error_time = sbi->s_last_error_time;
678 spin_unlock(&sbi->s_error_lock);
681 /* Deal with the reporting of failure conditions on a filesystem such as
682 * inconsistencies detected or read IO failures.
684 * On ext2, we can store the error state of the filesystem in the
685 * superblock. That is not possible on ext4, because we may have other
686 * write ordering constraints on the superblock which prevent us from
687 * writing it out straight away; and given that the journal is about to
688 * be aborted, we can't rely on the current, or future, transactions to
689 * write out the superblock safely.
691 * We'll just use the jbd2_journal_abort() error code to record an error in
692 * the journal instead. On recovery, the journal will complain about
693 * that error until we've noted it down and cleared it.
695 * If force_ro is set, we unconditionally force the filesystem into an
696 * ABORT|READONLY state, unless the error response on the fs has been set to
697 * panic in which case we take the easy way out and panic immediately. This is
698 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
699 * at a critical moment in log management.
701 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
702 __u32 ino, __u64 block,
703 const char *func, unsigned int line)
705 journal_t *journal = EXT4_SB(sb)->s_journal;
706 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
708 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
709 if (test_opt(sb, WARN_ON_ERROR))
712 if (!continue_fs && !sb_rdonly(sb)) {
713 set_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
715 jbd2_journal_abort(journal, -EIO);
718 if (!bdev_read_only(sb->s_bdev)) {
719 save_error_info(sb, error, ino, block, func, line);
721 * In case the fs should keep running, we need to writeout
722 * superblock through the journal. Due to lock ordering
723 * constraints, it may not be safe to do it right here so we
724 * defer superblock flushing to a workqueue.
726 if (continue_fs && journal)
727 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
729 ext4_commit_super(sb);
733 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
734 * could panic during 'reboot -f' as the underlying device got already
737 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
738 panic("EXT4-fs (device %s): panic forced after error\n",
742 if (sb_rdonly(sb) || continue_fs)
745 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
747 * Make sure updated value of ->s_mount_flags will be visible before
751 sb->s_flags |= SB_RDONLY;
754 static void update_super_work(struct work_struct *work)
756 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
758 journal_t *journal = sbi->s_journal;
762 * If the journal is still running, we have to write out superblock
763 * through the journal to avoid collisions of other journalled sb
766 * We use directly jbd2 functions here to avoid recursing back into
767 * ext4 error handling code during handling of previous errors.
769 if (!sb_rdonly(sbi->s_sb) && journal) {
770 struct buffer_head *sbh = sbi->s_sbh;
771 bool call_notify_err = false;
773 handle = jbd2_journal_start(journal, 1);
776 if (jbd2_journal_get_write_access(handle, sbh)) {
777 jbd2_journal_stop(handle);
781 if (sbi->s_add_error_count > 0)
782 call_notify_err = true;
784 ext4_update_super(sbi->s_sb);
785 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
786 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
787 "superblock detected");
788 clear_buffer_write_io_error(sbh);
789 set_buffer_uptodate(sbh);
792 if (jbd2_journal_dirty_metadata(handle, sbh)) {
793 jbd2_journal_stop(handle);
796 jbd2_journal_stop(handle);
799 ext4_notify_error_sysfs(sbi);
805 * Write through journal failed. Write sb directly to get error info
806 * out and hope for the best.
808 ext4_commit_super(sbi->s_sb);
809 ext4_notify_error_sysfs(sbi);
812 #define ext4_error_ratelimit(sb) \
813 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
816 void __ext4_error(struct super_block *sb, const char *function,
817 unsigned int line, bool force_ro, int error, __u64 block,
818 const char *fmt, ...)
820 struct va_format vaf;
823 if (unlikely(ext4_forced_shutdown(sb)))
826 trace_ext4_error(sb, function, line);
827 if (ext4_error_ratelimit(sb)) {
832 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
833 sb->s_id, function, line, current->comm, &vaf);
836 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
838 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
841 void __ext4_error_inode(struct inode *inode, const char *function,
842 unsigned int line, ext4_fsblk_t block, int error,
843 const char *fmt, ...)
846 struct va_format vaf;
848 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
851 trace_ext4_error(inode->i_sb, function, line);
852 if (ext4_error_ratelimit(inode->i_sb)) {
857 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
858 "inode #%lu: block %llu: comm %s: %pV\n",
859 inode->i_sb->s_id, function, line, inode->i_ino,
860 block, current->comm, &vaf);
862 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
863 "inode #%lu: comm %s: %pV\n",
864 inode->i_sb->s_id, function, line, inode->i_ino,
865 current->comm, &vaf);
868 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
870 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
874 void __ext4_error_file(struct file *file, const char *function,
875 unsigned int line, ext4_fsblk_t block,
876 const char *fmt, ...)
879 struct va_format vaf;
880 struct inode *inode = file_inode(file);
881 char pathname[80], *path;
883 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
886 trace_ext4_error(inode->i_sb, function, line);
887 if (ext4_error_ratelimit(inode->i_sb)) {
888 path = file_path(file, pathname, sizeof(pathname));
896 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
897 "block %llu: comm %s: path %s: %pV\n",
898 inode->i_sb->s_id, function, line, inode->i_ino,
899 block, current->comm, path, &vaf);
902 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
903 "comm %s: path %s: %pV\n",
904 inode->i_sb->s_id, function, line, inode->i_ino,
905 current->comm, path, &vaf);
908 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
910 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
914 const char *ext4_decode_error(struct super_block *sb, int errno,
921 errstr = "Corrupt filesystem";
924 errstr = "Filesystem failed CRC";
927 errstr = "IO failure";
930 errstr = "Out of memory";
933 if (!sb || (EXT4_SB(sb)->s_journal &&
934 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
935 errstr = "Journal has aborted";
937 errstr = "Readonly filesystem";
940 /* If the caller passed in an extra buffer for unknown
941 * errors, textualise them now. Else we just return
944 /* Check for truncated error codes... */
945 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
954 /* __ext4_std_error decodes expected errors from journaling functions
955 * automatically and invokes the appropriate error response. */
957 void __ext4_std_error(struct super_block *sb, const char *function,
958 unsigned int line, int errno)
963 if (unlikely(ext4_forced_shutdown(sb)))
966 /* Special case: if the error is EROFS, and we're not already
967 * inside a transaction, then there's really no point in logging
969 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
972 if (ext4_error_ratelimit(sb)) {
973 errstr = ext4_decode_error(sb, errno, nbuf);
974 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
975 sb->s_id, function, line, errstr);
977 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
979 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
982 void __ext4_msg(struct super_block *sb,
983 const char *prefix, const char *fmt, ...)
985 struct va_format vaf;
989 atomic_inc(&EXT4_SB(sb)->s_msg_count);
990 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
999 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
1001 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
1005 static int ext4_warning_ratelimit(struct super_block *sb)
1007 atomic_inc(&EXT4_SB(sb)->s_warning_count);
1008 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
1012 void __ext4_warning(struct super_block *sb, const char *function,
1013 unsigned int line, const char *fmt, ...)
1015 struct va_format vaf;
1018 if (!ext4_warning_ratelimit(sb))
1021 va_start(args, fmt);
1024 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
1025 sb->s_id, function, line, &vaf);
1029 void __ext4_warning_inode(const struct inode *inode, const char *function,
1030 unsigned int line, const char *fmt, ...)
1032 struct va_format vaf;
1035 if (!ext4_warning_ratelimit(inode->i_sb))
1038 va_start(args, fmt);
1041 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
1042 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
1043 function, line, inode->i_ino, current->comm, &vaf);
1047 void __ext4_grp_locked_error(const char *function, unsigned int line,
1048 struct super_block *sb, ext4_group_t grp,
1049 unsigned long ino, ext4_fsblk_t block,
1050 const char *fmt, ...)
1054 struct va_format vaf;
1057 if (unlikely(ext4_forced_shutdown(sb)))
1060 trace_ext4_error(sb, function, line);
1061 if (ext4_error_ratelimit(sb)) {
1062 va_start(args, fmt);
1065 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1066 sb->s_id, function, line, grp);
1068 printk(KERN_CONT "inode %lu: ", ino);
1070 printk(KERN_CONT "block %llu:",
1071 (unsigned long long) block);
1072 printk(KERN_CONT "%pV\n", &vaf);
1076 if (test_opt(sb, ERRORS_CONT)) {
1077 if (test_opt(sb, WARN_ON_ERROR))
1079 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1080 if (!bdev_read_only(sb->s_bdev)) {
1081 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1083 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
1087 ext4_unlock_group(sb, grp);
1088 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1090 * We only get here in the ERRORS_RO case; relocking the group
1091 * may be dangerous, but nothing bad will happen since the
1092 * filesystem will have already been marked read/only and the
1093 * journal has been aborted. We return 1 as a hint to callers
1094 * who might what to use the return value from
1095 * ext4_grp_locked_error() to distinguish between the
1096 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1097 * aggressively from the ext4 function in question, with a
1098 * more appropriate error code.
1100 ext4_lock_group(sb, grp);
1104 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1108 struct ext4_sb_info *sbi = EXT4_SB(sb);
1109 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1110 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1115 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1116 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1119 percpu_counter_sub(&sbi->s_freeclusters_counter,
1123 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1124 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1129 count = ext4_free_inodes_count(sb, gdp);
1130 percpu_counter_sub(&sbi->s_freeinodes_counter,
1136 void ext4_update_dynamic_rev(struct super_block *sb)
1138 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1140 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1144 "updating to rev %d because of new feature flag, "
1145 "running e2fsck is recommended",
1148 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1149 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1150 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1151 /* leave es->s_feature_*compat flags alone */
1152 /* es->s_uuid will be set by e2fsck if empty */
1155 * The rest of the superblock fields should be zero, and if not it
1156 * means they are likely already in use, so leave them alone. We
1157 * can leave it up to e2fsck to clean up any inconsistencies there.
1161 static inline struct inode *orphan_list_entry(struct list_head *l)
1163 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1166 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1168 struct list_head *l;
1170 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1171 le32_to_cpu(sbi->s_es->s_last_orphan));
1173 printk(KERN_ERR "sb_info orphan list:\n");
1174 list_for_each(l, &sbi->s_orphan) {
1175 struct inode *inode = orphan_list_entry(l);
1177 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1178 inode->i_sb->s_id, inode->i_ino, inode,
1179 inode->i_mode, inode->i_nlink,
1180 NEXT_ORPHAN(inode));
1185 static int ext4_quota_off(struct super_block *sb, int type);
1187 static inline void ext4_quotas_off(struct super_block *sb, int type)
1189 BUG_ON(type > EXT4_MAXQUOTAS);
1191 /* Use our quota_off function to clear inode flags etc. */
1192 for (type--; type >= 0; type--)
1193 ext4_quota_off(sb, type);
1197 * This is a helper function which is used in the mount/remount
1198 * codepaths (which holds s_umount) to fetch the quota file name.
1200 static inline char *get_qf_name(struct super_block *sb,
1201 struct ext4_sb_info *sbi,
1204 return rcu_dereference_protected(sbi->s_qf_names[type],
1205 lockdep_is_held(&sb->s_umount));
1208 static inline void ext4_quotas_off(struct super_block *sb, int type)
1213 static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1218 block = ext4_count_free_clusters(sbi->s_sb);
1219 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1220 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1223 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1224 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1225 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1229 err = percpu_counter_init(&sbi->s_dirs_counter,
1230 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1232 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1235 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1238 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1241 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1246 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1248 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1249 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1250 percpu_counter_destroy(&sbi->s_dirs_counter);
1251 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1252 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1253 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1256 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1258 struct buffer_head **group_desc;
1262 group_desc = rcu_dereference(sbi->s_group_desc);
1263 for (i = 0; i < sbi->s_gdb_count; i++)
1264 brelse(group_desc[i]);
1269 static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1271 struct flex_groups **flex_groups;
1275 flex_groups = rcu_dereference(sbi->s_flex_groups);
1277 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1278 kvfree(flex_groups[i]);
1279 kvfree(flex_groups);
1284 static void ext4_put_super(struct super_block *sb)
1286 struct ext4_sb_info *sbi = EXT4_SB(sb);
1287 struct ext4_super_block *es = sbi->s_es;
1292 * Unregister sysfs before destroying jbd2 journal.
1293 * Since we could still access attr_journal_task attribute via sysfs
1294 * path which could have sbi->s_journal->j_task as NULL
1295 * Unregister sysfs before flush sbi->s_sb_upd_work.
1296 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1297 * read metadata verify failed then will queue error work.
1298 * update_super_work will call start_this_handle may trigger
1301 ext4_unregister_sysfs(sb);
1303 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1304 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1307 ext4_unregister_li_request(sb);
1308 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1310 flush_work(&sbi->s_sb_upd_work);
1311 destroy_workqueue(sbi->rsv_conversion_wq);
1312 ext4_release_orphan_info(sb);
1314 if (sbi->s_journal) {
1315 aborted = is_journal_aborted(sbi->s_journal);
1316 err = jbd2_journal_destroy(sbi->s_journal);
1317 sbi->s_journal = NULL;
1318 if ((err < 0) && !aborted) {
1319 ext4_abort(sb, -err, "Couldn't clean up the journal");
1323 ext4_es_unregister_shrinker(sbi);
1324 timer_shutdown_sync(&sbi->s_err_report);
1325 ext4_release_system_zone(sb);
1326 ext4_mb_release(sb);
1327 ext4_ext_release(sb);
1329 if (!sb_rdonly(sb) && !aborted) {
1330 ext4_clear_feature_journal_needs_recovery(sb);
1331 ext4_clear_feature_orphan_present(sb);
1332 es->s_state = cpu_to_le16(sbi->s_mount_state);
1335 ext4_commit_super(sb);
1337 ext4_group_desc_free(sbi);
1338 ext4_flex_groups_free(sbi);
1339 ext4_percpu_param_destroy(sbi);
1341 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1342 kfree(get_qf_name(sb, sbi, i));
1345 /* Debugging code just in case the in-memory inode orphan list
1346 * isn't empty. The on-disk one can be non-empty if we've
1347 * detected an error and taken the fs readonly, but the
1348 * in-memory list had better be clean by this point. */
1349 if (!list_empty(&sbi->s_orphan))
1350 dump_orphan_list(sb, sbi);
1351 ASSERT(list_empty(&sbi->s_orphan));
1353 sync_blockdev(sb->s_bdev);
1354 invalidate_bdev(sb->s_bdev);
1355 if (sbi->s_journal_bdev) {
1357 * Invalidate the journal device's buffers. We don't want them
1358 * floating about in memory - the physical journal device may
1359 * hotswapped, and it breaks the `ro-after' testing code.
1361 sync_blockdev(sbi->s_journal_bdev);
1362 invalidate_bdev(sbi->s_journal_bdev);
1365 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1366 sbi->s_ea_inode_cache = NULL;
1368 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1369 sbi->s_ea_block_cache = NULL;
1371 ext4_stop_mmpd(sbi);
1374 sb->s_fs_info = NULL;
1376 * Now that we are completely done shutting down the
1377 * superblock, we need to actually destroy the kobject.
