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 = ext4_normal_submit_inode_data_buffers(jinode);
547 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
551 if (!ext4_should_journal_data(jinode->i_vfs_inode))
552 ret = jbd2_journal_finish_inode_data_buffers(jinode);
557 static bool system_going_down(void)
559 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
560 || system_state == SYSTEM_RESTART;
563 struct ext4_err_translation {
568 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
570 static struct ext4_err_translation err_translation[] = {
571 EXT4_ERR_TRANSLATE(EIO),
572 EXT4_ERR_TRANSLATE(ENOMEM),
573 EXT4_ERR_TRANSLATE(EFSBADCRC),
574 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
575 EXT4_ERR_TRANSLATE(ENOSPC),
576 EXT4_ERR_TRANSLATE(ENOKEY),
577 EXT4_ERR_TRANSLATE(EROFS),
578 EXT4_ERR_TRANSLATE(EFBIG),
579 EXT4_ERR_TRANSLATE(EEXIST),
580 EXT4_ERR_TRANSLATE(ERANGE),
581 EXT4_ERR_TRANSLATE(EOVERFLOW),
582 EXT4_ERR_TRANSLATE(EBUSY),
583 EXT4_ERR_TRANSLATE(ENOTDIR),
584 EXT4_ERR_TRANSLATE(ENOTEMPTY),
585 EXT4_ERR_TRANSLATE(ESHUTDOWN),
586 EXT4_ERR_TRANSLATE(EFAULT),
589 static int ext4_errno_to_code(int errno)
593 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
594 if (err_translation[i].errno == errno)
595 return err_translation[i].code;
596 return EXT4_ERR_UNKNOWN;
599 static void save_error_info(struct super_block *sb, int error,
600 __u32 ino, __u64 block,
601 const char *func, unsigned int line)
603 struct ext4_sb_info *sbi = EXT4_SB(sb);
605 /* We default to EFSCORRUPTED error... */
607 error = EFSCORRUPTED;
609 spin_lock(&sbi->s_error_lock);
610 sbi->s_add_error_count++;
611 sbi->s_last_error_code = error;
612 sbi->s_last_error_line = line;
613 sbi->s_last_error_ino = ino;
614 sbi->s_last_error_block = block;
615 sbi->s_last_error_func = func;
616 sbi->s_last_error_time = ktime_get_real_seconds();
617 if (!sbi->s_first_error_time) {
618 sbi->s_first_error_code = error;
619 sbi->s_first_error_line = line;
620 sbi->s_first_error_ino = ino;
621 sbi->s_first_error_block = block;
622 sbi->s_first_error_func = func;
623 sbi->s_first_error_time = sbi->s_last_error_time;
625 spin_unlock(&sbi->s_error_lock);
628 /* Deal with the reporting of failure conditions on a filesystem such as
629 * inconsistencies detected or read IO failures.
631 * On ext2, we can store the error state of the filesystem in the
632 * superblock. That is not possible on ext4, because we may have other
633 * write ordering constraints on the superblock which prevent us from
634 * writing it out straight away; and given that the journal is about to
635 * be aborted, we can't rely on the current, or future, transactions to
636 * write out the superblock safely.
638 * We'll just use the jbd2_journal_abort() error code to record an error in
639 * the journal instead. On recovery, the journal will complain about
640 * that error until we've noted it down and cleared it.
642 * If force_ro is set, we unconditionally force the filesystem into an
643 * ABORT|READONLY state, unless the error response on the fs has been set to
644 * panic in which case we take the easy way out and panic immediately. This is
645 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
646 * at a critical moment in log management.
648 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
649 __u32 ino, __u64 block,
650 const char *func, unsigned int line)
652 journal_t *journal = EXT4_SB(sb)->s_journal;
653 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
655 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
656 if (test_opt(sb, WARN_ON_ERROR))
659 if (!continue_fs && !sb_rdonly(sb)) {
660 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
662 jbd2_journal_abort(journal, -EIO);
665 if (!bdev_read_only(sb->s_bdev)) {
666 save_error_info(sb, error, ino, block, func, line);
668 * In case the fs should keep running, we need to writeout
669 * superblock through the journal. Due to lock ordering
670 * constraints, it may not be safe to do it right here so we
671 * defer superblock flushing to a workqueue.
673 if (continue_fs && journal)
674 schedule_work(&EXT4_SB(sb)->s_error_work);
676 ext4_commit_super(sb);
680 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
681 * could panic during 'reboot -f' as the underlying device got already
684 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
685 panic("EXT4-fs (device %s): panic forced after error\n",
689 if (sb_rdonly(sb) || continue_fs)
692 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
694 * Make sure updated value of ->s_mount_flags will be visible before
698 sb->s_flags |= SB_RDONLY;
701 static void flush_stashed_error_work(struct work_struct *work)
703 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
705 journal_t *journal = sbi->s_journal;
709 * If the journal is still running, we have to write out superblock
710 * through the journal to avoid collisions of other journalled sb
713 * We use directly jbd2 functions here to avoid recursing back into
714 * ext4 error handling code during handling of previous errors.
716 if (!sb_rdonly(sbi->s_sb) && journal) {
717 struct buffer_head *sbh = sbi->s_sbh;
718 handle = jbd2_journal_start(journal, 1);
721 if (jbd2_journal_get_write_access(handle, sbh)) {
722 jbd2_journal_stop(handle);
725 ext4_update_super(sbi->s_sb);
726 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
727 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
728 "superblock detected");
729 clear_buffer_write_io_error(sbh);
730 set_buffer_uptodate(sbh);
733 if (jbd2_journal_dirty_metadata(handle, sbh)) {
734 jbd2_journal_stop(handle);
737 jbd2_journal_stop(handle);
738 ext4_notify_error_sysfs(sbi);
743 * Write through journal failed. Write sb directly to get error info
744 * out and hope for the best.
746 ext4_commit_super(sbi->s_sb);
747 ext4_notify_error_sysfs(sbi);
750 #define ext4_error_ratelimit(sb) \
751 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
754 void __ext4_error(struct super_block *sb, const char *function,
755 unsigned int line, bool force_ro, int error, __u64 block,
756 const char *fmt, ...)
758 struct va_format vaf;
761 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
764 trace_ext4_error(sb, function, line);
765 if (ext4_error_ratelimit(sb)) {
770 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
771 sb->s_id, function, line, current->comm, &vaf);
774 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
776 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
779 void __ext4_error_inode(struct inode *inode, const char *function,
780 unsigned int line, ext4_fsblk_t block, int error,
781 const char *fmt, ...)
784 struct va_format vaf;
786 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
789 trace_ext4_error(inode->i_sb, function, line);
790 if (ext4_error_ratelimit(inode->i_sb)) {
795 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
796 "inode #%lu: block %llu: comm %s: %pV\n",
797 inode->i_sb->s_id, function, line, inode->i_ino,
798 block, current->comm, &vaf);
800 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
801 "inode #%lu: comm %s: %pV\n",
802 inode->i_sb->s_id, function, line, inode->i_ino,
803 current->comm, &vaf);
806 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
808 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
812 void __ext4_error_file(struct file *file, const char *function,
813 unsigned int line, ext4_fsblk_t block,
814 const char *fmt, ...)
