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, blk_opf_t 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, blk_opf_t 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, blk_opf_t 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, blk_opf_t op_flags, bool wait)
210 ext4_read_bh_nowait(bh, op_flags, NULL);
213 return ext4_read_bh(bh, op_flags, NULL);
217 * This works like __bread_gfp() except it uses ERR_PTR for error
218 * returns. Currently with sb_bread it's impossible to distinguish
219 * between ENOMEM and EIO situations (since both result in a NULL
222 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
224 blk_opf_t op_flags, gfp_t gfp)
226 struct buffer_head *bh;
229 bh = sb_getblk_gfp(sb, block, gfp);
231 return ERR_PTR(-ENOMEM);
232 if (ext4_buffer_uptodate(bh))
235 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
243 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
246 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
249 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
252 return __ext4_sb_bread_gfp(sb, block, 0, 0);
255 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
257 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
260 if (trylock_buffer(bh))
261 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
266 static int ext4_verify_csum_type(struct super_block *sb,
267 struct ext4_super_block *es)
269 if (!ext4_has_feature_metadata_csum(sb))
272 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
275 __le32 ext4_superblock_csum(struct super_block *sb,
276 struct ext4_super_block *es)
278 struct ext4_sb_info *sbi = EXT4_SB(sb);
279 int offset = offsetof(struct ext4_super_block, s_checksum);
282 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
284 return cpu_to_le32(csum);
287 static int ext4_superblock_csum_verify(struct super_block *sb,
288 struct ext4_super_block *es)
290 if (!ext4_has_metadata_csum(sb))
293 return es->s_checksum == ext4_superblock_csum(sb, es);
296 void ext4_superblock_csum_set(struct super_block *sb)
298 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
300 if (!ext4_has_metadata_csum(sb))
303 es->s_checksum = ext4_superblock_csum(sb, es);
306 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
307 struct ext4_group_desc *bg)
309 return le32_to_cpu(bg->bg_block_bitmap_lo) |
310 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
311 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
314 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
317 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
319 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
322 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
323 struct ext4_group_desc *bg)
325 return le32_to_cpu(bg->bg_inode_table_lo) |
326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
330 __u32 ext4_free_group_clusters(struct super_block *sb,
331 struct ext4_group_desc *bg)
333 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
335 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
338 __u32 ext4_free_inodes_count(struct super_block *sb,
339 struct ext4_group_desc *bg)
341 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
343 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
346 __u32 ext4_used_dirs_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
349 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
351 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
354 __u32 ext4_itable_unused_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
357 return le16_to_cpu(bg->bg_itable_unused_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
359 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
362 void ext4_block_bitmap_set(struct super_block *sb,
363 struct ext4_group_desc *bg, ext4_fsblk_t blk)
365 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
366 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
367 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
370 void ext4_inode_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
373 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
378 void ext4_inode_table_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
381 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
386 void ext4_free_group_clusters_set(struct super_block *sb,
387 struct ext4_group_desc *bg, __u32 count)
389 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
394 void ext4_free_inodes_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
397 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
402 void ext4_used_dirs_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
405 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
410 void ext4_itable_unused_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
413 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
418 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
420 now = clamp_val(now, 0, (1ull << 40) - 1);
422 *lo = cpu_to_le32(lower_32_bits(now));
423 *hi = upper_32_bits(now);
426 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
428 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
430 #define ext4_update_tstamp(es, tstamp) \
431 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
432 ktime_get_real_seconds())
433 #define ext4_get_tstamp(es, tstamp) \
434 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
437 * The del_gendisk() function uninitializes the disk-specific data
438 * structures, including the bdi structure, without telling anyone
439 * else. Once this happens, any attempt to call mark_buffer_dirty()
440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
441 * This is a kludge to prevent these oops until we can put in a proper
442 * hook in del_gendisk() to inform the VFS and file system layers.
444 static int block_device_ejected(struct super_block *sb)
446 struct inode *bd_inode = sb->s_bdev->bd_inode;
447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
449 return bdi->dev == NULL;
452 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
454 struct super_block *sb = journal->j_private;
455 struct ext4_sb_info *sbi = EXT4_SB(sb);
456 int error = is_journal_aborted(journal);
457 struct ext4_journal_cb_entry *jce;
459 BUG_ON(txn->t_state == T_FINISHED);
461 ext4_process_freed_data(sb, txn->t_tid);
463 spin_lock(&sbi->s_md_lock);
464 while (!list_empty(&txn->t_private_list)) {
465 jce = list_entry(txn->t_private_list.next,
466 struct ext4_journal_cb_entry, jce_list);
467 list_del_init(&jce->jce_list);
468 spin_unlock(&sbi->s_md_lock);
469 jce->jce_func(sb, jce, error);
470 spin_lock(&sbi->s_md_lock);
472 spin_unlock(&sbi->s_md_lock);
476 * This writepage callback for write_cache_pages()
477 * takes care of a few cases after page cleaning.
479 * write_cache_pages() already checks for dirty pages
480 * and calls clear_page_dirty_for_io(), which we want,
481 * to write protect the pages.
483 * However, we may have to redirty a page (see below.)
485 static int ext4_journalled_writepage_callback(struct page *page,
486 struct writeback_control *wbc,
489 transaction_t *transaction = (transaction_t *) data;
490 struct buffer_head *bh, *head;
491 struct journal_head *jh;
493 bh = head = page_buffers(page);
496 * We have to redirty a page in these cases:
497 * 1) If buffer is dirty, it means the page was dirty because it
498 * contains a buffer that needs checkpointing. So the dirty bit
499 * needs to be preserved so that checkpointing writes the buffer
501 * 2) If buffer is not part of the committing transaction
502 * (we may have just accidentally come across this buffer because
503 * inode range tracking is not exact) or if the currently running
504 * transaction already contains this buffer as well, dirty bit
505 * needs to be preserved so that the buffer gets writeprotected
506 * properly on running transaction's commit.
509 if (buffer_dirty(bh) ||
510 (jh && (jh->b_transaction != transaction ||
511 jh->b_next_transaction))) {
512 redirty_page_for_writepage(wbc, page);
515 } while ((bh = bh->b_this_page) != head);
518 return AOP_WRITEPAGE_ACTIVATE;
521 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
523 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
524 struct writeback_control wbc = {
525 .sync_mode = WB_SYNC_ALL,
526 .nr_to_write = LONG_MAX,
527 .range_start = jinode->i_dirty_start,
528 .range_end = jinode->i_dirty_end,
531 return write_cache_pages(mapping, &wbc,
532 ext4_journalled_writepage_callback,
533 jinode->i_transaction);
536 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
540 if (ext4_should_journal_data(jinode->i_vfs_inode))
541 ret = ext4_journalled_submit_inode_data_buffers(jinode);
543 ret = jbd2_journal_submit_inode_data_buffers(jinode);
548 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
552 if (!ext4_should_journal_data(jinode->i_vfs_inode))
553 ret = jbd2_journal_finish_inode_data_buffers(jinode);
558 static bool system_going_down(void)
560 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
561 || system_state == SYSTEM_RESTART;
564 struct ext4_err_translation {
569 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
571 static struct ext4_err_translation err_translation[] = {
572 EXT4_ERR_TRANSLATE(EIO),
573 EXT4_ERR_TRANSLATE(ENOMEM),
574 EXT4_ERR_TRANSLATE(EFSBADCRC),
575 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
576 EXT4_ERR_TRANSLATE(ENOSPC),
577 EXT4_ERR_TRANSLATE(ENOKEY),
578 EXT4_ERR_TRANSLATE(EROFS),
579 EXT4_ERR_TRANSLATE(EFBIG),
580 EXT4_ERR_TRANSLATE(EEXIST),
581 EXT4_ERR_TRANSLATE(ERANGE),
582 EXT4_ERR_TRANSLATE(EOVERFLOW),
583 EXT4_ERR_TRANSLATE(EBUSY),
584 EXT4_ERR_TRANSLATE(ENOTDIR),
585 EXT4_ERR_TRANSLATE(ENOTEMPTY),
586 EXT4_ERR_TRANSLATE(ESHUTDOWN),
587 EXT4_ERR_TRANSLATE(EFAULT),
590 static int ext4_errno_to_code(int errno)
594 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
595 if (err_translation[i].errno == errno)
596 return err_translation[i].code;
597 return EXT4_ERR_UNKNOWN;
600 static void save_error_info(struct super_block *sb, int error,
601 __u32 ino, __u64 block,
602 const char *func, unsigned int line)
604 struct ext4_sb_info *sbi = EXT4_SB(sb);
606 /* We default to EFSCORRUPTED error... */
608 error = EFSCORRUPTED;
610 spin_lock(&sbi->s_error_lock);
611 sbi->s_add_error_count++;
612 sbi->s_last_error_code = error;
613 sbi->s_last_error_line = line;
614 sbi->s_last_error_ino = ino;
615 sbi->s_last_error_block = block;
616 sbi->s_last_error_func = func;
617 sbi->s_last_error_time = ktime_get_real_seconds();
618 if (!sbi->s_first_error_time) {
619 sbi->s_first_error_code = error;
620 sbi->s_first_error_line = line;
621 sbi->s_first_error_ino = ino;
622 sbi->s_first_error_block = block;
623 sbi->s_first_error_func = func;
624 sbi->s_first_error_time = sbi->s_last_error_time;
626 spin_unlock(&sbi->s_error_lock);
629 /* Deal with the reporting of failure conditions on a filesystem such as
630 * inconsistencies detected or read IO failures.
632 * On ext2, we can store the error state of the filesystem in the
633 * superblock. That is not possible on ext4, because we may have other
634 * write ordering constraints on the superblock which prevent us from
635 * writing it out straight away; and given that the journal is about to
636 * be aborted, we can't rely on the current, or future, transactions to
637 * write out the superblock safely.
639 * We'll just use the jbd2_journal_abort() error code to record an error in
640 * the journal instead. On recovery, the journal will complain about
641 * that error until we've noted it down and cleared it.
643 * If force_ro is set, we unconditionally force the filesystem into an
644 * ABORT|READONLY state, unless the error response on the fs has been set to
645 * panic in which case we take the easy way out and panic immediately. This is
646 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
647 * at a critical moment in log management.
649 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
650 __u32 ino, __u64 block,
651 const char *func, unsigned int line)
653 journal_t *journal = EXT4_SB(sb)->s_journal;
654 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
656 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
657 if (test_opt(sb, WARN_ON_ERROR))
660 if (!continue_fs && !sb_rdonly(sb)) {
661 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
663 jbd2_journal_abort(journal, -EIO);
666 if (!bdev_read_only(sb->s_bdev)) {
667 save_error_info(sb, error, ino, block, func, line);
669 * In case the fs should keep running, we need to writeout
670 * superblock through the journal. Due to lock ordering
671 * constraints, it may not be safe to do it right here so we
672 * defer superblock flushing to a workqueue.
674 if (continue_fs && journal)
675 schedule_work(&EXT4_SB(sb)->s_error_work);
677 ext4_commit_super(sb);
681 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
682 * could panic during 'reboot -f' as the underlying device got already
685 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
686 panic("EXT4-fs (device %s): panic forced after error\n",
690 if (sb_rdonly(sb) || continue_fs)
693 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
695 * Make sure updated value of ->s_mount_flags will be visible before
699 sb->s_flags |= SB_RDONLY;
702 static void flush_stashed_error_work(struct work_struct *work)
704 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
706 journal_t *journal = sbi->s_journal;
710 * If the journal is still running, we have to write out superblock
711 * through the journal to avoid collisions of other journalled sb
714 * We use directly jbd2 functions here to avoid recursing back into
715 * ext4 error handling code during handling of previous errors.
717 if (!sb_rdonly(sbi->s_sb) && journal) {
718 struct buffer_head *sbh = sbi->s_sbh;
719 handle = jbd2_journal_start(journal, 1);
722 if (jbd2_journal_get_write_access(handle, sbh)) {
723 jbd2_journal_stop(handle);
726 ext4_update_super(sbi->s_sb);
727 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
728 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
729 "superblock detected");
730 clear_buffer_write_io_error(sbh);
731 set_buffer_uptodate(sbh);
734 if (jbd2_journal_dirty_metadata(handle, sbh)) {
735 jbd2_journal_stop(handle);
738 jbd2_journal_stop(handle);
739 ext4_notify_error_sysfs(sbi);
744 * Write through journal failed. Write sb directly to get error info
745 * out and hope for the best.
747 ext4_commit_super(sbi->s_sb);
748 ext4_notify_error_sysfs(sbi);
751 #define ext4_error_ratelimit(sb) \
752 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
755 void __ext4_error(struct super_block *sb, const char *function,
756 unsigned int line, bool force_ro, int error, __u64 block,
757 const char *fmt, ...)
759 struct va_format vaf;
762 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
765 trace_ext4_error(sb, function, line);
766 if (ext4_error_ratelimit(sb)) {
771 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
772 sb->s_id, function, line, current->comm, &vaf);
775 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
777 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
780 void __ext4_error_inode(struct inode *inode, const char *function,
781 unsigned int line, ext4_fsblk_t block, int error,
782 const char *fmt, ...)
785 struct va_format vaf;
787 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
790 trace_ext4_error(inode->i_sb, function, line);
791 if (ext4_error_ratelimit(inode->i_sb)) {
796 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
797 "inode #%lu: block %llu: comm %s: %pV\n",
798 inode->i_sb->s_id, function, line, inode->i_ino,
799 block, current->comm, &vaf);
801 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
802 "inode #%lu: comm %s: %pV\n",
803 inode->i_sb->s_id, function, line, inode->i_ino,
804 current->comm, &vaf);
807 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
809 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
813 void __ext4_error_file(struct file *file, const char *function,
814 unsigned int line, ext4_fsblk_t block,
815 const char *fmt, ...)
818 struct va_format vaf;
819 struct inode *inode = file_inode(file);
820 char pathname[80], *path;
822 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
825 trace_ext4_error(inode->i_sb, function, line);
826 if (ext4_error_ratelimit(inode->i_sb)) {
827 path = file_path(file, pathname, sizeof(pathname));
835 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
836 "block %llu: comm %s: path %s: %pV\n",
837 inode->i_sb->s_id, function, line, inode->i_ino,
838 block, current->comm, path, &vaf);
841 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
842 "comm %s: path %s: %pV\n",
843 inode->i_sb->s_id, function, line, inode->i_ino,
844 current->comm, path, &vaf);
847 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
849 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
853 const char *ext4_decode_error(struct super_block *sb, int errno,
860 errstr = "Corrupt filesystem";
863 errstr = "Filesystem failed CRC";
866 errstr = "IO failure";
869 errstr = "Out of memory";
872 if (!sb || (EXT4_SB(sb)->s_journal &&
873 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
874 errstr = "Journal has aborted";
876 errstr = "Readonly filesystem";
879 /* If the caller passed in an extra buffer for unknown
880 * errors, textualise them now. Else we just return
883 /* Check for truncated error codes... */
884 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
893 /* __ext4_std_error decodes expected errors from journaling functions
894 * automatically and invokes the appropriate error response. */
896 void __ext4_std_error(struct super_block *sb, const char *function,
897 unsigned int line, int errno)
902 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
905 /* Special case: if the error is EROFS, and we're not already
906 * inside a transaction, then there's really no point in logging
908 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
911 if (ext4_error_ratelimit(sb)) {
912 errstr = ext4_decode_error(sb, errno, nbuf);
913 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
914 sb->s_id, function, line, errstr);
916 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
918 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
921 void __ext4_msg(struct super_block *sb,
922 const char *prefix, const char *fmt, ...)
924 struct va_format vaf;
928 atomic_inc(&EXT4_SB(sb)->s_msg_count);
929 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
938 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
940 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
944 static int ext4_warning_ratelimit(struct super_block *sb)
946 atomic_inc(&EXT4_SB(sb)->s_warning_count);
947 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
951 void __ext4_warning(struct super_block *sb, const char *function,
952 unsigned int line, const char *fmt, ...)
954 struct va_format vaf;
957 if (!ext4_warning_ratelimit(sb))
963 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
964 sb->s_id, function, line, &vaf);
968 void __ext4_warning_inode(const struct inode *inode, const char *function,
969 unsigned int line, const char *fmt, ...)
971 struct va_format vaf;
974 if (!ext4_warning_ratelimit(inode->i_sb))
980 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
981 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
982 function, line, inode->i_ino, current->comm, &vaf);
986 void __ext4_grp_locked_error(const char *function, unsigned int line,
987 struct super_block *sb, ext4_group_t grp,
988 unsigned long ino, ext4_fsblk_t block,
989 const char *fmt, ...)
993 struct va_format vaf;
996 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
999 trace_ext4_error(sb, function, line);
1000 if (ext4_error_ratelimit(sb)) {
1001 va_start(args, fmt);
1004 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1005 sb->s_id, function, line, grp);
1007 printk(KERN_CONT "inode %lu: ", ino);
1009 printk(KERN_CONT "block %llu:",
1010 (unsigned long long) block);
1011 printk(KERN_CONT "%pV\n", &vaf);
1015 if (test_opt(sb, ERRORS_CONT)) {
1016 if (test_opt(sb, WARN_ON_ERROR))
1018 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1019 if (!bdev_read_only(sb->s_bdev)) {
1020 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1022 schedule_work(&EXT4_SB(sb)->s_error_work);
1026 ext4_unlock_group(sb, grp);
1027 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1029 * We only get here in the ERRORS_RO case; relocking the group
1030 * may be dangerous, but nothing bad will happen since the
1031 * filesystem will have already been marked read/only and the
1032 * journal has been aborted. We return 1 as a hint to callers
1033 * who might what to use the return value from
1034 * ext4_grp_locked_error() to distinguish between the
1035 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1036 * aggressively from the ext4 function in question, with a
1037 * more appropriate error code.
1039 ext4_lock_group(sb, grp);
1043 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1047 struct ext4_sb_info *sbi = EXT4_SB(sb);
1048 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1049 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1052 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1053 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1056 percpu_counter_sub(&sbi->s_freeclusters_counter,
1060 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1061 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1066 count = ext4_free_inodes_count(sb, gdp);
1067 percpu_counter_sub(&sbi->s_freeinodes_counter,
1073 void ext4_update_dynamic_rev(struct super_block *sb)
1075 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1077 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1081 "updating to rev %d because of new feature flag, "
1082 "running e2fsck is recommended",
1085 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1086 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1087 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1088 /* leave es->s_feature_*compat flags alone */
1089 /* es->s_uuid will be set by e2fsck if empty */
1092 * The rest of the superblock fields should be zero, and if not it
1093 * means they are likely already in use, so leave them alone. We
1094 * can leave it up to e2fsck to clean up any inconsistencies there.
1099 * Open the external journal device
1101 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1103 struct block_device *bdev;
1105 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1111 ext4_msg(sb, KERN_ERR,
1112 "failed to open journal device unknown-block(%u,%u) %ld",
1113 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1118 * Release the journal device
1120 static void ext4_blkdev_put(struct block_device *bdev)
1122 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1125 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1127 struct block_device *bdev;
1128 bdev = sbi->s_journal_bdev;
1130 ext4_blkdev_put(bdev);
1131 sbi->s_journal_bdev = NULL;
1135 static inline struct inode *orphan_list_entry(struct list_head *l)
1137 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1140 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1142 struct list_head *l;
1144 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1145 le32_to_cpu(sbi->s_es->s_last_orphan));
1147 printk(KERN_ERR "sb_info orphan list:\n");
1148 list_for_each(l, &sbi->s_orphan) {
1149 struct inode *inode = orphan_list_entry(l);
1151 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1152 inode->i_sb->s_id, inode->i_ino, inode,
1153 inode->i_mode, inode->i_nlink,
1154 NEXT_ORPHAN(inode));
1159 static int ext4_quota_off(struct super_block *sb, int type);
1161 static inline void ext4_quota_off_umount(struct super_block *sb)
1165 /* Use our quota_off function to clear inode flags etc. */
1166 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1167 ext4_quota_off(sb, type);
1171 * This is a helper function which is used in the mount/remount
1172 * codepaths (which holds s_umount) to fetch the quota file name.