1379 kobject_put(&sbi->s_kobj);
1380 wait_for_completion(&sbi->s_kobj_unregister);
1381 if (sbi->s_chksum_driver)
1382 crypto_free_shash(sbi->s_chksum_driver);
1383 kfree(sbi->s_blockgroup_lock);
1384 fs_put_dax(sbi->s_daxdev, NULL);
1385 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1386 #if IS_ENABLED(CONFIG_UNICODE)
1387 utf8_unload(sb->s_encoding);
1392 static struct kmem_cache *ext4_inode_cachep;
1395 * Called inside transaction, so use GFP_NOFS
1397 static struct inode *ext4_alloc_inode(struct super_block *sb)
1399 struct ext4_inode_info *ei;
1401 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1405 inode_set_iversion(&ei->vfs_inode, 1);
1407 spin_lock_init(&ei->i_raw_lock);
1408 ei->i_prealloc_node = RB_ROOT;
1409 atomic_set(&ei->i_prealloc_active, 0);
1410 rwlock_init(&ei->i_prealloc_lock);
1411 ext4_es_init_tree(&ei->i_es_tree);
1412 rwlock_init(&ei->i_es_lock);
1413 INIT_LIST_HEAD(&ei->i_es_list);
1414 ei->i_es_all_nr = 0;
1415 ei->i_es_shk_nr = 0;
1416 ei->i_es_shrink_lblk = 0;
1417 ei->i_reserved_data_blocks = 0;
1418 spin_lock_init(&(ei->i_block_reservation_lock));
1419 ext4_init_pending_tree(&ei->i_pending_tree);
1421 ei->i_reserved_quota = 0;
1422 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1425 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1426 spin_lock_init(&ei->i_completed_io_lock);
1428 ei->i_datasync_tid = 0;
1429 atomic_set(&ei->i_unwritten, 0);
1430 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1431 ext4_fc_init_inode(&ei->vfs_inode);
1432 mutex_init(&ei->i_fc_lock);
1433 return &ei->vfs_inode;
1436 static int ext4_drop_inode(struct inode *inode)
1438 int drop = generic_drop_inode(inode);
1441 drop = fscrypt_drop_inode(inode);
1443 trace_ext4_drop_inode(inode, drop);
1447 static void ext4_free_in_core_inode(struct inode *inode)
1449 fscrypt_free_inode(inode);
1450 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1451 pr_warn("%s: inode %ld still in fc list",
1452 __func__, inode->i_ino);
1454 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1457 static void ext4_destroy_inode(struct inode *inode)
1459 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1460 ext4_msg(inode->i_sb, KERN_ERR,
1461 "Inode %lu (%p): orphan list check failed!",
1462 inode->i_ino, EXT4_I(inode));
1463 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1464 EXT4_I(inode), sizeof(struct ext4_inode_info),
1469 if (EXT4_I(inode)->i_reserved_data_blocks)
1470 ext4_msg(inode->i_sb, KERN_ERR,
1471 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1472 inode->i_ino, EXT4_I(inode),
1473 EXT4_I(inode)->i_reserved_data_blocks);
1476 static void ext4_shutdown(struct super_block *sb)
1478 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1481 static void init_once(void *foo)
1483 struct ext4_inode_info *ei = foo;
1485 INIT_LIST_HEAD(&ei->i_orphan);
1486 init_rwsem(&ei->xattr_sem);
1487 init_rwsem(&ei->i_data_sem);
1488 inode_init_once(&ei->vfs_inode);
1489 ext4_fc_init_inode(&ei->vfs_inode);
1492 static int __init init_inodecache(void)
1494 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1495 sizeof(struct ext4_inode_info), 0,
1496 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1498 offsetof(struct ext4_inode_info, i_data),
1499 sizeof_field(struct ext4_inode_info, i_data),
1501 if (ext4_inode_cachep == NULL)
1506 static void destroy_inodecache(void)
1509 * Make sure all delayed rcu free inodes are flushed before we
1513 kmem_cache_destroy(ext4_inode_cachep);
1516 void ext4_clear_inode(struct inode *inode)
1519 invalidate_inode_buffers(inode);
1521 ext4_discard_preallocations(inode, 0);
1522 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1524 if (EXT4_I(inode)->jinode) {
1525 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1526 EXT4_I(inode)->jinode);
1527 jbd2_free_inode(EXT4_I(inode)->jinode);
1528 EXT4_I(inode)->jinode = NULL;
1530 fscrypt_put_encryption_info(inode);
1531 fsverity_cleanup_inode(inode);
1534 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1535 u64 ino, u32 generation)
1537 struct inode *inode;
1540 * Currently we don't know the generation for parent directory, so
1541 * a generation of 0 means "accept any"
1543 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1545 return ERR_CAST(inode);
1546 if (generation && inode->i_generation != generation) {
1548 return ERR_PTR(-ESTALE);
1554 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1555 int fh_len, int fh_type)
1557 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1558 ext4_nfs_get_inode);
1561 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1562 int fh_len, int fh_type)
1564 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1565 ext4_nfs_get_inode);
1568 static int ext4_nfs_commit_metadata(struct inode *inode)
1570 struct writeback_control wbc = {
1571 .sync_mode = WB_SYNC_ALL
1574 trace_ext4_nfs_commit_metadata(inode);
1575 return ext4_write_inode(inode, &wbc);
1579 static const char * const quotatypes[] = INITQFNAMES;
1580 #define QTYPE2NAME(t) (quotatypes[t])
1582 static int ext4_write_dquot(struct dquot *dquot);
1583 static int ext4_acquire_dquot(struct dquot *dquot);
1584 static int ext4_release_dquot(struct dquot *dquot);
1585 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1586 static int ext4_write_info(struct super_block *sb, int type);
1587 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1588 const struct path *path);
1589 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1590 size_t len, loff_t off);
1591 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1592 const char *data, size_t len, loff_t off);
1593 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1594 unsigned int flags);
1596 static struct dquot **ext4_get_dquots(struct inode *inode)
1598 return EXT4_I(inode)->i_dquot;
1601 static const struct dquot_operations ext4_quota_operations = {
1602 .get_reserved_space = ext4_get_reserved_space,
1603 .write_dquot = ext4_write_dquot,
1604 .acquire_dquot = ext4_acquire_dquot,
1605 .release_dquot = ext4_release_dquot,
1606 .mark_dirty = ext4_mark_dquot_dirty,
1607 .write_info = ext4_write_info,
1608 .alloc_dquot = dquot_alloc,
1609 .destroy_dquot = dquot_destroy,
1610 .get_projid = ext4_get_projid,
1611 .get_inode_usage = ext4_get_inode_usage,
1612 .get_next_id = dquot_get_next_id,
1615 static const struct quotactl_ops ext4_qctl_operations = {
1616 .quota_on = ext4_quota_on,
1617 .quota_off = ext4_quota_off,
1618 .quota_sync = dquot_quota_sync,
1619 .get_state = dquot_get_state,
1620 .set_info = dquot_set_dqinfo,
1621 .get_dqblk = dquot_get_dqblk,
1622 .set_dqblk = dquot_set_dqblk,
1623 .get_nextdqblk = dquot_get_next_dqblk,
1627 static const struct super_operations ext4_sops = {
1628 .alloc_inode = ext4_alloc_inode,
1629 .free_inode = ext4_free_in_core_inode,
1630 .destroy_inode = ext4_destroy_inode,
1631 .write_inode = ext4_write_inode,
1632 .dirty_inode = ext4_dirty_inode,
1633 .drop_inode = ext4_drop_inode,
1634 .evict_inode = ext4_evict_inode,
1635 .put_super = ext4_put_super,
1636 .sync_fs = ext4_sync_fs,
1637 .freeze_fs = ext4_freeze,
1638 .unfreeze_fs = ext4_unfreeze,
1639 .statfs = ext4_statfs,
1640 .show_options = ext4_show_options,
1641 .shutdown = ext4_shutdown,
1643 .quota_read = ext4_quota_read,
1644 .quota_write = ext4_quota_write,
1645 .get_dquots = ext4_get_dquots,
1649 static const struct export_operations ext4_export_ops = {
1650 .fh_to_dentry = ext4_fh_to_dentry,
1651 .fh_to_parent = ext4_fh_to_parent,
1652 .get_parent = ext4_get_parent,
1653 .commit_metadata = ext4_nfs_commit_metadata,
1657 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1658 Opt_resgid, Opt_resuid, Opt_sb,
1659 Opt_nouid32, Opt_debug, Opt_removed,
1660 Opt_user_xattr, Opt_acl,
1661 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1662 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1663 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1664 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1665 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1667 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1668 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1669 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1670 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1671 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1672 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1673 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1674 Opt_inode_readahead_blks, Opt_journal_ioprio,
1675 Opt_dioread_nolock, Opt_dioread_lock,
1676 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1677 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1678 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1679 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1680 #ifdef CONFIG_EXT4_DEBUG
1681 Opt_fc_debug_max_replay, Opt_fc_debug_force
1685 static const struct constant_table ext4_param_errors[] = {
1686 {"continue", EXT4_MOUNT_ERRORS_CONT},
1687 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1688 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1692 static const struct constant_table ext4_param_data[] = {
1693 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1694 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1695 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1699 static const struct constant_table ext4_param_data_err[] = {
1700 {"abort", Opt_data_err_abort},
1701 {"ignore", Opt_data_err_ignore},
1705 static const struct constant_table ext4_param_jqfmt[] = {
1706 {"vfsold", QFMT_VFS_OLD},
1707 {"vfsv0", QFMT_VFS_V0},
1708 {"vfsv1", QFMT_VFS_V1},
1712 static const struct constant_table ext4_param_dax[] = {
1713 {"always", Opt_dax_always},
1714 {"inode", Opt_dax_inode},
1715 {"never", Opt_dax_never},
1719 /* String parameter that allows empty argument */
1720 #define fsparam_string_empty(NAME, OPT) \
1721 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1724 * Mount option specification
1725 * We don't use fsparam_flag_no because of the way we set the
1726 * options and the way we show them in _ext4_show_options(). To
1727 * keep the changes to a minimum, let's keep the negative options
1730 static const struct fs_parameter_spec ext4_param_specs[] = {
1731 fsparam_flag ("bsddf", Opt_bsd_df),
1732 fsparam_flag ("minixdf", Opt_minix_df),
1733 fsparam_flag ("grpid", Opt_grpid),
1734 fsparam_flag ("bsdgroups", Opt_grpid),
1735 fsparam_flag ("nogrpid", Opt_nogrpid),
1736 fsparam_flag ("sysvgroups", Opt_nogrpid),
1737 fsparam_u32 ("resgid", Opt_resgid),
1738 fsparam_u32 ("resuid", Opt_resuid),
1739 fsparam_u32 ("sb", Opt_sb),
1740 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1741 fsparam_flag ("nouid32", Opt_nouid32),
1742 fsparam_flag ("debug", Opt_debug),
1743 fsparam_flag ("oldalloc", Opt_removed),
1744 fsparam_flag ("orlov", Opt_removed),
1745 fsparam_flag ("user_xattr", Opt_user_xattr),
1746 fsparam_flag ("acl", Opt_acl),
1747 fsparam_flag ("norecovery", Opt_noload),
1748 fsparam_flag ("noload", Opt_noload),
1749 fsparam_flag ("bh", Opt_removed),
1750 fsparam_flag ("nobh", Opt_removed),
1751 fsparam_u32 ("commit", Opt_commit),
1752 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1753 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1754 fsparam_u32 ("journal_dev", Opt_journal_dev),
1755 fsparam_bdev ("journal_path", Opt_journal_path),
1756 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1757 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1758 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1759 fsparam_flag ("abort", Opt_abort),
1760 fsparam_enum ("data", Opt_data, ext4_param_data),
1761 fsparam_enum ("data_err", Opt_data_err,
1762 ext4_param_data_err),
1763 fsparam_string_empty
1764 ("usrjquota", Opt_usrjquota),
1765 fsparam_string_empty
1766 ("grpjquota", Opt_grpjquota),
1767 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1768 fsparam_flag ("grpquota", Opt_grpquota),
1769 fsparam_flag ("quota", Opt_quota),
1770 fsparam_flag ("noquota", Opt_noquota),
1771 fsparam_flag ("usrquota", Opt_usrquota),
1772 fsparam_flag ("prjquota", Opt_prjquota),
1773 fsparam_flag ("barrier", Opt_barrier),
1774 fsparam_u32 ("barrier", Opt_barrier),
1775 fsparam_flag ("nobarrier", Opt_nobarrier),
1776 fsparam_flag ("i_version", Opt_removed),
1777 fsparam_flag ("dax", Opt_dax),
1778 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1779 fsparam_u32 ("stripe", Opt_stripe),
1780 fsparam_flag ("delalloc", Opt_delalloc),
1781 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1782 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1783 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1784 fsparam_u32 ("debug_want_extra_isize",
1785 Opt_debug_want_extra_isize),
1786 fsparam_flag ("mblk_io_submit", Opt_removed),
1787 fsparam_flag ("nomblk_io_submit", Opt_removed),
1788 fsparam_flag ("block_validity", Opt_block_validity),
1789 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1790 fsparam_u32 ("inode_readahead_blks",
1791 Opt_inode_readahead_blks),
1792 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1793 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1794 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1795 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1796 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1797 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1798 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1799 fsparam_flag ("discard", Opt_discard),
1800 fsparam_flag ("nodiscard", Opt_nodiscard),
1801 fsparam_u32 ("init_itable", Opt_init_itable),
1802 fsparam_flag ("init_itable", Opt_init_itable),
1803 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1804 #ifdef CONFIG_EXT4_DEBUG
1805 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1806 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1808 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1809 fsparam_flag ("test_dummy_encryption",
1810 Opt_test_dummy_encryption),
1811 fsparam_string ("test_dummy_encryption",
1812 Opt_test_dummy_encryption),
1813 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1814 fsparam_flag ("nombcache", Opt_nombcache),
1815 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1816 fsparam_flag ("prefetch_block_bitmaps",
1818 fsparam_flag ("no_prefetch_block_bitmaps",
1819 Opt_no_prefetch_block_bitmaps),
1820 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1821 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1822 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1823 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1824 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1825 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1829 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1831 #define MOPT_SET 0x0001
1832 #define MOPT_CLEAR 0x0002
1833 #define MOPT_NOSUPPORT 0x0004
1834 #define MOPT_EXPLICIT 0x0008
1837 #define MOPT_QFMT 0x0010
1839 #define MOPT_Q MOPT_NOSUPPORT
1840 #define MOPT_QFMT MOPT_NOSUPPORT
1842 #define MOPT_NO_EXT2 0x0020
1843 #define MOPT_NO_EXT3 0x0040
1844 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1845 #define MOPT_SKIP 0x0080
1846 #define MOPT_2 0x0100
1848 static const struct mount_opts {
1852 } ext4_mount_opts[] = {
1853 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1854 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1855 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1856 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1857 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1858 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1859 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1860 MOPT_EXT4_ONLY | MOPT_SET},
1861 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1862 MOPT_EXT4_ONLY | MOPT_CLEAR},
1863 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1864 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1865 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1866 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1867 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1868 MOPT_EXT4_ONLY | MOPT_CLEAR},
1869 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1870 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1871 {Opt_commit, 0, MOPT_NO_EXT2},
1872 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1873 MOPT_EXT4_ONLY | MOPT_CLEAR},
1874 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1875 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1876 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1877 EXT4_MOUNT_JOURNAL_CHECKSUM),
1878 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1879 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1880 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1881 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1882 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1883 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1884 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1885 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1886 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1887 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1888 {Opt_journal_path, 0, MOPT_NO_EXT2},
1889 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1890 {Opt_data, 0, MOPT_NO_EXT2},
1891 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1892 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1893 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1895 {Opt_acl, 0, MOPT_NOSUPPORT},
1897 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1898 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1899 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1900 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1902 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1904 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1906 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1907 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1908 MOPT_CLEAR | MOPT_Q},
1909 {Opt_usrjquota, 0, MOPT_Q},
1910 {Opt_grpjquota, 0, MOPT_Q},
1911 {Opt_jqfmt, 0, MOPT_QFMT},
1912 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1913 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1915 #ifdef CONFIG_EXT4_DEBUG
1916 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1917 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1919 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2},
1923 #if IS_ENABLED(CONFIG_UNICODE)
1924 static const struct ext4_sb_encodings {
1927 unsigned int version;
1928 } ext4_sb_encoding_map[] = {
1929 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1932 static const struct ext4_sb_encodings *
1933 ext4_sb_read_encoding(const struct ext4_super_block *es)
1935 __u16 magic = le16_to_cpu(es->s_encoding);
1938 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1939 if (magic == ext4_sb_encoding_map[i].magic)
1940 return &ext4_sb_encoding_map[i];
1946 #define EXT4_SPEC_JQUOTA (1 << 0)
1947 #define EXT4_SPEC_JQFMT (1 << 1)
1948 #define EXT4_SPEC_DATAJ (1 << 2)
1949 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1950 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1951 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1952 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1953 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1954 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1955 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1956 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1957 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1958 #define EXT4_SPEC_s_stripe (1 << 13)
1959 #define EXT4_SPEC_s_resuid (1 << 14)
1960 #define EXT4_SPEC_s_resgid (1 << 15)
1961 #define EXT4_SPEC_s_commit_interval (1 << 16)
1962 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1963 #define EXT4_SPEC_s_sb_block (1 << 18)
1964 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1966 struct ext4_fs_context {
1967 char *s_qf_names[EXT4_MAXQUOTAS];
1968 struct fscrypt_dummy_policy dummy_enc_policy;
1969 int s_jquota_fmt; /* Format of quota to use */
1970 #ifdef CONFIG_EXT4_DEBUG
1971 int s_fc_debug_max_replay;
1973 unsigned short qname_spec;
1974 unsigned long vals_s_flags; /* Bits to set in s_flags */
1975 unsigned long mask_s_flags; /* Bits changed in s_flags */
1976 unsigned long journal_devnum;
1977 unsigned long s_commit_interval;
1978 unsigned long s_stripe;
1979 unsigned int s_inode_readahead_blks;
1980 unsigned int s_want_extra_isize;
1981 unsigned int s_li_wait_mult;
1982 unsigned int s_max_dir_size_kb;
1983 unsigned int journal_ioprio;
1984 unsigned int vals_s_mount_opt;
1985 unsigned int mask_s_mount_opt;
1986 unsigned int vals_s_mount_opt2;
1987 unsigned int mask_s_mount_opt2;
1988 unsigned int opt_flags; /* MOPT flags */
1990 u32 s_max_batch_time;
1991 u32 s_min_batch_time;
1994 ext4_fsblk_t s_sb_block;
1997 static void ext4_fc_free(struct fs_context *fc)
1999 struct ext4_fs_context *ctx = fc->fs_private;
2005 for (i = 0; i < EXT4_MAXQUOTAS; i++)
2006 kfree(ctx->s_qf_names[i]);
2008 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
2012 int ext4_init_fs_context(struct fs_context *fc)
2014 struct ext4_fs_context *ctx;
2016 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2020 fc->fs_private = ctx;
2021 fc->ops = &ext4_context_ops;
2028 * Note the name of the specified quota file.
2030 static int note_qf_name(struct fs_context *fc, int qtype,
2031 struct fs_parameter *param)
2033 struct ext4_fs_context *ctx = fc->fs_private;
2036 if (param->size < 1) {
2037 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2040 if (strchr(param->string, '/')) {
2041 ext4_msg(NULL, KERN_ERR,
2042 "quotafile must be on filesystem root");
2045 if (ctx->s_qf_names[qtype]) {
2046 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2047 ext4_msg(NULL, KERN_ERR,
2048 "%s quota file already specified",
2055 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2057 ext4_msg(NULL, KERN_ERR,
2058 "Not enough memory for storing quotafile name");
2061 ctx->s_qf_names[qtype] = qname;
2062 ctx->qname_spec |= 1 << qtype;
2063 ctx->spec |= EXT4_SPEC_JQUOTA;
2068 * Clear the name of the specified quota file.
2070 static int unnote_qf_name(struct fs_context *fc, int qtype)
2072 struct ext4_fs_context *ctx = fc->fs_private;
2074 if (ctx->s_qf_names[qtype])
2075 kfree(ctx->s_qf_names[qtype]);
2077 ctx->s_qf_names[qtype] = NULL;
2078 ctx->qname_spec |= 1 << qtype;
2079 ctx->spec |= EXT4_SPEC_JQUOTA;
2084 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2085 struct ext4_fs_context *ctx)
2089 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2090 ext4_msg(NULL, KERN_WARNING,
2091 "test_dummy_encryption option not supported");
2094 err = fscrypt_parse_test_dummy_encryption(param,
2095 &ctx->dummy_enc_policy);
2096 if (err == -EINVAL) {
2097 ext4_msg(NULL, KERN_WARNING,
2098 "Value of option \"%s\" is unrecognized", param->key);
2099 } else if (err == -EEXIST) {
2100 ext4_msg(NULL, KERN_WARNING,
2101 "Conflicting test_dummy_encryption options");
2107 #define EXT4_SET_CTX(name) \
2108 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2109 unsigned long flag) \
2111 ctx->mask_s_##name |= flag; \
2112 ctx->vals_s_##name |= flag; \
2115 #define EXT4_CLEAR_CTX(name) \
2116 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2117 unsigned long flag) \
2119 ctx->mask_s_##name |= flag; \
2120 ctx->vals_s_##name &= ~flag; \
2123 #define EXT4_TEST_CTX(name) \
2124 static inline unsigned long \
2125 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2127 return (ctx->vals_s_##name & flag); \
2130 EXT4_SET_CTX(flags); /* set only */
2131 EXT4_SET_CTX(mount_opt);
2132 EXT4_CLEAR_CTX(mount_opt);
2133 EXT4_TEST_CTX(mount_opt);
2134 EXT4_SET_CTX(mount_opt2);
2135 EXT4_CLEAR_CTX(mount_opt2);
2136 EXT4_TEST_CTX(mount_opt2);
2138 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2140 struct ext4_fs_context *ctx = fc->fs_private;
2141 struct fs_parse_result result;
2142 const struct mount_opts *m;
2148 token = fs_parse(fc, ext4_param_specs, param, &result);
2151 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2153 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2154 if (token == m->token)
2157 ctx->opt_flags |= m->flags;
2159 if (m->flags & MOPT_EXPLICIT) {
2160 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2161 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2162 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2163 ctx_set_mount_opt2(ctx,
2164 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2169 if (m->flags & MOPT_NOSUPPORT) {
2170 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2178 if (!*param->string)
2179 return unnote_qf_name(fc, USRQUOTA);
2181 return note_qf_name(fc, USRQUOTA, param);
2183 if (!*param->string)
2184 return unnote_qf_name(fc, GRPQUOTA);
2186 return note_qf_name(fc, GRPQUOTA, param);
2189 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2190 ext4_msg(NULL, KERN_WARNING,
2191 "Ignoring %s option on remount", param->key);
2193 ctx->s_sb_block = result.uint_32;
2194 ctx->spec |= EXT4_SPEC_s_sb_block;
2198 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2201 case Opt_inlinecrypt:
2202 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2203 ctx_set_flags(ctx, SB_INLINECRYPT);
2205 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2209 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2210 ctx_set_mount_opt(ctx, result.uint_32);
2214 ctx->s_jquota_fmt = result.uint_32;
2215 ctx->spec |= EXT4_SPEC_JQFMT;
2219 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2220 ctx_set_mount_opt(ctx, result.uint_32);
2221 ctx->spec |= EXT4_SPEC_DATAJ;
2224 if (result.uint_32 == 0)
2225 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2226 else if (result.uint_32 > INT_MAX / HZ) {
2227 ext4_msg(NULL, KERN_ERR,
2228 "Invalid commit interval %d, "
2229 "must be smaller than %d",
2230 result.uint_32, INT_MAX / HZ);
2233 ctx->s_commit_interval = HZ * result.uint_32;
2234 ctx->spec |= EXT4_SPEC_s_commit_interval;
2236 case Opt_debug_want_extra_isize:
2237 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2238 ext4_msg(NULL, KERN_ERR,
2239 "Invalid want_extra_isize %d", result.uint_32);
2242 ctx->s_want_extra_isize = result.uint_32;
2243 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2245 case Opt_max_batch_time:
2246 ctx->s_max_batch_time = result.uint_32;
2247 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2249 case Opt_min_batch_time:
2250 ctx->s_min_batch_time = result.uint_32;
2251 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2253 case Opt_inode_readahead_blks:
2254 if (result.uint_32 &&
2255 (result.uint_32 > (1 << 30) ||
2256 !is_power_of_2(result.uint_32))) {
2257 ext4_msg(NULL, KERN_ERR,
2258 "EXT4-fs: inode_readahead_blks must be "
2259 "0 or a power of 2 smaller than 2^31");
2262 ctx->s_inode_readahead_blks = result.uint_32;
2263 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2265 case Opt_init_itable:
2266 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2267 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2268 if (param->type == fs_value_is_string)
2269 ctx->s_li_wait_mult = result.uint_32;
2270 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2272 case Opt_max_dir_size_kb:
2273 ctx->s_max_dir_size_kb = result.uint_32;
2274 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2276 #ifdef CONFIG_EXT4_DEBUG
2277 case Opt_fc_debug_max_replay:
2278 ctx->s_fc_debug_max_replay = result.uint_32;
2279 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2283 ctx->s_stripe = result.uint_32;
2284 ctx->spec |= EXT4_SPEC_s_stripe;
2287 uid = make_kuid(current_user_ns(), result.uint_32);
2288 if (!uid_valid(uid)) {
2289 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2293 ctx->s_resuid = uid;
2294 ctx->spec |= EXT4_SPEC_s_resuid;
2297 gid = make_kgid(current_user_ns(), result.uint_32);
2298 if (!gid_valid(gid)) {
2299 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2303 ctx->s_resgid = gid;
2304 ctx->spec |= EXT4_SPEC_s_resgid;
2306 case Opt_journal_dev:
2308 ext4_msg(NULL, KERN_ERR,
2309 "Cannot specify journal on remount");
2312 ctx->journal_devnum = result.uint_32;
2313 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2315 case Opt_journal_path:
2317 struct inode *journal_inode;
2322 ext4_msg(NULL, KERN_ERR,
2323 "Cannot specify journal on remount");
2327 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2329 ext4_msg(NULL, KERN_ERR, "error: could not find "
2330 "journal device path");
2334 journal_inode = d_inode(path.dentry);
2335 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2336 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2340 case Opt_journal_ioprio:
2341 if (result.uint_32 > 7) {
2342 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2346 ctx->journal_ioprio =
2347 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2348 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2350 case Opt_test_dummy_encryption:
2351 return ext4_parse_test_dummy_encryption(param, ctx);
2354 #ifdef CONFIG_FS_DAX
2356 int type = (token == Opt_dax) ?