817 struct va_format vaf;
818 struct inode *inode = file_inode(file);
819 char pathname[80], *path;
821 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
824 trace_ext4_error(inode->i_sb, function, line);
825 if (ext4_error_ratelimit(inode->i_sb)) {
826 path = file_path(file, pathname, sizeof(pathname));
834 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
835 "block %llu: comm %s: path %s: %pV\n",
836 inode->i_sb->s_id, function, line, inode->i_ino,
837 block, current->comm, path, &vaf);
840 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
841 "comm %s: path %s: %pV\n",
842 inode->i_sb->s_id, function, line, inode->i_ino,
843 current->comm, path, &vaf);
846 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
848 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
852 const char *ext4_decode_error(struct super_block *sb, int errno,
859 errstr = "Corrupt filesystem";
862 errstr = "Filesystem failed CRC";
865 errstr = "IO failure";
868 errstr = "Out of memory";
871 if (!sb || (EXT4_SB(sb)->s_journal &&
872 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
873 errstr = "Journal has aborted";
875 errstr = "Readonly filesystem";
878 /* If the caller passed in an extra buffer for unknown
879 * errors, textualise them now. Else we just return
882 /* Check for truncated error codes... */
883 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
892 /* __ext4_std_error decodes expected errors from journaling functions
893 * automatically and invokes the appropriate error response. */
895 void __ext4_std_error(struct super_block *sb, const char *function,
896 unsigned int line, int errno)
901 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
904 /* Special case: if the error is EROFS, and we're not already
905 * inside a transaction, then there's really no point in logging
907 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
910 if (ext4_error_ratelimit(sb)) {
911 errstr = ext4_decode_error(sb, errno, nbuf);
912 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
913 sb->s_id, function, line, errstr);
915 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
917 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
920 void __ext4_msg(struct super_block *sb,
921 const char *prefix, const char *fmt, ...)
923 struct va_format vaf;
927 atomic_inc(&EXT4_SB(sb)->s_msg_count);
928 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
937 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
939 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
943 static int ext4_warning_ratelimit(struct super_block *sb)
945 atomic_inc(&EXT4_SB(sb)->s_warning_count);
946 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
950 void __ext4_warning(struct super_block *sb, const char *function,
951 unsigned int line, const char *fmt, ...)
953 struct va_format vaf;
956 if (!ext4_warning_ratelimit(sb))
962 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
963 sb->s_id, function, line, &vaf);
967 void __ext4_warning_inode(const struct inode *inode, const char *function,
968 unsigned int line, const char *fmt, ...)
970 struct va_format vaf;
973 if (!ext4_warning_ratelimit(inode->i_sb))
979 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
980 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
981 function, line, inode->i_ino, current->comm, &vaf);
985 void __ext4_grp_locked_error(const char *function, unsigned int line,
986 struct super_block *sb, ext4_group_t grp,
987 unsigned long ino, ext4_fsblk_t block,
988 const char *fmt, ...)
992 struct va_format vaf;
995 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
998 trace_ext4_error(sb, function, line);
999 if (ext4_error_ratelimit(sb)) {
1000 va_start(args, fmt);
1003 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1004 sb->s_id, function, line, grp);
1006 printk(KERN_CONT "inode %lu: ", ino);
1008 printk(KERN_CONT "block %llu:",
1009 (unsigned long long) block);
1010 printk(KERN_CONT "%pV\n", &vaf);
1014 if (test_opt(sb, ERRORS_CONT)) {
1015 if (test_opt(sb, WARN_ON_ERROR))
1017 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1018 if (!bdev_read_only(sb->s_bdev)) {
1019 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1021 schedule_work(&EXT4_SB(sb)->s_error_work);
1025 ext4_unlock_group(sb, grp);
1026 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1028 * We only get here in the ERRORS_RO case; relocking the group
1029 * may be dangerous, but nothing bad will happen since the
1030 * filesystem will have already been marked read/only and the
1031 * journal has been aborted. We return 1 as a hint to callers
1032 * who might what to use the return value from
1033 * ext4_grp_locked_error() to distinguish between the
1034 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1035 * aggressively from the ext4 function in question, with a
1036 * more appropriate error code.
1038 ext4_lock_group(sb, grp);
1042 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1046 struct ext4_sb_info *sbi = EXT4_SB(sb);
1047 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1048 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1051 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1052 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1055 percpu_counter_sub(&sbi->s_freeclusters_counter,
1059 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1060 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1065 count = ext4_free_inodes_count(sb, gdp);
1066 percpu_counter_sub(&sbi->s_freeinodes_counter,
1072 void ext4_update_dynamic_rev(struct super_block *sb)
1074 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1076 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1080 "updating to rev %d because of new feature flag, "
1081 "running e2fsck is recommended",
1084 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1085 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1086 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1087 /* leave es->s_feature_*compat flags alone */
1088 /* es->s_uuid will be set by e2fsck if empty */
1091 * The rest of the superblock fields should be zero, and if not it
1092 * means they are likely already in use, so leave them alone. We
1093 * can leave it up to e2fsck to clean up any inconsistencies there.
1098 * Open the external journal device
1100 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1102 struct block_device *bdev;
1104 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1110 ext4_msg(sb, KERN_ERR,
1111 "failed to open journal device unknown-block(%u,%u) %ld",
1112 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1117 * Release the journal device
1119 static void ext4_blkdev_put(struct block_device *bdev)
1121 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1124 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1126 struct block_device *bdev;
1127 bdev = sbi->s_journal_bdev;
1129 ext4_blkdev_put(bdev);
1130 sbi->s_journal_bdev = NULL;
1134 static inline struct inode *orphan_list_entry(struct list_head *l)
1136 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1139 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1141 struct list_head *l;
1143 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1144 le32_to_cpu(sbi->s_es->s_last_orphan));
1146 printk(KERN_ERR "sb_info orphan list:\n");
1147 list_for_each(l, &sbi->s_orphan) {
1148 struct inode *inode = orphan_list_entry(l);
1150 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1151 inode->i_sb->s_id, inode->i_ino, inode,
1152 inode->i_mode, inode->i_nlink,
1153 NEXT_ORPHAN(inode));
1158 static int ext4_quota_off(struct super_block *sb, int type);
1160 static inline void ext4_quota_off_umount(struct super_block *sb)
1164 /* Use our quota_off function to clear inode flags etc. */
1165 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1166 ext4_quota_off(sb, type);
1170 * This is a helper function which is used in the mount/remount
1171 * codepaths (which holds s_umount) to fetch the quota file name.
