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 const struct fs_parameter_spec ext4_param_specs[];
102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
103 * -> page lock -> i_data_sem (rw)
105 * buffered write path:
106 * sb_start_write -> i_mutex -> mmap_lock
107 * sb_start_write -> i_mutex -> transaction start -> page lock ->
111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
117 * sb_start_write -> i_mutex -> mmap_lock
118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
121 * transaction start -> page lock(s) -> i_data_sem (rw)
124 static const struct fs_context_operations ext4_context_ops = {
125 .parse_param = ext4_parse_param,
126 .get_tree = ext4_get_tree,
127 .reconfigure = ext4_reconfigure,
128 .free = ext4_fc_free,
132 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
133 static struct file_system_type ext2_fs_type = {
134 .owner = THIS_MODULE,
136 .init_fs_context = ext4_init_fs_context,
137 .parameters = ext4_param_specs,
138 .kill_sb = kill_block_super,
139 .fs_flags = FS_REQUIRES_DEV,
141 MODULE_ALIAS_FS("ext2");
142 MODULE_ALIAS("ext2");
143 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
145 #define IS_EXT2_SB(sb) (0)
149 static struct file_system_type ext3_fs_type = {
150 .owner = THIS_MODULE,
152 .init_fs_context = ext4_init_fs_context,
153 .parameters = ext4_param_specs,
154 .kill_sb = kill_block_super,
155 .fs_flags = FS_REQUIRES_DEV,
157 MODULE_ALIAS_FS("ext3");
158 MODULE_ALIAS("ext3");
159 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
162 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
166 * buffer's verified bit is no longer valid after reading from
167 * disk again due to write out error, clear it to make sure we
168 * recheck the buffer contents.
170 clear_buffer_verified(bh);
172 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
174 submit_bh(REQ_OP_READ, op_flags, bh);
177 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
180 BUG_ON(!buffer_locked(bh));
182 if (ext4_buffer_uptodate(bh)) {
186 __ext4_read_bh(bh, op_flags, end_io);
189 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
191 BUG_ON(!buffer_locked(bh));
193 if (ext4_buffer_uptodate(bh)) {
198 __ext4_read_bh(bh, op_flags, end_io);
201 if (buffer_uptodate(bh))
206 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
208 if (trylock_buffer(bh)) {
210 return ext4_read_bh(bh, op_flags, NULL);
211 ext4_read_bh_nowait(bh, op_flags, NULL);
216 if (buffer_uptodate(bh))
224 * This works like __bread_gfp() except it uses ERR_PTR for error
225 * returns. Currently with sb_bread it's impossible to distinguish
226 * between ENOMEM and EIO situations (since both result in a NULL
229 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
230 sector_t block, int op_flags,
233 struct buffer_head *bh;
236 bh = sb_getblk_gfp(sb, block, gfp);
238 return ERR_PTR(-ENOMEM);
239 if (ext4_buffer_uptodate(bh))
242 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
250 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
253 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
256 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
259 return __ext4_sb_bread_gfp(sb, block, 0, 0);
262 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
264 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
267 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
272 static int ext4_verify_csum_type(struct super_block *sb,
273 struct ext4_super_block *es)
275 if (!ext4_has_feature_metadata_csum(sb))
278 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
281 __le32 ext4_superblock_csum(struct super_block *sb,
282 struct ext4_super_block *es)
284 struct ext4_sb_info *sbi = EXT4_SB(sb);
285 int offset = offsetof(struct ext4_super_block, s_checksum);
288 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
290 return cpu_to_le32(csum);
293 static int ext4_superblock_csum_verify(struct super_block *sb,
294 struct ext4_super_block *es)
296 if (!ext4_has_metadata_csum(sb))
299 return es->s_checksum == ext4_superblock_csum(sb, es);
302 void ext4_superblock_csum_set(struct super_block *sb)
304 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306 if (!ext4_has_metadata_csum(sb))
309 es->s_checksum = ext4_superblock_csum(sb, es);
312 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
313 struct ext4_group_desc *bg)
315 return le32_to_cpu(bg->bg_block_bitmap_lo) |
316 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
317 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
320 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
321 struct ext4_group_desc *bg)
323 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
324 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
325 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
328 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
329 struct ext4_group_desc *bg)
331 return le32_to_cpu(bg->bg_inode_table_lo) |
332 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
333 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
336 __u32 ext4_free_group_clusters(struct super_block *sb,
337 struct ext4_group_desc *bg)
339 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
340 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
341 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
344 __u32 ext4_free_inodes_count(struct super_block *sb,
345 struct ext4_group_desc *bg)
347 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
348 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
349 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
352 __u32 ext4_used_dirs_count(struct super_block *sb,
353 struct ext4_group_desc *bg)
355 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
356 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
357 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
360 __u32 ext4_itable_unused_count(struct super_block *sb,
361 struct ext4_group_desc *bg)
363 return le16_to_cpu(bg->bg_itable_unused_lo) |
364 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
365 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
368 void ext4_block_bitmap_set(struct super_block *sb,
369 struct ext4_group_desc *bg, ext4_fsblk_t blk)
371 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
372 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
373 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
376 void ext4_inode_bitmap_set(struct super_block *sb,
377 struct ext4_group_desc *bg, ext4_fsblk_t blk)
379 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
380 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
381 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
384 void ext4_inode_table_set(struct super_block *sb,
385 struct ext4_group_desc *bg, ext4_fsblk_t blk)
387 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
388 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
389 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
392 void ext4_free_group_clusters_set(struct super_block *sb,
393 struct ext4_group_desc *bg, __u32 count)
395 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
396 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
397 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
400 void ext4_free_inodes_set(struct super_block *sb,
401 struct ext4_group_desc *bg, __u32 count)
403 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
404 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
405 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
408 void ext4_used_dirs_set(struct super_block *sb,
409 struct ext4_group_desc *bg, __u32 count)
411 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
412 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
413 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
416 void ext4_itable_unused_set(struct super_block *sb,
417 struct ext4_group_desc *bg, __u32 count)
419 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
420 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
421 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
424 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
426 now = clamp_val(now, 0, (1ull << 40) - 1);
428 *lo = cpu_to_le32(lower_32_bits(now));
429 *hi = upper_32_bits(now);
432 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
434 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
436 #define ext4_update_tstamp(es, tstamp) \
437 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
438 ktime_get_real_seconds())
439 #define ext4_get_tstamp(es, tstamp) \
440 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
443 * The del_gendisk() function uninitializes the disk-specific data
444 * structures, including the bdi structure, without telling anyone
445 * else. Once this happens, any attempt to call mark_buffer_dirty()
446 * (for example, by ext4_commit_super), will cause a kernel OOPS.
447 * This is a kludge to prevent these oops until we can put in a proper
448 * hook in del_gendisk() to inform the VFS and file system layers.
450 static int block_device_ejected(struct super_block *sb)
452 struct inode *bd_inode = sb->s_bdev->bd_inode;
453 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
455 return bdi->dev == NULL;
458 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
460 struct super_block *sb = journal->j_private;
461 struct ext4_sb_info *sbi = EXT4_SB(sb);
462 int error = is_journal_aborted(journal);
463 struct ext4_journal_cb_entry *jce;
465 BUG_ON(txn->t_state == T_FINISHED);
467 ext4_process_freed_data(sb, txn->t_tid);
469 spin_lock(&sbi->s_md_lock);
470 while (!list_empty(&txn->t_private_list)) {
471 jce = list_entry(txn->t_private_list.next,
472 struct ext4_journal_cb_entry, jce_list);
473 list_del_init(&jce->jce_list);
474 spin_unlock(&sbi->s_md_lock);
475 jce->jce_func(sb, jce, error);
476 spin_lock(&sbi->s_md_lock);
478 spin_unlock(&sbi->s_md_lock);
482 * This writepage callback for write_cache_pages()
483 * takes care of a few cases after page cleaning.
485 * write_cache_pages() already checks for dirty pages
486 * and calls clear_page_dirty_for_io(), which we want,
487 * to write protect the pages.
489 * However, we may have to redirty a page (see below.)
491 static int ext4_journalled_writepage_callback(struct page *page,
492 struct writeback_control *wbc,
495 transaction_t *transaction = (transaction_t *) data;
496 struct buffer_head *bh, *head;
497 struct journal_head *jh;
499 bh = head = page_buffers(page);
502 * We have to redirty a page in these cases:
503 * 1) If buffer is dirty, it means the page was dirty because it
504 * contains a buffer that needs checkpointing. So the dirty bit
505 * needs to be preserved so that checkpointing writes the buffer
507 * 2) If buffer is not part of the committing transaction
508 * (we may have just accidentally come across this buffer because
509 * inode range tracking is not exact) or if the currently running
510 * transaction already contains this buffer as well, dirty bit
511 * needs to be preserved so that the buffer gets writeprotected
512 * properly on running transaction's commit.
515 if (buffer_dirty(bh) ||
516 (jh && (jh->b_transaction != transaction ||
517 jh->b_next_transaction))) {
518 redirty_page_for_writepage(wbc, page);
521 } while ((bh = bh->b_this_page) != head);
524 return AOP_WRITEPAGE_ACTIVATE;
527 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
529 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
530 struct writeback_control wbc = {
531 .sync_mode = WB_SYNC_ALL,
532 .nr_to_write = LONG_MAX,
533 .range_start = jinode->i_dirty_start,
534 .range_end = jinode->i_dirty_end,
537 return write_cache_pages(mapping, &wbc,
538 ext4_journalled_writepage_callback,
539 jinode->i_transaction);
542 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
546 if (ext4_should_journal_data(jinode->i_vfs_inode))
547 ret = ext4_journalled_submit_inode_data_buffers(jinode);
549 ret = jbd2_journal_submit_inode_data_buffers(jinode);
554 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
558 if (!ext4_should_journal_data(jinode->i_vfs_inode))
559 ret = jbd2_journal_finish_inode_data_buffers(jinode);
564 static bool system_going_down(void)
566 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
567 || system_state == SYSTEM_RESTART;
570 struct ext4_err_translation {
575 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
577 static struct ext4_err_translation err_translation[] = {
578 EXT4_ERR_TRANSLATE(EIO),
579 EXT4_ERR_TRANSLATE(ENOMEM),
580 EXT4_ERR_TRANSLATE(EFSBADCRC),
581 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
582 EXT4_ERR_TRANSLATE(ENOSPC),
583 EXT4_ERR_TRANSLATE(ENOKEY),
584 EXT4_ERR_TRANSLATE(EROFS),
585 EXT4_ERR_TRANSLATE(EFBIG),
586 EXT4_ERR_TRANSLATE(EEXIST),
587 EXT4_ERR_TRANSLATE(ERANGE),
588 EXT4_ERR_TRANSLATE(EOVERFLOW),
589 EXT4_ERR_TRANSLATE(EBUSY),
590 EXT4_ERR_TRANSLATE(ENOTDIR),
591 EXT4_ERR_TRANSLATE(ENOTEMPTY),
592 EXT4_ERR_TRANSLATE(ESHUTDOWN),
593 EXT4_ERR_TRANSLATE(EFAULT),
596 static int ext4_errno_to_code(int errno)
600 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
601 if (err_translation[i].errno == errno)
602 return err_translation[i].code;
603 return EXT4_ERR_UNKNOWN;
606 static void save_error_info(struct super_block *sb, int error,
607 __u32 ino, __u64 block,
608 const char *func, unsigned int line)
610 struct ext4_sb_info *sbi = EXT4_SB(sb);
612 /* We default to EFSCORRUPTED error... */
614 error = EFSCORRUPTED;
616 spin_lock(&sbi->s_error_lock);
617 sbi->s_add_error_count++;
618 sbi->s_last_error_code = error;
619 sbi->s_last_error_line = line;
620 sbi->s_last_error_ino = ino;
621 sbi->s_last_error_block = block;
622 sbi->s_last_error_func = func;
623 sbi->s_last_error_time = ktime_get_real_seconds();
624 if (!sbi->s_first_error_time) {
625 sbi->s_first_error_code = error;
626 sbi->s_first_error_line = line;
627 sbi->s_first_error_ino = ino;
628 sbi->s_first_error_block = block;
629 sbi->s_first_error_func = func;
630 sbi->s_first_error_time = sbi->s_last_error_time;
632 spin_unlock(&sbi->s_error_lock);
635 /* Deal with the reporting of failure conditions on a filesystem such as
636 * inconsistencies detected or read IO failures.
638 * On ext2, we can store the error state of the filesystem in the
639 * superblock. That is not possible on ext4, because we may have other
640 * write ordering constraints on the superblock which prevent us from
641 * writing it out straight away; and given that the journal is about to
642 * be aborted, we can't rely on the current, or future, transactions to
643 * write out the superblock safely.
645 * We'll just use the jbd2_journal_abort() error code to record an error in
646 * the journal instead. On recovery, the journal will complain about
647 * that error until we've noted it down and cleared it.
649 * If force_ro is set, we unconditionally force the filesystem into an
650 * ABORT|READONLY state, unless the error response on the fs has been set to
651 * panic in which case we take the easy way out and panic immediately. This is
652 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
653 * at a critical moment in log management.
655 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
656 __u32 ino, __u64 block,
657 const char *func, unsigned int line)
659 journal_t *journal = EXT4_SB(sb)->s_journal;
660 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
662 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
663 if (test_opt(sb, WARN_ON_ERROR))
666 if (!continue_fs && !sb_rdonly(sb)) {
667 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
669 jbd2_journal_abort(journal, -EIO);
672 if (!bdev_read_only(sb->s_bdev)) {
673 save_error_info(sb, error, ino, block, func, line);
675 * In case the fs should keep running, we need to writeout
676 * superblock through the journal. Due to lock ordering
677 * constraints, it may not be safe to do it right here so we
678 * defer superblock flushing to a workqueue.
680 if (continue_fs && journal)
681 schedule_work(&EXT4_SB(sb)->s_error_work);
683 ext4_commit_super(sb);
687 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
688 * could panic during 'reboot -f' as the underlying device got already
691 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
692 panic("EXT4-fs (device %s): panic forced after error\n",
696 if (sb_rdonly(sb) || continue_fs)
699 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
701 * Make sure updated value of ->s_mount_flags will be visible before
705 sb->s_flags |= SB_RDONLY;
708 static void flush_stashed_error_work(struct work_struct *work)
710 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
712 journal_t *journal = sbi->s_journal;
716 * If the journal is still running, we have to write out superblock
717 * through the journal to avoid collisions of other journalled sb
720 * We use directly jbd2 functions here to avoid recursing back into
721 * ext4 error handling code during handling of previous errors.
723 if (!sb_rdonly(sbi->s_sb) && journal) {
724 struct buffer_head *sbh = sbi->s_sbh;
725 handle = jbd2_journal_start(journal, 1);
728 if (jbd2_journal_get_write_access(handle, sbh)) {
729 jbd2_journal_stop(handle);
732 ext4_update_super(sbi->s_sb);
733 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
734 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
735 "superblock detected");
736 clear_buffer_write_io_error(sbh);
737 set_buffer_uptodate(sbh);
740 if (jbd2_journal_dirty_metadata(handle, sbh)) {
741 jbd2_journal_stop(handle);
744 jbd2_journal_stop(handle);
745 ext4_notify_error_sysfs(sbi);
750 * Write through journal failed. Write sb directly to get error info
751 * out and hope for the best.
753 ext4_commit_super(sbi->s_sb);
754 ext4_notify_error_sysfs(sbi);
757 #define ext4_error_ratelimit(sb) \
758 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
761 void __ext4_error(struct super_block *sb, const char *function,
762 unsigned int line, bool force_ro, int error, __u64 block,
763 const char *fmt, ...)
765 struct va_format vaf;
768 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
771 trace_ext4_error(sb, function, line);
772 if (ext4_error_ratelimit(sb)) {
777 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
778 sb->s_id, function, line, current->comm, &vaf);
781 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
783 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
786 void __ext4_error_inode(struct inode *inode, const char *function,
787 unsigned int line, ext4_fsblk_t block, int error,
788 const char *fmt, ...)
791 struct va_format vaf;
793 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
796 trace_ext4_error(inode->i_sb, function, line);
797 if (ext4_error_ratelimit(inode->i_sb)) {
802 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
803 "inode #%lu: block %llu: comm %s: %pV\n",
804 inode->i_sb->s_id, function, line, inode->i_ino,
805 block, current->comm, &vaf);
807 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
808 "inode #%lu: comm %s: %pV\n",
809 inode->i_sb->s_id, function, line, inode->i_ino,
810 current->comm, &vaf);
813 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
815 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
819 void __ext4_error_file(struct file *file, const char *function,
820 unsigned int line, ext4_fsblk_t block,
821 const char *fmt, ...)
824 struct va_format vaf;
825 struct inode *inode = file_inode(file);
826 char pathname[80], *path;
828 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
831 trace_ext4_error(inode->i_sb, function, line);
832 if (ext4_error_ratelimit(inode->i_sb)) {
833 path = file_path(file, pathname, sizeof(pathname));
841 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
842 "block %llu: comm %s: path %s: %pV\n",
843 inode->i_sb->s_id, function, line, inode->i_ino,
844 block, current->comm, path, &vaf);
847 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
848 "comm %s: path %s: %pV\n",
849 inode->i_sb->s_id, function, line, inode->i_ino,
850 current->comm, path, &vaf);
853 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
855 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
859 const char *ext4_decode_error(struct super_block *sb, int errno,
866 errstr = "Corrupt filesystem";
869 errstr = "Filesystem failed CRC";
872 errstr = "IO failure";
875 errstr = "Out of memory";
878 if (!sb || (EXT4_SB(sb)->s_journal &&
879 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
880 errstr = "Journal has aborted";
882 errstr = "Readonly filesystem";
885 /* If the caller passed in an extra buffer for unknown
886 * errors, textualise them now. Else we just return
889 /* Check for truncated error codes... */
890 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
899 /* __ext4_std_error decodes expected errors from journaling functions
900 * automatically and invokes the appropriate error response. */
902 void __ext4_std_error(struct super_block *sb, const char *function,
903 unsigned int line, int errno)
908 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
911 /* Special case: if the error is EROFS, and we're not already
912 * inside a transaction, then there's really no point in logging
914 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
917 if (ext4_error_ratelimit(sb)) {
918 errstr = ext4_decode_error(sb, errno, nbuf);
919 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
920 sb->s_id, function, line, errstr);
922 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
924 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
927 void __ext4_msg(struct super_block *sb,
928 const char *prefix, const char *fmt, ...)
930 struct va_format vaf;
934 atomic_inc(&EXT4_SB(sb)->s_msg_count);
935 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
944 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
946 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
950 static int ext4_warning_ratelimit(struct super_block *sb)
952 atomic_inc(&EXT4_SB(sb)->s_warning_count);
953 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
957 void __ext4_warning(struct super_block *sb, const char *function,
958 unsigned int line, const char *fmt, ...)
960 struct va_format vaf;
963 if (!ext4_warning_ratelimit(sb))
969 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
970 sb->s_id, function, line, &vaf);
974 void __ext4_warning_inode(const struct inode *inode, const char *function,
975 unsigned int line, const char *fmt, ...)
977 struct va_format vaf;
980 if (!ext4_warning_ratelimit(inode->i_sb))
986 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
987 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
988 function, line, inode->i_ino, current->comm, &vaf);
992 void __ext4_grp_locked_error(const char *function, unsigned int line,
993 struct super_block *sb, ext4_group_t grp,
994 unsigned long ino, ext4_fsblk_t block,
995 const char *fmt, ...)
999 struct va_format vaf;
1002 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
1005 trace_ext4_error(sb, function, line);
1006 if (ext4_error_ratelimit(sb)) {
1007 va_start(args, fmt);
1010 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1011 sb->s_id, function, line, grp);
1013 printk(KERN_CONT "inode %lu: ", ino);
1015 printk(KERN_CONT "block %llu:",
1016 (unsigned long long) block);
1017 printk(KERN_CONT "%pV\n", &vaf);
1021 if (test_opt(sb, ERRORS_CONT)) {
1022 if (test_opt(sb, WARN_ON_ERROR))
1024 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1025 if (!bdev_read_only(sb->s_bdev)) {
1026 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1028 schedule_work(&EXT4_SB(sb)->s_error_work);
1032 ext4_unlock_group(sb, grp);
1033 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1035 * We only get here in the ERRORS_RO case; relocking the group
1036 * may be dangerous, but nothing bad will happen since the
1037 * filesystem will have already been marked read/only and the
1038 * journal has been aborted. We return 1 as a hint to callers
1039 * who might what to use the return value from
1040 * ext4_grp_locked_error() to distinguish between the
1041 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1042 * aggressively from the ext4 function in question, with a
1043 * more appropriate error code.