2357 Opt_dax : result.uint_32;
2361 case Opt_dax_always:
2362 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2363 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2366 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2367 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2370 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2371 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2372 /* Strictly for printing options */
2373 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2379 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2383 if (result.uint_32 == Opt_data_err_abort)
2384 ctx_set_mount_opt(ctx, m->mount_opt);
2385 else if (result.uint_32 == Opt_data_err_ignore)
2386 ctx_clear_mount_opt(ctx, m->mount_opt);
2388 case Opt_mb_optimize_scan:
2389 if (result.int_32 == 1) {
2390 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2391 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2392 } else if (result.int_32 == 0) {
2393 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2394 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2396 ext4_msg(NULL, KERN_WARNING,
2397 "mb_optimize_scan should be set to 0 or 1.");
2404 * At this point we should only be getting options requiring MOPT_SET,
2405 * or MOPT_CLEAR. Anything else is a bug
2407 if (m->token == Opt_err) {
2408 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2415 unsigned int set = 0;
2417 if ((param->type == fs_value_is_flag) ||
2421 if (m->flags & MOPT_CLEAR)
2423 else if (unlikely(!(m->flags & MOPT_SET))) {
2424 ext4_msg(NULL, KERN_WARNING,
2425 "buggy handling of option %s",
2430 if (m->flags & MOPT_2) {
2432 ctx_set_mount_opt2(ctx, m->mount_opt);
2434 ctx_clear_mount_opt2(ctx, m->mount_opt);
2437 ctx_set_mount_opt(ctx, m->mount_opt);
2439 ctx_clear_mount_opt(ctx, m->mount_opt);
2446 static int parse_options(struct fs_context *fc, char *options)
2448 struct fs_parameter param;
2455 while ((key = strsep(&options, ",")) != NULL) {
2458 char *value = strchr(key, '=');
2460 param.type = fs_value_is_flag;
2461 param.string = NULL;
2468 v_len = strlen(value);
2469 param.string = kmemdup_nul(value, v_len,
2473 param.type = fs_value_is_string;
2479 ret = ext4_parse_param(fc, ¶m);
2481 kfree(param.string);
2487 ret = ext4_validate_options(fc);
2494 static int parse_apply_sb_mount_options(struct super_block *sb,
2495 struct ext4_fs_context *m_ctx)
2497 struct ext4_sb_info *sbi = EXT4_SB(sb);
2498 char *s_mount_opts = NULL;
2499 struct ext4_fs_context *s_ctx = NULL;
2500 struct fs_context *fc = NULL;
2503 if (!sbi->s_es->s_mount_opts[0])
2506 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2507 sizeof(sbi->s_es->s_mount_opts),
2512 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2516 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2520 fc->fs_private = s_ctx;
2521 fc->s_fs_info = sbi;
2523 ret = parse_options(fc, s_mount_opts);
2527 ret = ext4_check_opt_consistency(fc, sb);
2530 ext4_msg(sb, KERN_WARNING,
2531 "failed to parse options in superblock: %s",
2537 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2538 m_ctx->journal_devnum = s_ctx->journal_devnum;
2539 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2540 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2542 ext4_apply_options(fc, sb);
2550 kfree(s_mount_opts);
2554 static void ext4_apply_quota_options(struct fs_context *fc,
2555 struct super_block *sb)
2558 bool quota_feature = ext4_has_feature_quota(sb);
2559 struct ext4_fs_context *ctx = fc->fs_private;
2560 struct ext4_sb_info *sbi = EXT4_SB(sb);
2567 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2568 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2569 if (!(ctx->qname_spec & (1 << i)))
2572 qname = ctx->s_qf_names[i]; /* May be NULL */
2575 ctx->s_qf_names[i] = NULL;
2576 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2577 lockdep_is_held(&sb->s_umount));
2579 kfree_rcu_mightsleep(qname);
2583 if (ctx->spec & EXT4_SPEC_JQFMT)
2584 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2589 * Check quota settings consistency.
2591 static int ext4_check_quota_consistency(struct fs_context *fc,
2592 struct super_block *sb)
2595 struct ext4_fs_context *ctx = fc->fs_private;
2596 struct ext4_sb_info *sbi = EXT4_SB(sb);
2597 bool quota_feature = ext4_has_feature_quota(sb);
2598 bool quota_loaded = sb_any_quota_loaded(sb);
2599 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2603 * We do the test below only for project quotas. 'usrquota' and
2604 * 'grpquota' mount options are allowed even without quota feature
2605 * to support legacy quotas in quota files.
2607 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2608 !ext4_has_feature_project(sb)) {
2609 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2610 "Cannot enable project quota enforcement.");
2614 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2615 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2617 ctx->mask_s_mount_opt & quota_flags &&
2618 !ctx_test_mount_opt(ctx, quota_flags))
2619 goto err_quota_change;
2621 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2623 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2624 if (!(ctx->qname_spec & (1 << i)))
2628 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2629 goto err_jquota_change;
2631 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2632 strcmp(get_qf_name(sb, sbi, i),
2633 ctx->s_qf_names[i]) != 0)
2634 goto err_jquota_specified;
2637 if (quota_feature) {
2638 ext4_msg(NULL, KERN_INFO,
2639 "Journaled quota options ignored when "
2640 "QUOTA feature is enabled");
2645 if (ctx->spec & EXT4_SPEC_JQFMT) {
2646 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2647 goto err_jquota_change;
2648 if (quota_feature) {
2649 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2650 "ignored when QUOTA feature is enabled");
2655 /* Make sure we don't mix old and new quota format */
2656 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2657 ctx->s_qf_names[USRQUOTA]);
2658 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2659 ctx->s_qf_names[GRPQUOTA]);
2661 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2662 test_opt(sb, USRQUOTA));
2664 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2665 test_opt(sb, GRPQUOTA));
2668 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2672 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2676 if (usr_qf_name || grp_qf_name) {
2677 if (usrquota || grpquota) {
2678 ext4_msg(NULL, KERN_ERR, "old and new quota "
2683 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2684 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2693 ext4_msg(NULL, KERN_ERR,
2694 "Cannot change quota options when quota turned on");
2697 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2698 "options when quota turned on");
2700 err_jquota_specified:
2701 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2709 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2710 struct super_block *sb)
2712 const struct ext4_fs_context *ctx = fc->fs_private;
2713 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2715 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2718 if (!ext4_has_feature_encrypt(sb)) {
2719 ext4_msg(NULL, KERN_WARNING,
2720 "test_dummy_encryption requires encrypt feature");
2724 * This mount option is just for testing, and it's not worthwhile to
2725 * implement the extra complexity (e.g. RCU protection) that would be
2726 * needed to allow it to be set or changed during remount. We do allow
2727 * it to be specified during remount, but only if there is no change.
2729 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2730 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2731 &ctx->dummy_enc_policy))
2733 ext4_msg(NULL, KERN_WARNING,
2734 "Can't set or change test_dummy_encryption on remount");
2737 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2738 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2739 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2740 &ctx->dummy_enc_policy))
2742 ext4_msg(NULL, KERN_WARNING,
2743 "Conflicting test_dummy_encryption options");
2749 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2750 struct super_block *sb)
2752 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2753 /* if already set, it was already verified to be the same */
2754 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2756 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2757 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2758 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2761 static int ext4_check_opt_consistency(struct fs_context *fc,
2762 struct super_block *sb)
2764 struct ext4_fs_context *ctx = fc->fs_private;
2765 struct ext4_sb_info *sbi = fc->s_fs_info;
2766 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2769 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2770 ext4_msg(NULL, KERN_ERR,
2771 "Mount option(s) incompatible with ext2");
2774 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2775 ext4_msg(NULL, KERN_ERR,
2776 "Mount option(s) incompatible with ext3");
2780 if (ctx->s_want_extra_isize >
2781 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2782 ext4_msg(NULL, KERN_ERR,
2783 "Invalid want_extra_isize %d",
2784 ctx->s_want_extra_isize);
2788 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2790 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2791 if (blocksize < PAGE_SIZE)
2792 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2793 "experimental mount option 'dioread_nolock' "
2794 "for blocksize < PAGE_SIZE");
2797 err = ext4_check_test_dummy_encryption(fc, sb);
2801 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2802 if (!sbi->s_journal) {
2803 ext4_msg(NULL, KERN_WARNING,
2804 "Remounting file system with no journal "
2805 "so ignoring journalled data option");
2806 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2807 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2808 test_opt(sb, DATA_FLAGS)) {
2809 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2816 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2817 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2818 ext4_msg(NULL, KERN_ERR, "can't mount with "
2819 "both data=journal and dax");
2823 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2824 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2825 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2826 fail_dax_change_remount:
2827 ext4_msg(NULL, KERN_ERR, "can't change "
2828 "dax mount option while remounting");
2830 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2831 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2832 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2833 goto fail_dax_change_remount;
2834 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2835 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2836 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2837 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2838 goto fail_dax_change_remount;
2842 return ext4_check_quota_consistency(fc, sb);
2845 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2847 struct ext4_fs_context *ctx = fc->fs_private;
2848 struct ext4_sb_info *sbi = fc->s_fs_info;
2850 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2851 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2852 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2853 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2854 sb->s_flags &= ~ctx->mask_s_flags;
2855 sb->s_flags |= ctx->vals_s_flags;
2857 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2858 APPLY(s_commit_interval);
2860 APPLY(s_max_batch_time);
2861 APPLY(s_min_batch_time);
2862 APPLY(s_want_extra_isize);
2863 APPLY(s_inode_readahead_blks);
2864 APPLY(s_max_dir_size_kb);
2865 APPLY(s_li_wait_mult);
2869 #ifdef CONFIG_EXT4_DEBUG
2870 APPLY(s_fc_debug_max_replay);
2873 ext4_apply_quota_options(fc, sb);
2874 ext4_apply_test_dummy_encryption(ctx, sb);
2878 static int ext4_validate_options(struct fs_context *fc)
2881 struct ext4_fs_context *ctx = fc->fs_private;
2882 char *usr_qf_name, *grp_qf_name;
2884 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2885 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2887 if (usr_qf_name || grp_qf_name) {
2888 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2889 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2891 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2892 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2894 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2895 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2896 ext4_msg(NULL, KERN_ERR, "old and new quota "
2905 static inline void ext4_show_quota_options(struct seq_file *seq,
2906 struct super_block *sb)
2908 #if defined(CONFIG_QUOTA)
2909 struct ext4_sb_info *sbi = EXT4_SB(sb);
2910 char *usr_qf_name, *grp_qf_name;
2912 if (sbi->s_jquota_fmt) {
2915 switch (sbi->s_jquota_fmt) {
2926 seq_printf(seq, ",jqfmt=%s", fmtname);
2930 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2931 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2933 seq_show_option(seq, "usrjquota", usr_qf_name);
2935 seq_show_option(seq, "grpjquota", grp_qf_name);
2940 static const char *token2str(int token)
2942 const struct fs_parameter_spec *spec;
2944 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2945 if (spec->opt == token && !spec->type)
2952 * - it's set to a non-default value OR
2953 * - if the per-sb default is different from the global default
2955 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2958 struct ext4_sb_info *sbi = EXT4_SB(sb);
2959 struct ext4_super_block *es = sbi->s_es;
2961 const struct mount_opts *m;
2962 char sep = nodefs ? '\n' : ',';
2964 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2965 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2967 if (sbi->s_sb_block != 1)
2968 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2970 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2971 int want_set = m->flags & MOPT_SET;
2972 int opt_2 = m->flags & MOPT_2;
2973 unsigned int mount_opt, def_mount_opt;
2975 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2976 m->flags & MOPT_SKIP)
2980 mount_opt = sbi->s_mount_opt2;
2981 def_mount_opt = sbi->s_def_mount_opt2;
2983 mount_opt = sbi->s_mount_opt;
2984 def_mount_opt = sbi->s_def_mount_opt;
2986 /* skip if same as the default */
2987 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2989 /* select Opt_noFoo vs Opt_Foo */
2991 (mount_opt & m->mount_opt) != m->mount_opt) ||
2992 (!want_set && (mount_opt & m->mount_opt)))
2994 SEQ_OPTS_PRINT("%s", token2str(m->token));
2997 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2998 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2999 SEQ_OPTS_PRINT("resuid=%u",
3000 from_kuid_munged(&init_user_ns, sbi->s_resuid));
3001 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
3002 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
3003 SEQ_OPTS_PRINT("resgid=%u",
3004 from_kgid_munged(&init_user_ns, sbi->s_resgid));
3005 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
3006 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3007 SEQ_OPTS_PUTS("errors=remount-ro");
3008 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
3009 SEQ_OPTS_PUTS("errors=continue");
3010 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
3011 SEQ_OPTS_PUTS("errors=panic");
3012 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
3013 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
3014 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
3015 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
3016 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
3017 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3018 if (nodefs || sbi->s_stripe)
3019 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3020 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3021 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3022 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3023 SEQ_OPTS_PUTS("data=journal");
3024 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3025 SEQ_OPTS_PUTS("data=ordered");
3026 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3027 SEQ_OPTS_PUTS("data=writeback");
3030 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3031 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3032 sbi->s_inode_readahead_blks);
3034 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3035 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3036 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3037 if (nodefs || sbi->s_max_dir_size_kb)
3038 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3039 if (test_opt(sb, DATA_ERR_ABORT))
3040 SEQ_OPTS_PUTS("data_err=abort");
3042 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3044 if (sb->s_flags & SB_INLINECRYPT)
3045 SEQ_OPTS_PUTS("inlinecrypt");
3047 if (test_opt(sb, DAX_ALWAYS)) {
3049 SEQ_OPTS_PUTS("dax");
3051 SEQ_OPTS_PUTS("dax=always");
3052 } else if (test_opt2(sb, DAX_NEVER)) {
3053 SEQ_OPTS_PUTS("dax=never");
3054 } else if (test_opt2(sb, DAX_INODE)) {
3055 SEQ_OPTS_PUTS("dax=inode");
3058 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3059 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3060 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3061 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3062 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3063 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3066 ext4_show_quota_options(seq, sb);
3070 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3072 return _ext4_show_options(seq, root->d_sb, 0);
3075 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3077 struct super_block *sb = seq->private;
3080 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3081 rc = _ext4_show_options(seq, sb, 1);
3082 seq_puts(seq, "\n");
3086 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3089 struct ext4_sb_info *sbi = EXT4_SB(sb);
3092 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3093 ext4_msg(sb, KERN_ERR, "revision level too high, "
3094 "forcing read-only mode");
3100 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3101 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3102 "running e2fsck is recommended");
3103 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3104 ext4_msg(sb, KERN_WARNING,
3105 "warning: mounting fs with errors, "
3106 "running e2fsck is recommended");
3107 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3108 le16_to_cpu(es->s_mnt_count) >=
3109 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3110 ext4_msg(sb, KERN_WARNING,
3111 "warning: maximal mount count reached, "
3112 "running e2fsck is recommended");
3113 else if (le32_to_cpu(es->s_checkinterval) &&
3114 (ext4_get_tstamp(es, s_lastcheck) +
3115 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3116 ext4_msg(sb, KERN_WARNING,
3117 "warning: checktime reached, "
3118 "running e2fsck is recommended");
3119 if (!sbi->s_journal)
3120 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3121 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3122 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3123 le16_add_cpu(&es->s_mnt_count, 1);
3124 ext4_update_tstamp(es, s_mtime);
3125 if (sbi->s_journal) {
3126 ext4_set_feature_journal_needs_recovery(sb);
3127 if (ext4_has_feature_orphan_file(sb))
3128 ext4_set_feature_orphan_present(sb);
3131 err = ext4_commit_super(sb);
3133 if (test_opt(sb, DEBUG))
3134 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3135 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3137 sbi->s_groups_count,
3138 EXT4_BLOCKS_PER_GROUP(sb),
3139 EXT4_INODES_PER_GROUP(sb),
3140 sbi->s_mount_opt, sbi->s_mount_opt2);
3144 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3146 struct ext4_sb_info *sbi = EXT4_SB(sb);
3147 struct flex_groups **old_groups, **new_groups;
3150 if (!sbi->s_log_groups_per_flex)
3153 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3154 if (size <= sbi->s_flex_groups_allocated)
3157 new_groups = kvzalloc(roundup_pow_of_two(size *
3158 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3160 ext4_msg(sb, KERN_ERR,
3161 "not enough memory for %d flex group pointers", size);
3164 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3165 new_groups[i] = kvzalloc(roundup_pow_of_two(
3166 sizeof(struct flex_groups)),
3168 if (!new_groups[i]) {
3169 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3170 kvfree(new_groups[j]);
3172 ext4_msg(sb, KERN_ERR,
3173 "not enough memory for %d flex groups", size);
3178 old_groups = rcu_dereference(sbi->s_flex_groups);
3180 memcpy(new_groups, old_groups,
3181 (sbi->s_flex_groups_allocated *
3182 sizeof(struct flex_groups *)));
3184 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3185 sbi->s_flex_groups_allocated = size;
3187 ext4_kvfree_array_rcu(old_groups);
3191 static int ext4_fill_flex_info(struct super_block *sb)
3193 struct ext4_sb_info *sbi = EXT4_SB(sb);
3194 struct ext4_group_desc *gdp = NULL;
3195 struct flex_groups *fg;
3196 ext4_group_t flex_group;
3199 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3200 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3201 sbi->s_log_groups_per_flex = 0;
3205 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3209 for (i = 0; i < sbi->s_groups_count; i++) {
3210 gdp = ext4_get_group_desc(sb, i, NULL);
3212 flex_group = ext4_flex_group(sbi, i);
3213 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3214 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3215 atomic64_add(ext4_free_group_clusters(sb, gdp),
3216 &fg->free_clusters);
3217 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3225 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3226 struct ext4_group_desc *gdp)
3228 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3230 __le32 le_group = cpu_to_le32(block_group);
3231 struct ext4_sb_info *sbi = EXT4_SB(sb);
3233 if (ext4_has_metadata_csum(sbi->s_sb)) {
3234 /* Use new metadata_csum algorithm */
3236 __u16 dummy_csum = 0;
3238 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3240 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3241 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3242 sizeof(dummy_csum));
3243 offset += sizeof(dummy_csum);
3244 if (offset < sbi->s_desc_size)
3245 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3246 sbi->s_desc_size - offset);
3248 crc = csum32 & 0xFFFF;
3252 /* old crc16 code */
3253 if (!ext4_has_feature_gdt_csum(sb))
3256 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3257 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3258 crc = crc16(crc, (__u8 *)gdp, offset);
3259 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3260 /* for checksum of struct ext4_group_desc do the rest...*/
3261 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3262 crc = crc16(crc, (__u8 *)gdp + offset,
3263 sbi->s_desc_size - offset);
3266 return cpu_to_le16(crc);
3269 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3270 struct ext4_group_desc *gdp)
3272 if (ext4_has_group_desc_csum(sb) &&
3273 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3279 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3280 struct ext4_group_desc *gdp)
3282 if (!ext4_has_group_desc_csum(sb))
3284 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3287 /* Called at mount-time, super-block is locked */
3288 static int ext4_check_descriptors(struct super_block *sb,
3289 ext4_fsblk_t sb_block,
3290 ext4_group_t *first_not_zeroed)
3292 struct ext4_sb_info *sbi = EXT4_SB(sb);
3293 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3294 ext4_fsblk_t last_block;
3295 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3296 ext4_fsblk_t block_bitmap;
3297 ext4_fsblk_t inode_bitmap;
3298 ext4_fsblk_t inode_table;
3299 int flexbg_flag = 0;
3300 ext4_group_t i, grp = sbi->s_groups_count;
3302 if (ext4_has_feature_flex_bg(sb))
3305 ext4_debug("Checking group descriptors");
3307 for (i = 0; i < sbi->s_groups_count; i++) {
3308 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3310 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3311 last_block = ext4_blocks_count(sbi->s_es) - 1;
3313 last_block = first_block +
3314 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3316 if ((grp == sbi->s_groups_count) &&
3317 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3320 block_bitmap = ext4_block_bitmap(sb, gdp);
3321 if (block_bitmap == sb_block) {
3322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3323 "Block bitmap for group %u overlaps "
3328 if (block_bitmap >= sb_block + 1 &&
3329 block_bitmap <= last_bg_block) {
3330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3331 "Block bitmap for group %u overlaps "
3332 "block group descriptors", i);
3336 if (block_bitmap < first_block || block_bitmap > last_block) {
3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3338 "Block bitmap for group %u not in group "
3339 "(block %llu)!", i, block_bitmap);
3342 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3343 if (inode_bitmap == sb_block) {
3344 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3345 "Inode bitmap for group %u overlaps "
3350 if (inode_bitmap >= sb_block + 1 &&
3351 inode_bitmap <= last_bg_block) {
3352 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3353 "Inode bitmap for group %u overlaps "
3354 "block group descriptors", i);
3358 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3360 "Inode bitmap for group %u not in group "
3361 "(block %llu)!", i, inode_bitmap);
3364 inode_table = ext4_inode_table(sb, gdp);
3365 if (inode_table == sb_block) {
3366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3367 "Inode table for group %u overlaps "
3372 if (inode_table >= sb_block + 1 &&
3373 inode_table <= last_bg_block) {
3374 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3375 "Inode table for group %u overlaps "
3376 "block group descriptors", i);
3380 if (inode_table < first_block ||
3381 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3382 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3383 "Inode table for group %u not in group "
3384 "(block %llu)!", i, inode_table);
3387 ext4_lock_group(sb, i);
3388 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3389 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3390 "Checksum for group %u failed (%u!=%u)",
3391 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3392 gdp)), le16_to_cpu(gdp->bg_checksum));
3393 if (!sb_rdonly(sb)) {
3394 ext4_unlock_group(sb, i);
3398 ext4_unlock_group(sb, i);
3400 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3402 if (NULL != first_not_zeroed)
3403 *first_not_zeroed = grp;
3408 * Maximal extent format file size.