1173 static inline char *get_qf_name(struct super_block *sb,
1174 struct ext4_sb_info *sbi,
1177 return rcu_dereference_protected(sbi->s_qf_names[type],
1178 lockdep_is_held(&sb->s_umount));
1181 static inline void ext4_quota_off_umount(struct super_block *sb)
1186 static void ext4_put_super(struct super_block *sb)
1188 struct ext4_sb_info *sbi = EXT4_SB(sb);
1189 struct ext4_super_block *es = sbi->s_es;
1190 struct buffer_head **group_desc;
1191 struct flex_groups **flex_groups;
1196 * Unregister sysfs before destroying jbd2 journal.
1197 * Since we could still access attr_journal_task attribute via sysfs
1198 * path which could have sbi->s_journal->j_task as NULL
1199 * Unregister sysfs before flush sbi->s_error_work.
1200 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1201 * read metadata verify failed then will queue error work.
1202 * flush_stashed_error_work will call start_this_handle may trigger
1205 ext4_unregister_sysfs(sb);
1207 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1208 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1211 ext4_unregister_li_request(sb);
1212 ext4_quota_off_umount(sb);
1214 flush_work(&sbi->s_error_work);
1215 destroy_workqueue(sbi->rsv_conversion_wq);
1216 ext4_release_orphan_info(sb);
1218 if (sbi->s_journal) {
1219 aborted = is_journal_aborted(sbi->s_journal);
1220 err = jbd2_journal_destroy(sbi->s_journal);
1221 sbi->s_journal = NULL;
1222 if ((err < 0) && !aborted) {
1223 ext4_abort(sb, -err, "Couldn't clean up the journal");
1227 ext4_es_unregister_shrinker(sbi);
1228 timer_shutdown_sync(&sbi->s_err_report);
1229 ext4_release_system_zone(sb);
1230 ext4_mb_release(sb);
1231 ext4_ext_release(sb);
1233 if (!sb_rdonly(sb) && !aborted) {
1234 ext4_clear_feature_journal_needs_recovery(sb);
1235 ext4_clear_feature_orphan_present(sb);
1236 es->s_state = cpu_to_le16(sbi->s_mount_state);
1239 ext4_commit_super(sb);
1242 group_desc = rcu_dereference(sbi->s_group_desc);
1243 for (i = 0; i < sbi->s_gdb_count; i++)
1244 brelse(group_desc[i]);
1246 flex_groups = rcu_dereference(sbi->s_flex_groups);
1248 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1249 kvfree(flex_groups[i]);
1250 kvfree(flex_groups);
1253 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1254 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1255 percpu_counter_destroy(&sbi->s_dirs_counter);
1256 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1257 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1258 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1260 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1261 kfree(get_qf_name(sb, sbi, i));
1264 /* Debugging code just in case the in-memory inode orphan list
1265 * isn't empty. The on-disk one can be non-empty if we've
1266 * detected an error and taken the fs readonly, but the
1267 * in-memory list had better be clean by this point. */
1268 if (!list_empty(&sbi->s_orphan))
1269 dump_orphan_list(sb, sbi);
1270 ASSERT(list_empty(&sbi->s_orphan));
1272 sync_blockdev(sb->s_bdev);
1273 invalidate_bdev(sb->s_bdev);
1274 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1276 * Invalidate the journal device's buffers. We don't want them
1277 * floating about in memory - the physical journal device may
1278 * hotswapped, and it breaks the `ro-after' testing code.
1280 sync_blockdev(sbi->s_journal_bdev);
1281 invalidate_bdev(sbi->s_journal_bdev);
1282 ext4_blkdev_remove(sbi);
1285 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1286 sbi->s_ea_inode_cache = NULL;
1288 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1289 sbi->s_ea_block_cache = NULL;
1291 ext4_stop_mmpd(sbi);
1294 sb->s_fs_info = NULL;
1296 * Now that we are completely done shutting down the
1297 * superblock, we need to actually destroy the kobject.
1299 kobject_put(&sbi->s_kobj);
1300 wait_for_completion(&sbi->s_kobj_unregister);
1301 if (sbi->s_chksum_driver)
1302 crypto_free_shash(sbi->s_chksum_driver);
1303 kfree(sbi->s_blockgroup_lock);
1304 fs_put_dax(sbi->s_daxdev, NULL);
1305 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1306 #if IS_ENABLED(CONFIG_UNICODE)
1307 utf8_unload(sb->s_encoding);
1312 static struct kmem_cache *ext4_inode_cachep;
1315 * Called inside transaction, so use GFP_NOFS
1317 static struct inode *ext4_alloc_inode(struct super_block *sb)
1319 struct ext4_inode_info *ei;
1321 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1325 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 result.uint_32 = 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;
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 int opt_2 = m->flags & MOPT_2;
2910 unsigned int mount_opt, def_mount_opt;
2912 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2913 m->flags & MOPT_SKIP)
2917 mount_opt = sbi->s_mount_opt2;
2918 def_mount_opt = sbi->s_def_mount_opt2;
2920 mount_opt = sbi->s_mount_opt;
2921 def_mount_opt = sbi->s_def_mount_opt;
2923 /* skip if same as the default */
2924 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2926 /* select Opt_noFoo vs Opt_Foo */
2928 (mount_opt & m->mount_opt) != m->mount_opt) ||
2929 (!want_set && (mount_opt & m->mount_opt)))
2931 SEQ_OPTS_PRINT("%s", token2str(m->token));
2934 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2935 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2936 SEQ_OPTS_PRINT("resuid=%u",
2937 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2938 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2939 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2940 SEQ_OPTS_PRINT("resgid=%u",
2941 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2942 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2943 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2944 SEQ_OPTS_PUTS("errors=remount-ro");
2945 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2946 SEQ_OPTS_PUTS("errors=continue");
2947 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2948 SEQ_OPTS_PUTS("errors=panic");
2949 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2950 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2951 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2952 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2953 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2954 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2955 if (nodefs || sbi->s_stripe)
2956 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2957 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2958 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
2959 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2960 SEQ_OPTS_PUTS("data=journal");
2961 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2962 SEQ_OPTS_PUTS("data=ordered");
2963 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2964 SEQ_OPTS_PUTS("data=writeback");
2967 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2968 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2969 sbi->s_inode_readahead_blks);
2971 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2972 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2973 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2974 if (nodefs || sbi->s_max_dir_size_kb)
2975 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2976 if (test_opt(sb, DATA_ERR_ABORT))
2977 SEQ_OPTS_PUTS("data_err=abort");
2979 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2981 if (sb->s_flags & SB_INLINECRYPT)
2982 SEQ_OPTS_PUTS("inlinecrypt");
2984 if (test_opt(sb, DAX_ALWAYS)) {
2986 SEQ_OPTS_PUTS("dax");
2988 SEQ_OPTS_PUTS("dax=always");
2989 } else if (test_opt2(sb, DAX_NEVER)) {
2990 SEQ_OPTS_PUTS("dax=never");
2991 } else if (test_opt2(sb, DAX_INODE)) {
2992 SEQ_OPTS_PUTS("dax=inode");
2995 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2996 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2997 SEQ_OPTS_PUTS("mb_optimize_scan=0");
2998 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2999 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3000 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3003 ext4_show_quota_options(seq, sb);
3007 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3009 return _ext4_show_options(seq, root->d_sb, 0);
3012 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3014 struct super_block *sb = seq->private;
3017 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3018 rc = _ext4_show_options(seq, sb, 1);
3019 seq_puts(seq, "\n");
3023 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3026 struct ext4_sb_info *sbi = EXT4_SB(sb);
3029 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3030 ext4_msg(sb, KERN_ERR, "revision level too high, "
3031 "forcing read-only mode");