1174 static inline char *get_qf_name(struct super_block *sb,
1175 struct ext4_sb_info *sbi,
1178 return rcu_dereference_protected(sbi->s_qf_names[type],
1179 lockdep_is_held(&sb->s_umount));
1182 static inline void ext4_quota_off_umount(struct super_block *sb)
1187 static void ext4_put_super(struct super_block *sb)
1189 struct ext4_sb_info *sbi = EXT4_SB(sb);
1190 struct ext4_super_block *es = sbi->s_es;
1191 struct buffer_head **group_desc;
1192 struct flex_groups **flex_groups;
1197 * Unregister sysfs before destroying jbd2 journal.
1198 * Since we could still access attr_journal_task attribute via sysfs
1199 * path which could have sbi->s_journal->j_task as NULL
1200 * Unregister sysfs before flush sbi->s_error_work.
1201 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1202 * read metadata verify failed then will queue error work.
1203 * flush_stashed_error_work will call start_this_handle may trigger
1206 ext4_unregister_sysfs(sb);
1208 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1209 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1212 ext4_unregister_li_request(sb);
1213 ext4_quota_off_umount(sb);
1215 flush_work(&sbi->s_error_work);
1216 destroy_workqueue(sbi->rsv_conversion_wq);
1217 ext4_release_orphan_info(sb);
1219 if (sbi->s_journal) {
1220 aborted = is_journal_aborted(sbi->s_journal);
1221 err = jbd2_journal_destroy(sbi->s_journal);
1222 sbi->s_journal = NULL;
1223 if ((err < 0) && !aborted) {
1224 ext4_abort(sb, -err, "Couldn't clean up the journal");
1228 ext4_es_unregister_shrinker(sbi);
1229 del_timer_sync(&sbi->s_err_report);
1230 ext4_release_system_zone(sb);
1231 ext4_mb_release(sb);
1232 ext4_ext_release(sb);
1234 if (!sb_rdonly(sb) && !aborted) {
1235 ext4_clear_feature_journal_needs_recovery(sb);
1236 ext4_clear_feature_orphan_present(sb);
1237 es->s_state = cpu_to_le16(sbi->s_mount_state);
1240 ext4_commit_super(sb);
1243 group_desc = rcu_dereference(sbi->s_group_desc);
1244 for (i = 0; i < sbi->s_gdb_count; i++)
1245 brelse(group_desc[i]);
1247 flex_groups = rcu_dereference(sbi->s_flex_groups);
1249 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1250 kvfree(flex_groups[i]);
1251 kvfree(flex_groups);
1254 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1255 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1256 percpu_counter_destroy(&sbi->s_dirs_counter);
1257 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1258 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1259 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1261 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1262 kfree(get_qf_name(sb, sbi, i));
1265 /* Debugging code just in case the in-memory inode orphan list
1266 * isn't empty. The on-disk one can be non-empty if we've
1267 * detected an error and taken the fs readonly, but the
1268 * in-memory list had better be clean by this point. */
1269 if (!list_empty(&sbi->s_orphan))
1270 dump_orphan_list(sb, sbi);
1271 ASSERT(list_empty(&sbi->s_orphan));
1273 sync_blockdev(sb->s_bdev);
1274 invalidate_bdev(sb->s_bdev);
1275 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1277 * Invalidate the journal device's buffers. We don't want them
1278 * floating about in memory - the physical journal device may
1279 * hotswapped, and it breaks the `ro-after' testing code.
1281 sync_blockdev(sbi->s_journal_bdev);
1282 invalidate_bdev(sbi->s_journal_bdev);
1283 ext4_blkdev_remove(sbi);
1286 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1287 sbi->s_ea_inode_cache = NULL;
1289 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1290 sbi->s_ea_block_cache = NULL;
1292 ext4_stop_mmpd(sbi);
1295 sb->s_fs_info = NULL;
1297 * Now that we are completely done shutting down the
1298 * superblock, we need to actually destroy the kobject.
1300 kobject_put(&sbi->s_kobj);
1301 wait_for_completion(&sbi->s_kobj_unregister);
1302 if (sbi->s_chksum_driver)
1303 crypto_free_shash(sbi->s_chksum_driver);
1304 kfree(sbi->s_blockgroup_lock);
1305 fs_put_dax(sbi->s_daxdev, NULL);
1306 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1307 #if IS_ENABLED(CONFIG_UNICODE)
1308 utf8_unload(sb->s_encoding);
1313 static struct kmem_cache *ext4_inode_cachep;
1316 * Called inside transaction, so use GFP_NOFS
1318 static struct inode *ext4_alloc_inode(struct super_block *sb)
1320 struct ext4_inode_info *ei;
1322 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1326 inode_set_iversion(&ei->vfs_inode, 1);
1327 spin_lock_init(&ei->i_raw_lock);
1328 INIT_LIST_HEAD(&ei->i_prealloc_list);
1329 atomic_set(&ei->i_prealloc_active, 0);
1330 spin_lock_init(&ei->i_prealloc_lock);
1331 ext4_es_init_tree(&ei->i_es_tree);
1332 rwlock_init(&ei->i_es_lock);
1333 INIT_LIST_HEAD(&ei->i_es_list);
1334 ei->i_es_all_nr = 0;
1335 ei->i_es_shk_nr = 0;
1336 ei->i_es_shrink_lblk = 0;
1337 ei->i_reserved_data_blocks = 0;
1338 spin_lock_init(&(ei->i_block_reservation_lock));
1339 ext4_init_pending_tree(&ei->i_pending_tree);
1341 ei->i_reserved_quota = 0;
1342 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1345 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1346 spin_lock_init(&ei->i_completed_io_lock);
1348 ei->i_datasync_tid = 0;
1349 atomic_set(&ei->i_unwritten, 0);
1350 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1351 ext4_fc_init_inode(&ei->vfs_inode);
1352 mutex_init(&ei->i_fc_lock);
1353 return &ei->vfs_inode;
1356 static int ext4_drop_inode(struct inode *inode)
1358 int drop = generic_drop_inode(inode);
1361 drop = fscrypt_drop_inode(inode);
1363 trace_ext4_drop_inode(inode, drop);
1367 static void ext4_free_in_core_inode(struct inode *inode)
1369 fscrypt_free_inode(inode);
1370 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1371 pr_warn("%s: inode %ld still in fc list",
1372 __func__, inode->i_ino);
1374 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1377 static void ext4_destroy_inode(struct inode *inode)
1379 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1380 ext4_msg(inode->i_sb, KERN_ERR,
1381 "Inode %lu (%p): orphan list check failed!",
1382 inode->i_ino, EXT4_I(inode));
1383 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1384 EXT4_I(inode), sizeof(struct ext4_inode_info),
1389 if (EXT4_I(inode)->i_reserved_data_blocks)
1390 ext4_msg(inode->i_sb, KERN_ERR,
1391 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1392 inode->i_ino, EXT4_I(inode),
1393 EXT4_I(inode)->i_reserved_data_blocks);
1396 static void init_once(void *foo)
1398 struct ext4_inode_info *ei = foo;
1400 INIT_LIST_HEAD(&ei->i_orphan);
1401 init_rwsem(&ei->xattr_sem);
1402 init_rwsem(&ei->i_data_sem);
1403 inode_init_once(&ei->vfs_inode);
1404 ext4_fc_init_inode(&ei->vfs_inode);
1407 static int __init init_inodecache(void)
1409 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1410 sizeof(struct ext4_inode_info), 0,
1411 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1413 offsetof(struct ext4_inode_info, i_data),
1414 sizeof_field(struct ext4_inode_info, i_data),
1416 if (ext4_inode_cachep == NULL)
1421 static void destroy_inodecache(void)
1424 * Make sure all delayed rcu free inodes are flushed before we
1428 kmem_cache_destroy(ext4_inode_cachep);
1431 void ext4_clear_inode(struct inode *inode)
1434 invalidate_inode_buffers(inode);
1436 ext4_discard_preallocations(inode, 0);
1437 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1439 if (EXT4_I(inode)->jinode) {
1440 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1441 EXT4_I(inode)->jinode);
1442 jbd2_free_inode(EXT4_I(inode)->jinode);
1443 EXT4_I(inode)->jinode = NULL;
1445 fscrypt_put_encryption_info(inode);
1446 fsverity_cleanup_inode(inode);
1449 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1450 u64 ino, u32 generation)
1452 struct inode *inode;
1455 * Currently we don't know the generation for parent directory, so
1456 * a generation of 0 means "accept any"
1458 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1460 return ERR_CAST(inode);
1461 if (generation && inode->i_generation != generation) {
1463 return ERR_PTR(-ESTALE);
1469 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1470 int fh_len, int fh_type)
1472 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1473 ext4_nfs_get_inode);
1476 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1477 int fh_len, int fh_type)
1479 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1480 ext4_nfs_get_inode);
1483 static int ext4_nfs_commit_metadata(struct inode *inode)
1485 struct writeback_control wbc = {
1486 .sync_mode = WB_SYNC_ALL
1489 trace_ext4_nfs_commit_metadata(inode);
1490 return ext4_write_inode(inode, &wbc);
1494 static const char * const quotatypes[] = INITQFNAMES;
1495 #define QTYPE2NAME(t) (quotatypes[t])
1497 static int ext4_write_dquot(struct dquot *dquot);
1498 static int ext4_acquire_dquot(struct dquot *dquot);
1499 static int ext4_release_dquot(struct dquot *dquot);
1500 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1501 static int ext4_write_info(struct super_block *sb, int type);
1502 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1503 const struct path *path);
1504 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1505 size_t len, loff_t off);
1506 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1507 const char *data, size_t len, loff_t off);
1508 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1509 unsigned int flags);
1511 static struct dquot **ext4_get_dquots(struct inode *inode)
1513 return EXT4_I(inode)->i_dquot;
1516 static const struct dquot_operations ext4_quota_operations = {
1517 .get_reserved_space = ext4_get_reserved_space,
1518 .write_dquot = ext4_write_dquot,
1519 .acquire_dquot = ext4_acquire_dquot,
1520 .release_dquot = ext4_release_dquot,
1521 .mark_dirty = ext4_mark_dquot_dirty,
1522 .write_info = ext4_write_info,
1523 .alloc_dquot = dquot_alloc,
1524 .destroy_dquot = dquot_destroy,
1525 .get_projid = ext4_get_projid,
1526 .get_inode_usage = ext4_get_inode_usage,
1527 .get_next_id = dquot_get_next_id,
1530 static const struct quotactl_ops ext4_qctl_operations = {
1531 .quota_on = ext4_quota_on,
1532 .quota_off = ext4_quota_off,
1533 .quota_sync = dquot_quota_sync,
1534 .get_state = dquot_get_state,
1535 .set_info = dquot_set_dqinfo,
1536 .get_dqblk = dquot_get_dqblk,
1537 .set_dqblk = dquot_set_dqblk,
1538 .get_nextdqblk = dquot_get_next_dqblk,
1542 static const struct super_operations ext4_sops = {
1543 .alloc_inode = ext4_alloc_inode,
1544 .free_inode = ext4_free_in_core_inode,
1545 .destroy_inode = ext4_destroy_inode,
1546 .write_inode = ext4_write_inode,
1547 .dirty_inode = ext4_dirty_inode,
1548 .drop_inode = ext4_drop_inode,
1549 .evict_inode = ext4_evict_inode,
1550 .put_super = ext4_put_super,
1551 .sync_fs = ext4_sync_fs,
1552 .freeze_fs = ext4_freeze,
1553 .unfreeze_fs = ext4_unfreeze,
1554 .statfs = ext4_statfs,
1555 .show_options = ext4_show_options,
1557 .quota_read = ext4_quota_read,
1558 .quota_write = ext4_quota_write,
1559 .get_dquots = ext4_get_dquots,
1563 static const struct export_operations ext4_export_ops = {
1564 .fh_to_dentry = ext4_fh_to_dentry,
1565 .fh_to_parent = ext4_fh_to_parent,
1566 .get_parent = ext4_get_parent,
1567 .commit_metadata = ext4_nfs_commit_metadata,
1571 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1572 Opt_resgid, Opt_resuid, Opt_sb,
1573 Opt_nouid32, Opt_debug, Opt_removed,
1574 Opt_user_xattr, Opt_acl,
1575 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1576 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1577 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1578 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1579 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1581 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1582 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1583 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1584 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1585 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1586 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1587 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1588 Opt_inode_readahead_blks, Opt_journal_ioprio,
1589 Opt_dioread_nolock, Opt_dioread_lock,
1590 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1591 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1592 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1593 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1594 #ifdef CONFIG_EXT4_DEBUG
1595 Opt_fc_debug_max_replay, Opt_fc_debug_force
1599 static const struct constant_table ext4_param_errors[] = {
1600 {"continue", EXT4_MOUNT_ERRORS_CONT},
1601 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1602 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1606 static const struct constant_table ext4_param_data[] = {
1607 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1608 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1609 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1613 static const struct constant_table ext4_param_data_err[] = {
1614 {"abort", Opt_data_err_abort},
1615 {"ignore", Opt_data_err_ignore},
1619 static const struct constant_table ext4_param_jqfmt[] = {
1620 {"vfsold", QFMT_VFS_OLD},
1621 {"vfsv0", QFMT_VFS_V0},
1622 {"vfsv1", QFMT_VFS_V1},
1626 static const struct constant_table ext4_param_dax[] = {
1627 {"always", Opt_dax_always},
1628 {"inode", Opt_dax_inode},
1629 {"never", Opt_dax_never},
1633 /* String parameter that allows empty argument */
1634 #define fsparam_string_empty(NAME, OPT) \
1635 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1638 * Mount option specification
1639 * We don't use fsparam_flag_no because of the way we set the
1640 * options and the way we show them in _ext4_show_options(). To
1641 * keep the changes to a minimum, let's keep the negative options
1644 static const struct fs_parameter_spec ext4_param_specs[] = {
1645 fsparam_flag ("bsddf", Opt_bsd_df),
1646 fsparam_flag ("minixdf", Opt_minix_df),
1647 fsparam_flag ("grpid", Opt_grpid),
1648 fsparam_flag ("bsdgroups", Opt_grpid),
1649 fsparam_flag ("nogrpid", Opt_nogrpid),
1650 fsparam_flag ("sysvgroups", Opt_nogrpid),
1651 fsparam_u32 ("resgid", Opt_resgid),
1652 fsparam_u32 ("resuid", Opt_resuid),
1653 fsparam_u32 ("sb", Opt_sb),
1654 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1655 fsparam_flag ("nouid32", Opt_nouid32),
1656 fsparam_flag ("debug", Opt_debug),
1657 fsparam_flag ("oldalloc", Opt_removed),
1658 fsparam_flag ("orlov", Opt_removed),
1659 fsparam_flag ("user_xattr", Opt_user_xattr),
1660 fsparam_flag ("acl", Opt_acl),
1661 fsparam_flag ("norecovery", Opt_noload),
1662 fsparam_flag ("noload", Opt_noload),
1663 fsparam_flag ("bh", Opt_removed),
1664 fsparam_flag ("nobh", Opt_removed),
1665 fsparam_u32 ("commit", Opt_commit),
1666 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1667 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1668 fsparam_u32 ("journal_dev", Opt_journal_dev),
1669 fsparam_bdev ("journal_path", Opt_journal_path),
1670 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1671 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1672 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1673 fsparam_flag ("abort", Opt_abort),
1674 fsparam_enum ("data", Opt_data, ext4_param_data),
1675 fsparam_enum ("data_err", Opt_data_err,
1676 ext4_param_data_err),
1677 fsparam_string_empty
1678 ("usrjquota", Opt_usrjquota),
1679 fsparam_string_empty
1680 ("grpjquota", Opt_grpjquota),
1681 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1682 fsparam_flag ("grpquota", Opt_grpquota),
1683 fsparam_flag ("quota", Opt_quota),
1684 fsparam_flag ("noquota", Opt_noquota),
1685 fsparam_flag ("usrquota", Opt_usrquota),
1686 fsparam_flag ("prjquota", Opt_prjquota),
1687 fsparam_flag ("barrier", Opt_barrier),
1688 fsparam_u32 ("barrier", Opt_barrier),
1689 fsparam_flag ("nobarrier", Opt_nobarrier),
1690 fsparam_flag ("i_version", Opt_removed),
1691 fsparam_flag ("dax", Opt_dax),
1692 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1693 fsparam_u32 ("stripe", Opt_stripe),
1694 fsparam_flag ("delalloc", Opt_delalloc),
1695 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1696 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1697 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1698 fsparam_u32 ("debug_want_extra_isize",
1699 Opt_debug_want_extra_isize),
1700 fsparam_flag ("mblk_io_submit", Opt_removed),
1701 fsparam_flag ("nomblk_io_submit", Opt_removed),
1702 fsparam_flag ("block_validity", Opt_block_validity),
1703 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1704 fsparam_u32 ("inode_readahead_blks",
1705 Opt_inode_readahead_blks),
1706 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1707 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1708 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1709 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1710 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1711 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1712 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1713 fsparam_flag ("discard", Opt_discard),
1714 fsparam_flag ("nodiscard", Opt_nodiscard),
1715 fsparam_u32 ("init_itable", Opt_init_itable),
1716 fsparam_flag ("init_itable", Opt_init_itable),
1717 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1718 #ifdef CONFIG_EXT4_DEBUG
1719 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1720 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1722 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1723 fsparam_flag ("test_dummy_encryption",
1724 Opt_test_dummy_encryption),
1725 fsparam_string ("test_dummy_encryption",
1726 Opt_test_dummy_encryption),
1727 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1728 fsparam_flag ("nombcache", Opt_nombcache),
1729 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1730 fsparam_flag ("prefetch_block_bitmaps",
1732 fsparam_flag ("no_prefetch_block_bitmaps",
1733 Opt_no_prefetch_block_bitmaps),
1734 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1735 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1736 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1737 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1738 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1739 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1743 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1745 #define MOPT_SET 0x0001
1746 #define MOPT_CLEAR 0x0002
1747 #define MOPT_NOSUPPORT 0x0004
1748 #define MOPT_EXPLICIT 0x0008
1751 #define MOPT_QFMT 0x0010
1753 #define MOPT_Q MOPT_NOSUPPORT
1754 #define MOPT_QFMT MOPT_NOSUPPORT
1756 #define MOPT_NO_EXT2 0x0020
1757 #define MOPT_NO_EXT3 0x0040
1758 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1759 #define MOPT_SKIP 0x0080
1760 #define MOPT_2 0x0100
1762 static const struct mount_opts {
1766 } ext4_mount_opts[] = {
1767 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1768 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1769 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1770 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1771 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1772 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1773 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1774 MOPT_EXT4_ONLY | MOPT_SET},
1775 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1776 MOPT_EXT4_ONLY | MOPT_CLEAR},
1777 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1778 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1779 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1780 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1781 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1782 MOPT_EXT4_ONLY | MOPT_CLEAR},
1783 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1784 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1785 {Opt_commit, 0, MOPT_NO_EXT2},
1786 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1787 MOPT_EXT4_ONLY | MOPT_CLEAR},
1788 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1789 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1790 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1791 EXT4_MOUNT_JOURNAL_CHECKSUM),
1792 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1793 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1794 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1795 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1796 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1797 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1798 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1799 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1800 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1801 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1802 {Opt_journal_path, 0, MOPT_NO_EXT2},
1803 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1804 {Opt_data, 0, MOPT_NO_EXT2},
1805 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1806 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1807 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1809 {Opt_acl, 0, MOPT_NOSUPPORT},
1811 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1812 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1813 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1814 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1816 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1818 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1820 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1821 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1822 MOPT_CLEAR | MOPT_Q},
1823 {Opt_usrjquota, 0, MOPT_Q},
1824 {Opt_grpjquota, 0, MOPT_Q},
1825 {Opt_jqfmt, 0, MOPT_QFMT},
1826 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1827 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1829 #ifdef CONFIG_EXT4_DEBUG
1830 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1831 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1836 #if IS_ENABLED(CONFIG_UNICODE)
1837 static const struct ext4_sb_encodings {
1840 unsigned int version;
1841 } ext4_sb_encoding_map[] = {
1842 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1845 static const struct ext4_sb_encodings *
1846 ext4_sb_read_encoding(const struct ext4_super_block *es)
1848 __u16 magic = le16_to_cpu(es->s_encoding);
1851 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1852 if (magic == ext4_sb_encoding_map[i].magic)
1853 return &ext4_sb_encoding_map[i];
1859 #define EXT4_SPEC_JQUOTA (1 << 0)
1860 #define EXT4_SPEC_JQFMT (1 << 1)
1861 #define EXT4_SPEC_DATAJ (1 << 2)
1862 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1863 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1864 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1865 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1866 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1867 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1868 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1869 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1870 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1871 #define EXT4_SPEC_s_stripe (1 << 13)
1872 #define EXT4_SPEC_s_resuid (1 << 14)
1873 #define EXT4_SPEC_s_resgid (1 << 15)
1874 #define EXT4_SPEC_s_commit_interval (1 << 16)
1875 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1876 #define EXT4_SPEC_s_sb_block (1 << 18)
1877 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1879 struct ext4_fs_context {
1880 char *s_qf_names[EXT4_MAXQUOTAS];
1881 struct fscrypt_dummy_policy dummy_enc_policy;
1882 int s_jquota_fmt; /* Format of quota to use */
1883 #ifdef CONFIG_EXT4_DEBUG
1884 int s_fc_debug_max_replay;
1886 unsigned short qname_spec;
1887 unsigned long vals_s_flags; /* Bits to set in s_flags */
1888 unsigned long mask_s_flags; /* Bits changed in s_flags */
1889 unsigned long journal_devnum;
1890 unsigned long s_commit_interval;
1891 unsigned long s_stripe;
1892 unsigned int s_inode_readahead_blks;
1893 unsigned int s_want_extra_isize;
1894 unsigned int s_li_wait_mult;
1895 unsigned int s_max_dir_size_kb;
1896 unsigned int journal_ioprio;
1897 unsigned int vals_s_mount_opt;
1898 unsigned int mask_s_mount_opt;
1899 unsigned int vals_s_mount_opt2;
1900 unsigned int mask_s_mount_opt2;
1901 unsigned long vals_s_mount_flags;
1902 unsigned long mask_s_mount_flags;
1903 unsigned int opt_flags; /* MOPT flags */
1905 u32 s_max_batch_time;
1906 u32 s_min_batch_time;
1909 ext4_fsblk_t s_sb_block;
1912 static void ext4_fc_free(struct fs_context *fc)
1914 struct ext4_fs_context *ctx = fc->fs_private;
1920 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1921 kfree(ctx->s_qf_names[i]);
1923 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1927 int ext4_init_fs_context(struct fs_context *fc)
1929 struct ext4_fs_context *ctx;
1931 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1935 fc->fs_private = ctx;
1936 fc->ops = &ext4_context_ops;
1943 * Note the name of the specified quota file.