3409 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3410 * extent format containers, within a sector_t, and within i_blocks
3411 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3412 * so that won't be a limiting factor.
3414 * However there is other limiting factor. We do store extents in the form
3415 * of starting block and length, hence the resulting length of the extent
3416 * covering maximum file size must fit into on-disk format containers as
3417 * well. Given that length is always by 1 unit bigger than max unit (because
3418 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3420 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3422 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3425 loff_t upper_limit = MAX_LFS_FILESIZE;
3427 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3429 if (!has_huge_files) {
3430 upper_limit = (1LL << 32) - 1;
3432 /* total blocks in file system block size */
3433 upper_limit >>= (blkbits - 9);
3434 upper_limit <<= blkbits;
3438 * 32-bit extent-start container, ee_block. We lower the maxbytes
3439 * by one fs block, so ee_len can cover the extent of maximum file
3442 res = (1LL << 32) - 1;
3445 /* Sanity check against vm- & vfs- imposed limits */
3446 if (res > upper_limit)
3453 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3454 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3455 * We need to be 1 filesystem block less than the 2^48 sector limit.
3457 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3459 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3461 unsigned int ppb = 1 << (bits - 2);
3464 * This is calculated to be the largest file size for a dense, block
3465 * mapped file such that the file's total number of 512-byte sectors,
3466 * including data and all indirect blocks, does not exceed (2^48 - 1).
3468 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3469 * number of 512-byte sectors of the file.
3471 if (!has_huge_files) {
3473 * !has_huge_files or implies that the inode i_block field
3474 * represents total file blocks in 2^32 512-byte sectors ==
3475 * size of vfs inode i_blocks * 8
3477 upper_limit = (1LL << 32) - 1;
3479 /* total blocks in file system block size */
3480 upper_limit >>= (bits - 9);
3484 * We use 48 bit ext4_inode i_blocks
3485 * With EXT4_HUGE_FILE_FL set the i_blocks
3486 * represent total number of blocks in
3487 * file system block size
3489 upper_limit = (1LL << 48) - 1;
3493 /* Compute how many blocks we can address by block tree */
3496 res += ((loff_t)ppb) * ppb * ppb;
3497 /* Compute how many metadata blocks are needed */
3499 meta_blocks += 1 + ppb;
3500 meta_blocks += 1 + ppb + ppb * ppb;
3501 /* Does block tree limit file size? */
3502 if (res + meta_blocks <= upper_limit)
3506 /* How many metadata blocks are needed for addressing upper_limit? */
3507 upper_limit -= EXT4_NDIR_BLOCKS;
3508 /* indirect blocks */
3511 /* double indirect blocks */
3512 if (upper_limit < ppb * ppb) {
3513 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3517 meta_blocks += 1 + ppb;
3518 upper_limit -= ppb * ppb;
3519 /* tripple indirect blocks for the rest */
3520 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3521 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3525 if (res > MAX_LFS_FILESIZE)
3526 res = MAX_LFS_FILESIZE;
3531 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3532 ext4_fsblk_t logical_sb_block, int nr)
3534 struct ext4_sb_info *sbi = EXT4_SB(sb);
3535 ext4_group_t bg, first_meta_bg;
3538 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3540 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3541 return logical_sb_block + nr + 1;
3542 bg = sbi->s_desc_per_block * nr;
3543 if (ext4_bg_has_super(sb, bg))
3547 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3548 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3549 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3552 if (sb->s_blocksize == 1024 && nr == 0 &&
3553 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3556 return (has_super + ext4_group_first_block_no(sb, bg));
3560 * ext4_get_stripe_size: Get the stripe size.
3561 * @sbi: In memory super block info
3563 * If we have specified it via mount option, then
3564 * use the mount option value. If the value specified at mount time is
3565 * greater than the blocks per group use the super block value.
3566 * If the super block value is greater than blocks per group return 0.
3567 * Allocator needs it be less than blocks per group.
3570 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3572 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3573 unsigned long stripe_width =
3574 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3577 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3578 ret = sbi->s_stripe;
3579 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3581 else if (stride && stride <= sbi->s_blocks_per_group)
3587 * If the stripe width is 1, this makes no sense and
3588 * we set it to 0 to turn off stripe handling code.
3597 * Check whether this filesystem can be mounted based on
3598 * the features present and the RDONLY/RDWR mount requested.
3599 * Returns 1 if this filesystem can be mounted as requested,
3600 * 0 if it cannot be.
3602 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3604 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3605 ext4_msg(sb, KERN_ERR,
3606 "Couldn't mount because of "
3607 "unsupported optional features (%x)",
3608 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3609 ~EXT4_FEATURE_INCOMPAT_SUPP));
3613 #if !IS_ENABLED(CONFIG_UNICODE)
3614 if (ext4_has_feature_casefold(sb)) {
3615 ext4_msg(sb, KERN_ERR,
3616 "Filesystem with casefold feature cannot be "
3617 "mounted without CONFIG_UNICODE");
3625 if (ext4_has_feature_readonly(sb)) {
3626 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3627 sb->s_flags |= SB_RDONLY;
3631 /* Check that feature set is OK for a read-write mount */
3632 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3633 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3634 "unsupported optional features (%x)",
3635 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3636 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3639 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3640 ext4_msg(sb, KERN_ERR,
3641 "Can't support bigalloc feature without "
3642 "extents feature\n");
3646 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3647 if (!readonly && (ext4_has_feature_quota(sb) ||
3648 ext4_has_feature_project(sb))) {
3649 ext4_msg(sb, KERN_ERR,
3650 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3653 #endif /* CONFIG_QUOTA */
3658 * This function is called once a day if we have errors logged
3659 * on the file system
3661 static void print_daily_error_info(struct timer_list *t)
3663 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3664 struct super_block *sb = sbi->s_sb;
3665 struct ext4_super_block *es = sbi->s_es;
3667 if (es->s_error_count)
3668 /* fsck newer than v1.41.13 is needed to clean this condition. */
3669 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3670 le32_to_cpu(es->s_error_count));
3671 if (es->s_first_error_time) {
3672 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3674 ext4_get_tstamp(es, s_first_error_time),
3675 (int) sizeof(es->s_first_error_func),
3676 es->s_first_error_func,
3677 le32_to_cpu(es->s_first_error_line));
3678 if (es->s_first_error_ino)
3679 printk(KERN_CONT ": inode %u",
3680 le32_to_cpu(es->s_first_error_ino));
3681 if (es->s_first_error_block)
3682 printk(KERN_CONT ": block %llu", (unsigned long long)
3683 le64_to_cpu(es->s_first_error_block));
3684 printk(KERN_CONT "\n");
3686 if (es->s_last_error_time) {
3687 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3689 ext4_get_tstamp(es, s_last_error_time),
3690 (int) sizeof(es->s_last_error_func),
3691 es->s_last_error_func,
3692 le32_to_cpu(es->s_last_error_line));
3693 if (es->s_last_error_ino)
3694 printk(KERN_CONT ": inode %u",
3695 le32_to_cpu(es->s_last_error_ino));
3696 if (es->s_last_error_block)
3697 printk(KERN_CONT ": block %llu", (unsigned long long)
3698 le64_to_cpu(es->s_last_error_block));
3699 printk(KERN_CONT "\n");
3701 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3704 /* Find next suitable group and run ext4_init_inode_table */
3705 static int ext4_run_li_request(struct ext4_li_request *elr)
3707 struct ext4_group_desc *gdp = NULL;
3708 struct super_block *sb = elr->lr_super;
3709 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3710 ext4_group_t group = elr->lr_next_group;
3711 unsigned int prefetch_ios = 0;
3713 int nr = EXT4_SB(sb)->s_mb_prefetch;
3716 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3717 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3718 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3719 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3720 if (group >= elr->lr_next_group) {
3722 if (elr->lr_first_not_zeroed != ngroups &&
3723 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3724 elr->lr_next_group = elr->lr_first_not_zeroed;
3725 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3732 for (; group < ngroups; group++) {
3733 gdp = ext4_get_group_desc(sb, group, NULL);
3739 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3743 if (group >= ngroups)
3747 start_time = ktime_get_real_ns();
3748 ret = ext4_init_inode_table(sb, group,
3749 elr->lr_timeout ? 0 : 1);
3750 trace_ext4_lazy_itable_init(sb, group);
3751 if (elr->lr_timeout == 0) {
3752 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3753 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3755 elr->lr_next_sched = jiffies + elr->lr_timeout;
3756 elr->lr_next_group = group + 1;
3762 * Remove lr_request from the list_request and free the
3763 * request structure. Should be called with li_list_mtx held
3765 static void ext4_remove_li_request(struct ext4_li_request *elr)
3770 list_del(&elr->lr_request);
3771 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3775 static void ext4_unregister_li_request(struct super_block *sb)
3777 mutex_lock(&ext4_li_mtx);
3778 if (!ext4_li_info) {
3779 mutex_unlock(&ext4_li_mtx);
3783 mutex_lock(&ext4_li_info->li_list_mtx);
3784 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3785 mutex_unlock(&ext4_li_info->li_list_mtx);
3786 mutex_unlock(&ext4_li_mtx);
3789 static struct task_struct *ext4_lazyinit_task;
3792 * This is the function where ext4lazyinit thread lives. It walks
3793 * through the request list searching for next scheduled filesystem.
3794 * When such a fs is found, run the lazy initialization request
3795 * (ext4_rn_li_request) and keep track of the time spend in this
3796 * function. Based on that time we compute next schedule time of
3797 * the request. When walking through the list is complete, compute
3798 * next waking time and put itself into sleep.
3800 static int ext4_lazyinit_thread(void *arg)
3802 struct ext4_lazy_init *eli = arg;
3803 struct list_head *pos, *n;
3804 struct ext4_li_request *elr;
3805 unsigned long next_wakeup, cur;
3807 BUG_ON(NULL == eli);
3812 next_wakeup = MAX_JIFFY_OFFSET;
3814 mutex_lock(&eli->li_list_mtx);
3815 if (list_empty(&eli->li_request_list)) {
3816 mutex_unlock(&eli->li_list_mtx);
3819 list_for_each_safe(pos, n, &eli->li_request_list) {
3822 elr = list_entry(pos, struct ext4_li_request,
3825 if (time_before(jiffies, elr->lr_next_sched)) {
3826 if (time_before(elr->lr_next_sched, next_wakeup))
3827 next_wakeup = elr->lr_next_sched;
3830 if (down_read_trylock(&elr->lr_super->s_umount)) {
3831 if (sb_start_write_trylock(elr->lr_super)) {
3834 * We hold sb->s_umount, sb can not
3835 * be removed from the list, it is
3836 * now safe to drop li_list_mtx
3838 mutex_unlock(&eli->li_list_mtx);
3839 err = ext4_run_li_request(elr);
3840 sb_end_write(elr->lr_super);
3841 mutex_lock(&eli->li_list_mtx);
3844 up_read((&elr->lr_super->s_umount));
3846 /* error, remove the lazy_init job */
3848 ext4_remove_li_request(elr);
3852 elr->lr_next_sched = jiffies +
3853 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3855 if (time_before(elr->lr_next_sched, next_wakeup))
3856 next_wakeup = elr->lr_next_sched;
3858 mutex_unlock(&eli->li_list_mtx);
3863 if ((time_after_eq(cur, next_wakeup)) ||
3864 (MAX_JIFFY_OFFSET == next_wakeup)) {
3869 schedule_timeout_interruptible(next_wakeup - cur);
3871 if (kthread_should_stop()) {
3872 ext4_clear_request_list();
3879 * It looks like the request list is empty, but we need
3880 * to check it under the li_list_mtx lock, to prevent any
3881 * additions into it, and of course we should lock ext4_li_mtx
3882 * to atomically free the list and ext4_li_info, because at
3883 * this point another ext4 filesystem could be registering
3886 mutex_lock(&ext4_li_mtx);
3887 mutex_lock(&eli->li_list_mtx);
3888 if (!list_empty(&eli->li_request_list)) {
3889 mutex_unlock(&eli->li_list_mtx);
3890 mutex_unlock(&ext4_li_mtx);
3893 mutex_unlock(&eli->li_list_mtx);
3894 kfree(ext4_li_info);
3895 ext4_li_info = NULL;
3896 mutex_unlock(&ext4_li_mtx);
3901 static void ext4_clear_request_list(void)
3903 struct list_head *pos, *n;
3904 struct ext4_li_request *elr;
3906 mutex_lock(&ext4_li_info->li_list_mtx);
3907 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3908 elr = list_entry(pos, struct ext4_li_request,
3910 ext4_remove_li_request(elr);
3912 mutex_unlock(&ext4_li_info->li_list_mtx);
3915 static int ext4_run_lazyinit_thread(void)
3917 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3918 ext4_li_info, "ext4lazyinit");
3919 if (IS_ERR(ext4_lazyinit_task)) {
3920 int err = PTR_ERR(ext4_lazyinit_task);
3921 ext4_clear_request_list();
3922 kfree(ext4_li_info);
3923 ext4_li_info = NULL;
3924 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3925 "initialization thread\n",
3929 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3934 * Check whether it make sense to run itable init. thread or not.
3935 * If there is at least one uninitialized inode table, return
3936 * corresponding group number, else the loop goes through all
3937 * groups and return total number of groups.
3939 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3941 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3942 struct ext4_group_desc *gdp = NULL;
3944 if (!ext4_has_group_desc_csum(sb))
3947 for (group = 0; group < ngroups; group++) {
3948 gdp = ext4_get_group_desc(sb, group, NULL);
3952 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3959 static int ext4_li_info_new(void)
3961 struct ext4_lazy_init *eli = NULL;
3963 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3967 INIT_LIST_HEAD(&eli->li_request_list);
3968 mutex_init(&eli->li_list_mtx);
3970 eli->li_state |= EXT4_LAZYINIT_QUIT;
3977 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3980 struct ext4_li_request *elr;
3982 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3987 elr->lr_first_not_zeroed = start;
3988 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3989 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3990 elr->lr_next_group = start;
3992 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3996 * Randomize first schedule time of the request to
3997 * spread the inode table initialization requests
4000 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
4004 int ext4_register_li_request(struct super_block *sb,
4005 ext4_group_t first_not_zeroed)
4007 struct ext4_sb_info *sbi = EXT4_SB(sb);
4008 struct ext4_li_request *elr = NULL;
4009 ext4_group_t ngroups = sbi->s_groups_count;
4012 mutex_lock(&ext4_li_mtx);
4013 if (sbi->s_li_request != NULL) {
4015 * Reset timeout so it can be computed again, because
4016 * s_li_wait_mult might have changed.