3037 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3038 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3039 "running e2fsck is recommended");
3040 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3041 ext4_msg(sb, KERN_WARNING,
3042 "warning: mounting fs with errors, "
3043 "running e2fsck is recommended");
3044 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3045 le16_to_cpu(es->s_mnt_count) >=
3046 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3047 ext4_msg(sb, KERN_WARNING,
3048 "warning: maximal mount count reached, "
3049 "running e2fsck is recommended");
3050 else if (le32_to_cpu(es->s_checkinterval) &&
3051 (ext4_get_tstamp(es, s_lastcheck) +
3052 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3053 ext4_msg(sb, KERN_WARNING,
3054 "warning: checktime reached, "
3055 "running e2fsck is recommended");
3056 if (!sbi->s_journal)
3057 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3058 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3059 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3060 le16_add_cpu(&es->s_mnt_count, 1);
3061 ext4_update_tstamp(es, s_mtime);
3062 if (sbi->s_journal) {
3063 ext4_set_feature_journal_needs_recovery(sb);
3064 if (ext4_has_feature_orphan_file(sb))
3065 ext4_set_feature_orphan_present(sb);
3068 err = ext4_commit_super(sb);
3070 if (test_opt(sb, DEBUG))
3071 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3072 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3074 sbi->s_groups_count,
3075 EXT4_BLOCKS_PER_GROUP(sb),
3076 EXT4_INODES_PER_GROUP(sb),
3077 sbi->s_mount_opt, sbi->s_mount_opt2);
3081 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3083 struct ext4_sb_info *sbi = EXT4_SB(sb);
3084 struct flex_groups **old_groups, **new_groups;
3087 if (!sbi->s_log_groups_per_flex)
3090 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3091 if (size <= sbi->s_flex_groups_allocated)
3094 new_groups = kvzalloc(roundup_pow_of_two(size *
3095 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3097 ext4_msg(sb, KERN_ERR,
3098 "not enough memory for %d flex group pointers", size);
3101 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3102 new_groups[i] = kvzalloc(roundup_pow_of_two(
3103 sizeof(struct flex_groups)),
3105 if (!new_groups[i]) {
3106 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3107 kvfree(new_groups[j]);
3109 ext4_msg(sb, KERN_ERR,
3110 "not enough memory for %d flex groups", size);
3115 old_groups = rcu_dereference(sbi->s_flex_groups);
3117 memcpy(new_groups, old_groups,
3118 (sbi->s_flex_groups_allocated *
3119 sizeof(struct flex_groups *)));
3121 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3122 sbi->s_flex_groups_allocated = size;
3124 ext4_kvfree_array_rcu(old_groups);
3128 static int ext4_fill_flex_info(struct super_block *sb)
3130 struct ext4_sb_info *sbi = EXT4_SB(sb);
3131 struct ext4_group_desc *gdp = NULL;
3132 struct flex_groups *fg;
3133 ext4_group_t flex_group;
3136 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3137 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3138 sbi->s_log_groups_per_flex = 0;
3142 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3146 for (i = 0; i < sbi->s_groups_count; i++) {
3147 gdp = ext4_get_group_desc(sb, i, NULL);
3149 flex_group = ext4_flex_group(sbi, i);
3150 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3151 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3152 atomic64_add(ext4_free_group_clusters(sb, gdp),
3153 &fg->free_clusters);
3154 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3162 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3163 struct ext4_group_desc *gdp)
3165 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3167 __le32 le_group = cpu_to_le32(block_group);
3168 struct ext4_sb_info *sbi = EXT4_SB(sb);
3170 if (ext4_has_metadata_csum(sbi->s_sb)) {
3171 /* Use new metadata_csum algorithm */
3173 __u16 dummy_csum = 0;
3175 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3177 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3178 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3179 sizeof(dummy_csum));
3180 offset += sizeof(dummy_csum);
3181 if (offset < sbi->s_desc_size)
3182 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3183 sbi->s_desc_size - offset);
3185 crc = csum32 & 0xFFFF;
3189 /* old crc16 code */
3190 if (!ext4_has_feature_gdt_csum(sb))
3193 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3194 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3195 crc = crc16(crc, (__u8 *)gdp, offset);
3196 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3197 /* for checksum of struct ext4_group_desc do the rest...*/
3198 if (ext4_has_feature_64bit(sb) &&
3199 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3200 crc = crc16(crc, (__u8 *)gdp + offset,
3201 le16_to_cpu(sbi->s_es->s_desc_size) -
3205 return cpu_to_le16(crc);
3208 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3209 struct ext4_group_desc *gdp)
3211 if (ext4_has_group_desc_csum(sb) &&
3212 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3218 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3219 struct ext4_group_desc *gdp)
3221 if (!ext4_has_group_desc_csum(sb))
3223 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3226 /* Called at mount-time, super-block is locked */
3227 static int ext4_check_descriptors(struct super_block *sb,
3228 ext4_fsblk_t sb_block,
3229 ext4_group_t *first_not_zeroed)
3231 struct ext4_sb_info *sbi = EXT4_SB(sb);
3232 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3233 ext4_fsblk_t last_block;
3234 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3235 ext4_fsblk_t block_bitmap;
3236 ext4_fsblk_t inode_bitmap;
3237 ext4_fsblk_t inode_table;
3238 int flexbg_flag = 0;
3239 ext4_group_t i, grp = sbi->s_groups_count;
3241 if (ext4_has_feature_flex_bg(sb))
3244 ext4_debug("Checking group descriptors");
3246 for (i = 0; i < sbi->s_groups_count; i++) {
3247 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3249 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3250 last_block = ext4_blocks_count(sbi->s_es) - 1;
3252 last_block = first_block +
3253 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3255 if ((grp == sbi->s_groups_count) &&
3256 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3259 block_bitmap = ext4_block_bitmap(sb, gdp);
3260 if (block_bitmap == sb_block) {
3261 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3262 "Block bitmap for group %u overlaps "
3267 if (block_bitmap >= sb_block + 1 &&
3268 block_bitmap <= last_bg_block) {
3269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3270 "Block bitmap for group %u overlaps "
3271 "block group descriptors", i);
3275 if (block_bitmap < first_block || block_bitmap > last_block) {
3276 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3277 "Block bitmap for group %u not in group "
3278 "(block %llu)!", i, block_bitmap);
3281 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3282 if (inode_bitmap == sb_block) {
3283 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3284 "Inode bitmap for group %u overlaps "
3289 if (inode_bitmap >= sb_block + 1 &&
3290 inode_bitmap <= last_bg_block) {
3291 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3292 "Inode bitmap for group %u overlaps "
3293 "block group descriptors", i);
3297 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3298 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3299 "Inode bitmap for group %u not in group "
3300 "(block %llu)!", i, inode_bitmap);
3303 inode_table = ext4_inode_table(sb, gdp);
3304 if (inode_table == sb_block) {
3305 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3306 "Inode table for group %u overlaps "
3311 if (inode_table >= sb_block + 1 &&
3312 inode_table <= last_bg_block) {
3313 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3314 "Inode table for group %u overlaps "
3315 "block group descriptors", i);
3319 if (inode_table < first_block ||
3320 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3321 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3322 "Inode table for group %u not in group "
3323 "(block %llu)!", i, inode_table);
3326 ext4_lock_group(sb, i);
3327 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3328 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3329 "Checksum for group %u failed (%u!=%u)",
3330 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3331 gdp)), le16_to_cpu(gdp->bg_checksum));
3332 if (!sb_rdonly(sb)) {
3333 ext4_unlock_group(sb, i);
3337 ext4_unlock_group(sb, i);
3339 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3341 if (NULL != first_not_zeroed)
3342 *first_not_zeroed = grp;
3347 * Maximal extent format file size.