1945 static int note_qf_name(struct fs_context *fc, int qtype,
1946 struct fs_parameter *param)
1948 struct ext4_fs_context *ctx = fc->fs_private;
1951 if (param->size < 1) {
1952 ext4_msg(NULL, KERN_ERR, "Missing quota name");
1955 if (strchr(param->string, '/')) {
1956 ext4_msg(NULL, KERN_ERR,
1957 "quotafile must be on filesystem root");
1960 if (ctx->s_qf_names[qtype]) {
1961 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
1962 ext4_msg(NULL, KERN_ERR,
1963 "%s quota file already specified",
1970 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
1972 ext4_msg(NULL, KERN_ERR,
1973 "Not enough memory for storing quotafile name");
1976 ctx->s_qf_names[qtype] = qname;
1977 ctx->qname_spec |= 1 << qtype;
1978 ctx->spec |= EXT4_SPEC_JQUOTA;
1983 * Clear the name of the specified quota file.
1985 static int unnote_qf_name(struct fs_context *fc, int qtype)
1987 struct ext4_fs_context *ctx = fc->fs_private;
1989 if (ctx->s_qf_names[qtype])
1990 kfree(ctx->s_qf_names[qtype]);
1992 ctx->s_qf_names[qtype] = NULL;
1993 ctx->qname_spec |= 1 << qtype;
1994 ctx->spec |= EXT4_SPEC_JQUOTA;
1999 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2000 struct ext4_fs_context *ctx)
2004 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2005 ext4_msg(NULL, KERN_WARNING,
2006 "test_dummy_encryption option not supported");
2009 err = fscrypt_parse_test_dummy_encryption(param,
2010 &ctx->dummy_enc_policy);
2011 if (err == -EINVAL) {
2012 ext4_msg(NULL, KERN_WARNING,
2013 "Value of option \"%s\" is unrecognized", param->key);
2014 } else if (err == -EEXIST) {
2015 ext4_msg(NULL, KERN_WARNING,
2016 "Conflicting test_dummy_encryption options");
2022 #define EXT4_SET_CTX(name) \
2023 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2024 unsigned long flag) \
2026 ctx->mask_s_##name |= flag; \
2027 ctx->vals_s_##name |= flag; \
2030 #define EXT4_CLEAR_CTX(name) \
2031 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2032 unsigned long flag) \
2034 ctx->mask_s_##name |= flag; \
2035 ctx->vals_s_##name &= ~flag; \
2038 #define EXT4_TEST_CTX(name) \
2039 static inline unsigned long \
2040 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2042 return (ctx->vals_s_##name & flag); \
2045 EXT4_SET_CTX(flags); /* set only */
2046 EXT4_SET_CTX(mount_opt);
2047 EXT4_CLEAR_CTX(mount_opt);
2048 EXT4_TEST_CTX(mount_opt);
2049 EXT4_SET_CTX(mount_opt2);
2050 EXT4_CLEAR_CTX(mount_opt2);
2051 EXT4_TEST_CTX(mount_opt2);
2053 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2055 set_bit(bit, &ctx->mask_s_mount_flags);
2056 set_bit(bit, &ctx->vals_s_mount_flags);
2059 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2061 struct ext4_fs_context *ctx = fc->fs_private;
2062 struct fs_parse_result result;
2063 const struct mount_opts *m;
2069 token = fs_parse(fc, ext4_param_specs, param, &result);
2072 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2074 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2075 if (token == m->token)
2078 ctx->opt_flags |= m->flags;
2080 if (m->flags & MOPT_EXPLICIT) {
2081 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2082 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2083 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2084 ctx_set_mount_opt2(ctx,
2085 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2090 if (m->flags & MOPT_NOSUPPORT) {
2091 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2099 if (!*param->string)
2100 return unnote_qf_name(fc, USRQUOTA);
2102 return note_qf_name(fc, USRQUOTA, param);
2104 if (!*param->string)
2105 return unnote_qf_name(fc, GRPQUOTA);
2107 return note_qf_name(fc, GRPQUOTA, param);
2110 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2111 ext4_msg(NULL, KERN_WARNING,
2112 "Ignoring %s option on remount", param->key);
2114 ctx->s_sb_block = result.uint_32;
2115 ctx->spec |= EXT4_SPEC_s_sb_block;
2119 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2123 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2125 case Opt_inlinecrypt:
2126 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2127 ctx_set_flags(ctx, SB_INLINECRYPT);
2129 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2133 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2134 ctx_set_mount_opt(ctx, result.uint_32);
2138 ctx->s_jquota_fmt = result.uint_32;
2139 ctx->spec |= EXT4_SPEC_JQFMT;
2143 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2144 ctx_set_mount_opt(ctx, result.uint_32);
2145 ctx->spec |= EXT4_SPEC_DATAJ;
2148 if (result.uint_32 == 0)
2149 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
2150 else if (result.uint_32 > INT_MAX / HZ) {
2151 ext4_msg(NULL, KERN_ERR,
2152 "Invalid commit interval %d, "
2153 "must be smaller than %d",
2154 result.uint_32, INT_MAX / HZ);
2157 ctx->s_commit_interval = HZ * result.uint_32;
2158 ctx->spec |= EXT4_SPEC_s_commit_interval;
2160 case Opt_debug_want_extra_isize:
2161 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2162 ext4_msg(NULL, KERN_ERR,
2163 "Invalid want_extra_isize %d", result.uint_32);
2166 ctx->s_want_extra_isize = result.uint_32;
2167 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2169 case Opt_max_batch_time:
2170 ctx->s_max_batch_time = result.uint_32;
2171 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2173 case Opt_min_batch_time:
2174 ctx->s_min_batch_time = result.uint_32;
2175 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2177 case Opt_inode_readahead_blks:
2178 if (result.uint_32 &&
2179 (result.uint_32 > (1 << 30) ||
2180 !is_power_of_2(result.uint_32))) {
2181 ext4_msg(NULL, KERN_ERR,
2182 "EXT4-fs: inode_readahead_blks must be "
2183 "0 or a power of 2 smaller than 2^31");
2186 ctx->s_inode_readahead_blks = result.uint_32;
2187 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2189 case Opt_init_itable:
2190 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2191 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2192 if (param->type == fs_value_is_string)
2193 ctx->s_li_wait_mult = result.uint_32;
2194 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2196 case Opt_max_dir_size_kb:
2197 ctx->s_max_dir_size_kb = result.uint_32;
2198 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2200 #ifdef CONFIG_EXT4_DEBUG
2201 case Opt_fc_debug_max_replay:
2202 ctx->s_fc_debug_max_replay = result.uint_32;
2203 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2207 ctx->s_stripe = result.uint_32;
2208 ctx->spec |= EXT4_SPEC_s_stripe;
2211 uid = make_kuid(current_user_ns(), result.uint_32);
2212 if (!uid_valid(uid)) {
2213 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2217 ctx->s_resuid = uid;
2218 ctx->spec |= EXT4_SPEC_s_resuid;
2221 gid = make_kgid(current_user_ns(), result.uint_32);
2222 if (!gid_valid(gid)) {
2223 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2227 ctx->s_resgid = gid;
2228 ctx->spec |= EXT4_SPEC_s_resgid;
2230 case Opt_journal_dev:
2232 ext4_msg(NULL, KERN_ERR,
2233 "Cannot specify journal on remount");
2236 ctx->journal_devnum = result.uint_32;
2237 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2239 case Opt_journal_path:
2241 struct inode *journal_inode;
2246 ext4_msg(NULL, KERN_ERR,
2247 "Cannot specify journal on remount");
2251 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2253 ext4_msg(NULL, KERN_ERR, "error: could not find "
2254 "journal device path");
2258 journal_inode = d_inode(path.dentry);
2259 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2260 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2264 case Opt_journal_ioprio:
2265 if (result.uint_32 > 7) {
2266 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2270 ctx->journal_ioprio =
2271 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2272 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2274 case Opt_test_dummy_encryption:
2275 return ext4_parse_test_dummy_encryption(param, ctx);
2278 #ifdef CONFIG_FS_DAX
2280 int type = (token == Opt_dax) ?
2281 Opt_dax : result.uint_32;
2285 case Opt_dax_always:
2286 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2287 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2290 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2291 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2294 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2295 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2296 /* Strictly for printing options */
2297 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2303 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2307 if (result.uint_32 == Opt_data_err_abort)
2308 ctx_set_mount_opt(ctx, m->mount_opt);
2309 else if (result.uint_32 == Opt_data_err_ignore)
2310 ctx_clear_mount_opt(ctx, m->mount_opt);
2312 case Opt_mb_optimize_scan:
2313 if (result.int_32 == 1) {
2314 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2315 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2316 } else if (result.int_32 == 0) {
2317 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2318 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2320 ext4_msg(NULL, KERN_WARNING,
2321 "mb_optimize_scan should be set to 0 or 1.");
2328 * At this point we should only be getting options requiring MOPT_SET,
2329 * or MOPT_CLEAR. Anything else is a bug
2331 if (m->token == Opt_err) {
2332 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2339 unsigned int set = 0;
2341 if ((param->type == fs_value_is_flag) ||
2345 if (m->flags & MOPT_CLEAR)
2347 else if (unlikely(!(m->flags & MOPT_SET))) {
2348 ext4_msg(NULL, KERN_WARNING,
2349 "buggy handling of option %s",
2354 if (m->flags & MOPT_2) {
2356 ctx_set_mount_opt2(ctx, m->mount_opt);
2358 ctx_clear_mount_opt2(ctx, m->mount_opt);
2361 ctx_set_mount_opt(ctx, m->mount_opt);
2363 ctx_clear_mount_opt(ctx, m->mount_opt);
2370 static int parse_options(struct fs_context *fc, char *options)
2372 struct fs_parameter param;
2379 while ((key = strsep(&options, ",")) != NULL) {
2382 char *value = strchr(key, '=');
2384 param.type = fs_value_is_flag;
2385 param.string = NULL;
2392 v_len = strlen(value);
2393 param.string = kmemdup_nul(value, v_len,
2397 param.type = fs_value_is_string;
2403 ret = ext4_parse_param(fc, ¶m);
2405 kfree(param.string);
2411 ret = ext4_validate_options(fc);
2418 static int parse_apply_sb_mount_options(struct super_block *sb,
2419 struct ext4_fs_context *m_ctx)
2421 struct ext4_sb_info *sbi = EXT4_SB(sb);
2422 char *s_mount_opts = NULL;
2423 struct ext4_fs_context *s_ctx = NULL;
2424 struct fs_context *fc = NULL;
2427 if (!sbi->s_es->s_mount_opts[0])
2430 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2431 sizeof(sbi->s_es->s_mount_opts),
2436 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2440 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2444 fc->fs_private = s_ctx;
2445 fc->s_fs_info = sbi;
2447 ret = parse_options(fc, s_mount_opts);
2451 ret = ext4_check_opt_consistency(fc, sb);
2454 ext4_msg(sb, KERN_WARNING,
2455 "failed to parse options in superblock: %s",
2461 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2462 m_ctx->journal_devnum = s_ctx->journal_devnum;
2463 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2464 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2466 ext4_apply_options(fc, sb);
2474 kfree(s_mount_opts);
2478 static void ext4_apply_quota_options(struct fs_context *fc,
2479 struct super_block *sb)
2482 bool quota_feature = ext4_has_feature_quota(sb);
2483 struct ext4_fs_context *ctx = fc->fs_private;
2484 struct ext4_sb_info *sbi = EXT4_SB(sb);
2491 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2492 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2493 if (!(ctx->qname_spec & (1 << i)))
2496 qname = ctx->s_qf_names[i]; /* May be NULL */
2499 ctx->s_qf_names[i] = NULL;
2500 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2501 lockdep_is_held(&sb->s_umount));
2507 if (ctx->spec & EXT4_SPEC_JQFMT)
2508 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2513 * Check quota settings consistency.
2515 static int ext4_check_quota_consistency(struct fs_context *fc,
2516 struct super_block *sb)
2519 struct ext4_fs_context *ctx = fc->fs_private;
2520 struct ext4_sb_info *sbi = EXT4_SB(sb);
2521 bool quota_feature = ext4_has_feature_quota(sb);
2522 bool quota_loaded = sb_any_quota_loaded(sb);
2523 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2527 * We do the test below only for project quotas. 'usrquota' and
2528 * 'grpquota' mount options are allowed even without quota feature
2529 * to support legacy quotas in quota files.
2531 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2532 !ext4_has_feature_project(sb)) {
2533 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2534 "Cannot enable project quota enforcement.");
2538 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2539 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2541 ctx->mask_s_mount_opt & quota_flags &&
2542 !ctx_test_mount_opt(ctx, quota_flags))
2543 goto err_quota_change;
2545 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2547 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2548 if (!(ctx->qname_spec & (1 << i)))
2552 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2553 goto err_jquota_change;
2555 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2556 strcmp(get_qf_name(sb, sbi, i),
2557 ctx->s_qf_names[i]) != 0)
2558 goto err_jquota_specified;
2561 if (quota_feature) {
2562 ext4_msg(NULL, KERN_INFO,
2563 "Journaled quota options ignored when "
2564 "QUOTA feature is enabled");
2569 if (ctx->spec & EXT4_SPEC_JQFMT) {
2570 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2571 goto err_jquota_change;
2572 if (quota_feature) {
2573 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2574 "ignored when QUOTA feature is enabled");
2579 /* Make sure we don't mix old and new quota format */
2580 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2581 ctx->s_qf_names[USRQUOTA]);
2582 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2583 ctx->s_qf_names[GRPQUOTA]);
2585 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2586 test_opt(sb, USRQUOTA));
2588 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2589 test_opt(sb, GRPQUOTA));
2592 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2596 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2600 if (usr_qf_name || grp_qf_name) {
2601 if (usrquota || grpquota) {
2602 ext4_msg(NULL, KERN_ERR, "old and new quota "
2607 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2608 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2617 ext4_msg(NULL, KERN_ERR,
2618 "Cannot change quota options when quota turned on");
2621 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2622 "options when quota turned on");
2624 err_jquota_specified:
2625 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2633 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2634 struct super_block *sb)
2636 const struct ext4_fs_context *ctx = fc->fs_private;
2637 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2640 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2643 if (!ext4_has_feature_encrypt(sb)) {
2644 ext4_msg(NULL, KERN_WARNING,
2645 "test_dummy_encryption requires encrypt feature");
2649 * This mount option is just for testing, and it's not worthwhile to
2650 * implement the extra complexity (e.g. RCU protection) that would be
2651 * needed to allow it to be set or changed during remount. We do allow
2652 * it to be specified during remount, but only if there is no change.
2654 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2655 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2656 &ctx->dummy_enc_policy))
2658 ext4_msg(NULL, KERN_WARNING,
2659 "Can't set or change test_dummy_encryption on remount");
2662 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2663 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2664 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2665 &ctx->dummy_enc_policy))
2667 ext4_msg(NULL, KERN_WARNING,
2668 "Conflicting test_dummy_encryption options");
2672 * fscrypt_add_test_dummy_key() technically changes the super_block, so
2673 * technically it should be delayed until ext4_apply_options() like the
2674 * other changes. But since we never get here for remounts (see above),
2675 * and this is the last chance to report errors, we do it here.