4018 sbi->s_li_request->lr_timeout = 0;
4022 if (sb_rdonly(sb) ||
4023 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4024 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4027 elr = ext4_li_request_new(sb, first_not_zeroed);
4033 if (NULL == ext4_li_info) {
4034 ret = ext4_li_info_new();
4039 mutex_lock(&ext4_li_info->li_list_mtx);
4040 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4041 mutex_unlock(&ext4_li_info->li_list_mtx);
4043 sbi->s_li_request = elr;
4045 * set elr to NULL here since it has been inserted to
4046 * the request_list and the removal and free of it is
4047 * handled by ext4_clear_request_list from now on.
4051 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4052 ret = ext4_run_lazyinit_thread();
4057 mutex_unlock(&ext4_li_mtx);
4064 * We do not need to lock anything since this is called on
4067 static void ext4_destroy_lazyinit_thread(void)
4070 * If thread exited earlier
4071 * there's nothing to be done.
4073 if (!ext4_li_info || !ext4_lazyinit_task)
4076 kthread_stop(ext4_lazyinit_task);
4079 static int set_journal_csum_feature_set(struct super_block *sb)
4082 int compat, incompat;
4083 struct ext4_sb_info *sbi = EXT4_SB(sb);
4085 if (ext4_has_metadata_csum(sb)) {
4086 /* journal checksum v3 */
4088 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4090 /* journal checksum v1 */
4091 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4095 jbd2_journal_clear_features(sbi->s_journal,
4096 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4097 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4098 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4099 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4100 ret = jbd2_journal_set_features(sbi->s_journal,
4102 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4104 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4105 ret = jbd2_journal_set_features(sbi->s_journal,
4108 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4109 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4111 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4112 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4119 * Note: calculating the overhead so we can be compatible with
4120 * historical BSD practice is quite difficult in the face of
4121 * clusters/bigalloc. This is because multiple metadata blocks from
4122 * different block group can end up in the same allocation cluster.
4123 * Calculating the exact overhead in the face of clustered allocation
4124 * requires either O(all block bitmaps) in memory or O(number of block
4125 * groups**2) in time. We will still calculate the superblock for
4126 * older file systems --- and if we come across with a bigalloc file
4127 * system with zero in s_overhead_clusters the estimate will be close to
4128 * correct especially for very large cluster sizes --- but for newer
4129 * file systems, it's better to calculate this figure once at mkfs
4130 * time, and store it in the superblock. If the superblock value is
4131 * present (even for non-bigalloc file systems), we will use it.
4133 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4136 struct ext4_sb_info *sbi = EXT4_SB(sb);
4137 struct ext4_group_desc *gdp;
4138 ext4_fsblk_t first_block, last_block, b;
4139 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4140 int s, j, count = 0;
4141 int has_super = ext4_bg_has_super(sb, grp);
4143 if (!ext4_has_feature_bigalloc(sb))
4144 return (has_super + ext4_bg_num_gdb(sb, grp) +
4145 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4146 sbi->s_itb_per_group + 2);
4148 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4149 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4150 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4151 for (i = 0; i < ngroups; i++) {
4152 gdp = ext4_get_group_desc(sb, i, NULL);
4153 b = ext4_block_bitmap(sb, gdp);
4154 if (b >= first_block && b <= last_block) {
4155 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4158 b = ext4_inode_bitmap(sb, gdp);
4159 if (b >= first_block && b <= last_block) {
4160 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4163 b = ext4_inode_table(sb, gdp);
4164 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4165 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4166 int c = EXT4_B2C(sbi, b - first_block);
4167 ext4_set_bit(c, buf);
4173 if (ext4_bg_has_super(sb, grp)) {
4174 ext4_set_bit(s++, buf);
4177 j = ext4_bg_num_gdb(sb, grp);
4178 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4179 ext4_error(sb, "Invalid number of block group "
4180 "descriptor blocks: %d", j);
4181 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4185 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4189 return EXT4_CLUSTERS_PER_GROUP(sb) -
4190 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4194 * Compute the overhead and stash it in sbi->s_overhead
4196 int ext4_calculate_overhead(struct super_block *sb)
4198 struct ext4_sb_info *sbi = EXT4_SB(sb);
4199 struct ext4_super_block *es = sbi->s_es;
4200 struct inode *j_inode;
4201 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4202 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4203 ext4_fsblk_t overhead = 0;
4204 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4210 * Compute the overhead (FS structures). This is constant
4211 * for a given filesystem unless the number of block groups
4212 * changes so we cache the previous value until it does.
4216 * All of the blocks before first_data_block are overhead
4218 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4221 * Add the overhead found in each block group
4223 for (i = 0; i < ngroups; i++) {
4226 blks = count_overhead(sb, i, buf);
4229 memset(buf, 0, PAGE_SIZE);
4234 * Add the internal journal blocks whether the journal has been
4237 if (sbi->s_journal && !sbi->s_journal_bdev)
4238 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4239 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4240 /* j_inum for internal journal is non-zero */
4241 j_inode = ext4_get_journal_inode(sb, j_inum);
4242 if (!IS_ERR(j_inode)) {
4243 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4244 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4247 ext4_msg(sb, KERN_ERR, "can't get journal size");
4250 sbi->s_overhead = overhead;
4252 free_page((unsigned long) buf);
4256 static void ext4_set_resv_clusters(struct super_block *sb)
4258 ext4_fsblk_t resv_clusters;
4259 struct ext4_sb_info *sbi = EXT4_SB(sb);
4262 * There's no need to reserve anything when we aren't using extents.
4263 * The space estimates are exact, there are no unwritten extents,
4264 * hole punching doesn't need new metadata... This is needed especially
4265 * to keep ext2/3 backward compatibility.
4267 if (!ext4_has_feature_extents(sb))
4270 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4271 * This should cover the situations where we can not afford to run
4272 * out of space like for example punch hole, or converting
4273 * unwritten extents in delalloc path. In most cases such
4274 * allocation would require 1, or 2 blocks, higher numbers are
4277 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4278 sbi->s_cluster_bits);
4280 do_div(resv_clusters, 50);
4281 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4283 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4286 static const char *ext4_quota_mode(struct super_block *sb)
4289 if (!ext4_quota_capable(sb))
4292 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4293 return "journalled";
4301 static void ext4_setup_csum_trigger(struct super_block *sb,
4302 enum ext4_journal_trigger_type type,
4304 struct jbd2_buffer_trigger_type *type,
4305 struct buffer_head *bh,
4309 struct ext4_sb_info *sbi = EXT4_SB(sb);
4311 sbi->s_journal_triggers[type].sb = sb;
4312 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4315 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4320 kfree(sbi->s_blockgroup_lock);
4321 fs_put_dax(sbi->s_daxdev, NULL);
4325 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4327 struct ext4_sb_info *sbi;
4329 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4333 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4336 sbi->s_blockgroup_lock =
4337 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4339 if (!sbi->s_blockgroup_lock)
4342 sb->s_fs_info = sbi;
4346 fs_put_dax(sbi->s_daxdev, NULL);
4351 static void ext4_set_def_opts(struct super_block *sb,
4352 struct ext4_super_block *es)
4354 unsigned long def_mount_opts;
4356 /* Set defaults before we parse the mount options */
4357 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4358 set_opt(sb, INIT_INODE_TABLE);
4359 if (def_mount_opts & EXT4_DEFM_DEBUG)
4361 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4363 if (def_mount_opts & EXT4_DEFM_UID16)
4364 set_opt(sb, NO_UID32);
4365 /* xattr user namespace & acls are now defaulted on */
4366 set_opt(sb, XATTR_USER);
4367 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4368 set_opt(sb, POSIX_ACL);
4370 if (ext4_has_feature_fast_commit(sb))
4371 set_opt2(sb, JOURNAL_FAST_COMMIT);
4372 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4373 if (ext4_has_metadata_csum(sb))
4374 set_opt(sb, JOURNAL_CHECKSUM);
4376 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4377 set_opt(sb, JOURNAL_DATA);
4378 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4379 set_opt(sb, ORDERED_DATA);
4380 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4381 set_opt(sb, WRITEBACK_DATA);
4383 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4384 set_opt(sb, ERRORS_PANIC);
4385 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4386 set_opt(sb, ERRORS_CONT);
4388 set_opt(sb, ERRORS_RO);
4389 /* block_validity enabled by default; disable with noblock_validity */
4390 set_opt(sb, BLOCK_VALIDITY);
4391 if (def_mount_opts & EXT4_DEFM_DISCARD)
4392 set_opt(sb, DISCARD);
4394 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4395 set_opt(sb, BARRIER);
4398 * enable delayed allocation by default
4399 * Use -o nodelalloc to turn it off
4401 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4402 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4403 set_opt(sb, DELALLOC);
4405 if (sb->s_blocksize == PAGE_SIZE)
4406 set_opt(sb, DIOREAD_NOLOCK);
4409 static int ext4_handle_clustersize(struct super_block *sb)
4411 struct ext4_sb_info *sbi = EXT4_SB(sb);
4412 struct ext4_super_block *es = sbi->s_es;
4415 /* Handle clustersize */
4416 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4417 if (ext4_has_feature_bigalloc(sb)) {
4418 if (clustersize < sb->s_blocksize) {
4419 ext4_msg(sb, KERN_ERR,
4420 "cluster size (%d) smaller than "
4421 "block size (%lu)", clustersize, sb->s_blocksize);
4424 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4425 le32_to_cpu(es->s_log_block_size);
4426 sbi->s_clusters_per_group =
4427 le32_to_cpu(es->s_clusters_per_group);
4428 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4429 ext4_msg(sb, KERN_ERR,
4430 "#clusters per group too big: %lu",
4431 sbi->s_clusters_per_group);
4434 if (sbi->s_blocks_per_group !=
4435 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4436 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4437 "clusters per group (%lu) inconsistent",
4438 sbi->s_blocks_per_group,
4439 sbi->s_clusters_per_group);
4443 if (clustersize != sb->s_blocksize) {
4444 ext4_msg(sb, KERN_ERR,
4445 "fragment/cluster size (%d) != "
4446 "block size (%lu)", clustersize, sb->s_blocksize);
4449 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4450 ext4_msg(sb, KERN_ERR,
4451 "#blocks per group too big: %lu",
4452 sbi->s_blocks_per_group);
4455 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4456 sbi->s_cluster_bits = 0;
4458 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4460 /* Do we have standard group size of clustersize * 8 blocks ? */
4461 if (sbi->s_blocks_per_group == clustersize << 3)
4462 set_opt2(sb, STD_GROUP_SIZE);
4467 static void ext4_fast_commit_init(struct super_block *sb)
4469 struct ext4_sb_info *sbi = EXT4_SB(sb);
4471 /* Initialize fast commit stuff */
4472 atomic_set(&sbi->s_fc_subtid, 0);
4473 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4474 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4475 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4476 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4477 sbi->s_fc_bytes = 0;
4478 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4479 sbi->s_fc_ineligible_tid = 0;
4480 spin_lock_init(&sbi->s_fc_lock);
4481 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4482 sbi->s_fc_replay_state.fc_regions = NULL;
4483 sbi->s_fc_replay_state.fc_regions_size = 0;
4484 sbi->s_fc_replay_state.fc_regions_used = 0;
4485 sbi->s_fc_replay_state.fc_regions_valid = 0;
4486 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4487 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4488 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4491 static int ext4_inode_info_init(struct super_block *sb,
4492 struct ext4_super_block *es)
4494 struct ext4_sb_info *sbi = EXT4_SB(sb);
4496 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4497 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4498 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4500 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4501 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4502 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4503 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4507 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4508 (!is_power_of_2(sbi->s_inode_size)) ||
4509 (sbi->s_inode_size > sb->s_blocksize)) {
4510 ext4_msg(sb, KERN_ERR,
4511 "unsupported inode size: %d",
4513 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4517 * i_atime_extra is the last extra field available for
4518 * [acm]times in struct ext4_inode. Checking for that
4519 * field should suffice to ensure we have extra space
4522 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4523 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4524 sb->s_time_gran = 1;
4525 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4527 sb->s_time_gran = NSEC_PER_SEC;
4528 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4530 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4533 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4534 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4535 EXT4_GOOD_OLD_INODE_SIZE;
4536 if (ext4_has_feature_extra_isize(sb)) {
4537 unsigned v, max = (sbi->s_inode_size -
4538 EXT4_GOOD_OLD_INODE_SIZE);
4540 v = le16_to_cpu(es->s_want_extra_isize);
4542 ext4_msg(sb, KERN_ERR,
4543 "bad s_want_extra_isize: %d", v);
4546 if (sbi->s_want_extra_isize < v)
4547 sbi->s_want_extra_isize = v;
4549 v = le16_to_cpu(es->s_min_extra_isize);
4551 ext4_msg(sb, KERN_ERR,
4552 "bad s_min_extra_isize: %d", v);
4555 if (sbi->s_want_extra_isize < v)
4556 sbi->s_want_extra_isize = v;
4563 #if IS_ENABLED(CONFIG_UNICODE)
4564 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4566 const struct ext4_sb_encodings *encoding_info;
4567 struct unicode_map *encoding;
4568 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4570 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4573 encoding_info = ext4_sb_read_encoding(es);
4574 if (!encoding_info) {
4575 ext4_msg(sb, KERN_ERR,
4576 "Encoding requested by superblock is unknown");
4580 encoding = utf8_load(encoding_info->version);
4581 if (IS_ERR(encoding)) {
4582 ext4_msg(sb, KERN_ERR,
4583 "can't mount with superblock charset: %s-%u.%u.%u "
4584 "not supported by the kernel. flags: 0x%x.",
4585 encoding_info->name,
4586 unicode_major(encoding_info->version),
4587 unicode_minor(encoding_info->version),
4588 unicode_rev(encoding_info->version),
4592 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4593 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4594 unicode_major(encoding_info->version),
4595 unicode_minor(encoding_info->version),
4596 unicode_rev(encoding_info->version),
4599 sb->s_encoding = encoding;
4600 sb->s_encoding_flags = encoding_flags;
4605 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4611 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4613 struct ext4_sb_info *sbi = EXT4_SB(sb);
4615 /* Warn if metadata_csum and gdt_csum are both set. */
4616 if (ext4_has_feature_metadata_csum(sb) &&
4617 ext4_has_feature_gdt_csum(sb))
4618 ext4_warning(sb, "metadata_csum and uninit_bg are "
4619 "redundant flags; please run fsck.");
4621 /* Check for a known checksum algorithm */
4622 if (!ext4_verify_csum_type(sb, es)) {
4623 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4624 "unknown checksum algorithm.");
4627 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4628 ext4_orphan_file_block_trigger);
4630 /* Load the checksum driver */
4631 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4632 if (IS_ERR(sbi->s_chksum_driver)) {
4633 int ret = PTR_ERR(sbi->s_chksum_driver);
4634 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4635 sbi->s_chksum_driver = NULL;
4639 /* Check superblock checksum */
4640 if (!ext4_superblock_csum_verify(sb, es)) {
4641 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4642 "invalid superblock checksum. Run e2fsck?");
4646 /* Precompute checksum seed for all metadata */
4647 if (ext4_has_feature_csum_seed(sb))
4648 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4649 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4650 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4651 sizeof(es->s_uuid));
4655 static int ext4_check_feature_compatibility(struct super_block *sb,
4656 struct ext4_super_block *es,
4659 struct ext4_sb_info *sbi = EXT4_SB(sb);
4661 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4662 (ext4_has_compat_features(sb) ||
4663 ext4_has_ro_compat_features(sb) ||
4664 ext4_has_incompat_features(sb)))
4665 ext4_msg(sb, KERN_WARNING,
4666 "feature flags set on rev 0 fs, "
4667 "running e2fsck is recommended");
4669 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4670 set_opt2(sb, HURD_COMPAT);
4671 if (ext4_has_feature_64bit(sb)) {
4672 ext4_msg(sb, KERN_ERR,
4673 "The Hurd can't support 64-bit file systems");
4678 * ea_inode feature uses l_i_version field which is not
4679 * available in HURD_COMPAT mode.
4681 if (ext4_has_feature_ea_inode(sb)) {
4682 ext4_msg(sb, KERN_ERR,
4683 "ea_inode feature is not supported for Hurd");
4688 if (IS_EXT2_SB(sb)) {
4689 if (ext2_feature_set_ok(sb))
4690 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4691 "using the ext4 subsystem");
4694 * If we're probing be silent, if this looks like
4695 * it's actually an ext[34] filesystem.
4697 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4699 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4700 "to feature incompatibilities");
4705 if (IS_EXT3_SB(sb)) {
4706 if (ext3_feature_set_ok(sb))
4707 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4708 "using the ext4 subsystem");
4711 * If we're probing be silent, if this looks like
4712 * it's actually an ext4 filesystem.
4714 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4716 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4717 "to feature incompatibilities");
4723 * Check feature flags regardless of the revision level, since we
4724 * previously didn't change the revision level when setting the flags,
4725 * so there is a chance incompat flags are set on a rev 0 filesystem.
4727 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4730 if (sbi->s_daxdev) {
4731 if (sb->s_blocksize == PAGE_SIZE)
4732 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4734 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4737 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4738 if (ext4_has_feature_inline_data(sb)) {
4739 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4740 " that may contain inline data");
4743 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4744 ext4_msg(sb, KERN_ERR,
4745 "DAX unsupported by block device.");
4750 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4751 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4752 es->s_encryption_level);
4759 static int ext4_check_geometry(struct super_block *sb,
4760 struct ext4_super_block *es)
4762 struct ext4_sb_info *sbi = EXT4_SB(sb);
4766 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4767 ext4_msg(sb, KERN_ERR,
4768 "Number of reserved GDT blocks insanely large: %d",
4769 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4773 * Test whether we have more sectors than will fit in sector_t,
4774 * and whether the max offset is addressable by the page cache.
4776 err = generic_check_addressable(sb->s_blocksize_bits,
4777 ext4_blocks_count(es));
4779 ext4_msg(sb, KERN_ERR, "filesystem"
4780 " too large to mount safely on this system");
4784 /* check blocks count against device size */
4785 blocks_count = sb_bdev_nr_blocks(sb);
4786 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4787 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4788 "exceeds size of device (%llu blocks)",
4789 ext4_blocks_count(es), blocks_count);
4794 * It makes no sense for the first data block to be beyond the end
4795 * of the filesystem.