1045 ext4_lock_group(sb, grp);
1049 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1053 struct ext4_sb_info *sbi = EXT4_SB(sb);
1054 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1055 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1058 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1059 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1062 percpu_counter_sub(&sbi->s_freeclusters_counter,
1066 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1067 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1072 count = ext4_free_inodes_count(sb, gdp);
1073 percpu_counter_sub(&sbi->s_freeinodes_counter,
1079 void ext4_update_dynamic_rev(struct super_block *sb)
1081 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1083 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1087 "updating to rev %d because of new feature flag, "
1088 "running e2fsck is recommended",
1091 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1092 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1093 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1094 /* leave es->s_feature_*compat flags alone */
1095 /* es->s_uuid will be set by e2fsck if empty */
1098 * The rest of the superblock fields should be zero, and if not it
1099 * means they are likely already in use, so leave them alone. We
1100 * can leave it up to e2fsck to clean up any inconsistencies there.
1105 * Open the external journal device
1107 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1109 struct block_device *bdev;
1111 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1117 ext4_msg(sb, KERN_ERR,
1118 "failed to open journal device unknown-block(%u,%u) %ld",
1119 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1124 * Release the journal device
1126 static void ext4_blkdev_put(struct block_device *bdev)
1128 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1131 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1133 struct block_device *bdev;
1134 bdev = sbi->s_journal_bdev;
1136 ext4_blkdev_put(bdev);
1137 sbi->s_journal_bdev = NULL;
1141 static inline struct inode *orphan_list_entry(struct list_head *l)
1143 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1146 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1148 struct list_head *l;
1150 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1151 le32_to_cpu(sbi->s_es->s_last_orphan));
1153 printk(KERN_ERR "sb_info orphan list:\n");
1154 list_for_each(l, &sbi->s_orphan) {
1155 struct inode *inode = orphan_list_entry(l);
1157 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1158 inode->i_sb->s_id, inode->i_ino, inode,
1159 inode->i_mode, inode->i_nlink,
1160 NEXT_ORPHAN(inode));
1165 static int ext4_quota_off(struct super_block *sb, int type);
1167 static inline void ext4_quota_off_umount(struct super_block *sb)
1171 /* Use our quota_off function to clear inode flags etc. */
1172 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1173 ext4_quota_off(sb, type);
1177 * This is a helper function which is used in the mount/remount
1178 * codepaths (which holds s_umount) to fetch the quota file name.
1180 static inline char *get_qf_name(struct super_block *sb,
1181 struct ext4_sb_info *sbi,
1184 return rcu_dereference_protected(sbi->s_qf_names[type],
1185 lockdep_is_held(&sb->s_umount));
1188 static inline void ext4_quota_off_umount(struct super_block *sb)
1193 static void ext4_put_super(struct super_block *sb)
1195 struct ext4_sb_info *sbi = EXT4_SB(sb);
1196 struct ext4_super_block *es = sbi->s_es;
1197 struct buffer_head **group_desc;
1198 struct flex_groups **flex_groups;
1203 * Unregister sysfs before destroying jbd2 journal.
1204 * Since we could still access attr_journal_task attribute via sysfs
1205 * path which could have sbi->s_journal->j_task as NULL
1206 * Unregister sysfs before flush sbi->s_error_work.
1207 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1208 * read metadata verify failed then will queue error work.
1209 * flush_stashed_error_work will call start_this_handle may trigger
1212 ext4_unregister_sysfs(sb);
1214 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1215 ext4_msg(sb, KERN_INFO, "unmounting filesystem.");
1217 ext4_unregister_li_request(sb);
1218 ext4_quota_off_umount(sb);
1220 flush_work(&sbi->s_error_work);
1221 destroy_workqueue(sbi->rsv_conversion_wq);
1222 ext4_release_orphan_info(sb);
1224 if (sbi->s_journal) {
1225 aborted = is_journal_aborted(sbi->s_journal);
1226 err = jbd2_journal_destroy(sbi->s_journal);
1227 sbi->s_journal = NULL;
1228 if ((err < 0) && !aborted) {
1229 ext4_abort(sb, -err, "Couldn't clean up the journal");
1233 ext4_es_unregister_shrinker(sbi);
1234 del_timer_sync(&sbi->s_err_report);
1235 ext4_release_system_zone(sb);
1236 ext4_mb_release(sb);
1237 ext4_ext_release(sb);
1239 if (!sb_rdonly(sb) && !aborted) {
1240 ext4_clear_feature_journal_needs_recovery(sb);
1241 ext4_clear_feature_orphan_present(sb);
1242 es->s_state = cpu_to_le16(sbi->s_mount_state);
1245 ext4_commit_super(sb);
1248 group_desc = rcu_dereference(sbi->s_group_desc);
1249 for (i = 0; i < sbi->s_gdb_count; i++)
1250 brelse(group_desc[i]);
1252 flex_groups = rcu_dereference(sbi->s_flex_groups);
1254 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1255 kvfree(flex_groups[i]);
1256 kvfree(flex_groups);
1259 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1260 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1261 percpu_counter_destroy(&sbi->s_dirs_counter);
1262 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1263 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1264 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1266 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1267 kfree(get_qf_name(sb, sbi, i));
1270 /* Debugging code just in case the in-memory inode orphan list
1271 * isn't empty. The on-disk one can be non-empty if we've
1272 * detected an error and taken the fs readonly, but the
1273 * in-memory list had better be clean by this point. */
1274 if (!list_empty(&sbi->s_orphan))
1275 dump_orphan_list(sb, sbi);
1276 ASSERT(list_empty(&sbi->s_orphan));
1278 sync_blockdev(sb->s_bdev);
1279 invalidate_bdev(sb->s_bdev);
1280 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1282 * Invalidate the journal device's buffers. We don't want them
1283 * floating about in memory - the physical journal device may
1284 * hotswapped, and it breaks the `ro-after' testing code.
1286 sync_blockdev(sbi->s_journal_bdev);
1287 invalidate_bdev(sbi->s_journal_bdev);
1288 ext4_blkdev_remove(sbi);
1291 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1292 sbi->s_ea_inode_cache = NULL;
1294 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1295 sbi->s_ea_block_cache = NULL;
1297 ext4_stop_mmpd(sbi);
1300 sb->s_fs_info = NULL;
1302 * Now that we are completely done shutting down the
1303 * superblock, we need to actually destroy the kobject.
1305 kobject_put(&sbi->s_kobj);
1306 wait_for_completion(&sbi->s_kobj_unregister);
1307 if (sbi->s_chksum_driver)
1308 crypto_free_shash(sbi->s_chksum_driver);
1309 kfree(sbi->s_blockgroup_lock);
1310 fs_put_dax(sbi->s_daxdev);
1311 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1312 #if IS_ENABLED(CONFIG_UNICODE)
1313 utf8_unload(sb->s_encoding);
1318 static struct kmem_cache *ext4_inode_cachep;
1321 * Called inside transaction, so use GFP_NOFS
1323 static struct inode *ext4_alloc_inode(struct super_block *sb)
1325 struct ext4_inode_info *ei;
1327 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1331 inode_set_iversion(&ei->vfs_inode, 1);
1332 spin_lock_init(&ei->i_raw_lock);
1333 INIT_LIST_HEAD(&ei->i_prealloc_list);
1334 atomic_set(&ei->i_prealloc_active, 0);
1335 spin_lock_init(&ei->i_prealloc_lock);
1336 ext4_es_init_tree(&ei->i_es_tree);
1337 rwlock_init(&ei->i_es_lock);
1338 INIT_LIST_HEAD(&ei->i_es_list);
1339 ei->i_es_all_nr = 0;
1340 ei->i_es_shk_nr = 0;
1341 ei->i_es_shrink_lblk = 0;
1342 ei->i_reserved_data_blocks = 0;
1343 spin_lock_init(&(ei->i_block_reservation_lock));
1344 ext4_init_pending_tree(&ei->i_pending_tree);
1346 ei->i_reserved_quota = 0;
1347 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1350 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1351 spin_lock_init(&ei->i_completed_io_lock);
1353 ei->i_datasync_tid = 0;
1354 atomic_set(&ei->i_unwritten, 0);
1355 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1356 ext4_fc_init_inode(&ei->vfs_inode);
1357 mutex_init(&ei->i_fc_lock);
1358 return &ei->vfs_inode;
1361 static int ext4_drop_inode(struct inode *inode)
1363 int drop = generic_drop_inode(inode);
1366 drop = fscrypt_drop_inode(inode);
1368 trace_ext4_drop_inode(inode, drop);
1372 static void ext4_free_in_core_inode(struct inode *inode)
1374 fscrypt_free_inode(inode);
1375 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1376 pr_warn("%s: inode %ld still in fc list",
1377 __func__, inode->i_ino);
1379 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1382 static void ext4_destroy_inode(struct inode *inode)
1384 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1385 ext4_msg(inode->i_sb, KERN_ERR,
1386 "Inode %lu (%p): orphan list check failed!",
1387 inode->i_ino, EXT4_I(inode));
1388 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1389 EXT4_I(inode), sizeof(struct ext4_inode_info),
1394 if (EXT4_I(inode)->i_reserved_data_blocks)
1395 ext4_msg(inode->i_sb, KERN_ERR,
1396 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1397 inode->i_ino, EXT4_I(inode),
1398 EXT4_I(inode)->i_reserved_data_blocks);
1401 static void init_once(void *foo)
1403 struct ext4_inode_info *ei = foo;
1405 INIT_LIST_HEAD(&ei->i_orphan);
1406 init_rwsem(&ei->xattr_sem);
1407 init_rwsem(&ei->i_data_sem);
1408 inode_init_once(&ei->vfs_inode);
1409 ext4_fc_init_inode(&ei->vfs_inode);
1412 static int __init init_inodecache(void)
1414 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1415 sizeof(struct ext4_inode_info), 0,
1416 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1418 offsetof(struct ext4_inode_info, i_data),
1419 sizeof_field(struct ext4_inode_info, i_data),
1421 if (ext4_inode_cachep == NULL)
1426 static void destroy_inodecache(void)
1429 * Make sure all delayed rcu free inodes are flushed before we
1433 kmem_cache_destroy(ext4_inode_cachep);
1436 void ext4_clear_inode(struct inode *inode)
1439 invalidate_inode_buffers(inode);
1441 ext4_discard_preallocations(inode, 0);
1442 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1444 if (EXT4_I(inode)->jinode) {
1445 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1446 EXT4_I(inode)->jinode);
1447 jbd2_free_inode(EXT4_I(inode)->jinode);
1448 EXT4_I(inode)->jinode = NULL;
1450 fscrypt_put_encryption_info(inode);
1451 fsverity_cleanup_inode(inode);
1454 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1455 u64 ino, u32 generation)
1457 struct inode *inode;
1460 * Currently we don't know the generation for parent directory, so
1461 * a generation of 0 means "accept any"
1463 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1465 return ERR_CAST(inode);
1466 if (generation && inode->i_generation != generation) {
1468 return ERR_PTR(-ESTALE);
1474 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1475 int fh_len, int fh_type)
1477 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1478 ext4_nfs_get_inode);
1481 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1482 int fh_len, int fh_type)
1484 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1485 ext4_nfs_get_inode);
1488 static int ext4_nfs_commit_metadata(struct inode *inode)
1490 struct writeback_control wbc = {
1491 .sync_mode = WB_SYNC_ALL
1494 trace_ext4_nfs_commit_metadata(inode);
1495 return ext4_write_inode(inode, &wbc);
1499 static const char * const quotatypes[] = INITQFNAMES;
1500 #define QTYPE2NAME(t) (quotatypes[t])
1502 static int ext4_write_dquot(struct dquot *dquot);
1503 static int ext4_acquire_dquot(struct dquot *dquot);
1504 static int ext4_release_dquot(struct dquot *dquot);
1505 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1506 static int ext4_write_info(struct super_block *sb, int type);
1507 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1508 const struct path *path);
1509 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1510 size_t len, loff_t off);
1511 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1512 const char *data, size_t len, loff_t off);
1513 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1514 unsigned int flags);
1516 static struct dquot **ext4_get_dquots(struct inode *inode)
1518 return EXT4_I(inode)->i_dquot;
1521 static const struct dquot_operations ext4_quota_operations = {
1522 .get_reserved_space = ext4_get_reserved_space,
1523 .write_dquot = ext4_write_dquot,
1524 .acquire_dquot = ext4_acquire_dquot,
1525 .release_dquot = ext4_release_dquot,
1526 .mark_dirty = ext4_mark_dquot_dirty,
1527 .write_info = ext4_write_info,
1528 .alloc_dquot = dquot_alloc,
1529 .destroy_dquot = dquot_destroy,
1530 .get_projid = ext4_get_projid,
1531 .get_inode_usage = ext4_get_inode_usage,
1532 .get_next_id = dquot_get_next_id,
1535 static const struct quotactl_ops ext4_qctl_operations = {
1536 .quota_on = ext4_quota_on,
1537 .quota_off = ext4_quota_off,
1538 .quota_sync = dquot_quota_sync,
1539 .get_state = dquot_get_state,
1540 .set_info = dquot_set_dqinfo,
1541 .get_dqblk = dquot_get_dqblk,
1542 .set_dqblk = dquot_set_dqblk,
1543 .get_nextdqblk = dquot_get_next_dqblk,
1547 static const struct super_operations ext4_sops = {
1548 .alloc_inode = ext4_alloc_inode,
1549 .free_inode = ext4_free_in_core_inode,
1550 .destroy_inode = ext4_destroy_inode,
1551 .write_inode = ext4_write_inode,
1552 .dirty_inode = ext4_dirty_inode,
1553 .drop_inode = ext4_drop_inode,
1554 .evict_inode = ext4_evict_inode,
1555 .put_super = ext4_put_super,
1556 .sync_fs = ext4_sync_fs,
1557 .freeze_fs = ext4_freeze,
1558 .unfreeze_fs = ext4_unfreeze,
1559 .statfs = ext4_statfs,
1560 .show_options = ext4_show_options,
1562 .quota_read = ext4_quota_read,
1563 .quota_write = ext4_quota_write,
1564 .get_dquots = ext4_get_dquots,
1568 static const struct export_operations ext4_export_ops = {
1569 .fh_to_dentry = ext4_fh_to_dentry,
1570 .fh_to_parent = ext4_fh_to_parent,
1571 .get_parent = ext4_get_parent,
1572 .commit_metadata = ext4_nfs_commit_metadata,
1576 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1577 Opt_resgid, Opt_resuid, Opt_sb,
1578 Opt_nouid32, Opt_debug, Opt_removed,
1579 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1580 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1581 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1582 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1583 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1584 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1586 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1587 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1588 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1589 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1590 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1591 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1592 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1593 Opt_inode_readahead_blks, Opt_journal_ioprio,
1594 Opt_dioread_nolock, Opt_dioread_lock,
1595 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1596 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1597 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1598 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1599 #ifdef CONFIG_EXT4_DEBUG
1600 Opt_fc_debug_max_replay, Opt_fc_debug_force
1604 static const struct constant_table ext4_param_errors[] = {
1605 {"continue", EXT4_MOUNT_ERRORS_CONT},
1606 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1607 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1611 static const struct constant_table ext4_param_data[] = {
1612 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1613 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1614 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1618 static const struct constant_table ext4_param_data_err[] = {
1619 {"abort", Opt_data_err_abort},
1620 {"ignore", Opt_data_err_ignore},
1624 static const struct constant_table ext4_param_jqfmt[] = {
1625 {"vfsold", QFMT_VFS_OLD},
1626 {"vfsv0", QFMT_VFS_V0},
1627 {"vfsv1", QFMT_VFS_V1},
1631 static const struct constant_table ext4_param_dax[] = {
1632 {"always", Opt_dax_always},
1633 {"inode", Opt_dax_inode},
1634 {"never", Opt_dax_never},
1638 /* String parameter that allows empty argument */
1639 #define fsparam_string_empty(NAME, OPT) \
1640 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1643 * Mount option specification
1644 * We don't use fsparam_flag_no because of the way we set the
1645 * options and the way we show them in _ext4_show_options(). To
1646 * keep the changes to a minimum, let's keep the negative options
1649 static const struct fs_parameter_spec ext4_param_specs[] = {
1650 fsparam_flag ("bsddf", Opt_bsd_df),
1651 fsparam_flag ("minixdf", Opt_minix_df),
1652 fsparam_flag ("grpid", Opt_grpid),
1653 fsparam_flag ("bsdgroups", Opt_grpid),
1654 fsparam_flag ("nogrpid", Opt_nogrpid),
1655 fsparam_flag ("sysvgroups", Opt_nogrpid),
1656 fsparam_u32 ("resgid", Opt_resgid),
1657 fsparam_u32 ("resuid", Opt_resuid),
1658 fsparam_u32 ("sb", Opt_sb),
1659 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1660 fsparam_flag ("nouid32", Opt_nouid32),
1661 fsparam_flag ("debug", Opt_debug),
1662 fsparam_flag ("oldalloc", Opt_removed),
1663 fsparam_flag ("orlov", Opt_removed),
1664 fsparam_flag ("user_xattr", Opt_user_xattr),
1665 fsparam_flag ("nouser_xattr", Opt_nouser_xattr),
1666 fsparam_flag ("acl", Opt_acl),
1667 fsparam_flag ("noacl", Opt_noacl),
1668 fsparam_flag ("norecovery", Opt_noload),
1669 fsparam_flag ("noload", Opt_noload),
1670 fsparam_flag ("bh", Opt_removed),
1671 fsparam_flag ("nobh", Opt_removed),
1672 fsparam_u32 ("commit", Opt_commit),
1673 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1674 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1675 fsparam_u32 ("journal_dev", Opt_journal_dev),
1676 fsparam_bdev ("journal_path", Opt_journal_path),
1677 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1678 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1679 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1680 fsparam_flag ("abort", Opt_abort),
1681 fsparam_enum ("data", Opt_data, ext4_param_data),
1682 fsparam_enum ("data_err", Opt_data_err,
1683 ext4_param_data_err),
1684 fsparam_string_empty
1685 ("usrjquota", Opt_usrjquota),
1686 fsparam_string_empty
1687 ("grpjquota", Opt_grpjquota),
1688 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1689 fsparam_flag ("grpquota", Opt_grpquota),
1690 fsparam_flag ("quota", Opt_quota),
1691 fsparam_flag ("noquota", Opt_noquota),
1692 fsparam_flag ("usrquota", Opt_usrquota),
1693 fsparam_flag ("prjquota", Opt_prjquota),
1694 fsparam_flag ("barrier", Opt_barrier),
1695 fsparam_u32 ("barrier", Opt_barrier),
1696 fsparam_flag ("nobarrier", Opt_nobarrier),
1697 fsparam_flag ("i_version", Opt_i_version),
1698 fsparam_flag ("dax", Opt_dax),
1699 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1700 fsparam_u32 ("stripe", Opt_stripe),
1701 fsparam_flag ("delalloc", Opt_delalloc),
1702 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1703 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1704 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1705 fsparam_u32 ("debug_want_extra_isize",
1706 Opt_debug_want_extra_isize),
1707 fsparam_flag ("mblk_io_submit", Opt_removed),
1708 fsparam_flag ("nomblk_io_submit", Opt_removed),
1709 fsparam_flag ("block_validity", Opt_block_validity),
1710 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1711 fsparam_u32 ("inode_readahead_blks",
1712 Opt_inode_readahead_blks),
1713 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1714 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1715 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1716 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1717 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1718 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1719 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1720 fsparam_flag ("discard", Opt_discard),
1721 fsparam_flag ("nodiscard", Opt_nodiscard),
1722 fsparam_u32 ("init_itable", Opt_init_itable),
1723 fsparam_flag ("init_itable", Opt_init_itable),
1724 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1725 #ifdef CONFIG_EXT4_DEBUG
1726 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1727 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1729 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1730 fsparam_flag ("test_dummy_encryption",
1731 Opt_test_dummy_encryption),
1732 fsparam_string ("test_dummy_encryption",
1733 Opt_test_dummy_encryption),
1734 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1735 fsparam_flag ("nombcache", Opt_nombcache),
1736 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1737 fsparam_flag ("prefetch_block_bitmaps",
1739 fsparam_flag ("no_prefetch_block_bitmaps",
1740 Opt_no_prefetch_block_bitmaps),
1741 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1742 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1743 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1744 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1745 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1746 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1750 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1752 static const char deprecated_msg[] =
1753 "Mount option \"%s\" will be removed by %s\n"
1754 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1756 #define MOPT_SET 0x0001
1757 #define MOPT_CLEAR 0x0002
1758 #define MOPT_NOSUPPORT 0x0004
1759 #define MOPT_EXPLICIT 0x0008
1762 #define MOPT_QFMT 0x0010
1764 #define MOPT_Q MOPT_NOSUPPORT
1765 #define MOPT_QFMT MOPT_NOSUPPORT
1767 #define MOPT_NO_EXT2 0x0020
1768 #define MOPT_NO_EXT3 0x0040
1769 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1770 #define MOPT_SKIP 0x0080
1771 #define MOPT_2 0x0100
1773 static const struct mount_opts {
1777 } ext4_mount_opts[] = {
1778 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1779 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1780 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1781 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1782 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1783 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1784 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1785 MOPT_EXT4_ONLY | MOPT_SET},
1786 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1787 MOPT_EXT4_ONLY | MOPT_CLEAR},
1788 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1789 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1790 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1791 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1792 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1793 MOPT_EXT4_ONLY | MOPT_CLEAR},
1794 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1795 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1796 {Opt_commit, 0, MOPT_NO_EXT2},
1797 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1798 MOPT_EXT4_ONLY | MOPT_CLEAR},
1799 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1800 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1801 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1802 EXT4_MOUNT_JOURNAL_CHECKSUM),
1803 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1804 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1805 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1806 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1807 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1808 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1809 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1810 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1811 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1812 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1813 {Opt_journal_path, 0, MOPT_NO_EXT2},
1814 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1815 {Opt_data, 0, MOPT_NO_EXT2},
1816 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1817 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1818 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1819 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1820 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1822 {Opt_acl, 0, MOPT_NOSUPPORT},
1823 {Opt_noacl, 0, MOPT_NOSUPPORT},
1825 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1826 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1827 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1828 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1830 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1832 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1834 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1835 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1836 MOPT_CLEAR | MOPT_Q},
1837 {Opt_usrjquota, 0, MOPT_Q},
1838 {Opt_grpjquota, 0, MOPT_Q},
1839 {Opt_jqfmt, 0, MOPT_QFMT},
1840 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1841 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1843 #ifdef CONFIG_EXT4_DEBUG
1844 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1845 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1850 #if IS_ENABLED(CONFIG_UNICODE)
1851 static const struct ext4_sb_encodings {
1854 unsigned int version;
1855 } ext4_sb_encoding_map[] = {
1856 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1859 static const struct ext4_sb_encodings *
1860 ext4_sb_read_encoding(const struct ext4_super_block *es)
1862 __u16 magic = le16_to_cpu(es->s_encoding);
1865 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1866 if (magic == ext4_sb_encoding_map[i].magic)
1867 return &ext4_sb_encoding_map[i];
1873 #define EXT4_SPEC_JQUOTA (1 << 0)
1874 #define EXT4_SPEC_JQFMT (1 << 1)
1875 #define EXT4_SPEC_DATAJ (1 << 2)
1876 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1877 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1878 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1879 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1880 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1881 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1882 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1883 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1884 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1885 #define EXT4_SPEC_s_stripe (1 << 13)
1886 #define EXT4_SPEC_s_resuid (1 << 14)
1887 #define EXT4_SPEC_s_resgid (1 << 15)
1888 #define EXT4_SPEC_s_commit_interval (1 << 16)
1889 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1890 #define EXT4_SPEC_s_sb_block (1 << 18)
1891 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1893 struct ext4_fs_context {
1894 char *s_qf_names[EXT4_MAXQUOTAS];
1895 struct fscrypt_dummy_policy dummy_enc_policy;
1896 int s_jquota_fmt; /* Format of quota to use */
1897 #ifdef CONFIG_EXT4_DEBUG
1898 int s_fc_debug_max_replay;
1900 unsigned short qname_spec;
1901 unsigned long vals_s_flags; /* Bits to set in s_flags */
1902 unsigned long mask_s_flags; /* Bits changed in s_flags */
1903 unsigned long journal_devnum;
1904 unsigned long s_commit_interval;
1905 unsigned long s_stripe;
1906 unsigned int s_inode_readahead_blks;
1907 unsigned int s_want_extra_isize;
1908 unsigned int s_li_wait_mult;
1909 unsigned int s_max_dir_size_kb;
1910 unsigned int journal_ioprio;
1911 unsigned int vals_s_mount_opt;
1912 unsigned int mask_s_mount_opt;
1913 unsigned int vals_s_mount_opt2;
1914 unsigned int mask_s_mount_opt2;
1915 unsigned long vals_s_mount_flags;
1916 unsigned long mask_s_mount_flags;
1917 unsigned int opt_flags; /* MOPT flags */
1919 u32 s_max_batch_time;
1920 u32 s_min_batch_time;
1923 ext4_fsblk_t s_sb_block;
1926 static void ext4_fc_free(struct fs_context *fc)
1928 struct ext4_fs_context *ctx = fc->fs_private;
1934 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1935 kfree(ctx->s_qf_names[i]);
1937 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1941 int ext4_init_fs_context(struct fs_context *fc)
1943 struct ext4_fs_context *ctx;
1945 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1949 fc->fs_private = ctx;
1950 fc->ops = &ext4_context_ops;
1957 * Note the name of the specified quota file.