2677 err = fscrypt_add_test_dummy_key(sb, &ctx->dummy_enc_policy);
2679 ext4_msg(NULL, KERN_WARNING,
2680 "Error adding test dummy encryption key [%d]", err);
2684 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2685 struct super_block *sb)
2687 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2688 /* if already set, it was already verified to be the same */
2689 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2691 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2692 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2693 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2696 static int ext4_check_opt_consistency(struct fs_context *fc,
2697 struct super_block *sb)
2699 struct ext4_fs_context *ctx = fc->fs_private;
2700 struct ext4_sb_info *sbi = fc->s_fs_info;
2701 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2704 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2705 ext4_msg(NULL, KERN_ERR,
2706 "Mount option(s) incompatible with ext2");
2709 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2710 ext4_msg(NULL, KERN_ERR,
2711 "Mount option(s) incompatible with ext3");
2715 if (ctx->s_want_extra_isize >
2716 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2717 ext4_msg(NULL, KERN_ERR,
2718 "Invalid want_extra_isize %d",
2719 ctx->s_want_extra_isize);
2723 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2725 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2726 if (blocksize < PAGE_SIZE)
2727 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2728 "experimental mount option 'dioread_nolock' "
2729 "for blocksize < PAGE_SIZE");
2732 err = ext4_check_test_dummy_encryption(fc, sb);
2736 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2737 if (!sbi->s_journal) {
2738 ext4_msg(NULL, KERN_WARNING,
2739 "Remounting file system with no journal "
2740 "so ignoring journalled data option");
2741 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2742 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2743 test_opt(sb, DATA_FLAGS)) {
2744 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2751 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2752 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2753 ext4_msg(NULL, KERN_ERR, "can't mount with "
2754 "both data=journal and dax");
2758 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2759 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2760 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2761 fail_dax_change_remount:
2762 ext4_msg(NULL, KERN_ERR, "can't change "
2763 "dax mount option while remounting");
2765 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2766 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2767 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2768 goto fail_dax_change_remount;
2769 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2770 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2771 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2772 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2773 goto fail_dax_change_remount;
2777 return ext4_check_quota_consistency(fc, sb);
2780 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2782 struct ext4_fs_context *ctx = fc->fs_private;
2783 struct ext4_sb_info *sbi = fc->s_fs_info;
2785 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2786 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2787 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2788 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2789 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2790 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2791 sb->s_flags &= ~ctx->mask_s_flags;
2792 sb->s_flags |= ctx->vals_s_flags;
2794 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2795 APPLY(s_commit_interval);
2797 APPLY(s_max_batch_time);
2798 APPLY(s_min_batch_time);
2799 APPLY(s_want_extra_isize);
2800 APPLY(s_inode_readahead_blks);
2801 APPLY(s_max_dir_size_kb);
2802 APPLY(s_li_wait_mult);
2806 #ifdef CONFIG_EXT4_DEBUG
2807 APPLY(s_fc_debug_max_replay);
2810 ext4_apply_quota_options(fc, sb);
2811 ext4_apply_test_dummy_encryption(ctx, sb);
2815 static int ext4_validate_options(struct fs_context *fc)
2818 struct ext4_fs_context *ctx = fc->fs_private;
2819 char *usr_qf_name, *grp_qf_name;
2821 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2822 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2824 if (usr_qf_name || grp_qf_name) {
2825 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2826 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2828 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2829 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2831 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2832 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2833 ext4_msg(NULL, KERN_ERR, "old and new quota "
2842 static inline void ext4_show_quota_options(struct seq_file *seq,
2843 struct super_block *sb)
2845 #if defined(CONFIG_QUOTA)
2846 struct ext4_sb_info *sbi = EXT4_SB(sb);
2847 char *usr_qf_name, *grp_qf_name;
2849 if (sbi->s_jquota_fmt) {
2852 switch (sbi->s_jquota_fmt) {
2863 seq_printf(seq, ",jqfmt=%s", fmtname);
2867 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2868 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2870 seq_show_option(seq, "usrjquota", usr_qf_name);
2872 seq_show_option(seq, "grpjquota", grp_qf_name);
2877 static const char *token2str(int token)
2879 const struct fs_parameter_spec *spec;
2881 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2882 if (spec->opt == token && !spec->type)
2889 * - it's set to a non-default value OR
2890 * - if the per-sb default is different from the global default
2892 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2895 struct ext4_sb_info *sbi = EXT4_SB(sb);
2896 struct ext4_super_block *es = sbi->s_es;
2897 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2898 const struct mount_opts *m;
2899 char sep = nodefs ? '\n' : ',';
2901 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2902 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2904 if (sbi->s_sb_block != 1)
2905 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2907 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2908 int want_set = m->flags & MOPT_SET;
2909 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2910 m->flags & MOPT_SKIP)
2912 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2913 continue; /* skip if same as the default */
2915 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2916 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2917 continue; /* select Opt_noFoo vs Opt_Foo */
2918 SEQ_OPTS_PRINT("%s", token2str(m->token));
2921 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2922 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2923 SEQ_OPTS_PRINT("resuid=%u",
2924 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2925 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2926 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2927 SEQ_OPTS_PRINT("resgid=%u",
2928 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2929 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2930 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2931 SEQ_OPTS_PUTS("errors=remount-ro");
2932 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2933 SEQ_OPTS_PUTS("errors=continue");
2934 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2935 SEQ_OPTS_PUTS("errors=panic");
2936 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2937 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2938 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2939 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2940 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2941 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2942 if (nodefs || sbi->s_stripe)
2943 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2944 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2945 (sbi->s_mount_opt ^ def_mount_opt)) {
2946 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2947 SEQ_OPTS_PUTS("data=journal");
2948 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2949 SEQ_OPTS_PUTS("data=ordered");
2950 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2951 SEQ_OPTS_PUTS("data=writeback");
2954 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2955 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2956 sbi->s_inode_readahead_blks);
2958 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2959 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2960 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2961 if (nodefs || sbi->s_max_dir_size_kb)
2962 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2963 if (test_opt(sb, DATA_ERR_ABORT))
2964 SEQ_OPTS_PUTS("data_err=abort");
2966 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2968 if (sb->s_flags & SB_INLINECRYPT)
2969 SEQ_OPTS_PUTS("inlinecrypt");
2971 if (test_opt(sb, DAX_ALWAYS)) {
2973 SEQ_OPTS_PUTS("dax");
2975 SEQ_OPTS_PUTS("dax=always");
2976 } else if (test_opt2(sb, DAX_NEVER)) {
2977 SEQ_OPTS_PUTS("dax=never");
2978 } else if (test_opt2(sb, DAX_INODE)) {
2979 SEQ_OPTS_PUTS("dax=inode");
2982 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2983 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2984 SEQ_OPTS_PUTS("mb_optimize_scan=0");
2985 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2986 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2987 SEQ_OPTS_PUTS("mb_optimize_scan=1");
2990 ext4_show_quota_options(seq, sb);
2994 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2996 return _ext4_show_options(seq, root->d_sb, 0);
2999 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3001 struct super_block *sb = seq->private;
3004 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3005 rc = _ext4_show_options(seq, sb, 1);
3006 seq_puts(seq, "\n");
3010 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3013 struct ext4_sb_info *sbi = EXT4_SB(sb);
3016 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3017 ext4_msg(sb, KERN_ERR, "revision level too high, "
3018 "forcing read-only mode");
3024 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3025 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3026 "running e2fsck is recommended");
3027 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3028 ext4_msg(sb, KERN_WARNING,
3029 "warning: mounting fs with errors, "
3030 "running e2fsck is recommended");
3031 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3032 le16_to_cpu(es->s_mnt_count) >=
3033 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3034 ext4_msg(sb, KERN_WARNING,
3035 "warning: maximal mount count reached, "
3036 "running e2fsck is recommended");
3037 else if (le32_to_cpu(es->s_checkinterval) &&
3038 (ext4_get_tstamp(es, s_lastcheck) +
3039 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3040 ext4_msg(sb, KERN_WARNING,
3041 "warning: checktime reached, "
3042 "running e2fsck is recommended");
3043 if (!sbi->s_journal)
3044 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3045 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3046 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3047 le16_add_cpu(&es->s_mnt_count, 1);
3048 ext4_update_tstamp(es, s_mtime);
3049 if (sbi->s_journal) {
3050 ext4_set_feature_journal_needs_recovery(sb);
3051 if (ext4_has_feature_orphan_file(sb))
3052 ext4_set_feature_orphan_present(sb);
3055 err = ext4_commit_super(sb);
3057 if (test_opt(sb, DEBUG))
3058 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3059 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3061 sbi->s_groups_count,
3062 EXT4_BLOCKS_PER_GROUP(sb),
3063 EXT4_INODES_PER_GROUP(sb),
3064 sbi->s_mount_opt, sbi->s_mount_opt2);
3068 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3070 struct ext4_sb_info *sbi = EXT4_SB(sb);
3071 struct flex_groups **old_groups, **new_groups;
3074 if (!sbi->s_log_groups_per_flex)
3077 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3078 if (size <= sbi->s_flex_groups_allocated)
3081 new_groups = kvzalloc(roundup_pow_of_two(size *
3082 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3084 ext4_msg(sb, KERN_ERR,
3085 "not enough memory for %d flex group pointers", size);
3088 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3089 new_groups[i] = kvzalloc(roundup_pow_of_two(
3090 sizeof(struct flex_groups)),
3092 if (!new_groups[i]) {
3093 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3094 kvfree(new_groups[j]);
3096 ext4_msg(sb, KERN_ERR,
3097 "not enough memory for %d flex groups", size);
3102 old_groups = rcu_dereference(sbi->s_flex_groups);
3104 memcpy(new_groups, old_groups,
3105 (sbi->s_flex_groups_allocated *
3106 sizeof(struct flex_groups *)));
3108 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3109 sbi->s_flex_groups_allocated = size;
3111 ext4_kvfree_array_rcu(old_groups);
3115 static int ext4_fill_flex_info(struct super_block *sb)
3117 struct ext4_sb_info *sbi = EXT4_SB(sb);
3118 struct ext4_group_desc *gdp = NULL;
3119 struct flex_groups *fg;
3120 ext4_group_t flex_group;
3123 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3124 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3125 sbi->s_log_groups_per_flex = 0;
3129 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3133 for (i = 0; i < sbi->s_groups_count; i++) {
3134 gdp = ext4_get_group_desc(sb, i, NULL);
3136 flex_group = ext4_flex_group(sbi, i);
3137 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3138 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3139 atomic64_add(ext4_free_group_clusters(sb, gdp),
3140 &fg->free_clusters);
3141 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3149 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3150 struct ext4_group_desc *gdp)
3152 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3154 __le32 le_group = cpu_to_le32(block_group);
3155 struct ext4_sb_info *sbi = EXT4_SB(sb);
3157 if (ext4_has_metadata_csum(sbi->s_sb)) {
3158 /* Use new metadata_csum algorithm */
3160 __u16 dummy_csum = 0;
3162 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3164 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3165 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3166 sizeof(dummy_csum));
3167 offset += sizeof(dummy_csum);
3168 if (offset < sbi->s_desc_size)
3169 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3170 sbi->s_desc_size - offset);
3172 crc = csum32 & 0xFFFF;
3176 /* old crc16 code */
3177 if (!ext4_has_feature_gdt_csum(sb))
3180 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3181 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3182 crc = crc16(crc, (__u8 *)gdp, offset);
3183 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3184 /* for checksum of struct ext4_group_desc do the rest...*/
3185 if (ext4_has_feature_64bit(sb) &&
3186 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3187 crc = crc16(crc, (__u8 *)gdp + offset,
3188 le16_to_cpu(sbi->s_es->s_desc_size) -
3192 return cpu_to_le16(crc);
3195 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3196 struct ext4_group_desc *gdp)
3198 if (ext4_has_group_desc_csum(sb) &&
3199 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3205 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3206 struct ext4_group_desc *gdp)
3208 if (!ext4_has_group_desc_csum(sb))
3210 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3213 /* Called at mount-time, super-block is locked */
3214 static int ext4_check_descriptors(struct super_block *sb,
3215 ext4_fsblk_t sb_block,
3216 ext4_group_t *first_not_zeroed)
3218 struct ext4_sb_info *sbi = EXT4_SB(sb);
3219 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3220 ext4_fsblk_t last_block;
3221 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3222 ext4_fsblk_t block_bitmap;
3223 ext4_fsblk_t inode_bitmap;
3224 ext4_fsblk_t inode_table;
3225 int flexbg_flag = 0;
3226 ext4_group_t i, grp = sbi->s_groups_count;
3228 if (ext4_has_feature_flex_bg(sb))
3231 ext4_debug("Checking group descriptors");
3233 for (i = 0; i < sbi->s_groups_count; i++) {
3234 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3236 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3237 last_block = ext4_blocks_count(sbi->s_es) - 1;
3239 last_block = first_block +
3240 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3242 if ((grp == sbi->s_groups_count) &&
3243 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3246 block_bitmap = ext4_block_bitmap(sb, gdp);
3247 if (block_bitmap == sb_block) {
3248 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3249 "Block bitmap for group %u overlaps "
3254 if (block_bitmap >= sb_block + 1 &&
3255 block_bitmap <= last_bg_block) {
3256 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3257 "Block bitmap for group %u overlaps "
3258 "block group descriptors", i);
3262 if (block_bitmap < first_block || block_bitmap > last_block) {
3263 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3264 "Block bitmap for group %u not in group "
3265 "(block %llu)!", i, block_bitmap);
3268 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3269 if (inode_bitmap == sb_block) {
3270 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3271 "Inode bitmap for group %u overlaps "
3276 if (inode_bitmap >= sb_block + 1 &&
3277 inode_bitmap <= last_bg_block) {
3278 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3279 "Inode bitmap for group %u overlaps "
3280 "block group descriptors", i);
3284 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3285 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3286 "Inode bitmap for group %u not in group "
3287 "(block %llu)!", i, inode_bitmap);
3290 inode_table = ext4_inode_table(sb, gdp);
3291 if (inode_table == sb_block) {
3292 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3293 "Inode table for group %u overlaps "
3298 if (inode_table >= sb_block + 1 &&
3299 inode_table <= last_bg_block) {
3300 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3301 "Inode table for group %u overlaps "
3302 "block group descriptors", i);
3306 if (inode_table < first_block ||
3307 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3308 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3309 "Inode table for group %u not in group "
3310 "(block %llu)!", i, inode_table);
3313 ext4_lock_group(sb, i);
3314 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3316 "Checksum for group %u failed (%u!=%u)",
3317 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3318 gdp)), le16_to_cpu(gdp->bg_checksum));
3319 if (!sb_rdonly(sb)) {
3320 ext4_unlock_group(sb, i);
3324 ext4_unlock_group(sb, i);
3326 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3328 if (NULL != first_not_zeroed)
3329 *first_not_zeroed = grp;
3334 * Maximal extent format file size.
3335 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3336 * extent format containers, within a sector_t, and within i_blocks
3337 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3338 * so that won't be a limiting factor.
3340 * However there is other limiting factor. We do store extents in the form
3341 * of starting block and length, hence the resulting length of the extent
3342 * covering maximum file size must fit into on-disk format containers as
3343 * well. Given that length is always by 1 unit bigger than max unit (because
3344 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3346 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3348 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3351 loff_t upper_limit = MAX_LFS_FILESIZE;
3353 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3355 if (!has_huge_files) {
3356 upper_limit = (1LL << 32) - 1;
3358 /* total blocks in file system block size */
3359 upper_limit >>= (blkbits - 9);
3360 upper_limit <<= blkbits;
3364 * 32-bit extent-start container, ee_block. We lower the maxbytes
3365 * by one fs block, so ee_len can cover the extent of maximum file
3368 res = (1LL << 32) - 1;
3371 /* Sanity check against vm- & vfs- imposed limits */
3372 if (res > upper_limit)
3379 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3380 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3381 * We need to be 1 filesystem block less than the 2^48 sector limit.
3383 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3385 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3387 unsigned int ppb = 1 << (bits - 2);
3390 * This is calculated to be the largest file size for a dense, block
3391 * mapped file such that the file's total number of 512-byte sectors,
3392 * including data and all indirect blocks, does not exceed (2^48 - 1).
3394 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3395 * number of 512-byte sectors of the file.
3397 if (!has_huge_files) {
3399 * !has_huge_files or implies that the inode i_block field
3400 * represents total file blocks in 2^32 512-byte sectors ==
3401 * size of vfs inode i_blocks * 8
3403 upper_limit = (1LL << 32) - 1;
3405 /* total blocks in file system block size */
3406 upper_limit >>= (bits - 9);
3410 * We use 48 bit ext4_inode i_blocks
3411 * With EXT4_HUGE_FILE_FL set the i_blocks
3412 * represent total number of blocks in
3413 * file system block size
3415 upper_limit = (1LL << 48) - 1;
3419 /* Compute how many blocks we can address by block tree */
3422 res += ((loff_t)ppb) * ppb * ppb;
3423 /* Compute how many metadata blocks are needed */
3425 meta_blocks += 1 + ppb;
3426 meta_blocks += 1 + ppb + ppb * ppb;
3427 /* Does block tree limit file size? */
3428 if (res + meta_blocks <= upper_limit)
3432 /* How many metadata blocks are needed for addressing upper_limit? */
3433 upper_limit -= EXT4_NDIR_BLOCKS;
3434 /* indirect blocks */
3437 /* double indirect blocks */
3438 if (upper_limit < ppb * ppb) {
3439 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3443 meta_blocks += 1 + ppb;
3444 upper_limit -= ppb * ppb;
3445 /* tripple indirect blocks for the rest */
3446 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3447 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3451 if (res > MAX_LFS_FILESIZE)
3452 res = MAX_LFS_FILESIZE;
3457 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3458 ext4_fsblk_t logical_sb_block, int nr)
3460 struct ext4_sb_info *sbi = EXT4_SB(sb);
3461 ext4_group_t bg, first_meta_bg;
3464 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3466 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3467 return logical_sb_block + nr + 1;
3468 bg = sbi->s_desc_per_block * nr;
3469 if (ext4_bg_has_super(sb, bg))
3473 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3474 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3475 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3478 if (sb->s_blocksize == 1024 && nr == 0 &&
3479 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3482 return (has_super + ext4_group_first_block_no(sb, bg));
3486 * ext4_get_stripe_size: Get the stripe size.
3487 * @sbi: In memory super block info
3489 * If we have specified it via mount option, then
3490 * use the mount option value. If the value specified at mount time is
3491 * greater than the blocks per group use the super block value.
3492 * If the super block value is greater than blocks per group return 0.
3493 * Allocator needs it be less than blocks per group.
3496 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3498 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3499 unsigned long stripe_width =
3500 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3503 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3504 ret = sbi->s_stripe;
3505 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3507 else if (stride && stride <= sbi->s_blocks_per_group)
3513 * If the stripe width is 1, this makes no sense and
3514 * we set it to 0 to turn off stripe handling code.
3523 * Check whether this filesystem can be mounted based on
3524 * the features present and the RDONLY/RDWR mount requested.
3525 * Returns 1 if this filesystem can be mounted as requested,
3526 * 0 if it cannot be.
3528 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3530 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3531 ext4_msg(sb, KERN_ERR,
3532 "Couldn't mount because of "
3533 "unsupported optional features (%x)",
3534 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3535 ~EXT4_FEATURE_INCOMPAT_SUPP));
3539 #if !IS_ENABLED(CONFIG_UNICODE)
3540 if (ext4_has_feature_casefold(sb)) {
3541 ext4_msg(sb, KERN_ERR,
3542 "Filesystem with casefold feature cannot be "
3543 "mounted without CONFIG_UNICODE");
3551 if (ext4_has_feature_readonly(sb)) {
3552 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3553 sb->s_flags |= SB_RDONLY;
3557 /* Check that feature set is OK for a read-write mount */
3558 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3559 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3560 "unsupported optional features (%x)",
3561 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3562 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3565 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3566 ext4_msg(sb, KERN_ERR,
3567 "Can't support bigalloc feature without "
3568 "extents feature\n");
3572 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3573 if (!readonly && (ext4_has_feature_quota(sb) ||
3574 ext4_has_feature_project(sb))) {
3575 ext4_msg(sb, KERN_ERR,
3576 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3579 #endif /* CONFIG_QUOTA */
3584 * This function is called once a day if we have errors logged
3585 * on the file system
3587 static void print_daily_error_info(struct timer_list *t)
3589 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3590 struct super_block *sb = sbi->s_sb;
3591 struct ext4_super_block *es = sbi->s_es;
3593 if (es->s_error_count)
3594 /* fsck newer than v1.41.13 is needed to clean this condition. */
3595 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3596 le32_to_cpu(es->s_error_count));
3597 if (es->s_first_error_time) {
3598 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3600 ext4_get_tstamp(es, s_first_error_time),
3601 (int) sizeof(es->s_first_error_func),
3602 es->s_first_error_func,
3603 le32_to_cpu(es->s_first_error_line));
3604 if (es->s_first_error_ino)
3605 printk(KERN_CONT ": inode %u",
3606 le32_to_cpu(es->s_first_error_ino));
3607 if (es->s_first_error_block)
3608 printk(KERN_CONT ": block %llu", (unsigned long long)
3609 le64_to_cpu(es->s_first_error_block));
3610 printk(KERN_CONT "\n");
3612 if (es->s_last_error_time) {
3613 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3615 ext4_get_tstamp(es, s_last_error_time),
3616 (int) sizeof(es->s_last_error_func),
3617 es->s_last_error_func,
3618 le32_to_cpu(es->s_last_error_line));
3619 if (es->s_last_error_ino)
3620 printk(KERN_CONT ": inode %u",
3621 le32_to_cpu(es->s_last_error_ino));
3622 if (es->s_last_error_block)
3623 printk(KERN_CONT ": block %llu", (unsigned long long)
3624 le64_to_cpu(es->s_last_error_block));
3625 printk(KERN_CONT "\n");
3627 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3630 /* Find next suitable group and run ext4_init_inode_table */
3631 static int ext4_run_li_request(struct ext4_li_request *elr)
3633 struct ext4_group_desc *gdp = NULL;
3634 struct super_block *sb = elr->lr_super;
3635 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3636 ext4_group_t group = elr->lr_next_group;
3637 unsigned int prefetch_ios = 0;
3641 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3642 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3643 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3645 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3647 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3649 if (group >= elr->lr_next_group) {
3651 if (elr->lr_first_not_zeroed != ngroups &&
3652 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3653 elr->lr_next_group = elr->lr_first_not_zeroed;
3654 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3661 for (; group < ngroups; group++) {
3662 gdp = ext4_get_group_desc(sb, group, NULL);
3668 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3672 if (group >= ngroups)
3676 start_time = ktime_get_real_ns();
3677 ret = ext4_init_inode_table(sb, group,
3678 elr->lr_timeout ? 0 : 1);
3679 trace_ext4_lazy_itable_init(sb, group);
3680 if (elr->lr_timeout == 0) {
3681 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3682 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3684 elr->lr_next_sched = jiffies + elr->lr_timeout;
3685 elr->lr_next_group = group + 1;
3691 * Remove lr_request from the list_request and free the
3692 * request structure. Should be called with li_list_mtx held
3694 static void ext4_remove_li_request(struct ext4_li_request *elr)
3699 list_del(&elr->lr_request);
3700 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3704 static void ext4_unregister_li_request(struct super_block *sb)
3706 mutex_lock(&ext4_li_mtx);
3707 if (!ext4_li_info) {
3708 mutex_unlock(&ext4_li_mtx);
3712 mutex_lock(&ext4_li_info->li_list_mtx);
3713 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3714 mutex_unlock(&ext4_li_info->li_list_mtx);
3715 mutex_unlock(&ext4_li_mtx);
3718 static struct task_struct *ext4_lazyinit_task;
3721 * This is the function where ext4lazyinit thread lives. It walks
3722 * through the request list searching for next scheduled filesystem.