4797 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4798 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4799 "block %u is beyond end of filesystem (%llu)",
4800 le32_to_cpu(es->s_first_data_block),
4801 ext4_blocks_count(es));
4804 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4805 (sbi->s_cluster_ratio == 1)) {
4806 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4807 "block is 0 with a 1k block and cluster size");
4811 blocks_count = (ext4_blocks_count(es) -
4812 le32_to_cpu(es->s_first_data_block) +
4813 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4814 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4815 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4816 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4817 "(block count %llu, first data block %u, "
4818 "blocks per group %lu)", blocks_count,
4819 ext4_blocks_count(es),
4820 le32_to_cpu(es->s_first_data_block),
4821 EXT4_BLOCKS_PER_GROUP(sb));
4824 sbi->s_groups_count = blocks_count;
4825 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4826 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4827 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4828 le32_to_cpu(es->s_inodes_count)) {
4829 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4830 le32_to_cpu(es->s_inodes_count),
4831 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4838 static int ext4_group_desc_init(struct super_block *sb,
4839 struct ext4_super_block *es,
4840 ext4_fsblk_t logical_sb_block,
4841 ext4_group_t *first_not_zeroed)
4843 struct ext4_sb_info *sbi = EXT4_SB(sb);
4844 unsigned int db_count;
4848 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4849 EXT4_DESC_PER_BLOCK(sb);
4850 if (ext4_has_feature_meta_bg(sb)) {
4851 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4852 ext4_msg(sb, KERN_WARNING,
4853 "first meta block group too large: %u "
4854 "(group descriptor block count %u)",
4855 le32_to_cpu(es->s_first_meta_bg), db_count);
4859 rcu_assign_pointer(sbi->s_group_desc,
4860 kvmalloc_array(db_count,
4861 sizeof(struct buffer_head *),
4863 if (sbi->s_group_desc == NULL) {
4864 ext4_msg(sb, KERN_ERR, "not enough memory");
4868 bgl_lock_init(sbi->s_blockgroup_lock);
4870 /* Pre-read the descriptors into the buffer cache */
4871 for (i = 0; i < db_count; i++) {
4872 block = descriptor_loc(sb, logical_sb_block, i);
4873 ext4_sb_breadahead_unmovable(sb, block);
4876 for (i = 0; i < db_count; i++) {
4877 struct buffer_head *bh;
4879 block = descriptor_loc(sb, logical_sb_block, i);
4880 bh = ext4_sb_bread_unmovable(sb, block);
4882 ext4_msg(sb, KERN_ERR,
4883 "can't read group descriptor %d", i);
4884 sbi->s_gdb_count = i;
4888 rcu_dereference(sbi->s_group_desc)[i] = bh;
4891 sbi->s_gdb_count = db_count;
4892 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4893 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4894 return -EFSCORRUPTED;
4900 static int ext4_load_and_init_journal(struct super_block *sb,
4901 struct ext4_super_block *es,
4902 struct ext4_fs_context *ctx)
4904 struct ext4_sb_info *sbi = EXT4_SB(sb);
4907 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4911 if (ext4_has_feature_64bit(sb) &&
4912 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4913 JBD2_FEATURE_INCOMPAT_64BIT)) {
4914 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4918 if (!set_journal_csum_feature_set(sb)) {
4919 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4924 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4925 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4926 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4927 ext4_msg(sb, KERN_ERR,
4928 "Failed to set fast commit journal feature");
4932 /* We have now updated the journal if required, so we can
4933 * validate the data journaling mode. */
4934 switch (test_opt(sb, DATA_FLAGS)) {
4936 /* No mode set, assume a default based on the journal
4937 * capabilities: ORDERED_DATA if the journal can
4938 * cope, else JOURNAL_DATA
4940 if (jbd2_journal_check_available_features
4941 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4942 set_opt(sb, ORDERED_DATA);
4943 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4945 set_opt(sb, JOURNAL_DATA);
4946 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4950 case EXT4_MOUNT_ORDERED_DATA:
4951 case EXT4_MOUNT_WRITEBACK_DATA:
4952 if (!jbd2_journal_check_available_features
4953 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4954 ext4_msg(sb, KERN_ERR, "Journal does not support "
4955 "requested data journaling mode");
4963 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4964 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4965 ext4_msg(sb, KERN_ERR, "can't mount with "
4966 "journal_async_commit in data=ordered mode");
4970 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4972 sbi->s_journal->j_submit_inode_data_buffers =
4973 ext4_journal_submit_inode_data_buffers;
4974 sbi->s_journal->j_finish_inode_data_buffers =
4975 ext4_journal_finish_inode_data_buffers;
4980 /* flush s_sb_upd_work before destroying the journal. */
4981 flush_work(&sbi->s_sb_upd_work);
4982 jbd2_journal_destroy(sbi->s_journal);
4983 sbi->s_journal = NULL;
4987 static int ext4_check_journal_data_mode(struct super_block *sb)
4989 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4990 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4991 "data=journal disables delayed allocation, "
4992 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4993 /* can't mount with both data=journal and dioread_nolock. */
4994 clear_opt(sb, DIOREAD_NOLOCK);
4995 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4996 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4997 ext4_msg(sb, KERN_ERR, "can't mount with "
4998 "both data=journal and delalloc");
5001 if (test_opt(sb, DAX_ALWAYS)) {
5002 ext4_msg(sb, KERN_ERR, "can't mount with "
5003 "both data=journal and dax");
5006 if (ext4_has_feature_encrypt(sb)) {
5007 ext4_msg(sb, KERN_WARNING,
5008 "encrypted files will use data=ordered "
5009 "instead of data journaling mode");
5011 if (test_opt(sb, DELALLOC))
5012 clear_opt(sb, DELALLOC);
5014 sb->s_iflags |= SB_I_CGROUPWB;
5020 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5023 struct ext4_sb_info *sbi = EXT4_SB(sb);
5024 struct ext4_super_block *es;
5025 ext4_fsblk_t logical_sb_block;
5026 unsigned long offset = 0;
5027 struct buffer_head *bh;
5031 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5033 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5038 * The ext4 superblock will not be buffer aligned for other than 1kB
5039 * block sizes. We need to calculate the offset from buffer start.
5041 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5042 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5043 offset = do_div(logical_sb_block, blocksize);
5045 logical_sb_block = sbi->s_sb_block;
5048 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5050 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5054 * Note: s_es must be initialized as soon as possible because
5055 * some ext4 macro-instructions depend on its value
5057 es = (struct ext4_super_block *) (bh->b_data + offset);
5059 sb->s_magic = le16_to_cpu(es->s_magic);
5060 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5062 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5066 if (le32_to_cpu(es->s_log_block_size) >
5067 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5068 ext4_msg(sb, KERN_ERR,
5069 "Invalid log block size: %u",
5070 le32_to_cpu(es->s_log_block_size));
5073 if (le32_to_cpu(es->s_log_cluster_size) >
5074 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5075 ext4_msg(sb, KERN_ERR,
5076 "Invalid log cluster size: %u",
5077 le32_to_cpu(es->s_log_cluster_size));
5081 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5084 * If the default block size is not the same as the real block size,
5085 * we need to reload it.
5087 if (sb->s_blocksize == blocksize) {
5088 *lsb = logical_sb_block;
5094 * bh must be released before kill_bdev(), otherwise
5095 * it won't be freed and its page also. kill_bdev()
5096 * is called by sb_set_blocksize().
5099 /* Validate the filesystem blocksize */
5100 if (!sb_set_blocksize(sb, blocksize)) {
5101 ext4_msg(sb, KERN_ERR, "bad block size %d",
5107 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5108 offset = do_div(logical_sb_block, blocksize);
5109 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5111 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5116 es = (struct ext4_super_block *)(bh->b_data + offset);
5118 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5119 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5122 *lsb = logical_sb_block;
5130 static void ext4_hash_info_init(struct super_block *sb)
5132 struct ext4_sb_info *sbi = EXT4_SB(sb);
5133 struct ext4_super_block *es = sbi->s_es;
5136 for (i = 0; i < 4; i++)
5137 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5139 sbi->s_def_hash_version = es->s_def_hash_version;
5140 if (ext4_has_feature_dir_index(sb)) {
5141 i = le32_to_cpu(es->s_flags);
5142 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5143 sbi->s_hash_unsigned = 3;
5144 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5145 #ifdef __CHAR_UNSIGNED__
5148 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5149 sbi->s_hash_unsigned = 3;
5153 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5159 static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5161 struct ext4_sb_info *sbi = EXT4_SB(sb);
5162 struct ext4_super_block *es = sbi->s_es;
5165 has_huge_files = ext4_has_feature_huge_file(sb);
5166 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5168 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5170 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5171 if (ext4_has_feature_64bit(sb)) {
5172 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5173 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5174 !is_power_of_2(sbi->s_desc_size)) {
5175 ext4_msg(sb, KERN_ERR,
5176 "unsupported descriptor size %lu",
5181 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5183 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5184 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5186 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5187 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5189 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5192 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5193 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5194 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5195 sbi->s_inodes_per_group);
5198 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5199 sbi->s_inodes_per_block;
5200 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5201 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5202 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5203 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5208 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5210 struct ext4_super_block *es = NULL;
5211 struct ext4_sb_info *sbi = EXT4_SB(sb);
5212 ext4_fsblk_t logical_sb_block;
5216 ext4_group_t first_not_zeroed;
5217 struct ext4_fs_context *ctx = fc->fs_private;
5218 int silent = fc->sb_flags & SB_SILENT;
5220 /* Set defaults for the variables that will be set during parsing */
5221 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5222 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5224 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5225 sbi->s_sectors_written_start =
5226 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5228 err = ext4_load_super(sb, &logical_sb_block, silent);
5233 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5235 err = ext4_init_metadata_csum(sb, es);
5239 ext4_set_def_opts(sb, es);
5241 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5242 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5243 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5244 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5245 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5248 * set default s_li_wait_mult for lazyinit, for the case there is
5249 * no mount option specified.
5251 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5253 err = ext4_inode_info_init(sb, es);
5257 err = parse_apply_sb_mount_options(sb, ctx);
5261 sbi->s_def_mount_opt = sbi->s_mount_opt;
5262 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5264 err = ext4_check_opt_consistency(fc, sb);
5268 ext4_apply_options(fc, sb);
5270 err = ext4_encoding_init(sb, es);
5274 err = ext4_check_journal_data_mode(sb);
5278 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5279 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5281 /* i_version is always enabled now */
5282 sb->s_flags |= SB_I_VERSION;
5284 err = ext4_check_feature_compatibility(sb, es, silent);
5288 err = ext4_block_group_meta_init(sb, silent);
5292 ext4_hash_info_init(sb);
5294 err = ext4_handle_clustersize(sb);
5298 err = ext4_check_geometry(sb, es);
5302 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5303 spin_lock_init(&sbi->s_error_lock);
5304 INIT_WORK(&sbi->s_sb_upd_work, update_super_work);
5306 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5310 err = ext4_es_register_shrinker(sbi);
5314 sbi->s_stripe = ext4_get_stripe_size(sbi);
5316 * It's hard to get stripe aligned blocks if stripe is not aligned with
5317 * cluster, just disable stripe and alert user to simpfy code and avoid
5318 * stripe aligned allocation which will rarely successes.
5320 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5321 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5322 ext4_msg(sb, KERN_WARNING,
5323 "stripe (%lu) is not aligned with cluster size (%u), "
5324 "stripe is disabled",
5325 sbi->s_stripe, sbi->s_cluster_ratio);
5328 sbi->s_extent_max_zeroout_kb = 32;
5331 * set up enough so that it can read an inode
5333 sb->s_op = &ext4_sops;
5334 sb->s_export_op = &ext4_export_ops;
5335 sb->s_xattr = ext4_xattr_handlers;
5336 #ifdef CONFIG_FS_ENCRYPTION
5337 sb->s_cop = &ext4_cryptops;
5339 #ifdef CONFIG_FS_VERITY
5340 sb->s_vop = &ext4_verityops;
5343 sb->dq_op = &ext4_quota_operations;
5344 if (ext4_has_feature_quota(sb))
5345 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5347 sb->s_qcop = &ext4_qctl_operations;
5348 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5350 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5352 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5353 mutex_init(&sbi->s_orphan_lock);
5355 ext4_fast_commit_init(sb);
5359 needs_recovery = (es->s_last_orphan != 0 ||
5360 ext4_has_feature_orphan_present(sb) ||
5361 ext4_has_feature_journal_needs_recovery(sb));
5363 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5364 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5366 goto failed_mount3a;
5371 * The first inode we look at is the journal inode. Don't try
5372 * root first: it may be modified in the journal!
5374 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5375 err = ext4_load_and_init_journal(sb, es, ctx);
5377 goto failed_mount3a;
5378 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5379 ext4_has_feature_journal_needs_recovery(sb)) {
5380 ext4_msg(sb, KERN_ERR, "required journal recovery "
5381 "suppressed and not mounted read-only");
5382 goto failed_mount3a;
5384 /* Nojournal mode, all journal mount options are illegal */
5385 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5386 ext4_msg(sb, KERN_ERR, "can't mount with "
5387 "journal_async_commit, fs mounted w/o journal");
5388 goto failed_mount3a;
5391 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5392 ext4_msg(sb, KERN_ERR, "can't mount with "
5393 "journal_checksum, fs mounted w/o journal");
5394 goto failed_mount3a;
5396 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5397 ext4_msg(sb, KERN_ERR, "can't mount with "
5398 "commit=%lu, fs mounted w/o journal",
5399 sbi->s_commit_interval / HZ);
5400 goto failed_mount3a;
5402 if (EXT4_MOUNT_DATA_FLAGS &
5403 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5404 ext4_msg(sb, KERN_ERR, "can't mount with "
5405 "data=, fs mounted w/o journal");
5406 goto failed_mount3a;
5408 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5409 clear_opt(sb, JOURNAL_CHECKSUM);
5410 clear_opt(sb, DATA_FLAGS);
5411 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5412 sbi->s_journal = NULL;
5416 if (!test_opt(sb, NO_MBCACHE)) {
5417 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5418 if (!sbi->s_ea_block_cache) {
5419 ext4_msg(sb, KERN_ERR,
5420 "Failed to create ea_block_cache");
5422 goto failed_mount_wq;
5425 if (ext4_has_feature_ea_inode(sb)) {
5426 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5427 if (!sbi->s_ea_inode_cache) {
5428 ext4_msg(sb, KERN_ERR,
5429 "Failed to create ea_inode_cache");
5431 goto failed_mount_wq;
5437 * Get the # of file system overhead blocks from the
5438 * superblock if present.
5440 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5441 /* ignore the precalculated value if it is ridiculous */
5442 if (sbi->s_overhead > ext4_blocks_count(es))
5443 sbi->s_overhead = 0;
5445 * If the bigalloc feature is not enabled recalculating the
5446 * overhead doesn't take long, so we might as well just redo
5447 * it to make sure we are using the correct value.
5449 if (!ext4_has_feature_bigalloc(sb))
5450 sbi->s_overhead = 0;
5451 if (sbi->s_overhead == 0) {
5452 err = ext4_calculate_overhead(sb);
5454 goto failed_mount_wq;
5458 * The maximum number of concurrent works can be high and
5459 * concurrency isn't really necessary. Limit it to 1.
5461 EXT4_SB(sb)->rsv_conversion_wq =
5462 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5463 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5464 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5470 * The jbd2_journal_load will have done any necessary log recovery,
5471 * so we can safely mount the rest of the filesystem now.
5474 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5476 ext4_msg(sb, KERN_ERR, "get root inode failed");
5477 err = PTR_ERR(root);
5481 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5482 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5484 err = -EFSCORRUPTED;
5488 sb->s_root = d_make_root(root);
5490 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5495 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5496 if (err == -EROFS) {
5497 sb->s_flags |= SB_RDONLY;
5499 goto failed_mount4a;
5501 ext4_set_resv_clusters(sb);
5503 if (test_opt(sb, BLOCK_VALIDITY)) {
5504 err = ext4_setup_system_zone(sb);
5506 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5508 goto failed_mount4a;
5511 ext4_fc_replay_cleanup(sb);
5516 * Enable optimize_scan if number of groups is > threshold. This can be
5517 * turned off by passing "mb_optimize_scan=0". This can also be
5518 * turned on forcefully by passing "mb_optimize_scan=1".
5520 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5521 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5522 set_opt2(sb, MB_OPTIMIZE_SCAN);
5524 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5527 err = ext4_mb_init(sb);
5529 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5535 * We can only set up the journal commit callback once
5536 * mballoc is initialized
5539 sbi->s_journal->j_commit_callback =
5540 ext4_journal_commit_callback;
5542 err = ext4_percpu_param_init(sbi);
5546 if (ext4_has_feature_flex_bg(sb))
5547 if (!ext4_fill_flex_info(sb)) {
5548 ext4_msg(sb, KERN_ERR,
5549 "unable to initialize "
5550 "flex_bg meta info!");
5555 err = ext4_register_li_request(sb, first_not_zeroed);
5559 err = ext4_register_sysfs(sb);
5563 err = ext4_init_orphan_info(sb);
5567 /* Enable quota usage during mount. */
5568 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5569 err = ext4_enable_quotas(sb);
5573 #endif /* CONFIG_QUOTA */
5576 * Save the original bdev mapping's wb_err value which could be
5577 * used to detect the metadata async write error.
5579 spin_lock_init(&sbi->s_bdev_wb_lock);
5580 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5581 &sbi->s_bdev_wb_err);
5582 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5583 ext4_orphan_cleanup(sb, es);
5584 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5586 * Update the checksum after updating free space/inode counters and
5587 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5588 * checksum in the buffer cache until it is written out and
5589 * e2fsprogs programs trying to open a file system immediately
5590 * after it is mounted can fail.
5592 ext4_superblock_csum_set(sb);
5593 if (needs_recovery) {
5594 ext4_msg(sb, KERN_INFO, "recovery complete");
5595 err = ext4_mark_recovery_complete(sb, es);
5597 goto failed_mount10;
5600 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5601 ext4_msg(sb, KERN_WARNING,
5602 "mounting with \"discard\" option, but the device does not support discard");
5604 if (es->s_error_count)
5605 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5607 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5608 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5609 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5610 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5611 atomic_set(&sbi->s_warning_count, 0);
5612 atomic_set(&sbi->s_msg_count, 0);
5617 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5618 failed_mount9: __maybe_unused
5619 ext4_release_orphan_info(sb);
5621 ext4_unregister_sysfs(sb);
5622 kobject_put(&sbi->s_kobj);
5624 ext4_unregister_li_request(sb);
5626 ext4_mb_release(sb);
5627 ext4_flex_groups_free(sbi);
5628 ext4_percpu_param_destroy(sbi);
5630 ext4_ext_release(sb);
5631 ext4_release_system_zone(sb);
5636 ext4_msg(sb, KERN_ERR, "mount failed");
5637 if (EXT4_SB(sb)->rsv_conversion_wq)
5638 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5640 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5641 sbi->s_ea_inode_cache = NULL;
5643 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5644 sbi->s_ea_block_cache = NULL;
5646 if (sbi->s_journal) {
5647 /* flush s_sb_upd_work before journal destroy. */
5648 flush_work(&sbi->s_sb_upd_work);
5649 jbd2_journal_destroy(sbi->s_journal);
5650 sbi->s_journal = NULL;
5653 ext4_es_unregister_shrinker(sbi);
5655 /* flush s_sb_upd_work before sbi destroy */
5656 flush_work(&sbi->s_sb_upd_work);
5657 del_timer_sync(&sbi->s_err_report);
5658 ext4_stop_mmpd(sbi);
5659 ext4_group_desc_free(sbi);
5661 if (sbi->s_chksum_driver)
5662 crypto_free_shash(sbi->s_chksum_driver);
5664 #if IS_ENABLED(CONFIG_UNICODE)
5665 utf8_unload(sb->s_encoding);
5669 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5670 kfree(get_qf_name(sb, sbi, i));
5672 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5674 if (sbi->s_journal_bdev) {
5675 invalidate_bdev(sbi->s_journal_bdev);
5676 blkdev_put(sbi->s_journal_bdev, sb);
5679 invalidate_bdev(sb->s_bdev);
5680 sb->s_fs_info = NULL;
5684 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5686 struct ext4_fs_context *ctx = fc->fs_private;
5687 struct ext4_sb_info *sbi;
5691 sbi = ext4_alloc_sbi(sb);
5695 fc->s_fs_info = sbi;
5697 /* Cleanup superblock name */
5698 strreplace(sb->s_id, '/', '!');
5700 sbi->s_sb_block = 1; /* Default super block location */
5701 if (ctx->spec & EXT4_SPEC_s_sb_block)
5702 sbi->s_sb_block = ctx->s_sb_block;
5704 ret = __ext4_fill_super(fc, sb);
5708 if (sbi->s_journal) {
5709 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5710 descr = " journalled data mode";
5711 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5712 descr = " ordered data mode";
5714 descr = " writeback data mode";
5716 descr = "out journal";
5718 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5719 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5720 "Quota mode: %s.", &sb->s_uuid,
5721 sb_rdonly(sb) ? "ro" : "r/w", descr,
5722 ext4_quota_mode(sb));
5724 /* Update the s_overhead_clusters if necessary */
5725 ext4_update_overhead(sb, false);
5730 fc->s_fs_info = NULL;
5734 static int ext4_get_tree(struct fs_context *fc)
5736 return get_tree_bdev(fc, ext4_fill_super);
5740 * Setup any per-fs journal parameters now. We'll do this both on
5741 * initial mount, once the journal has been initialised but before we've
5742 * done any recovery; and again on any subsequent remount.