1959 static int note_qf_name(struct fs_context *fc, int qtype,
1960 struct fs_parameter *param)
1962 struct ext4_fs_context *ctx = fc->fs_private;
1965 if (param->size < 1) {
1966 ext4_msg(NULL, KERN_ERR, "Missing quota name");
1969 if (strchr(param->string, '/')) {
1970 ext4_msg(NULL, KERN_ERR,
1971 "quotafile must be on filesystem root");
1974 if (ctx->s_qf_names[qtype]) {
1975 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
1976 ext4_msg(NULL, KERN_ERR,
1977 "%s quota file already specified",
1984 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
1986 ext4_msg(NULL, KERN_ERR,
1987 "Not enough memory for storing quotafile name");
1990 ctx->s_qf_names[qtype] = qname;
1991 ctx->qname_spec |= 1 << qtype;
1992 ctx->spec |= EXT4_SPEC_JQUOTA;
1997 * Clear the name of the specified quota file.
1999 static int unnote_qf_name(struct fs_context *fc, int qtype)
2001 struct ext4_fs_context *ctx = fc->fs_private;
2003 if (ctx->s_qf_names[qtype])
2004 kfree(ctx->s_qf_names[qtype]);
2006 ctx->s_qf_names[qtype] = NULL;
2007 ctx->qname_spec |= 1 << qtype;
2008 ctx->spec |= EXT4_SPEC_JQUOTA;
2013 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2014 struct ext4_fs_context *ctx)
2018 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2019 ext4_msg(NULL, KERN_WARNING,
2020 "test_dummy_encryption option not supported");
2023 err = fscrypt_parse_test_dummy_encryption(param,
2024 &ctx->dummy_enc_policy);
2025 if (err == -EINVAL) {
2026 ext4_msg(NULL, KERN_WARNING,
2027 "Value of option \"%s\" is unrecognized", param->key);
2028 } else if (err == -EEXIST) {
2029 ext4_msg(NULL, KERN_WARNING,
2030 "Conflicting test_dummy_encryption options");
2036 #define EXT4_SET_CTX(name) \
2037 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2038 unsigned long flag) \
2040 ctx->mask_s_##name |= flag; \
2041 ctx->vals_s_##name |= flag; \
2044 #define EXT4_CLEAR_CTX(name) \
2045 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2046 unsigned long flag) \
2048 ctx->mask_s_##name |= flag; \
2049 ctx->vals_s_##name &= ~flag; \
2052 #define EXT4_TEST_CTX(name) \
2053 static inline unsigned long \
2054 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2056 return (ctx->vals_s_##name & flag); \
2059 EXT4_SET_CTX(flags); /* set only */
2060 EXT4_SET_CTX(mount_opt);
2061 EXT4_CLEAR_CTX(mount_opt);
2062 EXT4_TEST_CTX(mount_opt);
2063 EXT4_SET_CTX(mount_opt2);
2064 EXT4_CLEAR_CTX(mount_opt2);
2065 EXT4_TEST_CTX(mount_opt2);
2067 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2069 set_bit(bit, &ctx->mask_s_mount_flags);
2070 set_bit(bit, &ctx->vals_s_mount_flags);
2073 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2075 struct ext4_fs_context *ctx = fc->fs_private;
2076 struct fs_parse_result result;
2077 const struct mount_opts *m;
2083 token = fs_parse(fc, ext4_param_specs, param, &result);
2086 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2088 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2089 if (token == m->token)
2092 ctx->opt_flags |= m->flags;
2094 if (m->flags & MOPT_EXPLICIT) {
2095 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2096 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2097 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2098 ctx_set_mount_opt2(ctx,
2099 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2104 if (m->flags & MOPT_NOSUPPORT) {
2105 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2113 if (!*param->string)
2114 return unnote_qf_name(fc, USRQUOTA);
2116 return note_qf_name(fc, USRQUOTA, param);
2118 if (!*param->string)
2119 return unnote_qf_name(fc, GRPQUOTA);
2121 return note_qf_name(fc, GRPQUOTA, param);
2124 case Opt_nouser_xattr:
2125 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5");
2128 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2129 ext4_msg(NULL, KERN_WARNING,
2130 "Ignoring %s option on remount", param->key);
2132 ctx->s_sb_block = result.uint_32;
2133 ctx->spec |= EXT4_SPEC_s_sb_block;
2137 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2141 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2144 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
2145 ext4_msg(NULL, KERN_WARNING, "Use iversion instead\n");
2146 ctx_set_flags(ctx, SB_I_VERSION);
2148 case Opt_inlinecrypt:
2149 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2150 ctx_set_flags(ctx, SB_INLINECRYPT);
2152 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2156 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2157 ctx_set_mount_opt(ctx, result.uint_32);
2161 ctx->s_jquota_fmt = result.uint_32;
2162 ctx->spec |= EXT4_SPEC_JQFMT;
2166 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2167 ctx_set_mount_opt(ctx, result.uint_32);
2168 ctx->spec |= EXT4_SPEC_DATAJ;
2171 if (result.uint_32 == 0)
2172 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
2173 else if (result.uint_32 > INT_MAX / HZ) {
2174 ext4_msg(NULL, KERN_ERR,
2175 "Invalid commit interval %d, "
2176 "must be smaller than %d",
2177 result.uint_32, INT_MAX / HZ);
2180 ctx->s_commit_interval = HZ * result.uint_32;
2181 ctx->spec |= EXT4_SPEC_s_commit_interval;
2183 case Opt_debug_want_extra_isize:
2184 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2185 ext4_msg(NULL, KERN_ERR,
2186 "Invalid want_extra_isize %d", result.uint_32);
2189 ctx->s_want_extra_isize = result.uint_32;
2190 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2192 case Opt_max_batch_time:
2193 ctx->s_max_batch_time = result.uint_32;
2194 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2196 case Opt_min_batch_time:
2197 ctx->s_min_batch_time = result.uint_32;
2198 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2200 case Opt_inode_readahead_blks:
2201 if (result.uint_32 &&
2202 (result.uint_32 > (1 << 30) ||
2203 !is_power_of_2(result.uint_32))) {
2204 ext4_msg(NULL, KERN_ERR,
2205 "EXT4-fs: inode_readahead_blks must be "
2206 "0 or a power of 2 smaller than 2^31");
2209 ctx->s_inode_readahead_blks = result.uint_32;
2210 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2212 case Opt_init_itable:
2213 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2214 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2215 if (param->type == fs_value_is_string)
2216 ctx->s_li_wait_mult = result.uint_32;
2217 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2219 case Opt_max_dir_size_kb:
2220 ctx->s_max_dir_size_kb = result.uint_32;
2221 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2223 #ifdef CONFIG_EXT4_DEBUG
2224 case Opt_fc_debug_max_replay:
2225 ctx->s_fc_debug_max_replay = result.uint_32;
2226 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2230 ctx->s_stripe = result.uint_32;
2231 ctx->spec |= EXT4_SPEC_s_stripe;
2234 uid = make_kuid(current_user_ns(), result.uint_32);
2235 if (!uid_valid(uid)) {
2236 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2240 ctx->s_resuid = uid;
2241 ctx->spec |= EXT4_SPEC_s_resuid;
2244 gid = make_kgid(current_user_ns(), result.uint_32);
2245 if (!gid_valid(gid)) {
2246 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2250 ctx->s_resgid = gid;
2251 ctx->spec |= EXT4_SPEC_s_resgid;
2253 case Opt_journal_dev:
2255 ext4_msg(NULL, KERN_ERR,
2256 "Cannot specify journal on remount");
2259 ctx->journal_devnum = result.uint_32;
2260 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2262 case Opt_journal_path:
2264 struct inode *journal_inode;
2269 ext4_msg(NULL, KERN_ERR,
2270 "Cannot specify journal on remount");
2274 error = fs_lookup_param(fc, param, 1, &path);
2276 ext4_msg(NULL, KERN_ERR, "error: could not find "
2277 "journal device path");
2281 journal_inode = d_inode(path.dentry);
2282 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2283 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2287 case Opt_journal_ioprio:
2288 if (result.uint_32 > 7) {
2289 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2293 ctx->journal_ioprio =
2294 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2295 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2297 case Opt_test_dummy_encryption:
2298 return ext4_parse_test_dummy_encryption(param, ctx);
2301 #ifdef CONFIG_FS_DAX
2303 int type = (token == Opt_dax) ?
2304 Opt_dax : result.uint_32;
2308 case Opt_dax_always:
2309 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2310 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2313 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2314 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2317 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2318 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2319 /* Strictly for printing options */
2320 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2326 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2330 if (result.uint_32 == Opt_data_err_abort)
2331 ctx_set_mount_opt(ctx, m->mount_opt);
2332 else if (result.uint_32 == Opt_data_err_ignore)
2333 ctx_clear_mount_opt(ctx, m->mount_opt);
2335 case Opt_mb_optimize_scan:
2336 if (result.int_32 == 1) {
2337 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2338 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2339 } else if (result.int_32 == 0) {
2340 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2341 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2343 ext4_msg(NULL, KERN_WARNING,
2344 "mb_optimize_scan should be set to 0 or 1.");
2351 * At this point we should only be getting options requiring MOPT_SET,
2352 * or MOPT_CLEAR. Anything else is a bug
2354 if (m->token == Opt_err) {
2355 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2362 unsigned int set = 0;
2364 if ((param->type == fs_value_is_flag) ||
2368 if (m->flags & MOPT_CLEAR)
2370 else if (unlikely(!(m->flags & MOPT_SET))) {
2371 ext4_msg(NULL, KERN_WARNING,
2372 "buggy handling of option %s",
2377 if (m->flags & MOPT_2) {
2379 ctx_set_mount_opt2(ctx, m->mount_opt);
2381 ctx_clear_mount_opt2(ctx, m->mount_opt);
2384 ctx_set_mount_opt(ctx, m->mount_opt);
2386 ctx_clear_mount_opt(ctx, m->mount_opt);
2393 static int parse_options(struct fs_context *fc, char *options)
2395 struct fs_parameter param;
2402 while ((key = strsep(&options, ",")) != NULL) {
2405 char *value = strchr(key, '=');
2407 param.type = fs_value_is_flag;
2408 param.string = NULL;
2415 v_len = strlen(value);
2416 param.string = kmemdup_nul(value, v_len,
2420 param.type = fs_value_is_string;
2426 ret = ext4_parse_param(fc, ¶m);
2428 kfree(param.string);
2434 ret = ext4_validate_options(fc);
2441 static int parse_apply_sb_mount_options(struct super_block *sb,
2442 struct ext4_fs_context *m_ctx)
2444 struct ext4_sb_info *sbi = EXT4_SB(sb);
2445 char *s_mount_opts = NULL;
2446 struct ext4_fs_context *s_ctx = NULL;
2447 struct fs_context *fc = NULL;
2450 if (!sbi->s_es->s_mount_opts[0])
2453 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2454 sizeof(sbi->s_es->s_mount_opts),
2459 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2463 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2467 fc->fs_private = s_ctx;
2468 fc->s_fs_info = sbi;
2470 ret = parse_options(fc, s_mount_opts);
2474 ret = ext4_check_opt_consistency(fc, sb);
2477 ext4_msg(sb, KERN_WARNING,
2478 "failed to parse options in superblock: %s",
2484 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2485 m_ctx->journal_devnum = s_ctx->journal_devnum;
2486 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2487 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2489 ext4_apply_options(fc, sb);
2497 kfree(s_mount_opts);
2501 static void ext4_apply_quota_options(struct fs_context *fc,
2502 struct super_block *sb)
2505 bool quota_feature = ext4_has_feature_quota(sb);
2506 struct ext4_fs_context *ctx = fc->fs_private;
2507 struct ext4_sb_info *sbi = EXT4_SB(sb);
2514 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2515 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2516 if (!(ctx->qname_spec & (1 << i)))
2519 qname = ctx->s_qf_names[i]; /* May be NULL */
2522 ctx->s_qf_names[i] = NULL;
2523 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2524 lockdep_is_held(&sb->s_umount));
2530 if (ctx->spec & EXT4_SPEC_JQFMT)
2531 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2536 * Check quota settings consistency.
2538 static int ext4_check_quota_consistency(struct fs_context *fc,
2539 struct super_block *sb)
2542 struct ext4_fs_context *ctx = fc->fs_private;
2543 struct ext4_sb_info *sbi = EXT4_SB(sb);
2544 bool quota_feature = ext4_has_feature_quota(sb);
2545 bool quota_loaded = sb_any_quota_loaded(sb);
2546 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2550 * We do the test below only for project quotas. 'usrquota' and
2551 * 'grpquota' mount options are allowed even without quota feature
2552 * to support legacy quotas in quota files.
2554 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2555 !ext4_has_feature_project(sb)) {
2556 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2557 "Cannot enable project quota enforcement.");
2561 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2562 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2564 ctx->mask_s_mount_opt & quota_flags &&
2565 !ctx_test_mount_opt(ctx, quota_flags))
2566 goto err_quota_change;
2568 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2570 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2571 if (!(ctx->qname_spec & (1 << i)))
2575 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2576 goto err_jquota_change;
2578 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2579 strcmp(get_qf_name(sb, sbi, i),
2580 ctx->s_qf_names[i]) != 0)
2581 goto err_jquota_specified;
2584 if (quota_feature) {
2585 ext4_msg(NULL, KERN_INFO,
2586 "Journaled quota options ignored when "
2587 "QUOTA feature is enabled");
2592 if (ctx->spec & EXT4_SPEC_JQFMT) {
2593 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2594 goto err_jquota_change;
2595 if (quota_feature) {
2596 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2597 "ignored when QUOTA feature is enabled");
2602 /* Make sure we don't mix old and new quota format */
2603 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2604 ctx->s_qf_names[USRQUOTA]);
2605 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2606 ctx->s_qf_names[GRPQUOTA]);
2608 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2609 test_opt(sb, USRQUOTA));
2611 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2612 test_opt(sb, GRPQUOTA));
2615 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2619 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2623 if (usr_qf_name || grp_qf_name) {
2624 if (usrquota || grpquota) {
2625 ext4_msg(NULL, KERN_ERR, "old and new quota "
2630 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2631 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2640 ext4_msg(NULL, KERN_ERR,
2641 "Cannot change quota options when quota turned on");
2644 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2645 "options when quota turned on");
2647 err_jquota_specified:
2648 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2656 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2657 struct super_block *sb)
2659 const struct ext4_fs_context *ctx = fc->fs_private;
2660 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2663 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2666 if (!ext4_has_feature_encrypt(sb)) {
2667 ext4_msg(NULL, KERN_WARNING,
2668 "test_dummy_encryption requires encrypt feature");
2672 * This mount option is just for testing, and it's not worthwhile to
2673 * implement the extra complexity (e.g. RCU protection) that would be
2674 * needed to allow it to be set or changed during remount. We do allow
2675 * it to be specified during remount, but only if there is no change.