3348 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3349 * extent format containers, within a sector_t, and within i_blocks
3350 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3351 * so that won't be a limiting factor.
3353 * However there is other limiting factor. We do store extents in the form
3354 * of starting block and length, hence the resulting length of the extent
3355 * covering maximum file size must fit into on-disk format containers as
3356 * well. Given that length is always by 1 unit bigger than max unit (because
3357 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3359 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3361 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3364 loff_t upper_limit = MAX_LFS_FILESIZE;
3366 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3368 if (!has_huge_files) {
3369 upper_limit = (1LL << 32) - 1;
3371 /* total blocks in file system block size */
3372 upper_limit >>= (blkbits - 9);
3373 upper_limit <<= blkbits;
3377 * 32-bit extent-start container, ee_block. We lower the maxbytes
3378 * by one fs block, so ee_len can cover the extent of maximum file
3381 res = (1LL << 32) - 1;
3384 /* Sanity check against vm- & vfs- imposed limits */
3385 if (res > upper_limit)
3392 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3393 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3394 * We need to be 1 filesystem block less than the 2^48 sector limit.
3396 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3398 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3400 unsigned int ppb = 1 << (bits - 2);
3403 * This is calculated to be the largest file size for a dense, block
3404 * mapped file such that the file's total number of 512-byte sectors,
3405 * including data and all indirect blocks, does not exceed (2^48 - 1).
3407 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3408 * number of 512-byte sectors of the file.
3410 if (!has_huge_files) {
3412 * !has_huge_files or implies that the inode i_block field
3413 * represents total file blocks in 2^32 512-byte sectors ==
3414 * size of vfs inode i_blocks * 8
3416 upper_limit = (1LL << 32) - 1;
3418 /* total blocks in file system block size */
3419 upper_limit >>= (bits - 9);
3423 * We use 48 bit ext4_inode i_blocks
3424 * With EXT4_HUGE_FILE_FL set the i_blocks
3425 * represent total number of blocks in
3426 * file system block size
3428 upper_limit = (1LL << 48) - 1;
3432 /* Compute how many blocks we can address by block tree */
3435 res += ((loff_t)ppb) * ppb * ppb;
3436 /* Compute how many metadata blocks are needed */
3438 meta_blocks += 1 + ppb;
3439 meta_blocks += 1 + ppb + ppb * ppb;
3440 /* Does block tree limit file size? */
3441 if (res + meta_blocks <= upper_limit)
3445 /* How many metadata blocks are needed for addressing upper_limit? */
3446 upper_limit -= EXT4_NDIR_BLOCKS;
3447 /* indirect blocks */
3450 /* double indirect blocks */
3451 if (upper_limit < ppb * ppb) {
3452 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3456 meta_blocks += 1 + ppb;
3457 upper_limit -= ppb * ppb;
3458 /* tripple indirect blocks for the rest */
3459 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3460 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3464 if (res > MAX_LFS_FILESIZE)
3465 res = MAX_LFS_FILESIZE;
3470 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3471 ext4_fsblk_t logical_sb_block, int nr)
3473 struct ext4_sb_info *sbi = EXT4_SB(sb);
3474 ext4_group_t bg, first_meta_bg;
3477 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3479 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3480 return logical_sb_block + nr + 1;
3481 bg = sbi->s_desc_per_block * nr;
3482 if (ext4_bg_has_super(sb, bg))
3486 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3487 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3488 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3491 if (sb->s_blocksize == 1024 && nr == 0 &&
3492 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3495 return (has_super + ext4_group_first_block_no(sb, bg));
3499 * ext4_get_stripe_size: Get the stripe size.
3500 * @sbi: In memory super block info
3502 * If we have specified it via mount option, then
3503 * use the mount option value. If the value specified at mount time is
3504 * greater than the blocks per group use the super block value.
3505 * If the super block value is greater than blocks per group return 0.
3506 * Allocator needs it be less than blocks per group.
3509 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3511 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3512 unsigned long stripe_width =
3513 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3516 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3517 ret = sbi->s_stripe;
3518 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3520 else if (stride && stride <= sbi->s_blocks_per_group)
3526 * If the stripe width is 1, this makes no sense and
3527 * we set it to 0 to turn off stripe handling code.
3536 * Check whether this filesystem can be mounted based on
3537 * the features present and the RDONLY/RDWR mount requested.
3538 * Returns 1 if this filesystem can be mounted as requested,
3539 * 0 if it cannot be.