5744 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5746 struct ext4_sb_info *sbi = EXT4_SB(sb);
5748 journal->j_commit_interval = sbi->s_commit_interval;
5749 journal->j_min_batch_time = sbi->s_min_batch_time;
5750 journal->j_max_batch_time = sbi->s_max_batch_time;
5751 ext4_fc_init(sb, journal);
5753 write_lock(&journal->j_state_lock);
5754 if (test_opt(sb, BARRIER))
5755 journal->j_flags |= JBD2_BARRIER;
5757 journal->j_flags &= ~JBD2_BARRIER;
5758 if (test_opt(sb, DATA_ERR_ABORT))
5759 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5761 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5763 * Always enable journal cycle record option, letting the journal
5764 * records log transactions continuously between each mount.
5766 journal->j_flags |= JBD2_CYCLE_RECORD;
5767 write_unlock(&journal->j_state_lock);
5770 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5771 unsigned int journal_inum)
5773 struct inode *journal_inode;
5776 * Test for the existence of a valid inode on disk. Bad things
5777 * happen if we iget() an unused inode, as the subsequent iput()
5778 * will try to delete it.
5780 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5781 if (IS_ERR(journal_inode)) {
5782 ext4_msg(sb, KERN_ERR, "no journal found");
5783 return ERR_CAST(journal_inode);
5785 if (!journal_inode->i_nlink) {
5786 make_bad_inode(journal_inode);
5787 iput(journal_inode);
5788 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5789 return ERR_PTR(-EFSCORRUPTED);
5791 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5792 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5793 iput(journal_inode);
5794 return ERR_PTR(-EFSCORRUPTED);
5797 ext4_debug("Journal inode found at %p: %lld bytes\n",
5798 journal_inode, journal_inode->i_size);
5799 return journal_inode;
5802 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5804 struct ext4_map_blocks map;
5807 if (journal->j_inode == NULL)
5810 map.m_lblk = *block;
5812 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5814 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5815 "journal bmap failed: block %llu ret %d\n",
5817 jbd2_journal_abort(journal, ret ? ret : -EIO);
5820 *block = map.m_pblk;
5824 static journal_t *ext4_open_inode_journal(struct super_block *sb,
5825 unsigned int journal_inum)
5827 struct inode *journal_inode;
5830 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5831 if (IS_ERR(journal_inode))
5832 return ERR_CAST(journal_inode);
5834 journal = jbd2_journal_init_inode(journal_inode);
5835 if (IS_ERR(journal)) {
5836 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5837 iput(journal_inode);
5838 return ERR_CAST(journal);
5840 journal->j_private = sb;
5841 journal->j_bmap = ext4_journal_bmap;
5842 ext4_init_journal_params(sb, journal);
5846 static struct block_device *ext4_get_journal_blkdev(struct super_block *sb,
5847 dev_t j_dev, ext4_fsblk_t *j_start,
5848 ext4_fsblk_t *j_len)
5850 struct buffer_head *bh;
5851 struct block_device *bdev;
5852 int hblock, blocksize;
5853 ext4_fsblk_t sb_block;
5854 unsigned long offset;
5855 struct ext4_super_block *es;
5858 /* see get_tree_bdev why this is needed and safe */
5859 up_write(&sb->s_umount);
5860 bdev = blkdev_get_by_dev(j_dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb,
5862 down_write(&sb->s_umount);
5864 ext4_msg(sb, KERN_ERR,
5865 "failed to open journal device unknown-block(%u,%u) %ld",
5866 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev));
5867 return ERR_CAST(bdev);
5870 blocksize = sb->s_blocksize;
5871 hblock = bdev_logical_block_size(bdev);
5872 if (blocksize < hblock) {
5873 ext4_msg(sb, KERN_ERR,
5874 "blocksize too small for journal device");
5879 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5880 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5881 set_blocksize(bdev, blocksize);
5882 bh = __bread(bdev, sb_block, blocksize);
5884 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5885 "external journal");
5890 es = (struct ext4_super_block *) (bh->b_data + offset);
5891 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5892 !(le32_to_cpu(es->s_feature_incompat) &
5893 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5894 ext4_msg(sb, KERN_ERR, "external journal has bad superblock");
5895 errno = -EFSCORRUPTED;
5899 if ((le32_to_cpu(es->s_feature_ro_compat) &
5900 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5901 es->s_checksum != ext4_superblock_csum(sb, es)) {
5902 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock");
5903 errno = -EFSCORRUPTED;
5907 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5908 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5909 errno = -EFSCORRUPTED;
5913 *j_start = sb_block + 1;
5914 *j_len = ext4_blocks_count(es);
5921 blkdev_put(bdev, sb);
5922 return ERR_PTR(errno);
5925 static journal_t *ext4_open_dev_journal(struct super_block *sb,
5929 ext4_fsblk_t j_start;
5931 struct block_device *journal_bdev;
5934 journal_bdev = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len);
5935 if (IS_ERR(journal_bdev))
5936 return ERR_CAST(journal_bdev);
5938 journal = jbd2_journal_init_dev(journal_bdev, sb->s_bdev, j_start,
5939 j_len, sb->s_blocksize);
5940 if (IS_ERR(journal)) {
5941 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5942 errno = PTR_ERR(journal);
5945 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5946 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5947 "user (unsupported) - %d",
5948 be32_to_cpu(journal->j_superblock->s_nr_users));
5952 journal->j_private = sb;
5953 EXT4_SB(sb)->s_journal_bdev = journal_bdev;
5954 ext4_init_journal_params(sb, journal);
5958 jbd2_journal_destroy(journal);
5960 blkdev_put(journal_bdev, sb);
5961 return ERR_PTR(errno);
5964 static int ext4_load_journal(struct super_block *sb,
5965 struct ext4_super_block *es,
5966 unsigned long journal_devnum)
5969 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5972 int really_read_only;
5975 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5976 return -EFSCORRUPTED;
5978 if (journal_devnum &&
5979 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5980 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5981 "numbers have changed");
5982 journal_dev = new_decode_dev(journal_devnum);
5984 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5986 if (journal_inum && journal_dev) {
5987 ext4_msg(sb, KERN_ERR,
5988 "filesystem has both journal inode and journal device!");
5993 journal = ext4_open_inode_journal(sb, journal_inum);
5994 if (IS_ERR(journal))
5995 return PTR_ERR(journal);
5997 journal = ext4_open_dev_journal(sb, journal_dev);
5998 if (IS_ERR(journal))
5999 return PTR_ERR(journal);
6002 journal_dev_ro = bdev_read_only(journal->j_dev);
6003 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
6005 if (journal_dev_ro && !sb_rdonly(sb)) {
6006 ext4_msg(sb, KERN_ERR,
6007 "journal device read-only, try mounting with '-o ro'");
6013 * Are we loading a blank journal or performing recovery after a
6014 * crash? For recovery, we need to check in advance whether we
6015 * can get read-write access to the device.
6017 if (ext4_has_feature_journal_needs_recovery(sb)) {
6018 if (sb_rdonly(sb)) {
6019 ext4_msg(sb, KERN_INFO, "INFO: recovery "
6020 "required on readonly filesystem");
6021 if (really_read_only) {
6022 ext4_msg(sb, KERN_ERR, "write access "
6023 "unavailable, cannot proceed "
6024 "(try mounting with noload)");
6028 ext4_msg(sb, KERN_INFO, "write access will "
6029 "be enabled during recovery");
6033 if (!(journal->j_flags & JBD2_BARRIER))
6034 ext4_msg(sb, KERN_INFO, "barriers disabled");
6036 if (!ext4_has_feature_journal_needs_recovery(sb))
6037 err = jbd2_journal_wipe(journal, !really_read_only);
6039 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6041 bool changed = false;
6044 memcpy(save, ((char *) es) +
6045 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6046 err = jbd2_journal_load(journal);
6047 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6048 save, EXT4_S_ERR_LEN)) {
6049 memcpy(((char *) es) + EXT4_S_ERR_START,
6050 save, EXT4_S_ERR_LEN);
6054 orig_state = es->s_state;
6055 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6057 if (orig_state != es->s_state)
6059 /* Write out restored error information to the superblock */
6060 if (changed && !really_read_only) {
6062 err2 = ext4_commit_super(sb);
6068 ext4_msg(sb, KERN_ERR, "error loading journal");
6072 EXT4_SB(sb)->s_journal = journal;
6073 err = ext4_clear_journal_err(sb, es);
6075 EXT4_SB(sb)->s_journal = NULL;
6076 jbd2_journal_destroy(journal);
6080 if (!really_read_only && journal_devnum &&
6081 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6082 es->s_journal_dev = cpu_to_le32(journal_devnum);
6083 ext4_commit_super(sb);
6085 if (!really_read_only && journal_inum &&
6086 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6087 es->s_journal_inum = cpu_to_le32(journal_inum);
6088 ext4_commit_super(sb);
6094 jbd2_journal_destroy(journal);
6098 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6099 static void ext4_update_super(struct super_block *sb)
6101 struct ext4_sb_info *sbi = EXT4_SB(sb);
6102 struct ext4_super_block *es = sbi->s_es;
6103 struct buffer_head *sbh = sbi->s_sbh;
6107 * If the file system is mounted read-only, don't update the
6108 * superblock write time. This avoids updating the superblock
6109 * write time when we are mounting the root file system
6110 * read/only but we need to replay the journal; at that point,
6111 * for people who are east of GMT and who make their clock
6112 * tick in localtime for Windows bug-for-bug compatibility,
6113 * the clock is set in the future, and this will cause e2fsck
6114 * to complain and force a full file system check.
6117 ext4_update_tstamp(es, s_wtime);
6118 es->s_kbytes_written =
6119 cpu_to_le64(sbi->s_kbytes_written +
6120 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6121 sbi->s_sectors_written_start) >> 1));
6122 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6123 ext4_free_blocks_count_set(es,
6124 EXT4_C2B(sbi, percpu_counter_sum_positive(
6125 &sbi->s_freeclusters_counter)));
6126 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6127 es->s_free_inodes_count =
6128 cpu_to_le32(percpu_counter_sum_positive(
6129 &sbi->s_freeinodes_counter));
6130 /* Copy error information to the on-disk superblock */
6131 spin_lock(&sbi->s_error_lock);
6132 if (sbi->s_add_error_count > 0) {
6133 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6134 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6135 __ext4_update_tstamp(&es->s_first_error_time,
6136 &es->s_first_error_time_hi,
6137 sbi->s_first_error_time);
6138 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6139 sizeof(es->s_first_error_func));
6140 es->s_first_error_line =
6141 cpu_to_le32(sbi->s_first_error_line);
6142 es->s_first_error_ino =
6143 cpu_to_le32(sbi->s_first_error_ino);
6144 es->s_first_error_block =
6145 cpu_to_le64(sbi->s_first_error_block);
6146 es->s_first_error_errcode =
6147 ext4_errno_to_code(sbi->s_first_error_code);
6149 __ext4_update_tstamp(&es->s_last_error_time,
6150 &es->s_last_error_time_hi,
6151 sbi->s_last_error_time);
6152 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6153 sizeof(es->s_last_error_func));
6154 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6155 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6156 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6157 es->s_last_error_errcode =
6158 ext4_errno_to_code(sbi->s_last_error_code);
6160 * Start the daily error reporting function if it hasn't been
6163 if (!es->s_error_count)
6164 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6165 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6166 sbi->s_add_error_count = 0;
6168 spin_unlock(&sbi->s_error_lock);
6170 ext4_superblock_csum_set(sb);
6174 static int ext4_commit_super(struct super_block *sb)
6176 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6180 if (block_device_ejected(sb))
6183 ext4_update_super(sb);
6186 /* Buffer got discarded which means block device got invalidated */
6187 if (!buffer_mapped(sbh)) {
6192 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6194 * Oh, dear. A previous attempt to write the
6195 * superblock failed. This could happen because the
6196 * USB device was yanked out. Or it could happen to
6197 * be a transient write error and maybe the block will
6198 * be remapped. Nothing we can do but to retry the
6199 * write and hope for the best.
6201 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6202 "superblock detected");
6203 clear_buffer_write_io_error(sbh);
6204 set_buffer_uptodate(sbh);
6207 /* Clear potential dirty bit if it was journalled update */
6208 clear_buffer_dirty(sbh);
6209 sbh->b_end_io = end_buffer_write_sync;
6210 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6211 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6212 wait_on_buffer(sbh);
6213 if (buffer_write_io_error(sbh)) {
6214 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6216 clear_buffer_write_io_error(sbh);
6217 set_buffer_uptodate(sbh);
6224 * Have we just finished recovery? If so, and if we are mounting (or
6225 * remounting) the filesystem readonly, then we will end up with a
6226 * consistent fs on disk. Record that fact.
6228 static int ext4_mark_recovery_complete(struct super_block *sb,
6229 struct ext4_super_block *es)
6232 journal_t *journal = EXT4_SB(sb)->s_journal;
6234 if (!ext4_has_feature_journal(sb)) {
6235 if (journal != NULL) {
6236 ext4_error(sb, "Journal got removed while the fs was "
6238 return -EFSCORRUPTED;
6242 jbd2_journal_lock_updates(journal);
6243 err = jbd2_journal_flush(journal, 0);
6247 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6248 ext4_has_feature_orphan_present(sb))) {
6249 if (!ext4_orphan_file_empty(sb)) {
6250 ext4_error(sb, "Orphan file not empty on read-only fs.");
6251 err = -EFSCORRUPTED;
6254 ext4_clear_feature_journal_needs_recovery(sb);
6255 ext4_clear_feature_orphan_present(sb);
6256 ext4_commit_super(sb);
6259 jbd2_journal_unlock_updates(journal);
6264 * If we are mounting (or read-write remounting) a filesystem whose journal
6265 * has recorded an error from a previous lifetime, move that error to the
6266 * main filesystem now.
6268 static int ext4_clear_journal_err(struct super_block *sb,
6269 struct ext4_super_block *es)
6275 if (!ext4_has_feature_journal(sb)) {
6276 ext4_error(sb, "Journal got removed while the fs was mounted!");
6277 return -EFSCORRUPTED;
6280 journal = EXT4_SB(sb)->s_journal;
6283 * Now check for any error status which may have been recorded in the
6284 * journal by a prior ext4_error() or ext4_abort()
6287 j_errno = jbd2_journal_errno(journal);
6291 errstr = ext4_decode_error(sb, j_errno, nbuf);
6292 ext4_warning(sb, "Filesystem error recorded "
6293 "from previous mount: %s", errstr);
6295 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6296 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6297 j_errno = ext4_commit_super(sb);
6300 ext4_warning(sb, "Marked fs in need of filesystem check.");
6302 jbd2_journal_clear_err(journal);
6303 jbd2_journal_update_sb_errno(journal);
6309 * Force the running and committing transactions to commit,
6310 * and wait on the commit.
6312 int ext4_force_commit(struct super_block *sb)
6314 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal);
6317 static int ext4_sync_fs(struct super_block *sb, int wait)
6321 bool needs_barrier = false;
6322 struct ext4_sb_info *sbi = EXT4_SB(sb);
6324 if (unlikely(ext4_forced_shutdown(sb)))
6327 trace_ext4_sync_fs(sb, wait);
6328 flush_workqueue(sbi->rsv_conversion_wq);
6330 * Writeback quota in non-journalled quota case - journalled quota has
6333 dquot_writeback_dquots(sb, -1);
6335 * Data writeback is possible w/o journal transaction, so barrier must
6336 * being sent at the end of the function. But we can skip it if
6337 * transaction_commit will do it for us.
6339 if (sbi->s_journal) {
6340 target = jbd2_get_latest_transaction(sbi->s_journal);
6341 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6342 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6343 needs_barrier = true;
6345 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6347 ret = jbd2_log_wait_commit(sbi->s_journal,
6350 } else if (wait && test_opt(sb, BARRIER))
6351 needs_barrier = true;
6352 if (needs_barrier) {
6354 err = blkdev_issue_flush(sb->s_bdev);
6363 * LVM calls this function before a (read-only) snapshot is created. This
6364 * gives us a chance to flush the journal completely and mark the fs clean.
6366 * Note that only this function cannot bring a filesystem to be in a clean
6367 * state independently. It relies on upper layer to stop all data & metadata
6370 static int ext4_freeze(struct super_block *sb)
6373 journal_t *journal = EXT4_SB(sb)->s_journal;
6376 /* Now we set up the journal barrier. */
6377 jbd2_journal_lock_updates(journal);
6380 * Don't clear the needs_recovery flag if we failed to
6381 * flush the journal.
6383 error = jbd2_journal_flush(journal, 0);
6387 /* Journal blocked and flushed, clear needs_recovery flag. */
6388 ext4_clear_feature_journal_needs_recovery(sb);
6389 if (ext4_orphan_file_empty(sb))
6390 ext4_clear_feature_orphan_present(sb);
6393 error = ext4_commit_super(sb);
6396 /* we rely on upper layer to stop further updates */
6397 jbd2_journal_unlock_updates(journal);
6402 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6403 * flag here, even though the filesystem is not technically dirty yet.