2677 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2678 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2679 &ctx->dummy_enc_policy))
2681 ext4_msg(NULL, KERN_WARNING,
2682 "Can't set or change test_dummy_encryption on remount");
2685 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2686 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2687 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2688 &ctx->dummy_enc_policy))
2690 ext4_msg(NULL, KERN_WARNING,
2691 "Conflicting test_dummy_encryption options");
2695 * fscrypt_add_test_dummy_key() technically changes the super_block, so
2696 * technically it should be delayed until ext4_apply_options() like the
2697 * other changes. But since we never get here for remounts (see above),
2698 * and this is the last chance to report errors, we do it here.
2700 err = fscrypt_add_test_dummy_key(sb, &ctx->dummy_enc_policy);
2702 ext4_msg(NULL, KERN_WARNING,
2703 "Error adding test dummy encryption key [%d]", err);
2707 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2708 struct super_block *sb)
2710 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2711 /* if already set, it was already verified to be the same */
2712 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2714 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2715 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2716 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2719 static int ext4_check_opt_consistency(struct fs_context *fc,
2720 struct super_block *sb)
2722 struct ext4_fs_context *ctx = fc->fs_private;
2723 struct ext4_sb_info *sbi = fc->s_fs_info;
2724 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2727 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2728 ext4_msg(NULL, KERN_ERR,
2729 "Mount option(s) incompatible with ext2");
2732 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2733 ext4_msg(NULL, KERN_ERR,
2734 "Mount option(s) incompatible with ext3");
2738 if (ctx->s_want_extra_isize >
2739 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2740 ext4_msg(NULL, KERN_ERR,
2741 "Invalid want_extra_isize %d",
2742 ctx->s_want_extra_isize);
2746 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2748 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2749 if (blocksize < PAGE_SIZE)
2750 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2751 "experimental mount option 'dioread_nolock' "
2752 "for blocksize < PAGE_SIZE");
2755 err = ext4_check_test_dummy_encryption(fc, sb);
2759 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2760 if (!sbi->s_journal) {
2761 ext4_msg(NULL, KERN_WARNING,
2762 "Remounting file system with no journal "
2763 "so ignoring journalled data option");
2764 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2765 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2766 test_opt(sb, DATA_FLAGS)) {
2767 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2774 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2775 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2776 ext4_msg(NULL, KERN_ERR, "can't mount with "
2777 "both data=journal and dax");
2781 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2782 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2783 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2784 fail_dax_change_remount:
2785 ext4_msg(NULL, KERN_ERR, "can't change "
2786 "dax mount option while remounting");
2788 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2789 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2790 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2791 goto fail_dax_change_remount;
2792 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2793 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2794 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2795 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2796 goto fail_dax_change_remount;
2800 return ext4_check_quota_consistency(fc, sb);
2803 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2805 struct ext4_fs_context *ctx = fc->fs_private;
2806 struct ext4_sb_info *sbi = fc->s_fs_info;
2808 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2809 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2810 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2811 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2812 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2813 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2814 sb->s_flags &= ~ctx->mask_s_flags;
2815 sb->s_flags |= ctx->vals_s_flags;
2818 * i_version differs from common mount option iversion so we have
2819 * to let vfs know that it was set, otherwise it would get cleared
2822 if (ctx->mask_s_flags & SB_I_VERSION)
2823 fc->sb_flags |= SB_I_VERSION;
2825 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2826 APPLY(s_commit_interval);
2828 APPLY(s_max_batch_time);
2829 APPLY(s_min_batch_time);
2830 APPLY(s_want_extra_isize);
2831 APPLY(s_inode_readahead_blks);
2832 APPLY(s_max_dir_size_kb);
2833 APPLY(s_li_wait_mult);
2837 #ifdef CONFIG_EXT4_DEBUG
2838 APPLY(s_fc_debug_max_replay);
2841 ext4_apply_quota_options(fc, sb);
2842 ext4_apply_test_dummy_encryption(ctx, sb);
2846 static int ext4_validate_options(struct fs_context *fc)
2849 struct ext4_fs_context *ctx = fc->fs_private;
2850 char *usr_qf_name, *grp_qf_name;
2852 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2853 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2855 if (usr_qf_name || grp_qf_name) {
2856 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2857 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2859 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2860 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2862 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2863 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2864 ext4_msg(NULL, KERN_ERR, "old and new quota "
2873 static inline void ext4_show_quota_options(struct seq_file *seq,
2874 struct super_block *sb)
2876 #if defined(CONFIG_QUOTA)
2877 struct ext4_sb_info *sbi = EXT4_SB(sb);
2878 char *usr_qf_name, *grp_qf_name;
2880 if (sbi->s_jquota_fmt) {
2883 switch (sbi->s_jquota_fmt) {
2894 seq_printf(seq, ",jqfmt=%s", fmtname);
2898 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2899 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2901 seq_show_option(seq, "usrjquota", usr_qf_name);
2903 seq_show_option(seq, "grpjquota", grp_qf_name);
2908 static const char *token2str(int token)
2910 const struct fs_parameter_spec *spec;
2912 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2913 if (spec->opt == token && !spec->type)
2920 * - it's set to a non-default value OR
2921 * - if the per-sb default is different from the global default
2923 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2926 struct ext4_sb_info *sbi = EXT4_SB(sb);
2927 struct ext4_super_block *es = sbi->s_es;
2928 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2929 const struct mount_opts *m;
2930 char sep = nodefs ? '\n' : ',';
2932 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2933 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2935 if (sbi->s_sb_block != 1)
2936 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2938 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2939 int want_set = m->flags & MOPT_SET;
2940 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2941 m->flags & MOPT_SKIP)
2943 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2944 continue; /* skip if same as the default */
2946 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2947 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2948 continue; /* select Opt_noFoo vs Opt_Foo */
2949 SEQ_OPTS_PRINT("%s", token2str(m->token));
2952 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2953 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2954 SEQ_OPTS_PRINT("resuid=%u",
2955 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2956 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2957 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2958 SEQ_OPTS_PRINT("resgid=%u",
2959 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2960 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2961 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2962 SEQ_OPTS_PUTS("errors=remount-ro");
2963 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2964 SEQ_OPTS_PUTS("errors=continue");
2965 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2966 SEQ_OPTS_PUTS("errors=panic");
2967 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2968 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2969 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2970 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2971 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2972 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2973 if (sb->s_flags & SB_I_VERSION)
2974 SEQ_OPTS_PUTS("i_version");
2975 if (nodefs || sbi->s_stripe)
2976 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2977 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2978 (sbi->s_mount_opt ^ def_mount_opt)) {
2979 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2980 SEQ_OPTS_PUTS("data=journal");
2981 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2982 SEQ_OPTS_PUTS("data=ordered");
2983 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2984 SEQ_OPTS_PUTS("data=writeback");
2987 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2988 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2989 sbi->s_inode_readahead_blks);
2991 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2992 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2993 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2994 if (nodefs || sbi->s_max_dir_size_kb)
2995 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2996 if (test_opt(sb, DATA_ERR_ABORT))
2997 SEQ_OPTS_PUTS("data_err=abort");
2999 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3001 if (sb->s_flags & SB_INLINECRYPT)
3002 SEQ_OPTS_PUTS("inlinecrypt");
3004 if (test_opt(sb, DAX_ALWAYS)) {
3006 SEQ_OPTS_PUTS("dax");
3008 SEQ_OPTS_PUTS("dax=always");
3009 } else if (test_opt2(sb, DAX_NEVER)) {
3010 SEQ_OPTS_PUTS("dax=never");
3011 } else if (test_opt2(sb, DAX_INODE)) {
3012 SEQ_OPTS_PUTS("dax=inode");
3014 ext4_show_quota_options(seq, sb);
3018 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3020 return _ext4_show_options(seq, root->d_sb, 0);
3023 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3025 struct super_block *sb = seq->private;
3028 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3029 rc = _ext4_show_options(seq, sb, 1);
3030 seq_puts(seq, "\n");
3034 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3037 struct ext4_sb_info *sbi = EXT4_SB(sb);
3040 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3041 ext4_msg(sb, KERN_ERR, "revision level too high, "
3042 "forcing read-only mode");
3048 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3049 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3050 "running e2fsck is recommended");
3051 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3052 ext4_msg(sb, KERN_WARNING,
3053 "warning: mounting fs with errors, "
3054 "running e2fsck is recommended");
3055 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3056 le16_to_cpu(es->s_mnt_count) >=
3057 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3058 ext4_msg(sb, KERN_WARNING,
3059 "warning: maximal mount count reached, "
3060 "running e2fsck is recommended");
3061 else if (le32_to_cpu(es->s_checkinterval) &&
3062 (ext4_get_tstamp(es, s_lastcheck) +
3063 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3064 ext4_msg(sb, KERN_WARNING,
3065 "warning: checktime reached, "
3066 "running e2fsck is recommended");
3067 if (!sbi->s_journal)
3068 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3069 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3070 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3071 le16_add_cpu(&es->s_mnt_count, 1);
3072 ext4_update_tstamp(es, s_mtime);
3073 if (sbi->s_journal) {
3074 ext4_set_feature_journal_needs_recovery(sb);
3075 if (ext4_has_feature_orphan_file(sb))
3076 ext4_set_feature_orphan_present(sb);
3079 err = ext4_commit_super(sb);
3081 if (test_opt(sb, DEBUG))
3082 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3083 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3085 sbi->s_groups_count,
3086 EXT4_BLOCKS_PER_GROUP(sb),
3087 EXT4_INODES_PER_GROUP(sb),
3088 sbi->s_mount_opt, sbi->s_mount_opt2);
3092 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
3095 struct flex_groups **old_groups, **new_groups;
3098 if (!sbi->s_log_groups_per_flex)
3101 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3102 if (size <= sbi->s_flex_groups_allocated)
3105 new_groups = kvzalloc(roundup_pow_of_two(size *
3106 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3108 ext4_msg(sb, KERN_ERR,
3109 "not enough memory for %d flex group pointers", size);
3112 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3113 new_groups[i] = kvzalloc(roundup_pow_of_two(
3114 sizeof(struct flex_groups)),
3116 if (!new_groups[i]) {
3117 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3118 kvfree(new_groups[j]);
3120 ext4_msg(sb, KERN_ERR,
3121 "not enough memory for %d flex groups", size);
3126 old_groups = rcu_dereference(sbi->s_flex_groups);
3128 memcpy(new_groups, old_groups,
3129 (sbi->s_flex_groups_allocated *
3130 sizeof(struct flex_groups *)));
3132 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3133 sbi->s_flex_groups_allocated = size;
3135 ext4_kvfree_array_rcu(old_groups);
3139 static int ext4_fill_flex_info(struct super_block *sb)
3141 struct ext4_sb_info *sbi = EXT4_SB(sb);
3142 struct ext4_group_desc *gdp = NULL;
3143 struct flex_groups *fg;
3144 ext4_group_t flex_group;
3147 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3148 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3149 sbi->s_log_groups_per_flex = 0;
3153 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3157 for (i = 0; i < sbi->s_groups_count; i++) {
3158 gdp = ext4_get_group_desc(sb, i, NULL);
3160 flex_group = ext4_flex_group(sbi, i);
3161 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3162 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3163 atomic64_add(ext4_free_group_clusters(sb, gdp),
3164 &fg->free_clusters);
3165 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3173 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3174 struct ext4_group_desc *gdp)
3176 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3178 __le32 le_group = cpu_to_le32(block_group);
3179 struct ext4_sb_info *sbi = EXT4_SB(sb);
3181 if (ext4_has_metadata_csum(sbi->s_sb)) {
3182 /* Use new metadata_csum algorithm */
3184 __u16 dummy_csum = 0;
3186 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3188 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3189 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3190 sizeof(dummy_csum));
3191 offset += sizeof(dummy_csum);
3192 if (offset < sbi->s_desc_size)
3193 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3194 sbi->s_desc_size - offset);
3196 crc = csum32 & 0xFFFF;
3200 /* old crc16 code */
3201 if (!ext4_has_feature_gdt_csum(sb))
3204 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3205 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3206 crc = crc16(crc, (__u8 *)gdp, offset);
3207 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3208 /* for checksum of struct ext4_group_desc do the rest...*/
3209 if (ext4_has_feature_64bit(sb) &&
3210 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3211 crc = crc16(crc, (__u8 *)gdp + offset,
3212 le16_to_cpu(sbi->s_es->s_desc_size) -
3216 return cpu_to_le16(crc);
3219 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3220 struct ext4_group_desc *gdp)
3222 if (ext4_has_group_desc_csum(sb) &&
3223 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3229 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3230 struct ext4_group_desc *gdp)
3232 if (!ext4_has_group_desc_csum(sb))
3234 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3237 /* Called at mount-time, super-block is locked */
3238 static int ext4_check_descriptors(struct super_block *sb,
3239 ext4_fsblk_t sb_block,
3240 ext4_group_t *first_not_zeroed)
3242 struct ext4_sb_info *sbi = EXT4_SB(sb);
3243 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3244 ext4_fsblk_t last_block;
3245 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3246 ext4_fsblk_t block_bitmap;
3247 ext4_fsblk_t inode_bitmap;
3248 ext4_fsblk_t inode_table;
3249 int flexbg_flag = 0;
3250 ext4_group_t i, grp = sbi->s_groups_count;
3252 if (ext4_has_feature_flex_bg(sb))
3255 ext4_debug("Checking group descriptors");
3257 for (i = 0; i < sbi->s_groups_count; i++) {
3258 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3260 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3261 last_block = ext4_blocks_count(sbi->s_es) - 1;
3263 last_block = first_block +
3264 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3266 if ((grp == sbi->s_groups_count) &&
3267 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3270 block_bitmap = ext4_block_bitmap(sb, gdp);
3271 if (block_bitmap == sb_block) {
3272 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3273 "Block bitmap for group %u overlaps "
3278 if (block_bitmap >= sb_block + 1 &&
3279 block_bitmap <= last_bg_block) {
3280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3281 "Block bitmap for group %u overlaps "
3282 "block group descriptors", i);
3286 if (block_bitmap < first_block || block_bitmap > last_block) {
3287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3288 "Block bitmap for group %u not in group "
3289 "(block %llu)!", i, block_bitmap);
3292 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3293 if (inode_bitmap == sb_block) {
3294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3295 "Inode bitmap for group %u overlaps "
3300 if (inode_bitmap >= sb_block + 1 &&
3301 inode_bitmap <= last_bg_block) {
3302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3303 "Inode bitmap for group %u overlaps "
3304 "block group descriptors", i);
3308 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3309 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3310 "Inode bitmap for group %u not in group "
3311 "(block %llu)!", i, inode_bitmap);
3314 inode_table = ext4_inode_table(sb, gdp);
3315 if (inode_table == sb_block) {
3316 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3317 "Inode table for group %u overlaps "
3322 if (inode_table >= sb_block + 1 &&
3323 inode_table <= last_bg_block) {
3324 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3325 "Inode table for group %u overlaps "
3326 "block group descriptors", i);
3330 if (inode_table < first_block ||
3331 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3333 "Inode table for group %u not in group "
3334 "(block %llu)!", i, inode_table);
3337 ext4_lock_group(sb, i);
3338 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3339 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3340 "Checksum for group %u failed (%u!=%u)",
3341 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3342 gdp)), le16_to_cpu(gdp->bg_checksum));
3343 if (!sb_rdonly(sb)) {
3344 ext4_unlock_group(sb, i);
3348 ext4_unlock_group(sb, i);
3350 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3352 if (NULL != first_not_zeroed)
3353 *first_not_zeroed = grp;
3358 * Maximal extent format file size.
3359 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3360 * extent format containers, within a sector_t, and within i_blocks
3361 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3362 * so that won't be a limiting factor.
3364 * However there is other limiting factor. We do store extents in the form
3365 * of starting block and length, hence the resulting length of the extent
3366 * covering maximum file size must fit into on-disk format containers as
3367 * well. Given that length is always by 1 unit bigger than max unit (because
3368 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3370 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3372 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3375 loff_t upper_limit = MAX_LFS_FILESIZE;
3377 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3379 if (!has_huge_files) {
3380 upper_limit = (1LL << 32) - 1;
3382 /* total blocks in file system block size */
3383 upper_limit >>= (blkbits - 9);
3384 upper_limit <<= blkbits;
3388 * 32-bit extent-start container, ee_block. We lower the maxbytes
3389 * by one fs block, so ee_len can cover the extent of maximum file
3392 res = (1LL << 32) - 1;
3395 /* Sanity check against vm- & vfs- imposed limits */
3396 if (res > upper_limit)
3403 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3404 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3405 * We need to be 1 filesystem block less than the 2^48 sector limit.
3407 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3409 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3411 unsigned int ppb = 1 << (bits - 2);
3414 * This is calculated to be the largest file size for a dense, block
3415 * mapped file such that the file's total number of 512-byte sectors,
3416 * including data and all indirect blocks, does not exceed (2^48 - 1).
3418 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3419 * number of 512-byte sectors of the file.
3421 if (!has_huge_files) {
3423 * !has_huge_files or implies that the inode i_block field
3424 * represents total file blocks in 2^32 512-byte sectors ==
3425 * size of vfs inode i_blocks * 8
3427 upper_limit = (1LL << 32) - 1;
3429 /* total blocks in file system block size */
3430 upper_limit >>= (bits - 9);
3434 * We use 48 bit ext4_inode i_blocks
3435 * With EXT4_HUGE_FILE_FL set the i_blocks
3436 * represent total number of blocks in
3437 * file system block size
3439 upper_limit = (1LL << 48) - 1;
3443 /* Compute how many blocks we can address by block tree */
3446 res += ((loff_t)ppb) * ppb * ppb;
3447 /* Compute how many metadata blocks are needed */
3449 meta_blocks += 1 + ppb;
3450 meta_blocks += 1 + ppb + ppb * ppb;
3451 /* Does block tree limit file size? */
3452 if (res + meta_blocks <= upper_limit)
3456 /* How many metadata blocks are needed for addressing upper_limit? */
3457 upper_limit -= EXT4_NDIR_BLOCKS;
3458 /* indirect blocks */
3461 /* double indirect blocks */
3462 if (upper_limit < ppb * ppb) {
3463 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3467 meta_blocks += 1 + ppb;
3468 upper_limit -= ppb * ppb;
3469 /* tripple indirect blocks for the rest */
3470 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3471 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3475 if (res > MAX_LFS_FILESIZE)
3476 res = MAX_LFS_FILESIZE;
3481 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3482 ext4_fsblk_t logical_sb_block, int nr)
3484 struct ext4_sb_info *sbi = EXT4_SB(sb);
3485 ext4_group_t bg, first_meta_bg;
3488 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3490 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3491 return logical_sb_block + nr + 1;
3492 bg = sbi->s_desc_per_block * nr;
3493 if (ext4_bg_has_super(sb, bg))
3497 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3498 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3499 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3502 if (sb->s_blocksize == 1024 && nr == 0 &&
3503 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3506 return (has_super + ext4_group_first_block_no(sb, bg));
3510 * ext4_get_stripe_size: Get the stripe size.
3511 * @sbi: In memory super block info
3513 * If we have specified it via mount option, then
3514 * use the mount option value. If the value specified at mount time is
3515 * greater than the blocks per group use the super block value.
3516 * If the super block value is greater than blocks per group return 0.
3517 * Allocator needs it be less than blocks per group.
3520 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3522 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3523 unsigned long stripe_width =
3524 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3527 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3528 ret = sbi->s_stripe;
3529 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3531 else if (stride && stride <= sbi->s_blocks_per_group)
3537 * If the stripe width is 1, this makes no sense and
3538 * we set it to 0 to turn off stripe handling code.
3547 * Check whether this filesystem can be mounted based on
3548 * the features present and the RDONLY/RDWR mount requested.
3549 * Returns 1 if this filesystem can be mounted as requested,
3550 * 0 if it cannot be.