3541 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3543 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3544 ext4_msg(sb, KERN_ERR,
3545 "Couldn't mount because of "
3546 "unsupported optional features (%x)",
3547 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3548 ~EXT4_FEATURE_INCOMPAT_SUPP));
3552 #if !IS_ENABLED(CONFIG_UNICODE)
3553 if (ext4_has_feature_casefold(sb)) {
3554 ext4_msg(sb, KERN_ERR,
3555 "Filesystem with casefold feature cannot be "
3556 "mounted without CONFIG_UNICODE");
3564 if (ext4_has_feature_readonly(sb)) {
3565 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3566 sb->s_flags |= SB_RDONLY;
3570 /* Check that feature set is OK for a read-write mount */
3571 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3572 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3573 "unsupported optional features (%x)",
3574 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3575 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3578 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3579 ext4_msg(sb, KERN_ERR,
3580 "Can't support bigalloc feature without "
3581 "extents feature\n");
3585 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3586 if (!readonly && (ext4_has_feature_quota(sb) ||
3587 ext4_has_feature_project(sb))) {
3588 ext4_msg(sb, KERN_ERR,
3589 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3592 #endif /* CONFIG_QUOTA */
3597 * This function is called once a day if we have errors logged
3598 * on the file system
3600 static void print_daily_error_info(struct timer_list *t)
3602 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3603 struct super_block *sb = sbi->s_sb;
3604 struct ext4_super_block *es = sbi->s_es;
3606 if (es->s_error_count)
3607 /* fsck newer than v1.41.13 is needed to clean this condition. */
3608 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3609 le32_to_cpu(es->s_error_count));
3610 if (es->s_first_error_time) {
3611 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3613 ext4_get_tstamp(es, s_first_error_time),
3614 (int) sizeof(es->s_first_error_func),
3615 es->s_first_error_func,
3616 le32_to_cpu(es->s_first_error_line));
3617 if (es->s_first_error_ino)
3618 printk(KERN_CONT ": inode %u",
3619 le32_to_cpu(es->s_first_error_ino));
3620 if (es->s_first_error_block)
3621 printk(KERN_CONT ": block %llu", (unsigned long long)
3622 le64_to_cpu(es->s_first_error_block));
3623 printk(KERN_CONT "\n");
3625 if (es->s_last_error_time) {
3626 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3628 ext4_get_tstamp(es, s_last_error_time),
3629 (int) sizeof(es->s_last_error_func),
3630 es->s_last_error_func,
3631 le32_to_cpu(es->s_last_error_line));
3632 if (es->s_last_error_ino)
3633 printk(KERN_CONT ": inode %u",
3634 le32_to_cpu(es->s_last_error_ino));
3635 if (es->s_last_error_block)
3636 printk(KERN_CONT ": block %llu", (unsigned long long)
3637 le64_to_cpu(es->s_last_error_block));
3638 printk(KERN_CONT "\n");
3640 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3643 /* Find next suitable group and run ext4_init_inode_table */
3644 static int ext4_run_li_request(struct ext4_li_request *elr)
3646 struct ext4_group_desc *gdp = NULL;
3647 struct super_block *sb = elr->lr_super;
3648 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3649 ext4_group_t group = elr->lr_next_group;
3650 unsigned int prefetch_ios = 0;
3654 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3655 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3656 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3658 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3660 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3662 if (group >= elr->lr_next_group) {
3664 if (elr->lr_first_not_zeroed != ngroups &&
3665 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3666 elr->lr_next_group = elr->lr_first_not_zeroed;
3667 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3674 for (; group < ngroups; group++) {
3675 gdp = ext4_get_group_desc(sb, group, NULL);
3681 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3685 if (group >= ngroups)
3689 start_time = ktime_get_real_ns();
3690 ret = ext4_init_inode_table(sb, group,
3691 elr->lr_timeout ? 0 : 1);
3692 trace_ext4_lazy_itable_init(sb, group);
3693 if (elr->lr_timeout == 0) {
3694 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3695 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3697 elr->lr_next_sched = jiffies + elr->lr_timeout;
3698 elr->lr_next_group = group + 1;
3704 * Remove lr_request from the list_request and free the
3705 * request structure. Should be called with li_list_mtx held
3707 static void ext4_remove_li_request(struct ext4_li_request *elr)
3712 list_del(&elr->lr_request);
3713 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3717 static void ext4_unregister_li_request(struct super_block *sb)
3719 mutex_lock(&ext4_li_mtx);
3720 if (!ext4_li_info) {
3721 mutex_unlock(&ext4_li_mtx);
3725 mutex_lock(&ext4_li_info->li_list_mtx);
3726 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3727 mutex_unlock(&ext4_li_info->li_list_mtx);
3728 mutex_unlock(&ext4_li_mtx);
3731 static struct task_struct *ext4_lazyinit_task;
3734 * This is the function where ext4lazyinit thread lives. It walks
3735 * through the request list searching for next scheduled filesystem.
3736 * When such a fs is found, run the lazy initialization request
3737 * (ext4_rn_li_request) and keep track of the time spend in this
3738 * function. Based on that time we compute next schedule time of
3739 * the request. When walking through the list is complete, compute
3740 * next waking time and put itself into sleep.
3742 static int ext4_lazyinit_thread(void *arg)
3744 struct ext4_lazy_init *eli = arg;
3745 struct list_head *pos, *n;
3746 struct ext4_li_request *elr;
3747 unsigned long next_wakeup, cur;
3749 BUG_ON(NULL == eli);
3754 next_wakeup = MAX_JIFFY_OFFSET;
3756 mutex_lock(&eli->li_list_mtx);
3757 if (list_empty(&eli->li_request_list)) {
3758 mutex_unlock(&eli->li_list_mtx);
3761 list_for_each_safe(pos, n, &eli->li_request_list) {
3764 elr = list_entry(pos, struct ext4_li_request,
3767 if (time_before(jiffies, elr->lr_next_sched)) {
3768 if (time_before(elr->lr_next_sched, next_wakeup))
3769 next_wakeup = elr->lr_next_sched;
3772 if (down_read_trylock(&elr->lr_super->s_umount)) {
3773 if (sb_start_write_trylock(elr->lr_super)) {
3776 * We hold sb->s_umount, sb can not
3777 * be removed from the list, it is
3778 * now safe to drop li_list_mtx
3780 mutex_unlock(&eli->li_list_mtx);
3781 err = ext4_run_li_request(elr);
3782 sb_end_write(elr->lr_super);
3783 mutex_lock(&eli->li_list_mtx);
3786 up_read((&elr->lr_super->s_umount));
3788 /* error, remove the lazy_init job */
3790 ext4_remove_li_request(elr);
3794 elr->lr_next_sched = jiffies +
3795 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3797 if (time_before(elr->lr_next_sched, next_wakeup))
3798 next_wakeup = elr->lr_next_sched;
3800 mutex_unlock(&eli->li_list_mtx);
3805 if ((time_after_eq(cur, next_wakeup)) ||
3806 (MAX_JIFFY_OFFSET == next_wakeup)) {
3811 schedule_timeout_interruptible(next_wakeup - cur);
3813 if (kthread_should_stop()) {
3814 ext4_clear_request_list();
3821 * It looks like the request list is empty, but we need
3822 * to check it under the li_list_mtx lock, to prevent any
3823 * additions into it, and of course we should lock ext4_li_mtx
3824 * to atomically free the list and ext4_li_info, because at
3825 * this point another ext4 filesystem could be registering
3828 mutex_lock(&ext4_li_mtx);
3829 mutex_lock(&eli->li_list_mtx);
3830 if (!list_empty(&eli->li_request_list)) {
3831 mutex_unlock(&eli->li_list_mtx);
3832 mutex_unlock(&ext4_li_mtx);
3835 mutex_unlock(&eli->li_list_mtx);
3836 kfree(ext4_li_info);
3837 ext4_li_info = NULL;
3838 mutex_unlock(&ext4_li_mtx);
3843 static void ext4_clear_request_list(void)
3845 struct list_head *pos, *n;
3846 struct ext4_li_request *elr;
3848 mutex_lock(&ext4_li_info->li_list_mtx);
3849 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3850 elr = list_entry(pos, struct ext4_li_request,
3852 ext4_remove_li_request(elr);
3854 mutex_unlock(&ext4_li_info->li_list_mtx);
3857 static int ext4_run_lazyinit_thread(void)
3859 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3860 ext4_li_info, "ext4lazyinit");
3861 if (IS_ERR(ext4_lazyinit_task)) {
3862 int err = PTR_ERR(ext4_lazyinit_task);
3863 ext4_clear_request_list();
3864 kfree(ext4_li_info);
3865 ext4_li_info = NULL;
3866 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3867 "initialization thread\n",
3871 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3876 * Check whether it make sense to run itable init. thread or not.