6405 static int ext4_unfreeze(struct super_block *sb)
6407 if (ext4_forced_shutdown(sb))
6410 if (EXT4_SB(sb)->s_journal) {
6411 /* Reset the needs_recovery flag before the fs is unlocked. */
6412 ext4_set_feature_journal_needs_recovery(sb);
6413 if (ext4_has_feature_orphan_file(sb))
6414 ext4_set_feature_orphan_present(sb);
6417 ext4_commit_super(sb);
6422 * Structure to save mount options for ext4_remount's benefit
6424 struct ext4_mount_options {
6425 unsigned long s_mount_opt;
6426 unsigned long s_mount_opt2;
6429 unsigned long s_commit_interval;
6430 u32 s_min_batch_time, s_max_batch_time;
6433 char *s_qf_names[EXT4_MAXQUOTAS];
6437 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6439 struct ext4_fs_context *ctx = fc->fs_private;
6440 struct ext4_super_block *es;
6441 struct ext4_sb_info *sbi = EXT4_SB(sb);
6442 unsigned long old_sb_flags;
6443 struct ext4_mount_options old_opts;
6447 int enable_quota = 0;
6449 char *to_free[EXT4_MAXQUOTAS];
6453 /* Store the original options */
6454 old_sb_flags = sb->s_flags;
6455 old_opts.s_mount_opt = sbi->s_mount_opt;
6456 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6457 old_opts.s_resuid = sbi->s_resuid;
6458 old_opts.s_resgid = sbi->s_resgid;
6459 old_opts.s_commit_interval = sbi->s_commit_interval;
6460 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6461 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6463 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6464 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6465 if (sbi->s_qf_names[i]) {
6466 char *qf_name = get_qf_name(sb, sbi, i);
6468 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6469 if (!old_opts.s_qf_names[i]) {
6470 for (j = 0; j < i; j++)
6471 kfree(old_opts.s_qf_names[j]);
6475 old_opts.s_qf_names[i] = NULL;
6477 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6478 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6479 ctx->journal_ioprio =
6480 sbi->s_journal->j_task->io_context->ioprio;
6482 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6486 ext4_apply_options(fc, sb);
6488 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6489 test_opt(sb, JOURNAL_CHECKSUM)) {
6490 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6491 "during remount not supported; ignoring");
6492 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6495 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6496 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6497 ext4_msg(sb, KERN_ERR, "can't mount with "
6498 "both data=journal and delalloc");
6502 if (test_opt(sb, DIOREAD_NOLOCK)) {
6503 ext4_msg(sb, KERN_ERR, "can't mount with "
6504 "both data=journal and dioread_nolock");
6508 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6509 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6510 ext4_msg(sb, KERN_ERR, "can't mount with "
6511 "journal_async_commit in data=ordered mode");
6517 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6518 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6523 if (test_opt2(sb, ABORT))
6524 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6526 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6527 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6531 if (sbi->s_journal) {
6532 ext4_init_journal_params(sb, sbi->s_journal);
6533 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6536 /* Flush outstanding errors before changing fs state */
6537 flush_work(&sbi->s_sb_upd_work);
6539 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6540 if (ext4_forced_shutdown(sb)) {
6545 if (fc->sb_flags & SB_RDONLY) {
6546 err = sync_filesystem(sb);
6549 err = dquot_suspend(sb, -1);
6554 * First of all, the unconditional stuff we have to do
6555 * to disable replay of the journal when we next remount
6557 sb->s_flags |= SB_RDONLY;
6560 * OK, test if we are remounting a valid rw partition
6561 * readonly, and if so set the rdonly flag and then
6562 * mark the partition as valid again.
6564 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6565 (sbi->s_mount_state & EXT4_VALID_FS))
6566 es->s_state = cpu_to_le16(sbi->s_mount_state);
6568 if (sbi->s_journal) {
6570 * We let remount-ro finish even if marking fs
6571 * as clean failed...
6573 ext4_mark_recovery_complete(sb, es);
6576 /* Make sure we can mount this feature set readwrite */
6577 if (ext4_has_feature_readonly(sb) ||
6578 !ext4_feature_set_ok(sb, 0)) {
6583 * Make sure the group descriptor checksums
6584 * are sane. If they aren't, refuse to remount r/w.
6586 for (g = 0; g < sbi->s_groups_count; g++) {
6587 struct ext4_group_desc *gdp =
6588 ext4_get_group_desc(sb, g, NULL);
6590 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6591 ext4_msg(sb, KERN_ERR,
6592 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6593 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6594 le16_to_cpu(gdp->bg_checksum));
6601 * If we have an unprocessed orphan list hanging
6602 * around from a previously readonly bdev mount,
6603 * require a full umount/remount for now.
6605 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6606 ext4_msg(sb, KERN_WARNING, "Couldn't "
6607 "remount RDWR because of unprocessed "
6608 "orphan inode list. Please "
6609 "umount/remount instead");
6615 * Mounting a RDONLY partition read-write, so reread
6616 * and store the current valid flag. (It may have
6617 * been changed by e2fsck since we originally mounted
6620 if (sbi->s_journal) {
6621 err = ext4_clear_journal_err(sb, es);
6625 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6628 err = ext4_setup_super(sb, es, 0);
6632 sb->s_flags &= ~SB_RDONLY;
6633 if (ext4_has_feature_mmp(sb)) {
6634 err = ext4_multi_mount_protect(sb,
6635 le64_to_cpu(es->s_mmp_block));
6646 * Handle creation of system zone data early because it can fail.
6647 * Releasing of existing data is done when we are sure remount will
6650 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6651 err = ext4_setup_system_zone(sb);
6656 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6657 err = ext4_commit_super(sb);
6664 if (sb_any_quota_suspended(sb))
6665 dquot_resume(sb, -1);
6666 else if (ext4_has_feature_quota(sb)) {
6667 err = ext4_enable_quotas(sb);
6672 /* Release old quota file names */
6673 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6674 kfree(old_opts.s_qf_names[i]);
6676 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6677 ext4_release_system_zone(sb);
6680 * Reinitialize lazy itable initialization thread based on
6683 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6684 ext4_unregister_li_request(sb);
6686 ext4_group_t first_not_zeroed;
6687 first_not_zeroed = ext4_has_uninit_itable(sb);
6688 ext4_register_li_request(sb, first_not_zeroed);
6691 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6692 ext4_stop_mmpd(sbi);
6698 * If there was a failing r/w to ro transition, we may need to
6701 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) &&
6702 sb_any_quota_suspended(sb))
6703 dquot_resume(sb, -1);
6704 sb->s_flags = old_sb_flags;
6705 sbi->s_mount_opt = old_opts.s_mount_opt;
6706 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6707 sbi->s_resuid = old_opts.s_resuid;
6708 sbi->s_resgid = old_opts.s_resgid;
6709 sbi->s_commit_interval = old_opts.s_commit_interval;
6710 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6711 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6712 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6713 ext4_release_system_zone(sb);
6715 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6716 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6717 to_free[i] = get_qf_name(sb, sbi, i);
6718 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6721 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6724 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6725 ext4_stop_mmpd(sbi);
6729 static int ext4_reconfigure(struct fs_context *fc)
6731 struct super_block *sb = fc->root->d_sb;
6734 fc->s_fs_info = EXT4_SB(sb);
6736 ret = ext4_check_opt_consistency(fc, sb);
6740 ret = __ext4_remount(fc, sb);
6744 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6745 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6746 ext4_quota_mode(sb));
6752 static int ext4_statfs_project(struct super_block *sb,
6753 kprojid_t projid, struct kstatfs *buf)
6756 struct dquot *dquot;
6760 qid = make_kqid_projid(projid);
6761 dquot = dqget(sb, qid);
6763 return PTR_ERR(dquot);
6764 spin_lock(&dquot->dq_dqb_lock);
6766 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6767 dquot->dq_dqb.dqb_bhardlimit);
6768 limit >>= sb->s_blocksize_bits;
6770 if (limit && buf->f_blocks > limit) {
6771 curblock = (dquot->dq_dqb.dqb_curspace +
6772 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6773 buf->f_blocks = limit;
6774 buf->f_bfree = buf->f_bavail =
6775 (buf->f_blocks > curblock) ?
6776 (buf->f_blocks - curblock) : 0;
6779 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6780 dquot->dq_dqb.dqb_ihardlimit);
6781 if (limit && buf->f_files > limit) {
6782 buf->f_files = limit;
6784 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6785 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6788 spin_unlock(&dquot->dq_dqb_lock);
6794 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6796 struct super_block *sb = dentry->d_sb;
6797 struct ext4_sb_info *sbi = EXT4_SB(sb);
6798 struct ext4_super_block *es = sbi->s_es;
6799 ext4_fsblk_t overhead = 0, resv_blocks;
6801 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6803 if (!test_opt(sb, MINIX_DF))
6804 overhead = sbi->s_overhead;
6806 buf->f_type = EXT4_SUPER_MAGIC;
6807 buf->f_bsize = sb->s_blocksize;
6808 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6809 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6810 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6811 /* prevent underflow in case that few free space is available */
6812 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6813 buf->f_bavail = buf->f_bfree -
6814 (ext4_r_blocks_count(es) + resv_blocks);
6815 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6817 buf->f_files = le32_to_cpu(es->s_inodes_count);
6818 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6819 buf->f_namelen = EXT4_NAME_LEN;
6820 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6823 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6824 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6825 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6834 * Helper functions so that transaction is started before we acquire dqio_sem
6835 * to keep correct lock ordering of transaction > dqio_sem
6837 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6839 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6842 static int ext4_write_dquot(struct dquot *dquot)
6846 struct inode *inode;
6848 inode = dquot_to_inode(dquot);
6849 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6850 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6852 return PTR_ERR(handle);
6853 ret = dquot_commit(dquot);
6854 err = ext4_journal_stop(handle);
6860 static int ext4_acquire_dquot(struct dquot *dquot)
6865 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6866 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6868 return PTR_ERR(handle);
6869 ret = dquot_acquire(dquot);
6870 err = ext4_journal_stop(handle);
6876 static int ext4_release_dquot(struct dquot *dquot)
6881 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6882 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6883 if (IS_ERR(handle)) {
6884 /* Release dquot anyway to avoid endless cycle in dqput() */
6885 dquot_release(dquot);
6886 return PTR_ERR(handle);
6888 ret = dquot_release(dquot);
6889 err = ext4_journal_stop(handle);
6895 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6897 struct super_block *sb = dquot->dq_sb;
6899 if (ext4_is_quota_journalled(sb)) {
6900 dquot_mark_dquot_dirty(dquot);
6901 return ext4_write_dquot(dquot);
6903 return dquot_mark_dquot_dirty(dquot);
6907 static int ext4_write_info(struct super_block *sb, int type)
6912 /* Data block + inode block */
6913 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6915 return PTR_ERR(handle);
6916 ret = dquot_commit_info(sb, type);
6917 err = ext4_journal_stop(handle);
6923 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6925 struct ext4_inode_info *ei = EXT4_I(inode);
6927 /* The first argument of lockdep_set_subclass has to be
6928 * *exactly* the same as the argument to init_rwsem() --- in
6929 * this case, in init_once() --- or lockdep gets unhappy
6930 * because the name of the lock is set using the
6931 * stringification of the argument to init_rwsem().
6933 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6934 lockdep_set_subclass(&ei->i_data_sem, subclass);
6938 * Standard function to be called on quota_on
6940 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6941 const struct path *path)
6945 if (!test_opt(sb, QUOTA))
6948 /* Quotafile not on the same filesystem? */
6949 if (path->dentry->d_sb != sb)
6952 /* Quota already enabled for this file? */
6953 if (IS_NOQUOTA(d_inode(path->dentry)))
6956 /* Journaling quota? */
6957 if (EXT4_SB(sb)->s_qf_names[type]) {
6958 /* Quotafile not in fs root? */
6959 if (path->dentry->d_parent != sb->s_root)
6960 ext4_msg(sb, KERN_WARNING,
6961 "Quota file not on filesystem root. "
6962 "Journaled quota will not work");
6963 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6966 * Clear the flag just in case mount options changed since
6969 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6972 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6973 err = dquot_quota_on(sb, type, format_id, path);
6975 struct inode *inode = d_inode(path->dentry);
6979 * Set inode flags to prevent userspace from messing with quota
6980 * files. If this fails, we return success anyway since quotas
6981 * are already enabled and this is not a hard failure.
6984 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6987 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6988 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6989 S_NOATIME | S_IMMUTABLE);
6990 err = ext4_mark_inode_dirty(handle, inode);
6991 ext4_journal_stop(handle);
6993 inode_unlock(inode);
6995 dquot_quota_off(sb, type);
6998 lockdep_set_quota_inode(path->dentry->d_inode,
7003 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7007 return qf_inum == EXT4_USR_QUOTA_INO;
7009 return qf_inum == EXT4_GRP_QUOTA_INO;
7011 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7017 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7021 struct inode *qf_inode;
7022 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7023 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7024 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7025 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7028 BUG_ON(!ext4_has_feature_quota(sb));
7030 if (!qf_inums[type])
7033 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7034 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7035 qf_inums[type], type);
7039 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7040 if (IS_ERR(qf_inode)) {
7041 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7042 qf_inums[type], type);
7043 return PTR_ERR(qf_inode);
7046 /* Don't account quota for quota files to avoid recursion */
7047 qf_inode->i_flags |= S_NOQUOTA;
7048 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7049 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
7051 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7057 /* Enable usage tracking for all quota types. */
7058 int ext4_enable_quotas(struct super_block *sb)
7061 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7062 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7063 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7064 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7066 bool quota_mopt[EXT4_MAXQUOTAS] = {
7067 test_opt(sb, USRQUOTA),
7068 test_opt(sb, GRPQUOTA),
7069 test_opt(sb, PRJQUOTA),
7072 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7073 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7074 if (qf_inums[type]) {
7075 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7076 DQUOT_USAGE_ENABLED |
7077 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7080 "Failed to enable quota tracking "
7081 "(type=%d, err=%d, ino=%lu). "
7082 "Please run e2fsck to fix.", type,
7083 err, qf_inums[type]);
7085 ext4_quotas_off(sb, type);
7093 static int ext4_quota_off(struct super_block *sb, int type)
7095 struct inode *inode = sb_dqopt(sb)->files[type];
7099 /* Force all delayed allocation blocks to be allocated.
7100 * Caller already holds s_umount sem */
7101 if (test_opt(sb, DELALLOC))
7102 sync_filesystem(sb);
7104 if (!inode || !igrab(inode))
7107 err = dquot_quota_off(sb, type);
7108 if (err || ext4_has_feature_quota(sb))
7111 * When the filesystem was remounted read-only first, we cannot cleanup
7112 * inode flags here. Bad luck but people should be using QUOTA feature
7113 * these days anyway.
7120 * Update modification times of quota files when userspace can
7121 * start looking at them. If we fail, we return success anyway since
7122 * this is not a hard failure and quotas are already disabled.
7124 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7125 if (IS_ERR(handle)) {
7126 err = PTR_ERR(handle);
7129 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7130 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7131 inode->i_mtime = inode_set_ctime_current(inode);
7132 err = ext4_mark_inode_dirty(handle, inode);
7133 ext4_journal_stop(handle);
7135 inode_unlock(inode);
7137 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7141 return dquot_quota_off(sb, type);
7144 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7145 * acquiring the locks... As quota files are never truncated and quota code
7146 * itself serializes the operations (and no one else should touch the files)
7147 * we don't have to be afraid of races */
7148 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7149 size_t len, loff_t off)
7151 struct inode *inode = sb_dqopt(sb)->files[type];
7152 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7153 int offset = off & (sb->s_blocksize - 1);
7156 struct buffer_head *bh;
7157 loff_t i_size = i_size_read(inode);
7161 if (off+len > i_size)
7164 while (toread > 0) {
7165 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7166 bh = ext4_bread(NULL, inode, blk, 0);
7169 if (!bh) /* A hole? */
7170 memset(data, 0, tocopy);
7172 memcpy(data, bh->b_data+offset, tocopy);
7182 /* Write to quotafile (we know the transaction is already started and has
7183 * enough credits) */
7184 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7185 const char *data, size_t len, loff_t off)
7187 struct inode *inode = sb_dqopt(sb)->files[type];
7188 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7189 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7191 struct buffer_head *bh;
7192 handle_t *handle = journal_current_handle();
7195 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7196 " cancelled because transaction is not started",
7197 (unsigned long long)off, (unsigned long long)len);
7201 * Since we account only one data block in transaction credits,
7202 * then it is impossible to cross a block boundary.
7204 if (sb->s_blocksize - offset < len) {
7205 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7206 " cancelled because not block aligned",
7207 (unsigned long long)off, (unsigned long long)len);
7212 bh = ext4_bread(handle, inode, blk,
7213 EXT4_GET_BLOCKS_CREATE |
7214 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7215 } while (PTR_ERR(bh) == -ENOSPC &&
7216 ext4_should_retry_alloc(inode->i_sb, &retries));
7221 BUFFER_TRACE(bh, "get write access");
7222 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7228 memcpy(bh->b_data+offset, data, len);
7229 flush_dcache_page(bh->b_page);
7231 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7234 if (inode->i_size < off + len) {
7235 i_size_write(inode, off + len);
7236 EXT4_I(inode)->i_disksize = inode->i_size;
7237 err2 = ext4_mark_inode_dirty(handle, inode);
7238 if (unlikely(err2 && !err))
7241 return err ? err : len;
7245 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7246 static inline void register_as_ext2(void)
7248 int err = register_filesystem(&ext2_fs_type);
7251 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7254 static inline void unregister_as_ext2(void)
7256 unregister_filesystem(&ext2_fs_type);
7259 static inline int ext2_feature_set_ok(struct super_block *sb)
7261 if (ext4_has_unknown_ext2_incompat_features(sb))
7265 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7270 static inline void register_as_ext2(void) { }
7271 static inline void unregister_as_ext2(void) { }
7272 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7275 static inline void register_as_ext3(void)
7277 int err = register_filesystem(&ext3_fs_type);
7280 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7283 static inline void unregister_as_ext3(void)
7285 unregister_filesystem(&ext3_fs_type);
7288 static inline int ext3_feature_set_ok(struct super_block *sb)
7290 if (ext4_has_unknown_ext3_incompat_features(sb))
7292 if (!ext4_has_feature_journal(sb))
7296 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7301 static void ext4_kill_sb(struct super_block *sb)
7303 struct ext4_sb_info *sbi = EXT4_SB(sb);
7304 struct block_device *journal_bdev = sbi ? sbi->s_journal_bdev : NULL;
7306 kill_block_super(sb);
7309 blkdev_put(journal_bdev, sb);
7312 static struct file_system_type ext4_fs_type = {
7313 .owner = THIS_MODULE,
7315 .init_fs_context = ext4_init_fs_context,
7316 .parameters = ext4_param_specs,
7317 .kill_sb = ext4_kill_sb,
7318 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7320 MODULE_ALIAS_FS("ext4");
7322 /* Shared across all ext4 file systems */
7323 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7325 static int __init ext4_init_fs(void)
7329 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7330 ext4_li_info = NULL;
7332 /* Build-time check for flags consistency */
7333 ext4_check_flag_values();
7335 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7336 init_waitqueue_head(&ext4__ioend_wq[i]);
7338 err = ext4_init_es();
7342 err = ext4_init_pending();
7346 err = ext4_init_post_read_processing();
7350 err = ext4_init_pageio();
7354 err = ext4_init_system_zone();
7358 err = ext4_init_sysfs();
7362 err = ext4_init_mballoc();
7365 err = init_inodecache();
7369 err = ext4_fc_init_dentry_cache();
7375 err = register_filesystem(&ext4_fs_type);
7381 unregister_as_ext2();
7382 unregister_as_ext3();
7383 ext4_fc_destroy_dentry_cache();
7385 destroy_inodecache();
7387 ext4_exit_mballoc();
7391 ext4_exit_system_zone();
7395 ext4_exit_post_read_processing();
7397 ext4_exit_pending();
7404 static void __exit ext4_exit_fs(void)
7406 ext4_destroy_lazyinit_thread();
7407 unregister_as_ext2();
7408 unregister_as_ext3();
7409 unregister_filesystem(&ext4_fs_type);
7410 ext4_fc_destroy_dentry_cache();
7411 destroy_inodecache();
7412 ext4_exit_mballoc();
7414 ext4_exit_system_zone();
7416 ext4_exit_post_read_processing();
7418 ext4_exit_pending();
7421 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7422 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7423 MODULE_LICENSE("GPL");
7424 MODULE_SOFTDEP("pre: crc32c");
7425 module_init(ext4_init_fs)
7426 module_exit(ext4_exit_fs)