3552 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3554 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3555 ext4_msg(sb, KERN_ERR,
3556 "Couldn't mount because of "
3557 "unsupported optional features (%x)",
3558 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3559 ~EXT4_FEATURE_INCOMPAT_SUPP));
3563 #if !IS_ENABLED(CONFIG_UNICODE)
3564 if (ext4_has_feature_casefold(sb)) {
3565 ext4_msg(sb, KERN_ERR,
3566 "Filesystem with casefold feature cannot be "
3567 "mounted without CONFIG_UNICODE");
3575 if (ext4_has_feature_readonly(sb)) {
3576 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3577 sb->s_flags |= SB_RDONLY;
3581 /* Check that feature set is OK for a read-write mount */
3582 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3583 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3584 "unsupported optional features (%x)",
3585 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3586 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3589 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3590 ext4_msg(sb, KERN_ERR,
3591 "Can't support bigalloc feature without "
3592 "extents feature\n");
3596 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3597 if (!readonly && (ext4_has_feature_quota(sb) ||
3598 ext4_has_feature_project(sb))) {
3599 ext4_msg(sb, KERN_ERR,
3600 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3603 #endif /* CONFIG_QUOTA */
3608 * This function is called once a day if we have errors logged
3609 * on the file system
3611 static void print_daily_error_info(struct timer_list *t)
3613 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3614 struct super_block *sb = sbi->s_sb;
3615 struct ext4_super_block *es = sbi->s_es;
3617 if (es->s_error_count)
3618 /* fsck newer than v1.41.13 is needed to clean this condition. */
3619 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3620 le32_to_cpu(es->s_error_count));
3621 if (es->s_first_error_time) {
3622 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3624 ext4_get_tstamp(es, s_first_error_time),
3625 (int) sizeof(es->s_first_error_func),
3626 es->s_first_error_func,
3627 le32_to_cpu(es->s_first_error_line));
3628 if (es->s_first_error_ino)
3629 printk(KERN_CONT ": inode %u",
3630 le32_to_cpu(es->s_first_error_ino));
3631 if (es->s_first_error_block)
3632 printk(KERN_CONT ": block %llu", (unsigned long long)
3633 le64_to_cpu(es->s_first_error_block));
3634 printk(KERN_CONT "\n");
3636 if (es->s_last_error_time) {
3637 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3639 ext4_get_tstamp(es, s_last_error_time),
3640 (int) sizeof(es->s_last_error_func),
3641 es->s_last_error_func,
3642 le32_to_cpu(es->s_last_error_line));
3643 if (es->s_last_error_ino)
3644 printk(KERN_CONT ": inode %u",
3645 le32_to_cpu(es->s_last_error_ino));
3646 if (es->s_last_error_block)
3647 printk(KERN_CONT ": block %llu", (unsigned long long)
3648 le64_to_cpu(es->s_last_error_block));
3649 printk(KERN_CONT "\n");
3651 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3654 /* Find next suitable group and run ext4_init_inode_table */
3655 static int ext4_run_li_request(struct ext4_li_request *elr)
3657 struct ext4_group_desc *gdp = NULL;
3658 struct super_block *sb = elr->lr_super;
3659 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3660 ext4_group_t group = elr->lr_next_group;
3661 unsigned int prefetch_ios = 0;
3665 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3666 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3667 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3669 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3671 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3673 if (group >= elr->lr_next_group) {
3675 if (elr->lr_first_not_zeroed != ngroups &&
3676 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3677 elr->lr_next_group = elr->lr_first_not_zeroed;
3678 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3685 for (; group < ngroups; group++) {
3686 gdp = ext4_get_group_desc(sb, group, NULL);
3692 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3696 if (group >= ngroups)
3700 start_time = ktime_get_real_ns();
3701 ret = ext4_init_inode_table(sb, group,
3702 elr->lr_timeout ? 0 : 1);
3703 trace_ext4_lazy_itable_init(sb, group);
3704 if (elr->lr_timeout == 0) {
3705 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3706 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3708 elr->lr_next_sched = jiffies + elr->lr_timeout;
3709 elr->lr_next_group = group + 1;
3715 * Remove lr_request from the list_request and free the
3716 * request structure. Should be called with li_list_mtx held
3718 static void ext4_remove_li_request(struct ext4_li_request *elr)
3723 list_del(&elr->lr_request);
3724 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3728 static void ext4_unregister_li_request(struct super_block *sb)
3730 mutex_lock(&ext4_li_mtx);
3731 if (!ext4_li_info) {
3732 mutex_unlock(&ext4_li_mtx);
3736 mutex_lock(&ext4_li_info->li_list_mtx);
3737 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3738 mutex_unlock(&ext4_li_info->li_list_mtx);
3739 mutex_unlock(&ext4_li_mtx);
3742 static struct task_struct *ext4_lazyinit_task;
3745 * This is the function where ext4lazyinit thread lives. It walks
3746 * through the request list searching for next scheduled filesystem.
3747 * When such a fs is found, run the lazy initialization request
3748 * (ext4_rn_li_request) and keep track of the time spend in this
3749 * function. Based on that time we compute next schedule time of
3750 * the request. When walking through the list is complete, compute
3751 * next waking time and put itself into sleep.
3753 static int ext4_lazyinit_thread(void *arg)
3755 struct ext4_lazy_init *eli = arg;
3756 struct list_head *pos, *n;
3757 struct ext4_li_request *elr;
3758 unsigned long next_wakeup, cur;
3760 BUG_ON(NULL == eli);
3764 next_wakeup = MAX_JIFFY_OFFSET;
3766 mutex_lock(&eli->li_list_mtx);
3767 if (list_empty(&eli->li_request_list)) {
3768 mutex_unlock(&eli->li_list_mtx);
3771 list_for_each_safe(pos, n, &eli->li_request_list) {
3774 elr = list_entry(pos, struct ext4_li_request,
3777 if (time_before(jiffies, elr->lr_next_sched)) {
3778 if (time_before(elr->lr_next_sched, next_wakeup))
3779 next_wakeup = elr->lr_next_sched;
3782 if (down_read_trylock(&elr->lr_super->s_umount)) {
3783 if (sb_start_write_trylock(elr->lr_super)) {
3786 * We hold sb->s_umount, sb can not
3787 * be removed from the list, it is
3788 * now safe to drop li_list_mtx
3790 mutex_unlock(&eli->li_list_mtx);
3791 err = ext4_run_li_request(elr);
3792 sb_end_write(elr->lr_super);
3793 mutex_lock(&eli->li_list_mtx);
3796 up_read((&elr->lr_super->s_umount));
3798 /* error, remove the lazy_init job */
3800 ext4_remove_li_request(elr);
3804 elr->lr_next_sched = jiffies +
3806 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3808 if (time_before(elr->lr_next_sched, next_wakeup))
3809 next_wakeup = elr->lr_next_sched;
3811 mutex_unlock(&eli->li_list_mtx);
3816 if ((time_after_eq(cur, next_wakeup)) ||
3817 (MAX_JIFFY_OFFSET == next_wakeup)) {
3822 schedule_timeout_interruptible(next_wakeup - cur);
3824 if (kthread_should_stop()) {
3825 ext4_clear_request_list();
3832 * It looks like the request list is empty, but we need
3833 * to check it under the li_list_mtx lock, to prevent any
3834 * additions into it, and of course we should lock ext4_li_mtx
3835 * to atomically free the list and ext4_li_info, because at
3836 * this point another ext4 filesystem could be registering
3839 mutex_lock(&ext4_li_mtx);
3840 mutex_lock(&eli->li_list_mtx);
3841 if (!list_empty(&eli->li_request_list)) {
3842 mutex_unlock(&eli->li_list_mtx);
3843 mutex_unlock(&ext4_li_mtx);
3846 mutex_unlock(&eli->li_list_mtx);
3847 kfree(ext4_li_info);
3848 ext4_li_info = NULL;
3849 mutex_unlock(&ext4_li_mtx);
3854 static void ext4_clear_request_list(void)
3856 struct list_head *pos, *n;
3857 struct ext4_li_request *elr;
3859 mutex_lock(&ext4_li_info->li_list_mtx);
3860 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3861 elr = list_entry(pos, struct ext4_li_request,
3863 ext4_remove_li_request(elr);
3865 mutex_unlock(&ext4_li_info->li_list_mtx);
3868 static int ext4_run_lazyinit_thread(void)
3870 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3871 ext4_li_info, "ext4lazyinit");
3872 if (IS_ERR(ext4_lazyinit_task)) {
3873 int err = PTR_ERR(ext4_lazyinit_task);
3874 ext4_clear_request_list();
3875 kfree(ext4_li_info);
3876 ext4_li_info = NULL;
3877 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3878 "initialization thread\n",
3882 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3887 * Check whether it make sense to run itable init. thread or not.
3888 * If there is at least one uninitialized inode table, return
3889 * corresponding group number, else the loop goes through all
3890 * groups and return total number of groups.
3892 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3894 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3895 struct ext4_group_desc *gdp = NULL;
3897 if (!ext4_has_group_desc_csum(sb))
3900 for (group = 0; group < ngroups; group++) {
3901 gdp = ext4_get_group_desc(sb, group, NULL);
3905 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3912 static int ext4_li_info_new(void)
3914 struct ext4_lazy_init *eli = NULL;
3916 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3920 INIT_LIST_HEAD(&eli->li_request_list);
3921 mutex_init(&eli->li_list_mtx);
3923 eli->li_state |= EXT4_LAZYINIT_QUIT;
3930 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3933 struct ext4_li_request *elr;
3935 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3940 elr->lr_first_not_zeroed = start;
3941 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3942 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3943 elr->lr_next_group = start;
3945 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3949 * Randomize first schedule time of the request to
3950 * spread the inode table initialization requests
3953 elr->lr_next_sched = jiffies + (prandom_u32() %
3954 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3958 int ext4_register_li_request(struct super_block *sb,
3959 ext4_group_t first_not_zeroed)
3961 struct ext4_sb_info *sbi = EXT4_SB(sb);
3962 struct ext4_li_request *elr = NULL;
3963 ext4_group_t ngroups = sbi->s_groups_count;
3966 mutex_lock(&ext4_li_mtx);
3967 if (sbi->s_li_request != NULL) {
3969 * Reset timeout so it can be computed again, because
3970 * s_li_wait_mult might have changed.
3972 sbi->s_li_request->lr_timeout = 0;
3976 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3977 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3978 !test_opt(sb, INIT_INODE_TABLE)))
3981 elr = ext4_li_request_new(sb, first_not_zeroed);
3987 if (NULL == ext4_li_info) {
3988 ret = ext4_li_info_new();
3993 mutex_lock(&ext4_li_info->li_list_mtx);
3994 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3995 mutex_unlock(&ext4_li_info->li_list_mtx);
3997 sbi->s_li_request = elr;
3999 * set elr to NULL here since it has been inserted to
4000 * the request_list and the removal and free of it is
4001 * handled by ext4_clear_request_list from now on.
4005 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4006 ret = ext4_run_lazyinit_thread();
4011 mutex_unlock(&ext4_li_mtx);
4018 * We do not need to lock anything since this is called on
4021 static void ext4_destroy_lazyinit_thread(void)
4024 * If thread exited earlier
4025 * there's nothing to be done.
4027 if (!ext4_li_info || !ext4_lazyinit_task)
4030 kthread_stop(ext4_lazyinit_task);
4033 static int set_journal_csum_feature_set(struct super_block *sb)
4036 int compat, incompat;
4037 struct ext4_sb_info *sbi = EXT4_SB(sb);
4039 if (ext4_has_metadata_csum(sb)) {
4040 /* journal checksum v3 */
4042 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4044 /* journal checksum v1 */
4045 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4049 jbd2_journal_clear_features(sbi->s_journal,
4050 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4051 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4052 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4053 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4054 ret = jbd2_journal_set_features(sbi->s_journal,
4056 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4058 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4059 ret = jbd2_journal_set_features(sbi->s_journal,
4062 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4063 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4065 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4066 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4073 * Note: calculating the overhead so we can be compatible with
4074 * historical BSD practice is quite difficult in the face of
4075 * clusters/bigalloc. This is because multiple metadata blocks from
4076 * different block group can end up in the same allocation cluster.
4077 * Calculating the exact overhead in the face of clustered allocation
4078 * requires either O(all block bitmaps) in memory or O(number of block
4079 * groups**2) in time. We will still calculate the superblock for
4080 * older file systems --- and if we come across with a bigalloc file
4081 * system with zero in s_overhead_clusters the estimate will be close to
4082 * correct especially for very large cluster sizes --- but for newer
4083 * file systems, it's better to calculate this figure once at mkfs
4084 * time, and store it in the superblock. If the superblock value is
4085 * present (even for non-bigalloc file systems), we will use it.
4087 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4090 struct ext4_sb_info *sbi = EXT4_SB(sb);
4091 struct ext4_group_desc *gdp;
4092 ext4_fsblk_t first_block, last_block, b;
4093 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4094 int s, j, count = 0;
4095 int has_super = ext4_bg_has_super(sb, grp);
4097 if (!ext4_has_feature_bigalloc(sb))
4098 return (has_super + ext4_bg_num_gdb(sb, grp) +
4099 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4100 sbi->s_itb_per_group + 2);
4102 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4103 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4104 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4105 for (i = 0; i < ngroups; i++) {
4106 gdp = ext4_get_group_desc(sb, i, NULL);
4107 b = ext4_block_bitmap(sb, gdp);
4108 if (b >= first_block && b <= last_block) {
4109 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4112 b = ext4_inode_bitmap(sb, gdp);
4113 if (b >= first_block && b <= last_block) {
4114 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4117 b = ext4_inode_table(sb, gdp);
4118 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4119 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4120 int c = EXT4_B2C(sbi, b - first_block);
4121 ext4_set_bit(c, buf);
4127 if (ext4_bg_has_super(sb, grp)) {
4128 ext4_set_bit(s++, buf);
4131 j = ext4_bg_num_gdb(sb, grp);
4132 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4133 ext4_error(sb, "Invalid number of block group "
4134 "descriptor blocks: %d", j);
4135 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4139 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4143 return EXT4_CLUSTERS_PER_GROUP(sb) -
4144 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4148 * Compute the overhead and stash it in sbi->s_overhead
4150 int ext4_calculate_overhead(struct super_block *sb)
4152 struct ext4_sb_info *sbi = EXT4_SB(sb);
4153 struct ext4_super_block *es = sbi->s_es;
4154 struct inode *j_inode;
4155 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4156 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4157 ext4_fsblk_t overhead = 0;
4158 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4164 * Compute the overhead (FS structures). This is constant
4165 * for a given filesystem unless the number of block groups
4166 * changes so we cache the previous value until it does.
4170 * All of the blocks before first_data_block are overhead
4172 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4175 * Add the overhead found in each block group
4177 for (i = 0; i < ngroups; i++) {
4180 blks = count_overhead(sb, i, buf);
4183 memset(buf, 0, PAGE_SIZE);
4188 * Add the internal journal blocks whether the journal has been
4191 if (sbi->s_journal && !sbi->s_journal_bdev)
4192 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4193 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4194 /* j_inum for internal journal is non-zero */
4195 j_inode = ext4_get_journal_inode(sb, j_inum);
4197 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4198 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4201 ext4_msg(sb, KERN_ERR, "can't get journal size");
4204 sbi->s_overhead = overhead;
4206 free_page((unsigned long) buf);
4210 static void ext4_set_resv_clusters(struct super_block *sb)
4212 ext4_fsblk_t resv_clusters;
4213 struct ext4_sb_info *sbi = EXT4_SB(sb);
4216 * There's no need to reserve anything when we aren't using extents.
4217 * The space estimates are exact, there are no unwritten extents,
4218 * hole punching doesn't need new metadata... This is needed especially
4219 * to keep ext2/3 backward compatibility.
4221 if (!ext4_has_feature_extents(sb))
4224 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4225 * This should cover the situations where we can not afford to run
4226 * out of space like for example punch hole, or converting
4227 * unwritten extents in delalloc path. In most cases such
4228 * allocation would require 1, or 2 blocks, higher numbers are
4231 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4232 sbi->s_cluster_bits);
4234 do_div(resv_clusters, 50);
4235 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4237 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4240 static const char *ext4_quota_mode(struct super_block *sb)
4243 if (!ext4_quota_capable(sb))
4246 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4247 return "journalled";
4255 static void ext4_setup_csum_trigger(struct super_block *sb,
4256 enum ext4_journal_trigger_type type,
4258 struct jbd2_buffer_trigger_type *type,
4259 struct buffer_head *bh,
4263 struct ext4_sb_info *sbi = EXT4_SB(sb);
4265 sbi->s_journal_triggers[type].sb = sb;
4266 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4269 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4274 kfree(sbi->s_blockgroup_lock);
4275 fs_put_dax(sbi->s_daxdev);
4279 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4281 struct ext4_sb_info *sbi;
4283 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4287 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off);
4289 sbi->s_blockgroup_lock =
4290 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4292 if (!sbi->s_blockgroup_lock)
4295 sb->s_fs_info = sbi;
4299 fs_put_dax(sbi->s_daxdev);
4304 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
4306 struct buffer_head *bh, **group_desc;
4307 struct ext4_super_block *es = NULL;
4308 struct ext4_sb_info *sbi = EXT4_SB(sb);
4309 struct flex_groups **flex_groups;
4311 ext4_fsblk_t logical_sb_block;
4312 unsigned long offset = 0;
4313 unsigned long def_mount_opts;
4316 int blocksize, clustersize;
4317 unsigned int db_count;
4319 int needs_recovery, has_huge_files;
4322 ext4_group_t first_not_zeroed;
4323 struct ext4_fs_context *ctx = fc->fs_private;
4324 int silent = fc->sb_flags & SB_SILENT;
4326 /* Set defaults for the variables that will be set during parsing */
4327 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
4328 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4330 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4331 sbi->s_sectors_written_start =
4332 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4334 /* -EINVAL is default */
4336 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4338 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4343 * The ext4 superblock will not be buffer aligned for other than 1kB
4344 * block sizes. We need to calculate the offset from buffer start.
4346 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4347 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4348 offset = do_div(logical_sb_block, blocksize);
4350 logical_sb_block = sbi->s_sb_block;
4353 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4355 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4360 * Note: s_es must be initialized as soon as possible because
4361 * some ext4 macro-instructions depend on its value
4363 es = (struct ext4_super_block *) (bh->b_data + offset);
4365 sb->s_magic = le16_to_cpu(es->s_magic);
4366 if (sb->s_magic != EXT4_SUPER_MAGIC)
4368 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4370 /* Warn if metadata_csum and gdt_csum are both set. */
4371 if (ext4_has_feature_metadata_csum(sb) &&
4372 ext4_has_feature_gdt_csum(sb))
4373 ext4_warning(sb, "metadata_csum and uninit_bg are "
4374 "redundant flags; please run fsck.");
4376 /* Check for a known checksum algorithm */
4377 if (!ext4_verify_csum_type(sb, es)) {
4378 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4379 "unknown checksum algorithm.");
4383 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4384 ext4_orphan_file_block_trigger);
4386 /* Load the checksum driver */
4387 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4388 if (IS_ERR(sbi->s_chksum_driver)) {
4389 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4390 ret = PTR_ERR(sbi->s_chksum_driver);
4391 sbi->s_chksum_driver = NULL;
4395 /* Check superblock checksum */
4396 if (!ext4_superblock_csum_verify(sb, es)) {
4397 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4398 "invalid superblock checksum. Run e2fsck?");
4404 /* Precompute checksum seed for all metadata */
4405 if (ext4_has_feature_csum_seed(sb))
4406 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4407 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4408 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4409 sizeof(es->s_uuid));
4411 /* Set defaults before we parse the mount options */
4412 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4413 set_opt(sb, INIT_INODE_TABLE);
4414 if (def_mount_opts & EXT4_DEFM_DEBUG)
4416 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4418 if (def_mount_opts & EXT4_DEFM_UID16)
4419 set_opt(sb, NO_UID32);
4420 /* xattr user namespace & acls are now defaulted on */
4421 set_opt(sb, XATTR_USER);
4422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4423 set_opt(sb, POSIX_ACL);
4425 if (ext4_has_feature_fast_commit(sb))
4426 set_opt2(sb, JOURNAL_FAST_COMMIT);
4427 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4428 if (ext4_has_metadata_csum(sb))
4429 set_opt(sb, JOURNAL_CHECKSUM);
4431 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4432 set_opt(sb, JOURNAL_DATA);
4433 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4434 set_opt(sb, ORDERED_DATA);
4435 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4436 set_opt(sb, WRITEBACK_DATA);
4438 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4439 set_opt(sb, ERRORS_PANIC);
4440 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4441 set_opt(sb, ERRORS_CONT);
4443 set_opt(sb, ERRORS_RO);
4444 /* block_validity enabled by default; disable with noblock_validity */
4445 set_opt(sb, BLOCK_VALIDITY);
4446 if (def_mount_opts & EXT4_DEFM_DISCARD)
4447 set_opt(sb, DISCARD);
4449 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4450 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4451 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4452 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4453 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4455 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4456 set_opt(sb, BARRIER);
4459 * enable delayed allocation by default
4460 * Use -o nodelalloc to turn it off
4462 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4463 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4464 set_opt(sb, DELALLOC);
4467 * set default s_li_wait_mult for lazyinit, for the case there is
4468 * no mount option specified.