3877 * If there is at least one uninitialized inode table, return
3878 * corresponding group number, else the loop goes through all
3879 * groups and return total number of groups.
3881 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3883 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3884 struct ext4_group_desc *gdp = NULL;
3886 if (!ext4_has_group_desc_csum(sb))
3889 for (group = 0; group < ngroups; group++) {
3890 gdp = ext4_get_group_desc(sb, group, NULL);
3894 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3901 static int ext4_li_info_new(void)
3903 struct ext4_lazy_init *eli = NULL;
3905 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3909 INIT_LIST_HEAD(&eli->li_request_list);
3910 mutex_init(&eli->li_list_mtx);
3912 eli->li_state |= EXT4_LAZYINIT_QUIT;
3919 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3922 struct ext4_li_request *elr;
3924 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3929 elr->lr_first_not_zeroed = start;
3930 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3931 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3932 elr->lr_next_group = start;
3934 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3938 * Randomize first schedule time of the request to
3939 * spread the inode table initialization requests
3942 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3946 int ext4_register_li_request(struct super_block *sb,
3947 ext4_group_t first_not_zeroed)
3949 struct ext4_sb_info *sbi = EXT4_SB(sb);
3950 struct ext4_li_request *elr = NULL;
3951 ext4_group_t ngroups = sbi->s_groups_count;
3954 mutex_lock(&ext4_li_mtx);
3955 if (sbi->s_li_request != NULL) {
3957 * Reset timeout so it can be computed again, because
3958 * s_li_wait_mult might have changed.
3960 sbi->s_li_request->lr_timeout = 0;
3964 if (sb_rdonly(sb) ||
3965 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3966 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
3969 elr = ext4_li_request_new(sb, first_not_zeroed);
3975 if (NULL == ext4_li_info) {
3976 ret = ext4_li_info_new();
3981 mutex_lock(&ext4_li_info->li_list_mtx);
3982 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3983 mutex_unlock(&ext4_li_info->li_list_mtx);
3985 sbi->s_li_request = elr;
3987 * set elr to NULL here since it has been inserted to
3988 * the request_list and the removal and free of it is
3989 * handled by ext4_clear_request_list from now on.
3993 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3994 ret = ext4_run_lazyinit_thread();
3999 mutex_unlock(&ext4_li_mtx);
4006 * We do not need to lock anything since this is called on
4009 static void ext4_destroy_lazyinit_thread(void)
4012 * If thread exited earlier
4013 * there's nothing to be done.
4015 if (!ext4_li_info || !ext4_lazyinit_task)
4018 kthread_stop(ext4_lazyinit_task);
4021 static int set_journal_csum_feature_set(struct super_block *sb)
4024 int compat, incompat;
4025 struct ext4_sb_info *sbi = EXT4_SB(sb);
4027 if (ext4_has_metadata_csum(sb)) {
4028 /* journal checksum v3 */
4030 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4032 /* journal checksum v1 */
4033 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4037 jbd2_journal_clear_features(sbi->s_journal,
4038 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4039 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4040 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4041 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4042 ret = jbd2_journal_set_features(sbi->s_journal,
4044 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4046 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4047 ret = jbd2_journal_set_features(sbi->s_journal,
4050 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4051 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4053 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4054 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4061 * Note: calculating the overhead so we can be compatible with
4062 * historical BSD practice is quite difficult in the face of
4063 * clusters/bigalloc. This is because multiple metadata blocks from
4064 * different block group can end up in the same allocation cluster.
4065 * Calculating the exact overhead in the face of clustered allocation
4066 * requires either O(all block bitmaps) in memory or O(number of block
4067 * groups**2) in time. We will still calculate the superblock for
4068 * older file systems --- and if we come across with a bigalloc file
4069 * system with zero in s_overhead_clusters the estimate will be close to
4070 * correct especially for very large cluster sizes --- but for newer
4071 * file systems, it's better to calculate this figure once at mkfs
4072 * time, and store it in the superblock. If the superblock value is
4073 * present (even for non-bigalloc file systems), we will use it.
4075 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4078 struct ext4_sb_info *sbi = EXT4_SB(sb);
4079 struct ext4_group_desc *gdp;
4080 ext4_fsblk_t first_block, last_block, b;
4081 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4082 int s, j, count = 0;
4083 int has_super = ext4_bg_has_super(sb, grp);
4085 if (!ext4_has_feature_bigalloc(sb))
4086 return (has_super + ext4_bg_num_gdb(sb, grp) +
4087 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4088 sbi->s_itb_per_group + 2);
4090 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4091 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4092 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4093 for (i = 0; i < ngroups; i++) {
4094 gdp = ext4_get_group_desc(sb, i, NULL);
4095 b = ext4_block_bitmap(sb, gdp);
4096 if (b >= first_block && b <= last_block) {
4097 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4100 b = ext4_inode_bitmap(sb, gdp);
4101 if (b >= first_block && b <= last_block) {
4102 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4105 b = ext4_inode_table(sb, gdp);
4106 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4107 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4108 int c = EXT4_B2C(sbi, b - first_block);
4109 ext4_set_bit(c, buf);
4115 if (ext4_bg_has_super(sb, grp)) {
4116 ext4_set_bit(s++, buf);
4119 j = ext4_bg_num_gdb(sb, grp);
4120 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4121 ext4_error(sb, "Invalid number of block group "
4122 "descriptor blocks: %d", j);
4123 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4127 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4131 return EXT4_CLUSTERS_PER_GROUP(sb) -
4132 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4136 * Compute the overhead and stash it in sbi->s_overhead
4138 int ext4_calculate_overhead(struct super_block *sb)
4140 struct ext4_sb_info *sbi = EXT4_SB(sb);
4141 struct ext4_super_block *es = sbi->s_es;
4142 struct inode *j_inode;
4143 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4144 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4145 ext4_fsblk_t overhead = 0;
4146 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4152 * Compute the overhead (FS structures). This is constant
4153 * for a given filesystem unless the number of block groups
4154 * changes so we cache the previous value until it does.