3723 * When such a fs is found, run the lazy initialization request
3724 * (ext4_rn_li_request) and keep track of the time spend in this
3725 * function. Based on that time we compute next schedule time of
3726 * the request. When walking through the list is complete, compute
3727 * next waking time and put itself into sleep.
3729 static int ext4_lazyinit_thread(void *arg)
3731 struct ext4_lazy_init *eli = arg;
3732 struct list_head *pos, *n;
3733 struct ext4_li_request *elr;
3734 unsigned long next_wakeup, cur;
3736 BUG_ON(NULL == eli);
3741 next_wakeup = MAX_JIFFY_OFFSET;
3743 mutex_lock(&eli->li_list_mtx);
3744 if (list_empty(&eli->li_request_list)) {
3745 mutex_unlock(&eli->li_list_mtx);
3748 list_for_each_safe(pos, n, &eli->li_request_list) {
3751 elr = list_entry(pos, struct ext4_li_request,
3754 if (time_before(jiffies, elr->lr_next_sched)) {
3755 if (time_before(elr->lr_next_sched, next_wakeup))
3756 next_wakeup = elr->lr_next_sched;
3759 if (down_read_trylock(&elr->lr_super->s_umount)) {
3760 if (sb_start_write_trylock(elr->lr_super)) {
3763 * We hold sb->s_umount, sb can not
3764 * be removed from the list, it is
3765 * now safe to drop li_list_mtx
3767 mutex_unlock(&eli->li_list_mtx);
3768 err = ext4_run_li_request(elr);
3769 sb_end_write(elr->lr_super);
3770 mutex_lock(&eli->li_list_mtx);
3773 up_read((&elr->lr_super->s_umount));
3775 /* error, remove the lazy_init job */
3777 ext4_remove_li_request(elr);
3781 elr->lr_next_sched = jiffies +
3782 prandom_u32_max(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3784 if (time_before(elr->lr_next_sched, next_wakeup))
3785 next_wakeup = elr->lr_next_sched;
3787 mutex_unlock(&eli->li_list_mtx);
3792 if ((time_after_eq(cur, next_wakeup)) ||
3793 (MAX_JIFFY_OFFSET == next_wakeup)) {
3798 schedule_timeout_interruptible(next_wakeup - cur);
3800 if (kthread_should_stop()) {
3801 ext4_clear_request_list();
3808 * It looks like the request list is empty, but we need
3809 * to check it under the li_list_mtx lock, to prevent any
3810 * additions into it, and of course we should lock ext4_li_mtx
3811 * to atomically free the list and ext4_li_info, because at
3812 * this point another ext4 filesystem could be registering
3815 mutex_lock(&ext4_li_mtx);
3816 mutex_lock(&eli->li_list_mtx);
3817 if (!list_empty(&eli->li_request_list)) {
3818 mutex_unlock(&eli->li_list_mtx);
3819 mutex_unlock(&ext4_li_mtx);
3822 mutex_unlock(&eli->li_list_mtx);
3823 kfree(ext4_li_info);
3824 ext4_li_info = NULL;
3825 mutex_unlock(&ext4_li_mtx);
3830 static void ext4_clear_request_list(void)
3832 struct list_head *pos, *n;
3833 struct ext4_li_request *elr;
3835 mutex_lock(&ext4_li_info->li_list_mtx);
3836 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3837 elr = list_entry(pos, struct ext4_li_request,
3839 ext4_remove_li_request(elr);
3841 mutex_unlock(&ext4_li_info->li_list_mtx);
3844 static int ext4_run_lazyinit_thread(void)
3846 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3847 ext4_li_info, "ext4lazyinit");
3848 if (IS_ERR(ext4_lazyinit_task)) {
3849 int err = PTR_ERR(ext4_lazyinit_task);
3850 ext4_clear_request_list();
3851 kfree(ext4_li_info);
3852 ext4_li_info = NULL;
3853 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3854 "initialization thread\n",
3858 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3863 * Check whether it make sense to run itable init. thread or not.
3864 * If there is at least one uninitialized inode table, return
3865 * corresponding group number, else the loop goes through all
3866 * groups and return total number of groups.
3868 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3870 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3871 struct ext4_group_desc *gdp = NULL;
3873 if (!ext4_has_group_desc_csum(sb))
3876 for (group = 0; group < ngroups; group++) {
3877 gdp = ext4_get_group_desc(sb, group, NULL);
3881 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3888 static int ext4_li_info_new(void)
3890 struct ext4_lazy_init *eli = NULL;
3892 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3896 INIT_LIST_HEAD(&eli->li_request_list);
3897 mutex_init(&eli->li_list_mtx);
3899 eli->li_state |= EXT4_LAZYINIT_QUIT;
3906 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3909 struct ext4_li_request *elr;
3911 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3916 elr->lr_first_not_zeroed = start;
3917 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3918 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3919 elr->lr_next_group = start;
3921 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3925 * Randomize first schedule time of the request to
3926 * spread the inode table initialization requests
3929 elr->lr_next_sched = jiffies + prandom_u32_max(
3930 EXT4_DEF_LI_MAX_START_DELAY * HZ);
3934 int ext4_register_li_request(struct super_block *sb,
3935 ext4_group_t first_not_zeroed)
3937 struct ext4_sb_info *sbi = EXT4_SB(sb);
3938 struct ext4_li_request *elr = NULL;
3939 ext4_group_t ngroups = sbi->s_groups_count;
3942 mutex_lock(&ext4_li_mtx);
3943 if (sbi->s_li_request != NULL) {
3945 * Reset timeout so it can be computed again, because
3946 * s_li_wait_mult might have changed.
3948 sbi->s_li_request->lr_timeout = 0;
3952 if (sb_rdonly(sb) ||
3953 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3954 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
3957 elr = ext4_li_request_new(sb, first_not_zeroed);
3963 if (NULL == ext4_li_info) {
3964 ret = ext4_li_info_new();
3969 mutex_lock(&ext4_li_info->li_list_mtx);
3970 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3971 mutex_unlock(&ext4_li_info->li_list_mtx);
3973 sbi->s_li_request = elr;
3975 * set elr to NULL here since it has been inserted to
3976 * the request_list and the removal and free of it is
3977 * handled by ext4_clear_request_list from now on.
3981 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3982 ret = ext4_run_lazyinit_thread();
3987 mutex_unlock(&ext4_li_mtx);
3994 * We do not need to lock anything since this is called on
3997 static void ext4_destroy_lazyinit_thread(void)
4000 * If thread exited earlier
4001 * there's nothing to be done.
4003 if (!ext4_li_info || !ext4_lazyinit_task)
4006 kthread_stop(ext4_lazyinit_task);
4009 static int set_journal_csum_feature_set(struct super_block *sb)
4012 int compat, incompat;
4013 struct ext4_sb_info *sbi = EXT4_SB(sb);
4015 if (ext4_has_metadata_csum(sb)) {
4016 /* journal checksum v3 */
4018 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4020 /* journal checksum v1 */
4021 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4025 jbd2_journal_clear_features(sbi->s_journal,
4026 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4027 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4028 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4029 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4030 ret = jbd2_journal_set_features(sbi->s_journal,
4032 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4034 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4035 ret = jbd2_journal_set_features(sbi->s_journal,
4038 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4039 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4041 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4042 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4049 * Note: calculating the overhead so we can be compatible with
4050 * historical BSD practice is quite difficult in the face of
4051 * clusters/bigalloc. This is because multiple metadata blocks from
4052 * different block group can end up in the same allocation cluster.
4053 * Calculating the exact overhead in the face of clustered allocation
4054 * requires either O(all block bitmaps) in memory or O(number of block
4055 * groups**2) in time. We will still calculate the superblock for
4056 * older file systems --- and if we come across with a bigalloc file
4057 * system with zero in s_overhead_clusters the estimate will be close to
4058 * correct especially for very large cluster sizes --- but for newer
4059 * file systems, it's better to calculate this figure once at mkfs
4060 * time, and store it in the superblock. If the superblock value is
4061 * present (even for non-bigalloc file systems), we will use it.
4063 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4066 struct ext4_sb_info *sbi = EXT4_SB(sb);
4067 struct ext4_group_desc *gdp;
4068 ext4_fsblk_t first_block, last_block, b;
4069 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4070 int s, j, count = 0;
4071 int has_super = ext4_bg_has_super(sb, grp);
4073 if (!ext4_has_feature_bigalloc(sb))
4074 return (has_super + ext4_bg_num_gdb(sb, grp) +
4075 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4076 sbi->s_itb_per_group + 2);
4078 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4079 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4080 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4081 for (i = 0; i < ngroups; i++) {
4082 gdp = ext4_get_group_desc(sb, i, NULL);
4083 b = ext4_block_bitmap(sb, gdp);
4084 if (b >= first_block && b <= last_block) {
4085 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4088 b = ext4_inode_bitmap(sb, gdp);
4089 if (b >= first_block && b <= last_block) {
4090 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4093 b = ext4_inode_table(sb, gdp);
4094 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4095 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4096 int c = EXT4_B2C(sbi, b - first_block);
4097 ext4_set_bit(c, buf);
4103 if (ext4_bg_has_super(sb, grp)) {
4104 ext4_set_bit(s++, buf);
4107 j = ext4_bg_num_gdb(sb, grp);
4108 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4109 ext4_error(sb, "Invalid number of block group "
4110 "descriptor blocks: %d", j);
4111 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4115 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4119 return EXT4_CLUSTERS_PER_GROUP(sb) -
4120 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4124 * Compute the overhead and stash it in sbi->s_overhead
4126 int ext4_calculate_overhead(struct super_block *sb)
4128 struct ext4_sb_info *sbi = EXT4_SB(sb);
4129 struct ext4_super_block *es = sbi->s_es;
4130 struct inode *j_inode;
4131 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4132 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4133 ext4_fsblk_t overhead = 0;
4134 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4140 * Compute the overhead (FS structures). This is constant
4141 * for a given filesystem unless the number of block groups
4142 * changes so we cache the previous value until it does.
4146 * All of the blocks before first_data_block are overhead
4148 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4151 * Add the overhead found in each block group
4153 for (i = 0; i < ngroups; i++) {
4156 blks = count_overhead(sb, i, buf);
4159 memset(buf, 0, PAGE_SIZE);
4164 * Add the internal journal blocks whether the journal has been
4167 if (sbi->s_journal && !sbi->s_journal_bdev)
4168 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4169 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4170 /* j_inum for internal journal is non-zero */
4171 j_inode = ext4_get_journal_inode(sb, j_inum);
4173 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4174 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4177 ext4_msg(sb, KERN_ERR, "can't get journal size");
4180 sbi->s_overhead = overhead;
4182 free_page((unsigned long) buf);
4186 static void ext4_set_resv_clusters(struct super_block *sb)
4188 ext4_fsblk_t resv_clusters;
4189 struct ext4_sb_info *sbi = EXT4_SB(sb);
4192 * There's no need to reserve anything when we aren't using extents.
4193 * The space estimates are exact, there are no unwritten extents,
4194 * hole punching doesn't need new metadata... This is needed especially
4195 * to keep ext2/3 backward compatibility.
4197 if (!ext4_has_feature_extents(sb))
4200 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4201 * This should cover the situations where we can not afford to run
4202 * out of space like for example punch hole, or converting
4203 * unwritten extents in delalloc path. In most cases such
4204 * allocation would require 1, or 2 blocks, higher numbers are
4207 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4208 sbi->s_cluster_bits);
4210 do_div(resv_clusters, 50);
4211 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4213 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4216 static const char *ext4_quota_mode(struct super_block *sb)
4219 if (!ext4_quota_capable(sb))
4222 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4223 return "journalled";
4231 static void ext4_setup_csum_trigger(struct super_block *sb,
4232 enum ext4_journal_trigger_type type,
4234 struct jbd2_buffer_trigger_type *type,
4235 struct buffer_head *bh,
4239 struct ext4_sb_info *sbi = EXT4_SB(sb);
4241 sbi->s_journal_triggers[type].sb = sb;
4242 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4245 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4250 kfree(sbi->s_blockgroup_lock);
4251 fs_put_dax(sbi->s_daxdev, NULL);
4255 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4257 struct ext4_sb_info *sbi;
4259 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4263 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4266 sbi->s_blockgroup_lock =
4267 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4269 if (!sbi->s_blockgroup_lock)
4272 sb->s_fs_info = sbi;
4276 fs_put_dax(sbi->s_daxdev, NULL);
4281 static void ext4_set_def_opts(struct super_block *sb,
4282 struct ext4_super_block *es)
4284 unsigned long def_mount_opts;
4286 /* Set defaults before we parse the mount options */
4287 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4288 set_opt(sb, INIT_INODE_TABLE);
4289 if (def_mount_opts & EXT4_DEFM_DEBUG)
4291 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4293 if (def_mount_opts & EXT4_DEFM_UID16)
4294 set_opt(sb, NO_UID32);
4295 /* xattr user namespace & acls are now defaulted on */
4296 set_opt(sb, XATTR_USER);
4297 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4298 set_opt(sb, POSIX_ACL);
4300 if (ext4_has_feature_fast_commit(sb))
4301 set_opt2(sb, JOURNAL_FAST_COMMIT);
4302 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4303 if (ext4_has_metadata_csum(sb))
4304 set_opt(sb, JOURNAL_CHECKSUM);
4306 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4307 set_opt(sb, JOURNAL_DATA);
4308 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4309 set_opt(sb, ORDERED_DATA);
4310 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4311 set_opt(sb, WRITEBACK_DATA);
4313 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4314 set_opt(sb, ERRORS_PANIC);
4315 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4316 set_opt(sb, ERRORS_CONT);
4318 set_opt(sb, ERRORS_RO);
4319 /* block_validity enabled by default; disable with noblock_validity */
4320 set_opt(sb, BLOCK_VALIDITY);
4321 if (def_mount_opts & EXT4_DEFM_DISCARD)
4322 set_opt(sb, DISCARD);
4324 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4325 set_opt(sb, BARRIER);
4328 * enable delayed allocation by default
4329 * Use -o nodelalloc to turn it off
4331 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4332 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4333 set_opt(sb, DELALLOC);
4335 if (sb->s_blocksize == PAGE_SIZE)
4336 set_opt(sb, DIOREAD_NOLOCK);
4339 static int ext4_handle_clustersize(struct super_block *sb)
4341 struct ext4_sb_info *sbi = EXT4_SB(sb);
4342 struct ext4_super_block *es = sbi->s_es;
4345 /* Handle clustersize */
4346 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4347 if (ext4_has_feature_bigalloc(sb)) {
4348 if (clustersize < sb->s_blocksize) {
4349 ext4_msg(sb, KERN_ERR,
4350 "cluster size (%d) smaller than "
4351 "block size (%lu)", clustersize, sb->s_blocksize);
4354 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4355 le32_to_cpu(es->s_log_block_size);
4356 sbi->s_clusters_per_group =
4357 le32_to_cpu(es->s_clusters_per_group);
4358 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4359 ext4_msg(sb, KERN_ERR,
4360 "#clusters per group too big: %lu",
4361 sbi->s_clusters_per_group);
4364 if (sbi->s_blocks_per_group !=
4365 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4366 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4367 "clusters per group (%lu) inconsistent",
4368 sbi->s_blocks_per_group,
4369 sbi->s_clusters_per_group);
4373 if (clustersize != sb->s_blocksize) {
4374 ext4_msg(sb, KERN_ERR,
4375 "fragment/cluster size (%d) != "
4376 "block size (%lu)", clustersize, sb->s_blocksize);
4379 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4380 ext4_msg(sb, KERN_ERR,
4381 "#blocks per group too big: %lu",
4382 sbi->s_blocks_per_group);
4385 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4386 sbi->s_cluster_bits = 0;
4388 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4390 /* Do we have standard group size of clustersize * 8 blocks ? */
4391 if (sbi->s_blocks_per_group == clustersize << 3)
4392 set_opt2(sb, STD_GROUP_SIZE);
4397 static void ext4_fast_commit_init(struct super_block *sb)
4399 struct ext4_sb_info *sbi = EXT4_SB(sb);
4401 /* Initialize fast commit stuff */
4402 atomic_set(&sbi->s_fc_subtid, 0);
4403 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4404 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4405 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4406 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4407 sbi->s_fc_bytes = 0;
4408 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4409 sbi->s_fc_ineligible_tid = 0;
4410 spin_lock_init(&sbi->s_fc_lock);
4411 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4412 sbi->s_fc_replay_state.fc_regions = NULL;
4413 sbi->s_fc_replay_state.fc_regions_size = 0;
4414 sbi->s_fc_replay_state.fc_regions_used = 0;
4415 sbi->s_fc_replay_state.fc_regions_valid = 0;
4416 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4417 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4418 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4421 static int ext4_inode_info_init(struct super_block *sb,
4422 struct ext4_super_block *es)
4424 struct ext4_sb_info *sbi = EXT4_SB(sb);
4426 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4427 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4428 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4430 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4431 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4432 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4433 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4437 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4438 (!is_power_of_2(sbi->s_inode_size)) ||
4439 (sbi->s_inode_size > sb->s_blocksize)) {
4440 ext4_msg(sb, KERN_ERR,
4441 "unsupported inode size: %d",
4443 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4447 * i_atime_extra is the last extra field available for
4448 * [acm]times in struct ext4_inode. Checking for that
4449 * field should suffice to ensure we have extra space
4452 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4453 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4454 sb->s_time_gran = 1;
4455 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4457 sb->s_time_gran = NSEC_PER_SEC;
4458 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4460 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4463 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4464 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4465 EXT4_GOOD_OLD_INODE_SIZE;
4466 if (ext4_has_feature_extra_isize(sb)) {
4467 unsigned v, max = (sbi->s_inode_size -
4468 EXT4_GOOD_OLD_INODE_SIZE);
4470 v = le16_to_cpu(es->s_want_extra_isize);
4472 ext4_msg(sb, KERN_ERR,
4473 "bad s_want_extra_isize: %d", v);
4476 if (sbi->s_want_extra_isize < v)
4477 sbi->s_want_extra_isize = v;
4479 v = le16_to_cpu(es->s_min_extra_isize);
4481 ext4_msg(sb, KERN_ERR,
4482 "bad s_min_extra_isize: %d", v);
4485 if (sbi->s_want_extra_isize < v)
4486 sbi->s_want_extra_isize = v;
4493 #if IS_ENABLED(CONFIG_UNICODE)
4494 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4496 const struct ext4_sb_encodings *encoding_info;
4497 struct unicode_map *encoding;
4498 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4500 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4503 encoding_info = ext4_sb_read_encoding(es);
4504 if (!encoding_info) {
4505 ext4_msg(sb, KERN_ERR,
4506 "Encoding requested by superblock is unknown");
4510 encoding = utf8_load(encoding_info->version);
4511 if (IS_ERR(encoding)) {
4512 ext4_msg(sb, KERN_ERR,
4513 "can't mount with superblock charset: %s-%u.%u.%u "
4514 "not supported by the kernel. flags: 0x%x.",
4515 encoding_info->name,
4516 unicode_major(encoding_info->version),
4517 unicode_minor(encoding_info->version),
4518 unicode_rev(encoding_info->version),
4522 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4523 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4524 unicode_major(encoding_info->version),
4525 unicode_minor(encoding_info->version),
4526 unicode_rev(encoding_info->version),
4529 sb->s_encoding = encoding;
4530 sb->s_encoding_flags = encoding_flags;
4535 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4541 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4543 struct ext4_sb_info *sbi = EXT4_SB(sb);
4545 /* Warn if metadata_csum and gdt_csum are both set. */
4546 if (ext4_has_feature_metadata_csum(sb) &&
4547 ext4_has_feature_gdt_csum(sb))
4548 ext4_warning(sb, "metadata_csum and uninit_bg are "
4549 "redundant flags; please run fsck.");
4551 /* Check for a known checksum algorithm */
4552 if (!ext4_verify_csum_type(sb, es)) {
4553 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4554 "unknown checksum algorithm.");
4557 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4558 ext4_orphan_file_block_trigger);
4560 /* Load the checksum driver */
4561 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4562 if (IS_ERR(sbi->s_chksum_driver)) {
4563 int ret = PTR_ERR(sbi->s_chksum_driver);
4564 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4565 sbi->s_chksum_driver = NULL;
4569 /* Check superblock checksum */
4570 if (!ext4_superblock_csum_verify(sb, es)) {
4571 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4572 "invalid superblock checksum. Run e2fsck?");
4576 /* Precompute checksum seed for all metadata */
4577 if (ext4_has_feature_csum_seed(sb))
4578 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4579 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4580 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4581 sizeof(es->s_uuid));
4585 static int ext4_check_feature_compatibility(struct super_block *sb,
4586 struct ext4_super_block *es,
4589 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4590 (ext4_has_compat_features(sb) ||
4591 ext4_has_ro_compat_features(sb) ||
4592 ext4_has_incompat_features(sb)))
4593 ext4_msg(sb, KERN_WARNING,
4594 "feature flags set on rev 0 fs, "
4595 "running e2fsck is recommended");
4597 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4598 set_opt2(sb, HURD_COMPAT);
4599 if (ext4_has_feature_64bit(sb)) {
4600 ext4_msg(sb, KERN_ERR,
4601 "The Hurd can't support 64-bit file systems");
4606 * ea_inode feature uses l_i_version field which is not
4607 * available in HURD_COMPAT mode.