4470 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4472 if (le32_to_cpu(es->s_log_block_size) >
4473 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4474 ext4_msg(sb, KERN_ERR,
4475 "Invalid log block size: %u",
4476 le32_to_cpu(es->s_log_block_size));
4479 if (le32_to_cpu(es->s_log_cluster_size) >
4480 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4481 ext4_msg(sb, KERN_ERR,
4482 "Invalid log cluster size: %u",
4483 le32_to_cpu(es->s_log_cluster_size));
4487 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4489 if (blocksize == PAGE_SIZE)
4490 set_opt(sb, DIOREAD_NOLOCK);
4492 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4493 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4494 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4496 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4497 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4498 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4499 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4503 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4504 (!is_power_of_2(sbi->s_inode_size)) ||
4505 (sbi->s_inode_size > blocksize)) {
4506 ext4_msg(sb, KERN_ERR,
4507 "unsupported inode size: %d",
4509 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4513 * i_atime_extra is the last extra field available for
4514 * [acm]times in struct ext4_inode. Checking for that
4515 * field should suffice to ensure we have extra space
4518 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4519 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4520 sb->s_time_gran = 1;
4521 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4523 sb->s_time_gran = NSEC_PER_SEC;
4524 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4526 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4528 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4529 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4530 EXT4_GOOD_OLD_INODE_SIZE;
4531 if (ext4_has_feature_extra_isize(sb)) {
4532 unsigned v, max = (sbi->s_inode_size -
4533 EXT4_GOOD_OLD_INODE_SIZE);
4535 v = le16_to_cpu(es->s_want_extra_isize);
4537 ext4_msg(sb, KERN_ERR,
4538 "bad s_want_extra_isize: %d", v);
4541 if (sbi->s_want_extra_isize < v)
4542 sbi->s_want_extra_isize = v;
4544 v = le16_to_cpu(es->s_min_extra_isize);
4546 ext4_msg(sb, KERN_ERR,
4547 "bad s_min_extra_isize: %d", v);
4550 if (sbi->s_want_extra_isize < v)
4551 sbi->s_want_extra_isize = v;
4555 err = parse_apply_sb_mount_options(sb, ctx);
4559 sbi->s_def_mount_opt = sbi->s_mount_opt;
4561 err = ext4_check_opt_consistency(fc, sb);
4565 ext4_apply_options(fc, sb);
4567 #if IS_ENABLED(CONFIG_UNICODE)
4568 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4569 const struct ext4_sb_encodings *encoding_info;
4570 struct unicode_map *encoding;
4571 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
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;
4604 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4605 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4606 /* can't mount with both data=journal and dioread_nolock. */
4607 clear_opt(sb, DIOREAD_NOLOCK);
4608 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4609 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4610 ext4_msg(sb, KERN_ERR, "can't mount with "
4611 "both data=journal and delalloc");
4614 if (test_opt(sb, DAX_ALWAYS)) {
4615 ext4_msg(sb, KERN_ERR, "can't mount with "
4616 "both data=journal and dax");
4619 if (ext4_has_feature_encrypt(sb)) {
4620 ext4_msg(sb, KERN_WARNING,
4621 "encrypted files will use data=ordered "
4622 "instead of data journaling mode");
4624 if (test_opt(sb, DELALLOC))
4625 clear_opt(sb, DELALLOC);
4627 sb->s_iflags |= SB_I_CGROUPWB;
4630 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4631 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4633 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4634 (ext4_has_compat_features(sb) ||
4635 ext4_has_ro_compat_features(sb) ||
4636 ext4_has_incompat_features(sb)))
4637 ext4_msg(sb, KERN_WARNING,
4638 "feature flags set on rev 0 fs, "
4639 "running e2fsck is recommended");
4641 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4642 set_opt2(sb, HURD_COMPAT);
4643 if (ext4_has_feature_64bit(sb)) {
4644 ext4_msg(sb, KERN_ERR,
4645 "The Hurd can't support 64-bit file systems");
4650 * ea_inode feature uses l_i_version field which is not
4651 * available in HURD_COMPAT mode.
4653 if (ext4_has_feature_ea_inode(sb)) {
4654 ext4_msg(sb, KERN_ERR,
4655 "ea_inode feature is not supported for Hurd");
4660 if (IS_EXT2_SB(sb)) {
4661 if (ext2_feature_set_ok(sb))
4662 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4663 "using the ext4 subsystem");
4666 * If we're probing be silent, if this looks like
4667 * it's actually an ext[34] filesystem.
4669 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4671 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4672 "to feature incompatibilities");
4677 if (IS_EXT3_SB(sb)) {
4678 if (ext3_feature_set_ok(sb))
4679 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4680 "using the ext4 subsystem");
4683 * If we're probing be silent, if this looks like
4684 * it's actually an ext4 filesystem.
4686 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4688 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4689 "to feature incompatibilities");
4695 * Check feature flags regardless of the revision level, since we
4696 * previously didn't change the revision level when setting the flags,
4697 * so there is a chance incompat flags are set on a rev 0 filesystem.
4699 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4702 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4703 ext4_msg(sb, KERN_ERR,
4704 "Number of reserved GDT blocks insanely large: %d",
4705 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4709 if (sbi->s_daxdev) {
4710 if (blocksize == PAGE_SIZE)
4711 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4713 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4716 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4717 if (ext4_has_feature_inline_data(sb)) {
4718 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4719 " that may contain inline data");
4722 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4723 ext4_msg(sb, KERN_ERR,
4724 "DAX unsupported by block device.");
4729 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4730 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4731 es->s_encryption_level);
4735 if (sb->s_blocksize != blocksize) {
4737 * bh must be released before kill_bdev(), otherwise
4738 * it won't be freed and its page also. kill_bdev()
4739 * is called by sb_set_blocksize().
4742 /* Validate the filesystem blocksize */
4743 if (!sb_set_blocksize(sb, blocksize)) {
4744 ext4_msg(sb, KERN_ERR, "bad block size %d",
4750 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4751 offset = do_div(logical_sb_block, blocksize);
4752 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4754 ext4_msg(sb, KERN_ERR,
4755 "Can't read superblock on 2nd try");
4760 es = (struct ext4_super_block *)(bh->b_data + offset);
4762 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4763 ext4_msg(sb, KERN_ERR,
4764 "Magic mismatch, very weird!");
4769 has_huge_files = ext4_has_feature_huge_file(sb);
4770 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4772 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4774 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4775 if (ext4_has_feature_64bit(sb)) {
4776 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4777 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4778 !is_power_of_2(sbi->s_desc_size)) {
4779 ext4_msg(sb, KERN_ERR,
4780 "unsupported descriptor size %lu",
4785 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4787 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4788 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4790 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4791 if (sbi->s_inodes_per_block == 0)
4793 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4794 sbi->s_inodes_per_group > blocksize * 8) {
4795 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4796 sbi->s_inodes_per_group);
4799 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4800 sbi->s_inodes_per_block;
4801 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4803 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
4804 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4805 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4807 for (i = 0; i < 4; i++)
4808 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4809 sbi->s_def_hash_version = es->s_def_hash_version;
4810 if (ext4_has_feature_dir_index(sb)) {
4811 i = le32_to_cpu(es->s_flags);
4812 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4813 sbi->s_hash_unsigned = 3;
4814 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4815 #ifdef __CHAR_UNSIGNED__
4818 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4819 sbi->s_hash_unsigned = 3;
4823 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4828 /* Handle clustersize */
4829 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4830 if (ext4_has_feature_bigalloc(sb)) {
4831 if (clustersize < blocksize) {
4832 ext4_msg(sb, KERN_ERR,
4833 "cluster size (%d) smaller than "
4834 "block size (%d)", clustersize, blocksize);
4837 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4838 le32_to_cpu(es->s_log_block_size);
4839 sbi->s_clusters_per_group =
4840 le32_to_cpu(es->s_clusters_per_group);
4841 if (sbi->s_clusters_per_group > blocksize * 8) {
4842 ext4_msg(sb, KERN_ERR,
4843 "#clusters per group too big: %lu",
4844 sbi->s_clusters_per_group);
4847 if (sbi->s_blocks_per_group !=
4848 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4849 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4850 "clusters per group (%lu) inconsistent",
4851 sbi->s_blocks_per_group,
4852 sbi->s_clusters_per_group);
4856 if (clustersize != blocksize) {
4857 ext4_msg(sb, KERN_ERR,
4858 "fragment/cluster size (%d) != "
4859 "block size (%d)", clustersize, blocksize);
4862 if (sbi->s_blocks_per_group > blocksize * 8) {
4863 ext4_msg(sb, KERN_ERR,
4864 "#blocks per group too big: %lu",
4865 sbi->s_blocks_per_group);
4868 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4869 sbi->s_cluster_bits = 0;
4871 sbi->s_cluster_ratio = clustersize / blocksize;
4873 /* Do we have standard group size of clustersize * 8 blocks ? */
4874 if (sbi->s_blocks_per_group == clustersize << 3)
4875 set_opt2(sb, STD_GROUP_SIZE);
4878 * Test whether we have more sectors than will fit in sector_t,
4879 * and whether the max offset is addressable by the page cache.
4881 err = generic_check_addressable(sb->s_blocksize_bits,
4882 ext4_blocks_count(es));
4884 ext4_msg(sb, KERN_ERR, "filesystem"
4885 " too large to mount safely on this system");
4889 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4892 /* check blocks count against device size */
4893 blocks_count = sb_bdev_nr_blocks(sb);
4894 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4895 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4896 "exceeds size of device (%llu blocks)",
4897 ext4_blocks_count(es), blocks_count);
4902 * It makes no sense for the first data block to be beyond the end
4903 * of the filesystem.
4905 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4906 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4907 "block %u is beyond end of filesystem (%llu)",
4908 le32_to_cpu(es->s_first_data_block),
4909 ext4_blocks_count(es));
4912 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4913 (sbi->s_cluster_ratio == 1)) {
4914 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4915 "block is 0 with a 1k block and cluster size");
4919 blocks_count = (ext4_blocks_count(es) -
4920 le32_to_cpu(es->s_first_data_block) +
4921 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4922 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4923 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4924 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4925 "(block count %llu, first data block %u, "
4926 "blocks per group %lu)", blocks_count,
4927 ext4_blocks_count(es),
4928 le32_to_cpu(es->s_first_data_block),
4929 EXT4_BLOCKS_PER_GROUP(sb));
4932 sbi->s_groups_count = blocks_count;
4933 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4934 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4935 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4936 le32_to_cpu(es->s_inodes_count)) {
4937 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4938 le32_to_cpu(es->s_inodes_count),
4939 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4943 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4944 EXT4_DESC_PER_BLOCK(sb);
4945 if (ext4_has_feature_meta_bg(sb)) {
4946 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4947 ext4_msg(sb, KERN_WARNING,
4948 "first meta block group too large: %u "
4949 "(group descriptor block count %u)",
4950 le32_to_cpu(es->s_first_meta_bg), db_count);
4954 rcu_assign_pointer(sbi->s_group_desc,
4955 kvmalloc_array(db_count,
4956 sizeof(struct buffer_head *),
4958 if (sbi->s_group_desc == NULL) {
4959 ext4_msg(sb, KERN_ERR, "not enough memory");
4964 bgl_lock_init(sbi->s_blockgroup_lock);
4966 /* Pre-read the descriptors into the buffer cache */
4967 for (i = 0; i < db_count; i++) {
4968 block = descriptor_loc(sb, logical_sb_block, i);
4969 ext4_sb_breadahead_unmovable(sb, block);
4972 for (i = 0; i < db_count; i++) {
4973 struct buffer_head *bh;
4975 block = descriptor_loc(sb, logical_sb_block, i);
4976 bh = ext4_sb_bread_unmovable(sb, block);
4978 ext4_msg(sb, KERN_ERR,
4979 "can't read group descriptor %d", i);
4985 rcu_dereference(sbi->s_group_desc)[i] = bh;
4988 sbi->s_gdb_count = db_count;
4989 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4990 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4991 ret = -EFSCORRUPTED;
4995 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4996 spin_lock_init(&sbi->s_error_lock);
4997 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4999 /* Register extent status tree shrinker */
5000 if (ext4_es_register_shrinker(sbi))
5003 sbi->s_stripe = ext4_get_stripe_size(sbi);
5004 sbi->s_extent_max_zeroout_kb = 32;
5007 * set up enough so that it can read an inode
5009 sb->s_op = &ext4_sops;
5010 sb->s_export_op = &ext4_export_ops;
5011 sb->s_xattr = ext4_xattr_handlers;
5012 #ifdef CONFIG_FS_ENCRYPTION
5013 sb->s_cop = &ext4_cryptops;
5015 #ifdef CONFIG_FS_VERITY
5016 sb->s_vop = &ext4_verityops;
5019 sb->dq_op = &ext4_quota_operations;
5020 if (ext4_has_feature_quota(sb))
5021 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5023 sb->s_qcop = &ext4_qctl_operations;
5024 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5026 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5028 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5029 mutex_init(&sbi->s_orphan_lock);
5031 /* Initialize fast commit stuff */
5032 atomic_set(&sbi->s_fc_subtid, 0);
5033 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
5034 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
5035 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
5036 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
5037 sbi->s_fc_bytes = 0;
5038 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
5039 sbi->s_fc_ineligible_tid = 0;
5040 spin_lock_init(&sbi->s_fc_lock);
5041 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
5042 sbi->s_fc_replay_state.fc_regions = NULL;
5043 sbi->s_fc_replay_state.fc_regions_size = 0;
5044 sbi->s_fc_replay_state.fc_regions_used = 0;
5045 sbi->s_fc_replay_state.fc_regions_valid = 0;
5046 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
5047 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
5048 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
5052 needs_recovery = (es->s_last_orphan != 0 ||
5053 ext4_has_feature_orphan_present(sb) ||
5054 ext4_has_feature_journal_needs_recovery(sb));
5056 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5057 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5058 goto failed_mount3a;
5061 * The first inode we look at is the journal inode. Don't try
5062 * root first: it may be modified in the journal!
5064 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5065 err = ext4_load_journal(sb, es, ctx->journal_devnum);
5067 goto failed_mount3a;
5068 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5069 ext4_has_feature_journal_needs_recovery(sb)) {
5070 ext4_msg(sb, KERN_ERR, "required journal recovery "
5071 "suppressed and not mounted read-only");
5072 goto failed_mount_wq;
5074 /* Nojournal mode, all journal mount options are illegal */
5075 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5076 ext4_msg(sb, KERN_ERR, "can't mount with "
5077 "journal_checksum, fs mounted w/o journal");
5078 goto failed_mount_wq;
5080 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5081 ext4_msg(sb, KERN_ERR, "can't mount with "
5082 "journal_async_commit, fs mounted w/o journal");
5083 goto failed_mount_wq;
5085 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5086 ext4_msg(sb, KERN_ERR, "can't mount with "
5087 "commit=%lu, fs mounted w/o journal",
5088 sbi->s_commit_interval / HZ);
5089 goto failed_mount_wq;
5091 if (EXT4_MOUNT_DATA_FLAGS &
5092 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5093 ext4_msg(sb, KERN_ERR, "can't mount with "
5094 "data=, fs mounted w/o journal");
5095 goto failed_mount_wq;
5097 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5098 clear_opt(sb, JOURNAL_CHECKSUM);
5099 clear_opt(sb, DATA_FLAGS);
5100 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5101 sbi->s_journal = NULL;
5106 if (ext4_has_feature_64bit(sb) &&
5107 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5108 JBD2_FEATURE_INCOMPAT_64BIT)) {
5109 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
5110 goto failed_mount_wq;
5113 if (!set_journal_csum_feature_set(sb)) {
5114 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
5116 goto failed_mount_wq;
5119 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
5120 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5121 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
5122 ext4_msg(sb, KERN_ERR,
5123 "Failed to set fast commit journal feature");
5124 goto failed_mount_wq;
5127 /* We have now updated the journal if required, so we can
5128 * validate the data journaling mode. */
5129 switch (test_opt(sb, DATA_FLAGS)) {
5131 /* No mode set, assume a default based on the journal
5132 * capabilities: ORDERED_DATA if the journal can
5133 * cope, else JOURNAL_DATA
5135 if (jbd2_journal_check_available_features
5136 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5137 set_opt(sb, ORDERED_DATA);
5138 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
5140 set_opt(sb, JOURNAL_DATA);
5141 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
5145 case EXT4_MOUNT_ORDERED_DATA:
5146 case EXT4_MOUNT_WRITEBACK_DATA:
5147 if (!jbd2_journal_check_available_features
5148 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5149 ext4_msg(sb, KERN_ERR, "Journal does not support "
5150 "requested data journaling mode");
5151 goto failed_mount_wq;
5158 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
5159 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5160 ext4_msg(sb, KERN_ERR, "can't mount with "
5161 "journal_async_commit in data=ordered mode");
5162 goto failed_mount_wq;
5165 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
5167 sbi->s_journal->j_submit_inode_data_buffers =
5168 ext4_journal_submit_inode_data_buffers;
5169 sbi->s_journal->j_finish_inode_data_buffers =
5170 ext4_journal_finish_inode_data_buffers;
5173 if (!test_opt(sb, NO_MBCACHE)) {
5174 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5175 if (!sbi->s_ea_block_cache) {
5176 ext4_msg(sb, KERN_ERR,
5177 "Failed to create ea_block_cache");
5178 goto failed_mount_wq;
5181 if (ext4_has_feature_ea_inode(sb)) {
5182 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5183 if (!sbi->s_ea_inode_cache) {
5184 ext4_msg(sb, KERN_ERR,
5185 "Failed to create ea_inode_cache");
5186 goto failed_mount_wq;
5191 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
5192 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5193 goto failed_mount_wq;
5197 * Get the # of file system overhead blocks from the
5198 * superblock if present.
5200 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5201 /* ignore the precalculated value if it is ridiculous */
5202 if (sbi->s_overhead > ext4_blocks_count(es))
5203 sbi->s_overhead = 0;
5205 * If the bigalloc feature is not enabled recalculating the
5206 * overhead doesn't take long, so we might as well just redo
5207 * it to make sure we are using the correct value.
5209 if (!ext4_has_feature_bigalloc(sb))
5210 sbi->s_overhead = 0;
5211 if (sbi->s_overhead == 0) {
5212 err = ext4_calculate_overhead(sb);
5214 goto failed_mount_wq;
5218 * The maximum number of concurrent works can be high and
5219 * concurrency isn't really necessary. Limit it to 1.
5221 EXT4_SB(sb)->rsv_conversion_wq =
5222 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5223 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5224 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5230 * The jbd2_journal_load will have done any necessary log recovery,
5231 * so we can safely mount the rest of the filesystem now.
5234 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5236 ext4_msg(sb, KERN_ERR, "get root inode failed");
5237 ret = PTR_ERR(root);
5241 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5242 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5247 sb->s_root = d_make_root(root);
5249 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5254 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5255 if (ret == -EROFS) {
5256 sb->s_flags |= SB_RDONLY;
5259 goto failed_mount4a;
5261 ext4_set_resv_clusters(sb);
5263 if (test_opt(sb, BLOCK_VALIDITY)) {
5264 err = ext4_setup_system_zone(sb);
5266 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5268 goto failed_mount4a;
5271 ext4_fc_replay_cleanup(sb);
5276 * Enable optimize_scan if number of groups is > threshold. This can be
5277 * turned off by passing "mb_optimize_scan=0". This can also be
5278 * turned on forcefully by passing "mb_optimize_scan=1".
5280 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5281 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5282 set_opt2(sb, MB_OPTIMIZE_SCAN);
5284 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5287 err = ext4_mb_init(sb);
5289 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5295 * We can only set up the journal commit callback once
5296 * mballoc is initialized
5299 sbi->s_journal->j_commit_callback =
5300 ext4_journal_commit_callback;
5302 block = ext4_count_free_clusters(sb);
5303 ext4_free_blocks_count_set(sbi->s_es,
5304 EXT4_C2B(sbi, block));
5305 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5308 unsigned long freei = ext4_count_free_inodes(sb);
5309 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5310 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5314 err = percpu_counter_init(&sbi->s_dirs_counter,
5315 ext4_count_dirs(sb), GFP_KERNEL);
5317 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5320 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5323 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5326 ext4_msg(sb, KERN_ERR, "insufficient memory");
5330 if (ext4_has_feature_flex_bg(sb))
5331 if (!ext4_fill_flex_info(sb)) {
5332 ext4_msg(sb, KERN_ERR,
5333 "unable to initialize "
5334 "flex_bg meta info!");
5339 err = ext4_register_li_request(sb, first_not_zeroed);
5343 err = ext4_register_sysfs(sb);
5347 err = ext4_init_orphan_info(sb);
5351 /* Enable quota usage during mount. */
5352 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5353 err = ext4_enable_quotas(sb);
5357 #endif /* CONFIG_QUOTA */
5360 * Save the original bdev mapping's wb_err value which could be
5361 * used to detect the metadata async write error.
5363 spin_lock_init(&sbi->s_bdev_wb_lock);
5364 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5365 &sbi->s_bdev_wb_err);
5366 sb->s_bdev->bd_super = sb;
5367 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5368 ext4_orphan_cleanup(sb, es);
5369 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5371 * Update the checksum after updating free space/inode counters and
5372 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5373 * checksum in the buffer cache until it is written out and
5374 * e2fsprogs programs trying to open a file system immediately
5375 * after it is mounted can fail.
5377 ext4_superblock_csum_set(sb);
5378 if (needs_recovery) {
5379 ext4_msg(sb, KERN_INFO, "recovery complete");
5380 err = ext4_mark_recovery_complete(sb, es);
5385 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5386 ext4_msg(sb, KERN_WARNING,
5387 "mounting with \"discard\" option, but the device does not support discard");
5389 if (es->s_error_count)
5390 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5392 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5393 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5394 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5395 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5396 atomic_set(&sbi->s_warning_count, 0);
5397 atomic_set(&sbi->s_msg_count, 0);
5403 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5407 ext4_release_orphan_info(sb);
5409 ext4_unregister_sysfs(sb);
5410 kobject_put(&sbi->s_kobj);
5412 ext4_unregister_li_request(sb);
5414 ext4_mb_release(sb);
5416 flex_groups = rcu_dereference(sbi->s_flex_groups);
5418 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5419 kvfree(flex_groups[i]);
5420 kvfree(flex_groups);
5423 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5424 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5425 percpu_counter_destroy(&sbi->s_dirs_counter);
5426 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5427 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5428 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5430 ext4_ext_release(sb);
5431 ext4_release_system_zone(sb);
5436 ext4_msg(sb, KERN_ERR, "mount failed");
5437 if (EXT4_SB(sb)->rsv_conversion_wq)
5438 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5440 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5441 sbi->s_ea_inode_cache = NULL;
5443 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5444 sbi->s_ea_block_cache = NULL;
5446 if (sbi->s_journal) {
5447 /* flush s_error_work before journal destroy. */
5448 flush_work(&sbi->s_error_work);
5449 jbd2_journal_destroy(sbi->s_journal);
5450 sbi->s_journal = NULL;
5453 ext4_es_unregister_shrinker(sbi);
5455 /* flush s_error_work before sbi destroy */
5456 flush_work(&sbi->s_error_work);
5457 del_timer_sync(&sbi->s_err_report);
5458 ext4_stop_mmpd(sbi);
5461 group_desc = rcu_dereference(sbi->s_group_desc);
5462 for (i = 0; i < db_count; i++)
5463 brelse(group_desc[i]);
5467 if (sbi->s_chksum_driver)
5468 crypto_free_shash(sbi->s_chksum_driver);
5470 #if IS_ENABLED(CONFIG_UNICODE)
5471 utf8_unload(sb->s_encoding);
5475 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5476 kfree(get_qf_name(sb, sbi, i));
5478 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5479 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5481 ext4_blkdev_remove(sbi);
5483 sb->s_fs_info = NULL;
5484 return err ? err : ret;
5487 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5489 struct ext4_fs_context *ctx = fc->fs_private;
5490 struct ext4_sb_info *sbi;
5494 sbi = ext4_alloc_sbi(sb);
5498 fc->s_fs_info = sbi;
5500 /* Cleanup superblock name */
5501 strreplace(sb->s_id, '/', '!');
5503 sbi->s_sb_block = 1; /* Default super block location */
5504 if (ctx->spec & EXT4_SPEC_s_sb_block)
5505 sbi->s_sb_block = ctx->s_sb_block;
5507 ret = __ext4_fill_super(fc, sb);
5511 if (sbi->s_journal) {
5512 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5513 descr = " journalled data mode";
5514 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5515 descr = " ordered data mode";
5517 descr = " writeback data mode";
5519 descr = "out journal";
5521 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5522 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5523 "Quota mode: %s.", descr, ext4_quota_mode(sb));
5525 /* Update the s_overhead_clusters if necessary */
5526 ext4_update_overhead(sb);
5531 fc->s_fs_info = NULL;
5535 static int ext4_get_tree(struct fs_context *fc)
5537 return get_tree_bdev(fc, ext4_fill_super);
5541 * Setup any per-fs journal parameters now. We'll do this both on
5542 * initial mount, once the journal has been initialised but before we've
5543 * done any recovery; and again on any subsequent remount.
5545 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5547 struct ext4_sb_info *sbi = EXT4_SB(sb);
5549 journal->j_commit_interval = sbi->s_commit_interval;
5550 journal->j_min_batch_time = sbi->s_min_batch_time;
5551 journal->j_max_batch_time = sbi->s_max_batch_time;
5552 ext4_fc_init(sb, journal);
5554 write_lock(&journal->j_state_lock);
5555 if (test_opt(sb, BARRIER))
5556 journal->j_flags |= JBD2_BARRIER;
5558 journal->j_flags &= ~JBD2_BARRIER;
5559 if (test_opt(sb, DATA_ERR_ABORT))
5560 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5562 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5563 write_unlock(&journal->j_state_lock);
5566 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5567 unsigned int journal_inum)
5569 struct inode *journal_inode;
5572 * Test for the existence of a valid inode on disk. Bad things
5573 * happen if we iget() an unused inode, as the subsequent iput()
5574 * will try to delete it.
5576 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5577 if (IS_ERR(journal_inode)) {
5578 ext4_msg(sb, KERN_ERR, "no journal found");
5581 if (!journal_inode->i_nlink) {
5582 make_bad_inode(journal_inode);
5583 iput(journal_inode);
5584 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5588 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5589 journal_inode, journal_inode->i_size);
5590 if (!S_ISREG(journal_inode->i_mode)) {
5591 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5592 iput(journal_inode);
5595 return journal_inode;
5598 static journal_t *ext4_get_journal(struct super_block *sb,
5599 unsigned int journal_inum)
5601 struct inode *journal_inode;
5604 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5607 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5611 journal = jbd2_journal_init_inode(journal_inode);
5613 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5614 iput(journal_inode);
5617 journal->j_private = sb;
5618 ext4_init_journal_params(sb, journal);
5622 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5625 struct buffer_head *bh;
5629 int hblock, blocksize;
5630 ext4_fsblk_t sb_block;
5631 unsigned long offset;
5632 struct ext4_super_block *es;
5633 struct block_device *bdev;
5635 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5638 bdev = ext4_blkdev_get(j_dev, sb);
5642 blocksize = sb->s_blocksize;
5643 hblock = bdev_logical_block_size(bdev);
5644 if (blocksize < hblock) {
5645 ext4_msg(sb, KERN_ERR,
5646 "blocksize too small for journal device");
5650 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5651 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5652 set_blocksize(bdev, blocksize);
5653 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5654 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5655 "external journal");
5659 es = (struct ext4_super_block *) (bh->b_data + offset);
5660 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5661 !(le32_to_cpu(es->s_feature_incompat) &
5662 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5663 ext4_msg(sb, KERN_ERR, "external journal has "
5669 if ((le32_to_cpu(es->s_feature_ro_compat) &
5670 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5671 es->s_checksum != ext4_superblock_csum(sb, es)) {
5672 ext4_msg(sb, KERN_ERR, "external journal has "
5673 "corrupt superblock");
5678 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5679 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5684 len = ext4_blocks_count(es);
5685 start = sb_block + 1;
5686 brelse(bh); /* we're done with the superblock */
5688 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5689 start, len, blocksize);
5691 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5694 journal->j_private = sb;
5695 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5696 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5699 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5700 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5701 "user (unsupported) - %d",
5702 be32_to_cpu(journal->j_superblock->s_nr_users));
5705 EXT4_SB(sb)->s_journal_bdev = bdev;
5706 ext4_init_journal_params(sb, journal);
5710 jbd2_journal_destroy(journal);
5712 ext4_blkdev_put(bdev);
5716 static int ext4_load_journal(struct super_block *sb,
5717 struct ext4_super_block *es,
5718 unsigned long journal_devnum)
5721 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5724 int really_read_only;
5727 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5728 return -EFSCORRUPTED;
5730 if (journal_devnum &&
5731 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5732 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5733 "numbers have changed");
5734 journal_dev = new_decode_dev(journal_devnum);
5736 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5738 if (journal_inum && journal_dev) {
5739 ext4_msg(sb, KERN_ERR,
5740 "filesystem has both journal inode and journal device!");
5745 journal = ext4_get_journal(sb, journal_inum);
5749 journal = ext4_get_dev_journal(sb, journal_dev);
5754 journal_dev_ro = bdev_read_only(journal->j_dev);
5755 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5757 if (journal_dev_ro && !sb_rdonly(sb)) {
5758 ext4_msg(sb, KERN_ERR,
5759 "journal device read-only, try mounting with '-o ro'");
5765 * Are we loading a blank journal or performing recovery after a
5766 * crash? For recovery, we need to check in advance whether we
5767 * can get read-write access to the device.
5769 if (ext4_has_feature_journal_needs_recovery(sb)) {
5770 if (sb_rdonly(sb)) {
5771 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5772 "required on readonly filesystem");
5773 if (really_read_only) {
5774 ext4_msg(sb, KERN_ERR, "write access "
5775 "unavailable, cannot proceed "
5776 "(try mounting with noload)");
5780 ext4_msg(sb, KERN_INFO, "write access will "
5781 "be enabled during recovery");
5785 if (!(journal->j_flags & JBD2_BARRIER))
5786 ext4_msg(sb, KERN_INFO, "barriers disabled");
5788 if (!ext4_has_feature_journal_needs_recovery(sb))
5789 err = jbd2_journal_wipe(journal, !really_read_only);
5791 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5793 memcpy(save, ((char *) es) +
5794 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5795 err = jbd2_journal_load(journal);
5797 memcpy(((char *) es) + EXT4_S_ERR_START,
5798 save, EXT4_S_ERR_LEN);
5803 ext4_msg(sb, KERN_ERR, "error loading journal");
5807 EXT4_SB(sb)->s_journal = journal;
5808 err = ext4_clear_journal_err(sb, es);
5810 EXT4_SB(sb)->s_journal = NULL;
5811 jbd2_journal_destroy(journal);
5815 if (!really_read_only && journal_devnum &&
5816 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5817 es->s_journal_dev = cpu_to_le32(journal_devnum);
5819 /* Make sure we flush the recovery flag to disk. */
5820 ext4_commit_super(sb);
5826 jbd2_journal_destroy(journal);
5830 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5831 static void ext4_update_super(struct super_block *sb)
5833 struct ext4_sb_info *sbi = EXT4_SB(sb);
5834 struct ext4_super_block *es = sbi->s_es;
5835 struct buffer_head *sbh = sbi->s_sbh;
5839 * If the file system is mounted read-only, don't update the
5840 * superblock write time. This avoids updating the superblock
5841 * write time when we are mounting the root file system
5842 * read/only but we need to replay the journal; at that point,
5843 * for people who are east of GMT and who make their clock
5844 * tick in localtime for Windows bug-for-bug compatibility,
5845 * the clock is set in the future, and this will cause e2fsck
5846 * to complain and force a full file system check.
5848 if (!(sb->s_flags & SB_RDONLY))
5849 ext4_update_tstamp(es, s_wtime);
5850 es->s_kbytes_written =
5851 cpu_to_le64(sbi->s_kbytes_written +
5852 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5853 sbi->s_sectors_written_start) >> 1));
5854 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5855 ext4_free_blocks_count_set(es,
5856 EXT4_C2B(sbi, percpu_counter_sum_positive(
5857 &sbi->s_freeclusters_counter)));
5858 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5859 es->s_free_inodes_count =
5860 cpu_to_le32(percpu_counter_sum_positive(
5861 &sbi->s_freeinodes_counter));
5862 /* Copy error information to the on-disk superblock */
5863 spin_lock(&sbi->s_error_lock);
5864 if (sbi->s_add_error_count > 0) {
5865 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5866 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5867 __ext4_update_tstamp(&es->s_first_error_time,
5868 &es->s_first_error_time_hi,
5869 sbi->s_first_error_time);
5870 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5871 sizeof(es->s_first_error_func));
5872 es->s_first_error_line =
5873 cpu_to_le32(sbi->s_first_error_line);
5874 es->s_first_error_ino =
5875 cpu_to_le32(sbi->s_first_error_ino);
5876 es->s_first_error_block =
5877 cpu_to_le64(sbi->s_first_error_block);
5878 es->s_first_error_errcode =
5879 ext4_errno_to_code(sbi->s_first_error_code);
5881 __ext4_update_tstamp(&es->s_last_error_time,
5882 &es->s_last_error_time_hi,
5883 sbi->s_last_error_time);
5884 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5885 sizeof(es->s_last_error_func));
5886 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5887 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5888 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5889 es->s_last_error_errcode =
5890 ext4_errno_to_code(sbi->s_last_error_code);
5892 * Start the daily error reporting function if it hasn't been
5895 if (!es->s_error_count)
5896 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5897 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5898 sbi->s_add_error_count = 0;
5900 spin_unlock(&sbi->s_error_lock);
5902 ext4_superblock_csum_set(sb);
5906 static int ext4_commit_super(struct super_block *sb)
5908 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5912 if (block_device_ejected(sb))
5915 ext4_update_super(sb);
5918 /* Buffer got discarded which means block device got invalidated */
5919 if (!buffer_mapped(sbh)) {
5924 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5926 * Oh, dear. A previous attempt to write the
5927 * superblock failed. This could happen because the
5928 * USB device was yanked out. Or it could happen to
5929 * be a transient write error and maybe the block will
5930 * be remapped. Nothing we can do but to retry the
5931 * write and hope for the best.
5933 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5934 "superblock detected");
5935 clear_buffer_write_io_error(sbh);
5936 set_buffer_uptodate(sbh);
5939 /* Clear potential dirty bit if it was journalled update */
5940 clear_buffer_dirty(sbh);
5941 sbh->b_end_io = end_buffer_write_sync;
5942 submit_bh(REQ_OP_WRITE,
5943 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
5944 wait_on_buffer(sbh);
5945 if (buffer_write_io_error(sbh)) {
5946 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5948 clear_buffer_write_io_error(sbh);
5949 set_buffer_uptodate(sbh);
5956 * Have we just finished recovery? If so, and if we are mounting (or
5957 * remounting) the filesystem readonly, then we will end up with a
5958 * consistent fs on disk. Record that fact.
5960 static int ext4_mark_recovery_complete(struct super_block *sb,
5961 struct ext4_super_block *es)
5964 journal_t *journal = EXT4_SB(sb)->s_journal;
5966 if (!ext4_has_feature_journal(sb)) {
5967 if (journal != NULL) {
5968 ext4_error(sb, "Journal got removed while the fs was "
5970 return -EFSCORRUPTED;
5974 jbd2_journal_lock_updates(journal);
5975 err = jbd2_journal_flush(journal, 0);
5979 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
5980 ext4_has_feature_orphan_present(sb))) {
5981 if (!ext4_orphan_file_empty(sb)) {
5982 ext4_error(sb, "Orphan file not empty on read-only fs.");
5983 err = -EFSCORRUPTED;
5986 ext4_clear_feature_journal_needs_recovery(sb);
5987 ext4_clear_feature_orphan_present(sb);
5988 ext4_commit_super(sb);
5991 jbd2_journal_unlock_updates(journal);
5996 * If we are mounting (or read-write remounting) a filesystem whose journal
5997 * has recorded an error from a previous lifetime, move that error to the
5998 * main filesystem now.
6000 static int ext4_clear_journal_err(struct super_block *sb,
6001 struct ext4_super_block *es)
6007 if (!ext4_has_feature_journal(sb)) {
6008 ext4_error(sb, "Journal got removed while the fs was mounted!");
6009 return -EFSCORRUPTED;
6012 journal = EXT4_SB(sb)->s_journal;
6015 * Now check for any error status which may have been recorded in the
6016 * journal by a prior ext4_error() or ext4_abort()
6019 j_errno = jbd2_journal_errno(journal);
6023 errstr = ext4_decode_error(sb, j_errno, nbuf);
6024 ext4_warning(sb, "Filesystem error recorded "
6025 "from previous mount: %s", errstr);
6026 ext4_warning(sb, "Marking fs in need of filesystem check.");
6028 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6029 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6030 ext4_commit_super(sb);
6032 jbd2_journal_clear_err(journal);
6033 jbd2_journal_update_sb_errno(journal);
6039 * Force the running and committing transactions to commit,
6040 * and wait on the commit.
6042 int ext4_force_commit(struct super_block *sb)
6049 journal = EXT4_SB(sb)->s_journal;
6050 return ext4_journal_force_commit(journal);
6053 static int ext4_sync_fs(struct super_block *sb, int wait)
6057 bool needs_barrier = false;
6058 struct ext4_sb_info *sbi = EXT4_SB(sb);
6060 if (unlikely(ext4_forced_shutdown(sbi)))
6063 trace_ext4_sync_fs(sb, wait);
6064 flush_workqueue(sbi->rsv_conversion_wq);
6066 * Writeback quota in non-journalled quota case - journalled quota has
6069 dquot_writeback_dquots(sb, -1);
6071 * Data writeback is possible w/o journal transaction, so barrier must
6072 * being sent at the end of the function. But we can skip it if
6073 * transaction_commit will do it for us.
6075 if (sbi->s_journal) {
6076 target = jbd2_get_latest_transaction(sbi->s_journal);
6077 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6078 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6079 needs_barrier = true;
6081 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6083 ret = jbd2_log_wait_commit(sbi->s_journal,
6086 } else if (wait && test_opt(sb, BARRIER))
6087 needs_barrier = true;
6088 if (needs_barrier) {
6090 err = blkdev_issue_flush(sb->s_bdev);
6099 * LVM calls this function before a (read-only) snapshot is created. This
6100 * gives us a chance to flush the journal completely and mark the fs clean.
6102 * Note that only this function cannot bring a filesystem to be in a clean
6103 * state independently. It relies on upper layer to stop all data & metadata
6106 static int ext4_freeze(struct super_block *sb)
6114 journal = EXT4_SB(sb)->s_journal;
6117 /* Now we set up the journal barrier. */
6118 jbd2_journal_lock_updates(journal);
6121 * Don't clear the needs_recovery flag if we failed to
6122 * flush the journal.
6124 error = jbd2_journal_flush(journal, 0);
6128 /* Journal blocked and flushed, clear needs_recovery flag. */
6129 ext4_clear_feature_journal_needs_recovery(sb);
6130 if (ext4_orphan_file_empty(sb))
6131 ext4_clear_feature_orphan_present(sb);
6134 error = ext4_commit_super(sb);
6137 /* we rely on upper layer to stop further updates */
6138 jbd2_journal_unlock_updates(journal);
6143 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6144 * flag here, even though the filesystem is not technically dirty yet.