4609 if (ext4_has_feature_ea_inode(sb)) {
4610 ext4_msg(sb, KERN_ERR,
4611 "ea_inode feature is not supported for Hurd");
4616 if (IS_EXT2_SB(sb)) {
4617 if (ext2_feature_set_ok(sb))
4618 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4619 "using the ext4 subsystem");
4622 * If we're probing be silent, if this looks like
4623 * it's actually an ext[34] filesystem.
4625 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4627 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4628 "to feature incompatibilities");
4633 if (IS_EXT3_SB(sb)) {
4634 if (ext3_feature_set_ok(sb))
4635 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4636 "using the ext4 subsystem");
4639 * If we're probing be silent, if this looks like
4640 * it's actually an ext4 filesystem.
4642 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4644 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4645 "to feature incompatibilities");
4651 * Check feature flags regardless of the revision level, since we
4652 * previously didn't change the revision level when setting the flags,
4653 * so there is a chance incompat flags are set on a rev 0 filesystem.
4655 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4661 static int ext4_geometry_check(struct super_block *sb,
4662 struct ext4_super_block *es)
4664 struct ext4_sb_info *sbi = EXT4_SB(sb);
4667 /* check blocks count against device size */
4668 blocks_count = sb_bdev_nr_blocks(sb);
4669 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4670 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4671 "exceeds size of device (%llu blocks)",
4672 ext4_blocks_count(es), blocks_count);
4677 * It makes no sense for the first data block to be beyond the end
4678 * of the filesystem.
4680 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4681 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4682 "block %u is beyond end of filesystem (%llu)",
4683 le32_to_cpu(es->s_first_data_block),
4684 ext4_blocks_count(es));
4687 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4688 (sbi->s_cluster_ratio == 1)) {
4689 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4690 "block is 0 with a 1k block and cluster size");
4694 blocks_count = (ext4_blocks_count(es) -
4695 le32_to_cpu(es->s_first_data_block) +
4696 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4697 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4698 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4699 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4700 "(block count %llu, first data block %u, "
4701 "blocks per group %lu)", blocks_count,
4702 ext4_blocks_count(es),
4703 le32_to_cpu(es->s_first_data_block),
4704 EXT4_BLOCKS_PER_GROUP(sb));
4707 sbi->s_groups_count = blocks_count;
4708 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4709 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4710 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4711 le32_to_cpu(es->s_inodes_count)) {
4712 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4713 le32_to_cpu(es->s_inodes_count),
4714 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4721 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
4723 struct buffer_head **group_desc;
4727 group_desc = rcu_dereference(sbi->s_group_desc);
4728 for (i = 0; i < sbi->s_gdb_count; i++)
4729 brelse(group_desc[i]);
4734 static int ext4_group_desc_init(struct super_block *sb,
4735 struct ext4_super_block *es,
4736 ext4_fsblk_t logical_sb_block,
4737 ext4_group_t *first_not_zeroed)
4739 struct ext4_sb_info *sbi = EXT4_SB(sb);
4740 unsigned int db_count;
4745 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4746 EXT4_DESC_PER_BLOCK(sb);
4747 if (ext4_has_feature_meta_bg(sb)) {
4748 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4749 ext4_msg(sb, KERN_WARNING,
4750 "first meta block group too large: %u "
4751 "(group descriptor block count %u)",
4752 le32_to_cpu(es->s_first_meta_bg), db_count);
4756 rcu_assign_pointer(sbi->s_group_desc,
4757 kvmalloc_array(db_count,
4758 sizeof(struct buffer_head *),
4760 if (sbi->s_group_desc == NULL) {
4761 ext4_msg(sb, KERN_ERR, "not enough memory");
4765 bgl_lock_init(sbi->s_blockgroup_lock);
4767 /* Pre-read the descriptors into the buffer cache */
4768 for (i = 0; i < db_count; i++) {
4769 block = descriptor_loc(sb, logical_sb_block, i);
4770 ext4_sb_breadahead_unmovable(sb, block);
4773 for (i = 0; i < db_count; i++) {
4774 struct buffer_head *bh;
4776 block = descriptor_loc(sb, logical_sb_block, i);
4777 bh = ext4_sb_bread_unmovable(sb, block);
4779 ext4_msg(sb, KERN_ERR,
4780 "can't read group descriptor %d", i);
4781 sbi->s_gdb_count = i;
4786 rcu_dereference(sbi->s_group_desc)[i] = bh;
4789 sbi->s_gdb_count = db_count;
4790 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4791 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4792 ret = -EFSCORRUPTED;
4797 ext4_group_desc_free(sbi);
4801 static int ext4_load_and_init_journal(struct super_block *sb,
4802 struct ext4_super_block *es,
4803 struct ext4_fs_context *ctx)
4805 struct ext4_sb_info *sbi = EXT4_SB(sb);
4808 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4812 if (ext4_has_feature_64bit(sb) &&
4813 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4814 JBD2_FEATURE_INCOMPAT_64BIT)) {
4815 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4819 if (!set_journal_csum_feature_set(sb)) {
4820 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4825 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4826 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4827 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4828 ext4_msg(sb, KERN_ERR,
4829 "Failed to set fast commit journal feature");
4833 /* We have now updated the journal if required, so we can
4834 * validate the data journaling mode. */
4835 switch (test_opt(sb, DATA_FLAGS)) {
4837 /* No mode set, assume a default based on the journal
4838 * capabilities: ORDERED_DATA if the journal can
4839 * cope, else JOURNAL_DATA
4841 if (jbd2_journal_check_available_features
4842 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4843 set_opt(sb, ORDERED_DATA);
4844 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4846 set_opt(sb, JOURNAL_DATA);
4847 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4851 case EXT4_MOUNT_ORDERED_DATA:
4852 case EXT4_MOUNT_WRITEBACK_DATA:
4853 if (!jbd2_journal_check_available_features
4854 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4855 ext4_msg(sb, KERN_ERR, "Journal does not support "
4856 "requested data journaling mode");
4864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4865 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4866 ext4_msg(sb, KERN_ERR, "can't mount with "
4867 "journal_async_commit in data=ordered mode");
4871 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4873 sbi->s_journal->j_submit_inode_data_buffers =
4874 ext4_journal_submit_inode_data_buffers;
4875 sbi->s_journal->j_finish_inode_data_buffers =
4876 ext4_journal_finish_inode_data_buffers;
4881 /* flush s_error_work before journal destroy. */
4882 flush_work(&sbi->s_error_work);
4883 jbd2_journal_destroy(sbi->s_journal);
4884 sbi->s_journal = NULL;
4888 static int ext4_journal_data_mode_check(struct super_block *sb)
4890 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4891 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4892 "data=journal disables delayed allocation, "
4893 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4894 /* can't mount with both data=journal and dioread_nolock. */
4895 clear_opt(sb, DIOREAD_NOLOCK);
4896 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4897 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4898 ext4_msg(sb, KERN_ERR, "can't mount with "
4899 "both data=journal and delalloc");
4902 if (test_opt(sb, DAX_ALWAYS)) {
4903 ext4_msg(sb, KERN_ERR, "can't mount with "
4904 "both data=journal and dax");
4907 if (ext4_has_feature_encrypt(sb)) {
4908 ext4_msg(sb, KERN_WARNING,
4909 "encrypted files will use data=ordered "
4910 "instead of data journaling mode");
4912 if (test_opt(sb, DELALLOC))
4913 clear_opt(sb, DELALLOC);
4915 sb->s_iflags |= SB_I_CGROUPWB;
4921 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
4924 struct ext4_sb_info *sbi = EXT4_SB(sb);
4925 struct ext4_super_block *es;
4926 ext4_fsblk_t logical_sb_block;
4927 unsigned long offset = 0;
4928 struct buffer_head *bh;
4932 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4934 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4939 * The ext4 superblock will not be buffer aligned for other than 1kB
4940 * block sizes. We need to calculate the offset from buffer start.
4942 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4943 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4944 offset = do_div(logical_sb_block, blocksize);
4946 logical_sb_block = sbi->s_sb_block;
4949 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4951 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4955 * Note: s_es must be initialized as soon as possible because
4956 * some ext4 macro-instructions depend on its value
4958 es = (struct ext4_super_block *) (bh->b_data + offset);
4960 sb->s_magic = le16_to_cpu(es->s_magic);
4961 if (sb->s_magic != EXT4_SUPER_MAGIC) {
4963 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4967 if (le32_to_cpu(es->s_log_block_size) >
4968 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4969 ext4_msg(sb, KERN_ERR,
4970 "Invalid log block size: %u",
4971 le32_to_cpu(es->s_log_block_size));
4974 if (le32_to_cpu(es->s_log_cluster_size) >
4975 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4976 ext4_msg(sb, KERN_ERR,
4977 "Invalid log cluster size: %u",
4978 le32_to_cpu(es->s_log_cluster_size));
4982 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4985 * If the default block size is not the same as the real block size,
4986 * we need to reload it.
4988 if (sb->s_blocksize == blocksize) {
4989 *lsb = logical_sb_block;
4995 * bh must be released before kill_bdev(), otherwise
4996 * it won't be freed and its page also. kill_bdev()
4997 * is called by sb_set_blocksize().
5000 /* Validate the filesystem blocksize */
5001 if (!sb_set_blocksize(sb, blocksize)) {
5002 ext4_msg(sb, KERN_ERR, "bad block size %d",
5008 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5009 offset = do_div(logical_sb_block, blocksize);
5010 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5012 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5017 es = (struct ext4_super_block *)(bh->b_data + offset);
5019 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5020 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5023 *lsb = logical_sb_block;
5031 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5033 struct ext4_super_block *es = NULL;
5034 struct ext4_sb_info *sbi = EXT4_SB(sb);
5035 struct flex_groups **flex_groups;
5037 ext4_fsblk_t logical_sb_block;
5041 int needs_recovery, has_huge_files;
5043 ext4_group_t first_not_zeroed;
5044 struct ext4_fs_context *ctx = fc->fs_private;
5045 int silent = fc->sb_flags & SB_SILENT;
5047 /* Set defaults for the variables that will be set during parsing */
5048 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5049 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5051 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5052 sbi->s_sectors_written_start =
5053 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5055 /* -EINVAL is default */
5057 err = ext4_load_super(sb, &logical_sb_block, silent);
5062 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5064 err = ext4_init_metadata_csum(sb, es);
5068 ext4_set_def_opts(sb, es);
5070 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5071 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5072 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5073 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5074 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5077 * set default s_li_wait_mult for lazyinit, for the case there is
5078 * no mount option specified.
5080 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5082 if (ext4_inode_info_init(sb, es))
5085 err = parse_apply_sb_mount_options(sb, ctx);
5089 sbi->s_def_mount_opt = sbi->s_mount_opt;
5091 err = ext4_check_opt_consistency(fc, sb);
5095 ext4_apply_options(fc, sb);
5097 if (ext4_encoding_init(sb, es))
5100 if (ext4_journal_data_mode_check(sb))
5103 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5104 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5106 /* i_version is always enabled now */
5107 sb->s_flags |= SB_I_VERSION;
5109 if (ext4_check_feature_compatibility(sb, es, silent))
5112 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
5113 ext4_msg(sb, KERN_ERR,
5114 "Number of reserved GDT blocks insanely large: %d",
5115 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
5119 if (sbi->s_daxdev) {
5120 if (sb->s_blocksize == PAGE_SIZE)
5121 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
5123 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
5126 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
5127 if (ext4_has_feature_inline_data(sb)) {
5128 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
5129 " that may contain inline data");
5132 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
5133 ext4_msg(sb, KERN_ERR,
5134 "DAX unsupported by block device.");
5139 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
5140 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
5141 es->s_encryption_level);
5145 has_huge_files = ext4_has_feature_huge_file(sb);
5146 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5148 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5150 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5151 if (ext4_has_feature_64bit(sb)) {
5152 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5153 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5154 !is_power_of_2(sbi->s_desc_size)) {
5155 ext4_msg(sb, KERN_ERR,
5156 "unsupported descriptor size %lu",
5161 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5163 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5164 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5166 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5167 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5169 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5172 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5173 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5174 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5175 sbi->s_inodes_per_group);
5178 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5179 sbi->s_inodes_per_block;
5180 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5181 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5182 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5183 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5185 for (i = 0; i < 4; i++)
5186 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5187 sbi->s_def_hash_version = es->s_def_hash_version;
5188 if (ext4_has_feature_dir_index(sb)) {
5189 i = le32_to_cpu(es->s_flags);
5190 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5191 sbi->s_hash_unsigned = 3;
5192 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5193 #ifdef __CHAR_UNSIGNED__
5196 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5197 sbi->s_hash_unsigned = 3;
5201 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5206 if (ext4_handle_clustersize(sb))
5210 * Test whether we have more sectors than will fit in sector_t,
5211 * and whether the max offset is addressable by the page cache.
5213 err = generic_check_addressable(sb->s_blocksize_bits,
5214 ext4_blocks_count(es));
5216 ext4_msg(sb, KERN_ERR, "filesystem"
5217 " too large to mount safely on this system");
5221 if (ext4_geometry_check(sb, es))
5224 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5228 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5229 spin_lock_init(&sbi->s_error_lock);
5230 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5232 /* Register extent status tree shrinker */
5233 if (ext4_es_register_shrinker(sbi))
5236 sbi->s_stripe = ext4_get_stripe_size(sbi);
5237 sbi->s_extent_max_zeroout_kb = 32;
5240 * set up enough so that it can read an inode
5242 sb->s_op = &ext4_sops;
5243 sb->s_export_op = &ext4_export_ops;
5244 sb->s_xattr = ext4_xattr_handlers;
5245 #ifdef CONFIG_FS_ENCRYPTION
5246 sb->s_cop = &ext4_cryptops;
5248 #ifdef CONFIG_FS_VERITY
5249 sb->s_vop = &ext4_verityops;
5252 sb->dq_op = &ext4_quota_operations;
5253 if (ext4_has_feature_quota(sb))
5254 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5256 sb->s_qcop = &ext4_qctl_operations;
5257 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5259 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5261 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5262 mutex_init(&sbi->s_orphan_lock);
5264 ext4_fast_commit_init(sb);
5268 needs_recovery = (es->s_last_orphan != 0 ||
5269 ext4_has_feature_orphan_present(sb) ||
5270 ext4_has_feature_journal_needs_recovery(sb));
5272 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5273 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5274 goto failed_mount3a;
5277 * The first inode we look at is the journal inode. Don't try
5278 * root first: it may be modified in the journal!
5280 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5281 err = ext4_load_and_init_journal(sb, es, ctx);
5283 goto failed_mount3a;
5284 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5285 ext4_has_feature_journal_needs_recovery(sb)) {
5286 ext4_msg(sb, KERN_ERR, "required journal recovery "
5287 "suppressed and not mounted read-only");
5288 goto failed_mount3a;
5290 /* Nojournal mode, all journal mount options are illegal */
5291 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5292 ext4_msg(sb, KERN_ERR, "can't mount with "
5293 "journal_async_commit, fs mounted w/o journal");
5294 goto failed_mount3a;
5297 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5298 ext4_msg(sb, KERN_ERR, "can't mount with "
5299 "journal_checksum, fs mounted w/o journal");
5300 goto failed_mount3a;
5302 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5303 ext4_msg(sb, KERN_ERR, "can't mount with "
5304 "commit=%lu, fs mounted w/o journal",
5305 sbi->s_commit_interval / HZ);
5306 goto failed_mount3a;
5308 if (EXT4_MOUNT_DATA_FLAGS &
5309 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5310 ext4_msg(sb, KERN_ERR, "can't mount with "
5311 "data=, fs mounted w/o journal");
5312 goto failed_mount3a;
5314 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5315 clear_opt(sb, JOURNAL_CHECKSUM);
5316 clear_opt(sb, DATA_FLAGS);
5317 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5318 sbi->s_journal = NULL;
5322 if (!test_opt(sb, NO_MBCACHE)) {
5323 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5324 if (!sbi->s_ea_block_cache) {
5325 ext4_msg(sb, KERN_ERR,
5326 "Failed to create ea_block_cache");
5327 goto failed_mount_wq;
5330 if (ext4_has_feature_ea_inode(sb)) {
5331 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5332 if (!sbi->s_ea_inode_cache) {
5333 ext4_msg(sb, KERN_ERR,
5334 "Failed to create ea_inode_cache");
5335 goto failed_mount_wq;
5340 if (ext4_has_feature_verity(sb) && sb->s_blocksize != PAGE_SIZE) {
5341 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5342 goto failed_mount_wq;
5346 * Get the # of file system overhead blocks from the
5347 * superblock if present.
5349 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5350 /* ignore the precalculated value if it is ridiculous */
5351 if (sbi->s_overhead > ext4_blocks_count(es))
5352 sbi->s_overhead = 0;
5354 * If the bigalloc feature is not enabled recalculating the
5355 * overhead doesn't take long, so we might as well just redo
5356 * it to make sure we are using the correct value.
5358 if (!ext4_has_feature_bigalloc(sb))
5359 sbi->s_overhead = 0;
5360 if (sbi->s_overhead == 0) {
5361 err = ext4_calculate_overhead(sb);
5363 goto failed_mount_wq;
5367 * The maximum number of concurrent works can be high and
5368 * concurrency isn't really necessary. Limit it to 1.
5370 EXT4_SB(sb)->rsv_conversion_wq =
5371 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5372 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5373 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5379 * The jbd2_journal_load will have done any necessary log recovery,
5380 * so we can safely mount the rest of the filesystem now.
5383 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5385 ext4_msg(sb, KERN_ERR, "get root inode failed");
5386 ret = PTR_ERR(root);
5390 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5391 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5396 sb->s_root = d_make_root(root);
5398 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5403 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5404 if (ret == -EROFS) {
5405 sb->s_flags |= SB_RDONLY;
5408 goto failed_mount4a;
5410 ext4_set_resv_clusters(sb);
5412 if (test_opt(sb, BLOCK_VALIDITY)) {
5413 err = ext4_setup_system_zone(sb);
5415 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5417 goto failed_mount4a;
5420 ext4_fc_replay_cleanup(sb);
5425 * Enable optimize_scan if number of groups is > threshold. This can be
5426 * turned off by passing "mb_optimize_scan=0". This can also be
5427 * turned on forcefully by passing "mb_optimize_scan=1".