6146 static int ext4_unfreeze(struct super_block *sb)
6148 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6151 if (EXT4_SB(sb)->s_journal) {
6152 /* Reset the needs_recovery flag before the fs is unlocked. */
6153 ext4_set_feature_journal_needs_recovery(sb);
6154 if (ext4_has_feature_orphan_file(sb))
6155 ext4_set_feature_orphan_present(sb);
6158 ext4_commit_super(sb);
6163 * Structure to save mount options for ext4_remount's benefit
6165 struct ext4_mount_options {
6166 unsigned long s_mount_opt;
6167 unsigned long s_mount_opt2;
6170 unsigned long s_commit_interval;
6171 u32 s_min_batch_time, s_max_batch_time;
6174 char *s_qf_names[EXT4_MAXQUOTAS];
6178 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6180 struct ext4_fs_context *ctx = fc->fs_private;
6181 struct ext4_super_block *es;
6182 struct ext4_sb_info *sbi = EXT4_SB(sb);
6183 unsigned long old_sb_flags;
6184 struct ext4_mount_options old_opts;
6188 int enable_quota = 0;
6190 char *to_free[EXT4_MAXQUOTAS];
6194 /* Store the original options */
6195 old_sb_flags = sb->s_flags;
6196 old_opts.s_mount_opt = sbi->s_mount_opt;
6197 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6198 old_opts.s_resuid = sbi->s_resuid;
6199 old_opts.s_resgid = sbi->s_resgid;
6200 old_opts.s_commit_interval = sbi->s_commit_interval;
6201 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6202 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6204 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6205 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6206 if (sbi->s_qf_names[i]) {
6207 char *qf_name = get_qf_name(sb, sbi, i);
6209 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6210 if (!old_opts.s_qf_names[i]) {
6211 for (j = 0; j < i; j++)
6212 kfree(old_opts.s_qf_names[j]);
6216 old_opts.s_qf_names[i] = NULL;
6218 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6219 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6220 ctx->journal_ioprio =
6221 sbi->s_journal->j_task->io_context->ioprio;
6223 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6227 ext4_apply_options(fc, sb);
6229 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6230 test_opt(sb, JOURNAL_CHECKSUM)) {
6231 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6232 "during remount not supported; ignoring");
6233 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6236 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6237 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6238 ext4_msg(sb, KERN_ERR, "can't mount with "
6239 "both data=journal and delalloc");
6243 if (test_opt(sb, DIOREAD_NOLOCK)) {
6244 ext4_msg(sb, KERN_ERR, "can't mount with "
6245 "both data=journal and dioread_nolock");
6249 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6250 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6251 ext4_msg(sb, KERN_ERR, "can't mount with "
6252 "journal_async_commit in data=ordered mode");
6258 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6259 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6264 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6265 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6267 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6268 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6272 if (sbi->s_journal) {
6273 ext4_init_journal_params(sb, sbi->s_journal);
6274 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6277 /* Flush outstanding errors before changing fs state */
6278 flush_work(&sbi->s_error_work);
6280 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6281 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6286 if (fc->sb_flags & SB_RDONLY) {
6287 err = sync_filesystem(sb);
6290 err = dquot_suspend(sb, -1);
6295 * First of all, the unconditional stuff we have to do
6296 * to disable replay of the journal when we next remount
6298 sb->s_flags |= SB_RDONLY;
6301 * OK, test if we are remounting a valid rw partition
6302 * readonly, and if so set the rdonly flag and then
6303 * mark the partition as valid again.
6305 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6306 (sbi->s_mount_state & EXT4_VALID_FS))
6307 es->s_state = cpu_to_le16(sbi->s_mount_state);
6309 if (sbi->s_journal) {
6311 * We let remount-ro finish even if marking fs
6312 * as clean failed...
6314 ext4_mark_recovery_complete(sb, es);
6317 /* Make sure we can mount this feature set readwrite */
6318 if (ext4_has_feature_readonly(sb) ||
6319 !ext4_feature_set_ok(sb, 0)) {
6324 * Make sure the group descriptor checksums
6325 * are sane. If they aren't, refuse to remount r/w.
6327 for (g = 0; g < sbi->s_groups_count; g++) {
6328 struct ext4_group_desc *gdp =
6329 ext4_get_group_desc(sb, g, NULL);
6331 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6332 ext4_msg(sb, KERN_ERR,
6333 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6334 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6335 le16_to_cpu(gdp->bg_checksum));
6342 * If we have an unprocessed orphan list hanging
6343 * around from a previously readonly bdev mount,
6344 * require a full umount/remount for now.
6346 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6347 ext4_msg(sb, KERN_WARNING, "Couldn't "
6348 "remount RDWR because of unprocessed "
6349 "orphan inode list. Please "
6350 "umount/remount instead");
6356 * Mounting a RDONLY partition read-write, so reread
6357 * and store the current valid flag. (It may have
6358 * been changed by e2fsck since we originally mounted
6361 if (sbi->s_journal) {
6362 err = ext4_clear_journal_err(sb, es);
6366 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6369 err = ext4_setup_super(sb, es, 0);
6373 sb->s_flags &= ~SB_RDONLY;
6374 if (ext4_has_feature_mmp(sb))
6375 if (ext4_multi_mount_protect(sb,
6376 le64_to_cpu(es->s_mmp_block))) {
6387 * Reinitialize lazy itable initialization thread based on
6390 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6391 ext4_unregister_li_request(sb);
6393 ext4_group_t first_not_zeroed;
6394 first_not_zeroed = ext4_has_uninit_itable(sb);
6395 ext4_register_li_request(sb, first_not_zeroed);
6399 * Handle creation of system zone data early because it can fail.
6400 * Releasing of existing data is done when we are sure remount will
6403 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6404 err = ext4_setup_system_zone(sb);
6409 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6410 err = ext4_commit_super(sb);
6416 /* Release old quota file names */
6417 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6418 kfree(old_opts.s_qf_names[i]);
6420 if (sb_any_quota_suspended(sb))
6421 dquot_resume(sb, -1);
6422 else if (ext4_has_feature_quota(sb)) {
6423 err = ext4_enable_quotas(sb);
6429 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6430 ext4_release_system_zone(sb);
6432 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6433 ext4_stop_mmpd(sbi);
6438 sb->s_flags = old_sb_flags;
6439 sbi->s_mount_opt = old_opts.s_mount_opt;
6440 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6441 sbi->s_resuid = old_opts.s_resuid;
6442 sbi->s_resgid = old_opts.s_resgid;
6443 sbi->s_commit_interval = old_opts.s_commit_interval;
6444 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6445 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6446 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6447 ext4_release_system_zone(sb);
6449 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6450 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6451 to_free[i] = get_qf_name(sb, sbi, i);
6452 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6455 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6458 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6459 ext4_stop_mmpd(sbi);
6463 static int ext4_reconfigure(struct fs_context *fc)
6465 struct super_block *sb = fc->root->d_sb;
6468 fc->s_fs_info = EXT4_SB(sb);
6470 ret = ext4_check_opt_consistency(fc, sb);
6474 ret = __ext4_remount(fc, sb);
6478 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.",
6479 ext4_quota_mode(sb));
6485 static int ext4_statfs_project(struct super_block *sb,
6486 kprojid_t projid, struct kstatfs *buf)
6489 struct dquot *dquot;
6493 qid = make_kqid_projid(projid);
6494 dquot = dqget(sb, qid);
6496 return PTR_ERR(dquot);
6497 spin_lock(&dquot->dq_dqb_lock);
6499 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6500 dquot->dq_dqb.dqb_bhardlimit);
6501 limit >>= sb->s_blocksize_bits;
6503 if (limit && buf->f_blocks > limit) {
6504 curblock = (dquot->dq_dqb.dqb_curspace +
6505 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6506 buf->f_blocks = limit;
6507 buf->f_bfree = buf->f_bavail =
6508 (buf->f_blocks > curblock) ?
6509 (buf->f_blocks - curblock) : 0;
6512 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6513 dquot->dq_dqb.dqb_ihardlimit);
6514 if (limit && buf->f_files > limit) {
6515 buf->f_files = limit;
6517 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6518 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6521 spin_unlock(&dquot->dq_dqb_lock);
6527 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6529 struct super_block *sb = dentry->d_sb;
6530 struct ext4_sb_info *sbi = EXT4_SB(sb);
6531 struct ext4_super_block *es = sbi->s_es;
6532 ext4_fsblk_t overhead = 0, resv_blocks;
6534 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6536 if (!test_opt(sb, MINIX_DF))
6537 overhead = sbi->s_overhead;
6539 buf->f_type = EXT4_SUPER_MAGIC;
6540 buf->f_bsize = sb->s_blocksize;
6541 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6542 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6543 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6544 /* prevent underflow in case that few free space is available */
6545 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6546 buf->f_bavail = buf->f_bfree -
6547 (ext4_r_blocks_count(es) + resv_blocks);
6548 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6550 buf->f_files = le32_to_cpu(es->s_inodes_count);
6551 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6552 buf->f_namelen = EXT4_NAME_LEN;
6553 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6556 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6557 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6558 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6567 * Helper functions so that transaction is started before we acquire dqio_sem
6568 * to keep correct lock ordering of transaction > dqio_sem
6570 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6572 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6575 static int ext4_write_dquot(struct dquot *dquot)
6579 struct inode *inode;
6581 inode = dquot_to_inode(dquot);
6582 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6583 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6585 return PTR_ERR(handle);
6586 ret = dquot_commit(dquot);
6587 err = ext4_journal_stop(handle);
6593 static int ext4_acquire_dquot(struct dquot *dquot)
6598 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6599 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6601 return PTR_ERR(handle);
6602 ret = dquot_acquire(dquot);
6603 err = ext4_journal_stop(handle);
6609 static int ext4_release_dquot(struct dquot *dquot)
6614 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6615 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6616 if (IS_ERR(handle)) {
6617 /* Release dquot anyway to avoid endless cycle in dqput() */
6618 dquot_release(dquot);
6619 return PTR_ERR(handle);
6621 ret = dquot_release(dquot);
6622 err = ext4_journal_stop(handle);
6628 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6630 struct super_block *sb = dquot->dq_sb;
6632 if (ext4_is_quota_journalled(sb)) {
6633 dquot_mark_dquot_dirty(dquot);
6634 return ext4_write_dquot(dquot);
6636 return dquot_mark_dquot_dirty(dquot);
6640 static int ext4_write_info(struct super_block *sb, int type)
6645 /* Data block + inode block */
6646 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6648 return PTR_ERR(handle);
6649 ret = dquot_commit_info(sb, type);
6650 err = ext4_journal_stop(handle);
6656 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6658 struct ext4_inode_info *ei = EXT4_I(inode);
6660 /* The first argument of lockdep_set_subclass has to be
6661 * *exactly* the same as the argument to init_rwsem() --- in
6662 * this case, in init_once() --- or lockdep gets unhappy
6663 * because the name of the lock is set using the
6664 * stringification of the argument to init_rwsem().
6666 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6667 lockdep_set_subclass(&ei->i_data_sem, subclass);
6671 * Standard function to be called on quota_on
6673 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6674 const struct path *path)
6678 if (!test_opt(sb, QUOTA))
6681 /* Quotafile not on the same filesystem? */
6682 if (path->dentry->d_sb != sb)
6685 /* Quota already enabled for this file? */
6686 if (IS_NOQUOTA(d_inode(path->dentry)))
6689 /* Journaling quota? */
6690 if (EXT4_SB(sb)->s_qf_names[type]) {
6691 /* Quotafile not in fs root? */
6692 if (path->dentry->d_parent != sb->s_root)
6693 ext4_msg(sb, KERN_WARNING,
6694 "Quota file not on filesystem root. "
6695 "Journaled quota will not work");
6696 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6699 * Clear the flag just in case mount options changed since
6702 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6706 * When we journal data on quota file, we have to flush journal to see
6707 * all updates to the file when we bypass pagecache...
6709 if (EXT4_SB(sb)->s_journal &&
6710 ext4_should_journal_data(d_inode(path->dentry))) {
6712 * We don't need to lock updates but journal_flush() could
6713 * otherwise be livelocked...
6715 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6716 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6717 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6722 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6723 err = dquot_quota_on(sb, type, format_id, path);
6725 struct inode *inode = d_inode(path->dentry);
6729 * Set inode flags to prevent userspace from messing with quota
6730 * files. If this fails, we return success anyway since quotas
6731 * are already enabled and this is not a hard failure.
6734 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6737 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6738 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6739 S_NOATIME | S_IMMUTABLE);
6740 err = ext4_mark_inode_dirty(handle, inode);
6741 ext4_journal_stop(handle);
6743 inode_unlock(inode);
6745 dquot_quota_off(sb, type);
6748 lockdep_set_quota_inode(path->dentry->d_inode,
6753 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6757 struct inode *qf_inode;
6758 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6759 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6760 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6761 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6764 BUG_ON(!ext4_has_feature_quota(sb));
6766 if (!qf_inums[type])
6769 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6770 if (IS_ERR(qf_inode)) {
6771 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6772 return PTR_ERR(qf_inode);
6775 /* Don't account quota for quota files to avoid recursion */
6776 qf_inode->i_flags |= S_NOQUOTA;
6777 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6778 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6780 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6786 /* Enable usage tracking for all quota types. */
6787 int ext4_enable_quotas(struct super_block *sb)
6790 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6791 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6792 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6793 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6795 bool quota_mopt[EXT4_MAXQUOTAS] = {
6796 test_opt(sb, USRQUOTA),
6797 test_opt(sb, GRPQUOTA),
6798 test_opt(sb, PRJQUOTA),
6801 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6802 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6803 if (qf_inums[type]) {
6804 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6805 DQUOT_USAGE_ENABLED |
6806 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6809 "Failed to enable quota tracking "
6810 "(type=%d, err=%d). Please run "
6811 "e2fsck to fix.", type, err);
6812 for (type--; type >= 0; type--) {
6813 struct inode *inode;
6815 inode = sb_dqopt(sb)->files[type];
6817 inode = igrab(inode);
6818 dquot_quota_off(sb, type);
6820 lockdep_set_quota_inode(inode,
6833 static int ext4_quota_off(struct super_block *sb, int type)
6835 struct inode *inode = sb_dqopt(sb)->files[type];
6839 /* Force all delayed allocation blocks to be allocated.
6840 * Caller already holds s_umount sem */
6841 if (test_opt(sb, DELALLOC))
6842 sync_filesystem(sb);
6844 if (!inode || !igrab(inode))
6847 err = dquot_quota_off(sb, type);
6848 if (err || ext4_has_feature_quota(sb))
6853 * Update modification times of quota files when userspace can
6854 * start looking at them. If we fail, we return success anyway since
6855 * this is not a hard failure and quotas are already disabled.
6857 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6858 if (IS_ERR(handle)) {
6859 err = PTR_ERR(handle);
6862 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6863 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6864 inode->i_mtime = inode->i_ctime = current_time(inode);
6865 err = ext4_mark_inode_dirty(handle, inode);
6866 ext4_journal_stop(handle);
6868 inode_unlock(inode);
6870 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6874 return dquot_quota_off(sb, type);
6877 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6878 * acquiring the locks... As quota files are never truncated and quota code
6879 * itself serializes the operations (and no one else should touch the files)
6880 * we don't have to be afraid of races */
6881 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6882 size_t len, loff_t off)
6884 struct inode *inode = sb_dqopt(sb)->files[type];
6885 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6886 int offset = off & (sb->s_blocksize - 1);
6889 struct buffer_head *bh;
6890 loff_t i_size = i_size_read(inode);
6894 if (off+len > i_size)
6897 while (toread > 0) {
6898 tocopy = sb->s_blocksize - offset < toread ?
6899 sb->s_blocksize - offset : toread;
6900 bh = ext4_bread(NULL, inode, blk, 0);
6903 if (!bh) /* A hole? */
6904 memset(data, 0, tocopy);
6906 memcpy(data, bh->b_data+offset, tocopy);
6916 /* Write to quotafile (we know the transaction is already started and has
6917 * enough credits) */
6918 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6919 const char *data, size_t len, loff_t off)
6921 struct inode *inode = sb_dqopt(sb)->files[type];
6922 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6923 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6925 struct buffer_head *bh;
6926 handle_t *handle = journal_current_handle();
6929 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6930 " cancelled because transaction is not started",
6931 (unsigned long long)off, (unsigned long long)len);
6935 * Since we account only one data block in transaction credits,
6936 * then it is impossible to cross a block boundary.
6938 if (sb->s_blocksize - offset < len) {
6939 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6940 " cancelled because not block aligned",
6941 (unsigned long long)off, (unsigned long long)len);
6946 bh = ext4_bread(handle, inode, blk,
6947 EXT4_GET_BLOCKS_CREATE |
6948 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6949 } while (PTR_ERR(bh) == -ENOSPC &&
6950 ext4_should_retry_alloc(inode->i_sb, &retries));
6955 BUFFER_TRACE(bh, "get write access");
6956 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6962 memcpy(bh->b_data+offset, data, len);
6963 flush_dcache_page(bh->b_page);
6965 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6968 if (inode->i_size < off + len) {
6969 i_size_write(inode, off + len);
6970 EXT4_I(inode)->i_disksize = inode->i_size;
6971 err2 = ext4_mark_inode_dirty(handle, inode);
6972 if (unlikely(err2 && !err))
6975 return err ? err : len;
6979 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6980 static inline void register_as_ext2(void)
6982 int err = register_filesystem(&ext2_fs_type);
6985 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6988 static inline void unregister_as_ext2(void)
6990 unregister_filesystem(&ext2_fs_type);
6993 static inline int ext2_feature_set_ok(struct super_block *sb)
6995 if (ext4_has_unknown_ext2_incompat_features(sb))
6999 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7004 static inline void register_as_ext2(void) { }
7005 static inline void unregister_as_ext2(void) { }
7006 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7009 static inline void register_as_ext3(void)
7011 int err = register_filesystem(&ext3_fs_type);
7014 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7017 static inline void unregister_as_ext3(void)
7019 unregister_filesystem(&ext3_fs_type);
7022 static inline int ext3_feature_set_ok(struct super_block *sb)
7024 if (ext4_has_unknown_ext3_incompat_features(sb))
7026 if (!ext4_has_feature_journal(sb))
7030 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7035 static struct file_system_type ext4_fs_type = {
7036 .owner = THIS_MODULE,
7038 .init_fs_context = ext4_init_fs_context,
7039 .parameters = ext4_param_specs,
7040 .kill_sb = kill_block_super,
7041 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7043 MODULE_ALIAS_FS("ext4");
7045 /* Shared across all ext4 file systems */
7046 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7048 static int __init ext4_init_fs(void)
7052 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7053 ext4_li_info = NULL;
7055 /* Build-time check for flags consistency */
7056 ext4_check_flag_values();
7058 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7059 init_waitqueue_head(&ext4__ioend_wq[i]);
7061 err = ext4_init_es();
7065 err = ext4_init_pending();
7069 err = ext4_init_post_read_processing();
7073 err = ext4_init_pageio();
7077 err = ext4_init_system_zone();
7081 err = ext4_init_sysfs();
7085 err = ext4_init_mballoc();
7088 err = init_inodecache();
7092 err = ext4_fc_init_dentry_cache();
7098 err = register_filesystem(&ext4_fs_type);
7104 unregister_as_ext2();
7105 unregister_as_ext3();
7106 ext4_fc_destroy_dentry_cache();
7108 destroy_inodecache();
7110 ext4_exit_mballoc();
7114 ext4_exit_system_zone();
7118 ext4_exit_post_read_processing();
7120 ext4_exit_pending();
7127 static void __exit ext4_exit_fs(void)
7129 ext4_destroy_lazyinit_thread();
7130 unregister_as_ext2();
7131 unregister_as_ext3();
7132 unregister_filesystem(&ext4_fs_type);
7133 ext4_fc_destroy_dentry_cache();
7134 destroy_inodecache();
7135 ext4_exit_mballoc();
7137 ext4_exit_system_zone();
7139 ext4_exit_post_read_processing();
7141 ext4_exit_pending();
7144 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7145 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7146 MODULE_LICENSE("GPL");
7147 MODULE_SOFTDEP("pre: crc32c");
7148 module_init(ext4_init_fs)
7149 module_exit(ext4_exit_fs)