5429 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5430 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5431 set_opt2(sb, MB_OPTIMIZE_SCAN);
5433 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5436 err = ext4_mb_init(sb);
5438 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5444 * We can only set up the journal commit callback once
5445 * mballoc is initialized
5448 sbi->s_journal->j_commit_callback =
5449 ext4_journal_commit_callback;
5451 block = ext4_count_free_clusters(sb);
5452 ext4_free_blocks_count_set(sbi->s_es,
5453 EXT4_C2B(sbi, block));
5454 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5457 unsigned long freei = ext4_count_free_inodes(sb);
5458 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5459 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5463 err = percpu_counter_init(&sbi->s_dirs_counter,
5464 ext4_count_dirs(sb), GFP_KERNEL);
5466 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5469 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5472 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5475 ext4_msg(sb, KERN_ERR, "insufficient memory");
5479 if (ext4_has_feature_flex_bg(sb))
5480 if (!ext4_fill_flex_info(sb)) {
5481 ext4_msg(sb, KERN_ERR,
5482 "unable to initialize "
5483 "flex_bg meta info!");
5488 err = ext4_register_li_request(sb, first_not_zeroed);
5492 err = ext4_register_sysfs(sb);
5496 err = ext4_init_orphan_info(sb);
5500 /* Enable quota usage during mount. */
5501 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5502 err = ext4_enable_quotas(sb);
5506 #endif /* CONFIG_QUOTA */
5509 * Save the original bdev mapping's wb_err value which could be
5510 * used to detect the metadata async write error.
5512 spin_lock_init(&sbi->s_bdev_wb_lock);
5513 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5514 &sbi->s_bdev_wb_err);
5515 sb->s_bdev->bd_super = sb;
5516 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5517 ext4_orphan_cleanup(sb, es);
5518 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5520 * Update the checksum after updating free space/inode counters and
5521 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5522 * checksum in the buffer cache until it is written out and
5523 * e2fsprogs programs trying to open a file system immediately
5524 * after it is mounted can fail.
5526 ext4_superblock_csum_set(sb);
5527 if (needs_recovery) {
5528 ext4_msg(sb, KERN_INFO, "recovery complete");
5529 err = ext4_mark_recovery_complete(sb, es);
5534 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5535 ext4_msg(sb, KERN_WARNING,
5536 "mounting with \"discard\" option, but the device does not support discard");
5538 if (es->s_error_count)
5539 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5541 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5542 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5543 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5544 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5545 atomic_set(&sbi->s_warning_count, 0);
5546 atomic_set(&sbi->s_msg_count, 0);
5551 ext4_release_orphan_info(sb);
5553 ext4_unregister_sysfs(sb);
5554 kobject_put(&sbi->s_kobj);
5556 ext4_unregister_li_request(sb);
5558 ext4_mb_release(sb);
5560 flex_groups = rcu_dereference(sbi->s_flex_groups);
5562 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5563 kvfree(flex_groups[i]);
5564 kvfree(flex_groups);
5567 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5568 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5569 percpu_counter_destroy(&sbi->s_dirs_counter);
5570 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5571 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5572 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5574 ext4_ext_release(sb);
5575 ext4_release_system_zone(sb);
5580 ext4_msg(sb, KERN_ERR, "mount failed");
5581 if (EXT4_SB(sb)->rsv_conversion_wq)
5582 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5584 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5585 sbi->s_ea_inode_cache = NULL;
5587 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5588 sbi->s_ea_block_cache = NULL;
5590 if (sbi->s_journal) {
5591 /* flush s_error_work before journal destroy. */
5592 flush_work(&sbi->s_error_work);
5593 jbd2_journal_destroy(sbi->s_journal);
5594 sbi->s_journal = NULL;
5597 ext4_es_unregister_shrinker(sbi);
5599 /* flush s_error_work before sbi destroy */
5600 flush_work(&sbi->s_error_work);
5601 del_timer_sync(&sbi->s_err_report);
5602 ext4_stop_mmpd(sbi);
5603 ext4_group_desc_free(sbi);
5605 if (sbi->s_chksum_driver)
5606 crypto_free_shash(sbi->s_chksum_driver);
5608 #if IS_ENABLED(CONFIG_UNICODE)
5609 utf8_unload(sb->s_encoding);
5613 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5614 kfree(get_qf_name(sb, sbi, i));
5616 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5617 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5619 ext4_blkdev_remove(sbi);
5621 sb->s_fs_info = NULL;
5622 return err ? err : ret;
5625 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5627 struct ext4_fs_context *ctx = fc->fs_private;
5628 struct ext4_sb_info *sbi;
5632 sbi = ext4_alloc_sbi(sb);
5636 fc->s_fs_info = sbi;
5638 /* Cleanup superblock name */
5639 strreplace(sb->s_id, '/', '!');
5641 sbi->s_sb_block = 1; /* Default super block location */
5642 if (ctx->spec & EXT4_SPEC_s_sb_block)
5643 sbi->s_sb_block = ctx->s_sb_block;
5645 ret = __ext4_fill_super(fc, sb);
5649 if (sbi->s_journal) {
5650 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5651 descr = " journalled data mode";
5652 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5653 descr = " ordered data mode";
5655 descr = " writeback data mode";
5657 descr = "out journal";
5659 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5660 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU with%s. "
5661 "Quota mode: %s.", &sb->s_uuid, descr,
5662 ext4_quota_mode(sb));
5664 /* Update the s_overhead_clusters if necessary */
5665 ext4_update_overhead(sb, false);
5670 fc->s_fs_info = NULL;
5674 static int ext4_get_tree(struct fs_context *fc)
5676 return get_tree_bdev(fc, ext4_fill_super);
5680 * Setup any per-fs journal parameters now. We'll do this both on
5681 * initial mount, once the journal has been initialised but before we've
5682 * done any recovery; and again on any subsequent remount.
5684 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5686 struct ext4_sb_info *sbi = EXT4_SB(sb);
5688 journal->j_commit_interval = sbi->s_commit_interval;
5689 journal->j_min_batch_time = sbi->s_min_batch_time;
5690 journal->j_max_batch_time = sbi->s_max_batch_time;
5691 ext4_fc_init(sb, journal);
5693 write_lock(&journal->j_state_lock);
5694 if (test_opt(sb, BARRIER))
5695 journal->j_flags |= JBD2_BARRIER;
5697 journal->j_flags &= ~JBD2_BARRIER;
5698 if (test_opt(sb, DATA_ERR_ABORT))
5699 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5701 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5702 write_unlock(&journal->j_state_lock);
5705 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5706 unsigned int journal_inum)
5708 struct inode *journal_inode;
5711 * Test for the existence of a valid inode on disk. Bad things
5712 * happen if we iget() an unused inode, as the subsequent iput()
5713 * will try to delete it.
5715 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5716 if (IS_ERR(journal_inode)) {
5717 ext4_msg(sb, KERN_ERR, "no journal found");
5720 if (!journal_inode->i_nlink) {
5721 make_bad_inode(journal_inode);
5722 iput(journal_inode);
5723 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5727 ext4_debug("Journal inode found at %p: %lld bytes\n",
5728 journal_inode, journal_inode->i_size);
5729 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5730 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5731 iput(journal_inode);
5734 return journal_inode;
5737 static journal_t *ext4_get_journal(struct super_block *sb,
5738 unsigned int journal_inum)
5740 struct inode *journal_inode;
5743 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5746 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5750 journal = jbd2_journal_init_inode(journal_inode);
5752 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5753 iput(journal_inode);
5756 journal->j_private = sb;
5757 ext4_init_journal_params(sb, journal);
5761 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5764 struct buffer_head *bh;
5768 int hblock, blocksize;
5769 ext4_fsblk_t sb_block;
5770 unsigned long offset;
5771 struct ext4_super_block *es;
5772 struct block_device *bdev;
5774 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5777 bdev = ext4_blkdev_get(j_dev, sb);
5781 blocksize = sb->s_blocksize;
5782 hblock = bdev_logical_block_size(bdev);
5783 if (blocksize < hblock) {
5784 ext4_msg(sb, KERN_ERR,
5785 "blocksize too small for journal device");
5789 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5790 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5791 set_blocksize(bdev, blocksize);
5792 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5793 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5794 "external journal");
5798 es = (struct ext4_super_block *) (bh->b_data + offset);
5799 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5800 !(le32_to_cpu(es->s_feature_incompat) &
5801 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5802 ext4_msg(sb, KERN_ERR, "external journal has "
5808 if ((le32_to_cpu(es->s_feature_ro_compat) &
5809 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5810 es->s_checksum != ext4_superblock_csum(sb, es)) {
5811 ext4_msg(sb, KERN_ERR, "external journal has "
5812 "corrupt superblock");
5817 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5818 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5823 len = ext4_blocks_count(es);
5824 start = sb_block + 1;
5825 brelse(bh); /* we're done with the superblock */
5827 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5828 start, len, blocksize);
5830 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5833 journal->j_private = sb;
5834 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5835 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5838 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5839 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5840 "user (unsupported) - %d",
5841 be32_to_cpu(journal->j_superblock->s_nr_users));
5844 EXT4_SB(sb)->s_journal_bdev = bdev;
5845 ext4_init_journal_params(sb, journal);
5849 jbd2_journal_destroy(journal);
5851 ext4_blkdev_put(bdev);
5855 static int ext4_load_journal(struct super_block *sb,
5856 struct ext4_super_block *es,
5857 unsigned long journal_devnum)
5860 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5863 int really_read_only;
5866 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5867 return -EFSCORRUPTED;
5869 if (journal_devnum &&
5870 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5871 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5872 "numbers have changed");
5873 journal_dev = new_decode_dev(journal_devnum);
5875 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5877 if (journal_inum && journal_dev) {
5878 ext4_msg(sb, KERN_ERR,
5879 "filesystem has both journal inode and journal device!");
5884 journal = ext4_get_journal(sb, journal_inum);
5888 journal = ext4_get_dev_journal(sb, journal_dev);
5893 journal_dev_ro = bdev_read_only(journal->j_dev);
5894 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5896 if (journal_dev_ro && !sb_rdonly(sb)) {
5897 ext4_msg(sb, KERN_ERR,
5898 "journal device read-only, try mounting with '-o ro'");
5904 * Are we loading a blank journal or performing recovery after a
5905 * crash? For recovery, we need to check in advance whether we
5906 * can get read-write access to the device.
5908 if (ext4_has_feature_journal_needs_recovery(sb)) {
5909 if (sb_rdonly(sb)) {
5910 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5911 "required on readonly filesystem");
5912 if (really_read_only) {
5913 ext4_msg(sb, KERN_ERR, "write access "
5914 "unavailable, cannot proceed "
5915 "(try mounting with noload)");
5919 ext4_msg(sb, KERN_INFO, "write access will "
5920 "be enabled during recovery");
5924 if (!(journal->j_flags & JBD2_BARRIER))
5925 ext4_msg(sb, KERN_INFO, "barriers disabled");
5927 if (!ext4_has_feature_journal_needs_recovery(sb))
5928 err = jbd2_journal_wipe(journal, !really_read_only);
5930 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5932 memcpy(save, ((char *) es) +
5933 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5934 err = jbd2_journal_load(journal);
5936 memcpy(((char *) es) + EXT4_S_ERR_START,
5937 save, EXT4_S_ERR_LEN);
5942 ext4_msg(sb, KERN_ERR, "error loading journal");
5946 EXT4_SB(sb)->s_journal = journal;
5947 err = ext4_clear_journal_err(sb, es);
5949 EXT4_SB(sb)->s_journal = NULL;
5950 jbd2_journal_destroy(journal);
5954 if (!really_read_only && journal_devnum &&
5955 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5956 es->s_journal_dev = cpu_to_le32(journal_devnum);
5958 /* Make sure we flush the recovery flag to disk. */
5959 ext4_commit_super(sb);
5965 jbd2_journal_destroy(journal);
5969 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5970 static void ext4_update_super(struct super_block *sb)
5972 struct ext4_sb_info *sbi = EXT4_SB(sb);
5973 struct ext4_super_block *es = sbi->s_es;
5974 struct buffer_head *sbh = sbi->s_sbh;
5978 * If the file system is mounted read-only, don't update the
5979 * superblock write time. This avoids updating the superblock
5980 * write time when we are mounting the root file system
5981 * read/only but we need to replay the journal; at that point,
5982 * for people who are east of GMT and who make their clock
5983 * tick in localtime for Windows bug-for-bug compatibility,
5984 * the clock is set in the future, and this will cause e2fsck
5985 * to complain and force a full file system check.
5987 if (!(sb->s_flags & SB_RDONLY))
5988 ext4_update_tstamp(es, s_wtime);
5989 es->s_kbytes_written =
5990 cpu_to_le64(sbi->s_kbytes_written +
5991 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5992 sbi->s_sectors_written_start) >> 1));
5993 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5994 ext4_free_blocks_count_set(es,
5995 EXT4_C2B(sbi, percpu_counter_sum_positive(
5996 &sbi->s_freeclusters_counter)));
5997 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5998 es->s_free_inodes_count =
5999 cpu_to_le32(percpu_counter_sum_positive(
6000 &sbi->s_freeinodes_counter));
6001 /* Copy error information to the on-disk superblock */
6002 spin_lock(&sbi->s_error_lock);
6003 if (sbi->s_add_error_count > 0) {
6004 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6005 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6006 __ext4_update_tstamp(&es->s_first_error_time,
6007 &es->s_first_error_time_hi,
6008 sbi->s_first_error_time);
6009 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6010 sizeof(es->s_first_error_func));
6011 es->s_first_error_line =
6012 cpu_to_le32(sbi->s_first_error_line);
6013 es->s_first_error_ino =
6014 cpu_to_le32(sbi->s_first_error_ino);
6015 es->s_first_error_block =
6016 cpu_to_le64(sbi->s_first_error_block);
6017 es->s_first_error_errcode =
6018 ext4_errno_to_code(sbi->s_first_error_code);
6020 __ext4_update_tstamp(&es->s_last_error_time,
6021 &es->s_last_error_time_hi,
6022 sbi->s_last_error_time);
6023 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6024 sizeof(es->s_last_error_func));
6025 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6026 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6027 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6028 es->s_last_error_errcode =
6029 ext4_errno_to_code(sbi->s_last_error_code);
6031 * Start the daily error reporting function if it hasn't been
6034 if (!es->s_error_count)
6035 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6036 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6037 sbi->s_add_error_count = 0;
6039 spin_unlock(&sbi->s_error_lock);
6041 ext4_superblock_csum_set(sb);
6045 static int ext4_commit_super(struct super_block *sb)
6047 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6051 if (block_device_ejected(sb))
6054 ext4_update_super(sb);
6057 /* Buffer got discarded which means block device got invalidated */
6058 if (!buffer_mapped(sbh)) {
6063 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6065 * Oh, dear. A previous attempt to write the
6066 * superblock failed. This could happen because the
6067 * USB device was yanked out. Or it could happen to
6068 * be a transient write error and maybe the block will
6069 * be remapped. Nothing we can do but to retry the
6070 * write and hope for the best.
6072 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6073 "superblock detected");
6074 clear_buffer_write_io_error(sbh);
6075 set_buffer_uptodate(sbh);
6078 /* Clear potential dirty bit if it was journalled update */
6079 clear_buffer_dirty(sbh);
6080 sbh->b_end_io = end_buffer_write_sync;
6081 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6082 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6083 wait_on_buffer(sbh);
6084 if (buffer_write_io_error(sbh)) {
6085 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6087 clear_buffer_write_io_error(sbh);
6088 set_buffer_uptodate(sbh);
6095 * Have we just finished recovery? If so, and if we are mounting (or
6096 * remounting) the filesystem readonly, then we will end up with a
6097 * consistent fs on disk. Record that fact.
6099 static int ext4_mark_recovery_complete(struct super_block *sb,
6100 struct ext4_super_block *es)
6103 journal_t *journal = EXT4_SB(sb)->s_journal;
6105 if (!ext4_has_feature_journal(sb)) {
6106 if (journal != NULL) {
6107 ext4_error(sb, "Journal got removed while the fs was "
6109 return -EFSCORRUPTED;
6113 jbd2_journal_lock_updates(journal);
6114 err = jbd2_journal_flush(journal, 0);
6118 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6119 ext4_has_feature_orphan_present(sb))) {
6120 if (!ext4_orphan_file_empty(sb)) {
6121 ext4_error(sb, "Orphan file not empty on read-only fs.");
6122 err = -EFSCORRUPTED;
6125 ext4_clear_feature_journal_needs_recovery(sb);
6126 ext4_clear_feature_orphan_present(sb);
6127 ext4_commit_super(sb);
6130 jbd2_journal_unlock_updates(journal);
6135 * If we are mounting (or read-write remounting) a filesystem whose journal
6136 * has recorded an error from a previous lifetime, move that error to the
6137 * main filesystem now.
6139 static int ext4_clear_journal_err(struct super_block *sb,
6140 struct ext4_super_block *es)
6146 if (!ext4_has_feature_journal(sb)) {
6147 ext4_error(sb, "Journal got removed while the fs was mounted!");
6148 return -EFSCORRUPTED;
6151 journal = EXT4_SB(sb)->s_journal;
6154 * Now check for any error status which may have been recorded in the
6155 * journal by a prior ext4_error() or ext4_abort()
6158 j_errno = jbd2_journal_errno(journal);
6162 errstr = ext4_decode_error(sb, j_errno, nbuf);
6163 ext4_warning(sb, "Filesystem error recorded "
6164 "from previous mount: %s", errstr);
6165 ext4_warning(sb, "Marking fs in need of filesystem check.");
6167 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6168 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6169 ext4_commit_super(sb);
6171 jbd2_journal_clear_err(journal);
6172 jbd2_journal_update_sb_errno(journal);
6178 * Force the running and committing transactions to commit,
6179 * and wait on the commit.
6181 int ext4_force_commit(struct super_block *sb)
6188 journal = EXT4_SB(sb)->s_journal;
6189 return ext4_journal_force_commit(journal);
6192 static int ext4_sync_fs(struct super_block *sb, int wait)
6196 bool needs_barrier = false;
6197 struct ext4_sb_info *sbi = EXT4_SB(sb);
6199 if (unlikely(ext4_forced_shutdown(sbi)))
6202 trace_ext4_sync_fs(sb, wait);
6203 flush_workqueue(sbi->rsv_conversion_wq);
6205 * Writeback quota in non-journalled quota case - journalled quota has
6208 dquot_writeback_dquots(sb, -1);
6210 * Data writeback is possible w/o journal transaction, so barrier must
6211 * being sent at the end of the function. But we can skip it if
6212 * transaction_commit will do it for us.
6214 if (sbi->s_journal) {
6215 target = jbd2_get_latest_transaction(sbi->s_journal);
6216 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6217 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6218 needs_barrier = true;
6220 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6222 ret = jbd2_log_wait_commit(sbi->s_journal,
6225 } else if (wait && test_opt(sb, BARRIER))
6226 needs_barrier = true;
6227 if (needs_barrier) {
6229 err = blkdev_issue_flush(sb->s_bdev);
6238 * LVM calls this function before a (read-only) snapshot is created. This
6239 * gives us a chance to flush the journal completely and mark the fs clean.
6241 * Note that only this function cannot bring a filesystem to be in a clean
6242 * state independently. It relies on upper layer to stop all data & metadata
6245 static int ext4_freeze(struct super_block *sb)
6253 journal = EXT4_SB(sb)->s_journal;
6256 /* Now we set up the journal barrier. */
6257 jbd2_journal_lock_updates(journal);
6260 * Don't clear the needs_recovery flag if we failed to
6261 * flush the journal.
6263 error = jbd2_journal_flush(journal, 0);
6267 /* Journal blocked and flushed, clear needs_recovery flag. */
6268 ext4_clear_feature_journal_needs_recovery(sb);
6269 if (ext4_orphan_file_empty(sb))
6270 ext4_clear_feature_orphan_present(sb);
6273 error = ext4_commit_super(sb);
6276 /* we rely on upper layer to stop further updates */
6277 jbd2_journal_unlock_updates(journal);
6282 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6283 * flag here, even though the filesystem is not technically dirty yet.
6285 static int ext4_unfreeze(struct super_block *sb)
6287 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6290 if (EXT4_SB(sb)->s_journal) {
6291 /* Reset the needs_recovery flag before the fs is unlocked. */
6292 ext4_set_feature_journal_needs_recovery(sb);
6293 if (ext4_has_feature_orphan_file(sb))
6294 ext4_set_feature_orphan_present(sb);
6297 ext4_commit_super(sb);
6302 * Structure to save mount options for ext4_remount's benefit
6304 struct ext4_mount_options {
6305 unsigned long s_mount_opt;
6306 unsigned long s_mount_opt2;
6309 unsigned long s_commit_interval;
6310 u32 s_min_batch_time, s_max_batch_time;
6313 char *s_qf_names[EXT4_MAXQUOTAS];
6317 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6319 struct ext4_fs_context *ctx = fc->fs_private;
6320 struct ext4_super_block *es;
6321 struct ext4_sb_info *sbi = EXT4_SB(sb);
6322 unsigned long old_sb_flags;
6323 struct ext4_mount_options old_opts;
6327 int enable_quota = 0;
6329 char *to_free[EXT4_MAXQUOTAS];
6333 /* Store the original options */
6334 old_sb_flags = sb->s_flags;
6335 old_opts.s_mount_opt = sbi->s_mount_opt;
6336 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6337 old_opts.s_resuid = sbi->s_resuid;
6338 old_opts.s_resgid = sbi->s_resgid;
6339 old_opts.s_commit_interval = sbi->s_commit_interval;
6340 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6341 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6343 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6344 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6345 if (sbi->s_qf_names[i]) {
6346 char *qf_name = get_qf_name(sb, sbi, i);
6348 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6349 if (!old_opts.s_qf_names[i]) {
6350 for (j = 0; j < i; j++)
6351 kfree(old_opts.s_qf_names[j]);
6355 old_opts.s_qf_names[i] = NULL;
6357 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6358 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6359 ctx->journal_ioprio =
6360 sbi->s_journal->j_task->io_context->ioprio;
6362 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6366 ext4_apply_options(fc, sb);
6368 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6369 test_opt(sb, JOURNAL_CHECKSUM)) {
6370 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6371 "during remount not supported; ignoring");
6372 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6375 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6376 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6377 ext4_msg(sb, KERN_ERR, "can't mount with "
6378 "both data=journal and delalloc");
6382 if (test_opt(sb, DIOREAD_NOLOCK)) {
6383 ext4_msg(sb, KERN_ERR, "can't mount with "
6384 "both data=journal and dioread_nolock");
6388 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6389 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6390 ext4_msg(sb, KERN_ERR, "can't mount with "
6391 "journal_async_commit in data=ordered mode");
6397 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6398 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6403 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6404 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6406 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6407 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6411 if (sbi->s_journal) {
6412 ext4_init_journal_params(sb, sbi->s_journal);
6413 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6416 /* Flush outstanding errors before changing fs state */
6417 flush_work(&sbi->s_error_work);
6419 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6420 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6425 if (fc->sb_flags & SB_RDONLY) {
6426 err = sync_filesystem(sb);
6429 err = dquot_suspend(sb, -1);
6434 * First of all, the unconditional stuff we have to do
6435 * to disable replay of the journal when we next remount
6437 sb->s_flags |= SB_RDONLY;
6440 * OK, test if we are remounting a valid rw partition
6441 * readonly, and if so set the rdonly flag and then
6442 * mark the partition as valid again.
6444 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6445 (sbi->s_mount_state & EXT4_VALID_FS))
6446 es->s_state = cpu_to_le16(sbi->s_mount_state);
6448 if (sbi->s_journal) {
6450 * We let remount-ro finish even if marking fs
6451 * as clean failed...
6453 ext4_mark_recovery_complete(sb, es);
6456 /* Make sure we can mount this feature set readwrite */
6457 if (ext4_has_feature_readonly(sb) ||
6458 !ext4_feature_set_ok(sb, 0)) {
6463 * Make sure the group descriptor checksums
6464 * are sane. If they aren't, refuse to remount r/w.
6466 for (g = 0; g < sbi->s_groups_count; g++) {
6467 struct ext4_group_desc *gdp =
6468 ext4_get_group_desc(sb, g, NULL);
6470 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6471 ext4_msg(sb, KERN_ERR,
6472 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6473 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6474 le16_to_cpu(gdp->bg_checksum));
6481 * If we have an unprocessed orphan list hanging
6482 * around from a previously readonly bdev mount,
6483 * require a full umount/remount for now.
6485 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6486 ext4_msg(sb, KERN_WARNING, "Couldn't "
6487 "remount RDWR because of unprocessed "
6488 "orphan inode list. Please "
6489 "umount/remount instead");
6495 * Mounting a RDONLY partition read-write, so reread
6496 * and store the current valid flag. (It may have
6497 * been changed by e2fsck since we originally mounted
6500 if (sbi->s_journal) {
6501 err = ext4_clear_journal_err(sb, es);
6505 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6508 err = ext4_setup_super(sb, es, 0);
6512 sb->s_flags &= ~SB_RDONLY;
6513 if (ext4_has_feature_mmp(sb))
6514 if (ext4_multi_mount_protect(sb,
6515 le64_to_cpu(es->s_mmp_block))) {
6526 * Reinitialize lazy itable initialization thread based on
6529 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6530 ext4_unregister_li_request(sb);
6532 ext4_group_t first_not_zeroed;
6533 first_not_zeroed = ext4_has_uninit_itable(sb);
6534 ext4_register_li_request(sb, first_not_zeroed);
6538 * Handle creation of system zone data early because it can fail.
6539 * Releasing of existing data is done when we are sure remount will
6542 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6543 err = ext4_setup_system_zone(sb);
6548 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6549 err = ext4_commit_super(sb);
6555 /* Release old quota file names */
6556 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6557 kfree(old_opts.s_qf_names[i]);
6559 if (sb_any_quota_suspended(sb))
6560 dquot_resume(sb, -1);
6561 else if (ext4_has_feature_quota(sb)) {
6562 err = ext4_enable_quotas(sb);
6568 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6569 ext4_release_system_zone(sb);
6571 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6572 ext4_stop_mmpd(sbi);
6577 sb->s_flags = old_sb_flags;
6578 sbi->s_mount_opt = old_opts.s_mount_opt;
6579 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6580 sbi->s_resuid = old_opts.s_resuid;
6581 sbi->s_resgid = old_opts.s_resgid;
6582 sbi->s_commit_interval = old_opts.s_commit_interval;
6583 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6584 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6585 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6586 ext4_release_system_zone(sb);
6588 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6589 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6590 to_free[i] = get_qf_name(sb, sbi, i);
6591 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6594 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6597 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6598 ext4_stop_mmpd(sbi);
6602 static int ext4_reconfigure(struct fs_context *fc)
6604 struct super_block *sb = fc->root->d_sb;
6607 fc->s_fs_info = EXT4_SB(sb);
6609 ret = ext4_check_opt_consistency(fc, sb);
6613 ret = __ext4_remount(fc, sb);
6617 ext4_msg(sb, KERN_INFO, "re-mounted %pU. Quota mode: %s.",
6618 &sb->s_uuid, ext4_quota_mode(sb));
6624 static int ext4_statfs_project(struct super_block *sb,
6625 kprojid_t projid, struct kstatfs *buf)
6628 struct dquot *dquot;
6632 qid = make_kqid_projid(projid);
6633 dquot = dqget(sb, qid);
6635 return PTR_ERR(dquot);
6636 spin_lock(&dquot->dq_dqb_lock);
6638 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6639 dquot->dq_dqb.dqb_bhardlimit);
6640 limit >>= sb->s_blocksize_bits;
6642 if (limit && buf->f_blocks > limit) {
6643 curblock = (dquot->dq_dqb.dqb_curspace +
6644 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6645 buf->f_blocks = limit;
6646 buf->f_bfree = buf->f_bavail =
6647 (buf->f_blocks > curblock) ?
6648 (buf->f_blocks - curblock) : 0;
6651 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6652 dquot->dq_dqb.dqb_ihardlimit);
6653 if (limit && buf->f_files > limit) {
6654 buf->f_files = limit;
6656 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6657 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6660 spin_unlock(&dquot->dq_dqb_lock);
6666 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6668 struct super_block *sb = dentry->d_sb;
6669 struct ext4_sb_info *sbi = EXT4_SB(sb);
6670 struct ext4_super_block *es = sbi->s_es;
6671 ext4_fsblk_t overhead = 0, resv_blocks;
6673 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6675 if (!test_opt(sb, MINIX_DF))
6676 overhead = sbi->s_overhead;
6678 buf->f_type = EXT4_SUPER_MAGIC;
6679 buf->f_bsize = sb->s_blocksize;
6680 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6681 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6682 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6683 /* prevent underflow in case that few free space is available */
6684 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6685 buf->f_bavail = buf->f_bfree -
6686 (ext4_r_blocks_count(es) + resv_blocks);
6687 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6689 buf->f_files = le32_to_cpu(es->s_inodes_count);
6690 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6691 buf->f_namelen = EXT4_NAME_LEN;
6692 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6695 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6696 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6697 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6706 * Helper functions so that transaction is started before we acquire dqio_sem
6707 * to keep correct lock ordering of transaction > dqio_sem
6709 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6711 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6714 static int ext4_write_dquot(struct dquot *dquot)
6718 struct inode *inode;
6720 inode = dquot_to_inode(dquot);
6721 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6722 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6724 return PTR_ERR(handle);
6725 ret = dquot_commit(dquot);
6726 err = ext4_journal_stop(handle);
6732 static int ext4_acquire_dquot(struct dquot *dquot)
6737 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6738 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6740 return PTR_ERR(handle);
6741 ret = dquot_acquire(dquot);
6742 err = ext4_journal_stop(handle);
6748 static int ext4_release_dquot(struct dquot *dquot)
6753 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6754 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6755 if (IS_ERR(handle)) {
6756 /* Release dquot anyway to avoid endless cycle in dqput() */
6757 dquot_release(dquot);
6758 return PTR_ERR(handle);
6760 ret = dquot_release(dquot);
6761 err = ext4_journal_stop(handle);
6767 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6769 struct super_block *sb = dquot->dq_sb;
6771 if (ext4_is_quota_journalled(sb)) {
6772 dquot_mark_dquot_dirty(dquot);
6773 return ext4_write_dquot(dquot);
6775 return dquot_mark_dquot_dirty(dquot);
6779 static int ext4_write_info(struct super_block *sb, int type)
6784 /* Data block + inode block */
6785 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6787 return PTR_ERR(handle);
6788 ret = dquot_commit_info(sb, type);
6789 err = ext4_journal_stop(handle);
6795 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6797 struct ext4_inode_info *ei = EXT4_I(inode);
6799 /* The first argument of lockdep_set_subclass has to be
6800 * *exactly* the same as the argument to init_rwsem() --- in
6801 * this case, in init_once() --- or lockdep gets unhappy
6802 * because the name of the lock is set using the
6803 * stringification of the argument to init_rwsem().
6805 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6806 lockdep_set_subclass(&ei->i_data_sem, subclass);
6810 * Standard function to be called on quota_on
6812 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6813 const struct path *path)
6817 if (!test_opt(sb, QUOTA))
6820 /* Quotafile not on the same filesystem? */
6821 if (path->dentry->d_sb != sb)
6824 /* Quota already enabled for this file? */
6825 if (IS_NOQUOTA(d_inode(path->dentry)))
6828 /* Journaling quota? */
6829 if (EXT4_SB(sb)->s_qf_names[type]) {
6830 /* Quotafile not in fs root? */
6831 if (path->dentry->d_parent != sb->s_root)
6832 ext4_msg(sb, KERN_WARNING,
6833 "Quota file not on filesystem root. "
6834 "Journaled quota will not work");
6835 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6838 * Clear the flag just in case mount options changed since
6841 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6845 * When we journal data on quota file, we have to flush journal to see
6846 * all updates to the file when we bypass pagecache...
6848 if (EXT4_SB(sb)->s_journal &&
6849 ext4_should_journal_data(d_inode(path->dentry))) {
6851 * We don't need to lock updates but journal_flush() could
6852 * otherwise be livelocked...
6854 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6855 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6856 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6861 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6862 err = dquot_quota_on(sb, type, format_id, path);
6864 struct inode *inode = d_inode(path->dentry);
6868 * Set inode flags to prevent userspace from messing with quota
6869 * files. If this fails, we return success anyway since quotas
6870 * are already enabled and this is not a hard failure.
6873 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6876 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6877 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6878 S_NOATIME | S_IMMUTABLE);
6879 err = ext4_mark_inode_dirty(handle, inode);
6880 ext4_journal_stop(handle);
6882 inode_unlock(inode);
6884 dquot_quota_off(sb, type);
6887 lockdep_set_quota_inode(path->dentry->d_inode,
6892 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6896 return qf_inum == EXT4_USR_QUOTA_INO;
6898 return qf_inum == EXT4_GRP_QUOTA_INO;
6900 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6906 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6910 struct inode *qf_inode;
6911 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6912 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6913 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6914 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6917 BUG_ON(!ext4_has_feature_quota(sb));
6919 if (!qf_inums[type])
6922 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6923 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6924 qf_inums[type], type);
6928 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6929 if (IS_ERR(qf_inode)) {
6930 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6931 qf_inums[type], type);
6932 return PTR_ERR(qf_inode);
6935 /* Don't account quota for quota files to avoid recursion */
6936 qf_inode->i_flags |= S_NOQUOTA;
6937 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6938 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6940 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6946 /* Enable usage tracking for all quota types. */
6947 int ext4_enable_quotas(struct super_block *sb)
6950 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6951 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6952 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6953 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6955 bool quota_mopt[EXT4_MAXQUOTAS] = {
6956 test_opt(sb, USRQUOTA),
6957 test_opt(sb, GRPQUOTA),
6958 test_opt(sb, PRJQUOTA),
6961 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6962 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6963 if (qf_inums[type]) {
6964 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6965 DQUOT_USAGE_ENABLED |
6966 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6969 "Failed to enable quota tracking "
6970 "(type=%d, err=%d, ino=%lu). "
6971 "Please run e2fsck to fix.", type,
6972 err, qf_inums[type]);
6973 for (type--; type >= 0; type--) {
6974 struct inode *inode;
6976 inode = sb_dqopt(sb)->files[type];
6978 inode = igrab(inode);
6979 dquot_quota_off(sb, type);
6981 lockdep_set_quota_inode(inode,
6994 static int ext4_quota_off(struct super_block *sb, int type)
6996 struct inode *inode = sb_dqopt(sb)->files[type];
7000 /* Force all delayed allocation blocks to be allocated.
7001 * Caller already holds s_umount sem */
7002 if (test_opt(sb, DELALLOC))
7003 sync_filesystem(sb);
7005 if (!inode || !igrab(inode))
7008 err = dquot_quota_off(sb, type);
7009 if (err || ext4_has_feature_quota(sb))
7014 * Update modification times of quota files when userspace can
7015 * start looking at them. If we fail, we return success anyway since
7016 * this is not a hard failure and quotas are already disabled.
7018 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7019 if (IS_ERR(handle)) {
7020 err = PTR_ERR(handle);
7023 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7024 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7025 inode->i_mtime = inode->i_ctime = current_time(inode);
7026 err = ext4_mark_inode_dirty(handle, inode);
7027 ext4_journal_stop(handle);
7029 inode_unlock(inode);
7031 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7035 return dquot_quota_off(sb, type);
7038 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7039 * acquiring the locks... As quota files are never truncated and quota code
7040 * itself serializes the operations (and no one else should touch the files)
7041 * we don't have to be afraid of races */
7042 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7043 size_t len, loff_t off)
7045 struct inode *inode = sb_dqopt(sb)->files[type];
7046 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7047 int offset = off & (sb->s_blocksize - 1);
7050 struct buffer_head *bh;
7051 loff_t i_size = i_size_read(inode);
7055 if (off+len > i_size)
7058 while (toread > 0) {
7059 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7060 bh = ext4_bread(NULL, inode, blk, 0);
7063 if (!bh) /* A hole? */
7064 memset(data, 0, tocopy);
7066 memcpy(data, bh->b_data+offset, tocopy);
7076 /* Write to quotafile (we know the transaction is already started and has
7077 * enough credits) */
7078 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7079 const char *data, size_t len, loff_t off)
7081 struct inode *inode = sb_dqopt(sb)->files[type];
7082 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7083 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7085 struct buffer_head *bh;
7086 handle_t *handle = journal_current_handle();
7089 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7090 " cancelled because transaction is not started",
7091 (unsigned long long)off, (unsigned long long)len);
7095 * Since we account only one data block in transaction credits,
7096 * then it is impossible to cross a block boundary.
7098 if (sb->s_blocksize - offset < len) {
7099 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7100 " cancelled because not block aligned",
7101 (unsigned long long)off, (unsigned long long)len);
7106 bh = ext4_bread(handle, inode, blk,
7107 EXT4_GET_BLOCKS_CREATE |
7108 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7109 } while (PTR_ERR(bh) == -ENOSPC &&
7110 ext4_should_retry_alloc(inode->i_sb, &retries));
7115 BUFFER_TRACE(bh, "get write access");
7116 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7122 memcpy(bh->b_data+offset, data, len);
7123 flush_dcache_page(bh->b_page);
7125 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7128 if (inode->i_size < off + len) {
7129 i_size_write(inode, off + len);
7130 EXT4_I(inode)->i_disksize = inode->i_size;
7131 err2 = ext4_mark_inode_dirty(handle, inode);
7132 if (unlikely(err2 && !err))
7135 return err ? err : len;
7139 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7140 static inline void register_as_ext2(void)
7142 int err = register_filesystem(&ext2_fs_type);
7145 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7148 static inline void unregister_as_ext2(void)
7150 unregister_filesystem(&ext2_fs_type);
7153 static inline int ext2_feature_set_ok(struct super_block *sb)
7155 if (ext4_has_unknown_ext2_incompat_features(sb))
7159 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7164 static inline void register_as_ext2(void) { }
7165 static inline void unregister_as_ext2(void) { }
7166 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7169 static inline void register_as_ext3(void)
7171 int err = register_filesystem(&ext3_fs_type);
7174 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7177 static inline void unregister_as_ext3(void)
7179 unregister_filesystem(&ext3_fs_type);
7182 static inline int ext3_feature_set_ok(struct super_block *sb)
7184 if (ext4_has_unknown_ext3_incompat_features(sb))
7186 if (!ext4_has_feature_journal(sb))
7190 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7195 static struct file_system_type ext4_fs_type = {
7196 .owner = THIS_MODULE,
7198 .init_fs_context = ext4_init_fs_context,
7199 .parameters = ext4_param_specs,
7200 .kill_sb = kill_block_super,
7201 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7203 MODULE_ALIAS_FS("ext4");
7205 /* Shared across all ext4 file systems */
7206 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7208 static int __init ext4_init_fs(void)
7212 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7213 ext4_li_info = NULL;
7215 /* Build-time check for flags consistency */
7216 ext4_check_flag_values();
7218 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7219 init_waitqueue_head(&ext4__ioend_wq[i]);
7221 err = ext4_init_es();
7225 err = ext4_init_pending();
7229 err = ext4_init_post_read_processing();
7233 err = ext4_init_pageio();
7237 err = ext4_init_system_zone();
7241 err = ext4_init_sysfs();
7245 err = ext4_init_mballoc();
7248 err = init_inodecache();
7252 err = ext4_fc_init_dentry_cache();
7258 err = register_filesystem(&ext4_fs_type);
7264 unregister_as_ext2();
7265 unregister_as_ext3();
7266 ext4_fc_destroy_dentry_cache();
7268 destroy_inodecache();
7270 ext4_exit_mballoc();
7274 ext4_exit_system_zone();
7278 ext4_exit_post_read_processing();
7280 ext4_exit_pending();
7287 static void __exit ext4_exit_fs(void)
7289 ext4_destroy_lazyinit_thread();
7290 unregister_as_ext2();
7291 unregister_as_ext3();
7292 unregister_filesystem(&ext4_fs_type);
7293 ext4_fc_destroy_dentry_cache();
7294 destroy_inodecache();
7295 ext4_exit_mballoc();
7297 ext4_exit_system_zone();
7299 ext4_exit_post_read_processing();
7301 ext4_exit_pending();
7304 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7305 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7306 MODULE_LICENSE("GPL");
7307 MODULE_SOFTDEP("pre: crc32c");
7308 module_init(ext4_init_fs)
7309 module_exit(ext4_exit_fs)