4158 * All of the blocks before first_data_block are overhead
4160 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4163 * Add the overhead found in each block group
4165 for (i = 0; i < ngroups; i++) {
4168 blks = count_overhead(sb, i, buf);
4171 memset(buf, 0, PAGE_SIZE);
4176 * Add the internal journal blocks whether the journal has been
4179 if (sbi->s_journal && !sbi->s_journal_bdev)
4180 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4181 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4182 /* j_inum for internal journal is non-zero */
4183 j_inode = ext4_get_journal_inode(sb, j_inum);
4185 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4186 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4189 ext4_msg(sb, KERN_ERR, "can't get journal size");
4192 sbi->s_overhead = overhead;
4194 free_page((unsigned long) buf);
4198 static void ext4_set_resv_clusters(struct super_block *sb)
4200 ext4_fsblk_t resv_clusters;
4201 struct ext4_sb_info *sbi = EXT4_SB(sb);
4204 * There's no need to reserve anything when we aren't using extents.
4205 * The space estimates are exact, there are no unwritten extents,
4206 * hole punching doesn't need new metadata... This is needed especially
4207 * to keep ext2/3 backward compatibility.
4209 if (!ext4_has_feature_extents(sb))
4212 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4213 * This should cover the situations where we can not afford to run
4214 * out of space like for example punch hole, or converting
4215 * unwritten extents in delalloc path. In most cases such
4216 * allocation would require 1, or 2 blocks, higher numbers are
4219 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4220 sbi->s_cluster_bits);
4222 do_div(resv_clusters, 50);
4223 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4225 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4228 static const char *ext4_quota_mode(struct super_block *sb)
4231 if (!ext4_quota_capable(sb))
4234 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4235 return "journalled";
4243 static void ext4_setup_csum_trigger(struct super_block *sb,
4244 enum ext4_journal_trigger_type type,
4246 struct jbd2_buffer_trigger_type *type,
4247 struct buffer_head *bh,
4251 struct ext4_sb_info *sbi = EXT4_SB(sb);
4253 sbi->s_journal_triggers[type].sb = sb;
4254 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4257 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4262 kfree(sbi->s_blockgroup_lock);
4263 fs_put_dax(sbi->s_daxdev, NULL);
4267 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4269 struct ext4_sb_info *sbi;
4271 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4275 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4278 sbi->s_blockgroup_lock =
4279 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4281 if (!sbi->s_blockgroup_lock)
4284 sb->s_fs_info = sbi;
4288 fs_put_dax(sbi->s_daxdev, NULL);
4293 static void ext4_set_def_opts(struct super_block *sb,
4294 struct ext4_super_block *es)
4296 unsigned long def_mount_opts;
4298 /* Set defaults before we parse the mount options */
4299 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4300 set_opt(sb, INIT_INODE_TABLE);
4301 if (def_mount_opts & EXT4_DEFM_DEBUG)
4303 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4305 if (def_mount_opts & EXT4_DEFM_UID16)
4306 set_opt(sb, NO_UID32);
4307 /* xattr user namespace & acls are now defaulted on */
4308 set_opt(sb, XATTR_USER);
4309 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4310 set_opt(sb, POSIX_ACL);
4312 if (ext4_has_feature_fast_commit(sb))
4313 set_opt2(sb, JOURNAL_FAST_COMMIT);
4314 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4315 if (ext4_has_metadata_csum(sb))
4316 set_opt(sb, JOURNAL_CHECKSUM);
4318 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4319 set_opt(sb, JOURNAL_DATA);
4320 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4321 set_opt(sb, ORDERED_DATA);
4322 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4323 set_opt(sb, WRITEBACK_DATA);
4325 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4326 set_opt(sb, ERRORS_PANIC);
4327 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4328 set_opt(sb, ERRORS_CONT);
4330 set_opt(sb, ERRORS_RO);
4331 /* block_validity enabled by default; disable with noblock_validity */
4332 set_opt(sb, BLOCK_VALIDITY);
4333 if (def_mount_opts & EXT4_DEFM_DISCARD)
4334 set_opt(sb, DISCARD);
4336 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4337 set_opt(sb, BARRIER);
4340 * enable delayed allocation by default
4341 * Use -o nodelalloc to turn it off
4343 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4344 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4345 set_opt(sb, DELALLOC);
4347 if (sb->s_blocksize == PAGE_SIZE)
4348 set_opt(sb, DIOREAD_NOLOCK);
4351 static int ext4_handle_clustersize(struct super_block *sb)
4353 struct ext4_sb_info *sbi = EXT4_SB(sb);
4354 struct ext4_super_block *es = sbi->s_es;
4357 /* Handle clustersize */
4358 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4359 if (ext4_has_feature_bigalloc(sb)) {
4360 if (clustersize < sb->s_blocksize) {
4361 ext4_msg(sb, KERN_ERR,
4362 "cluster size (%d) smaller than "
4363 "block size (%lu)", clustersize, sb->s_blocksize);
4366 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4367 le32_to_cpu(es->s_log_block_size);
4368 sbi->s_clusters_per_group =
4369 le32_to_cpu(es->s_clusters_per_group);
4370 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4371 ext4_msg(sb, KERN_ERR,
4372 "#clusters per group too big: %lu",
4373 sbi->s_clusters_per_group);
4376 if (sbi->s_blocks_per_group !=
4377 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4378 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4379 "clusters per group (%lu) inconsistent",
4380 sbi->s_blocks_per_group,
4381 sbi->s_clusters_per_group);
4385 if (clustersize != sb->s_blocksize) {
4386 ext4_msg(sb, KERN_ERR,
4387 "fragment/cluster size (%d) != "
4388 "block size (%lu)", clustersize, sb->s_blocksize);
4391 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4392 ext4_msg(sb, KERN_ERR,
4393 "#blocks per group too big: %lu",
4394 sbi->s_blocks_per_group);
4397 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4398 sbi->s_cluster_bits = 0;
4400 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4402 /* Do we have standard group size of clustersize * 8 blocks ? */
4403 if (sbi->s_blocks_per_group == clustersize << 3)
4404 set_opt2(sb, STD_GROUP_SIZE);
4409 static void ext4_fast_commit_init(struct super_block *sb)
4411 struct ext4_sb_info *sbi = EXT4_SB(sb);
4413 /* Initialize fast commit stuff */
4414 atomic_set(&sbi->s_fc_subtid, 0);
4415 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4416 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4417 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4418 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4419 sbi->s_fc_bytes = 0;
4420 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4421 sbi->s_fc_ineligible_tid = 0;
4422 spin_lock_init(&sbi->s_fc_lock);
4423 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4424 sbi->s_fc_replay_state.fc_regions = NULL;
4425 sbi->s_fc_replay_state.fc_regions_size = 0;
4426 sbi->s_fc_replay_state.fc_regions_used = 0;
4427 sbi->s_fc_replay_state.fc_regions_valid = 0;
4428 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4429 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4430 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4433 static int ext4_inode_info_init(struct super_block *sb,
4434 struct ext4_super_block *es)
4436 struct ext4_sb_info *sbi = EXT4_SB(sb);
4438 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4439 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4440 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4442 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4443 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4444 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4445 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4449 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4450 (!is_power_of_2(sbi->s_inode_size)) ||
4451 (sbi->s_inode_size > sb->s_blocksize)) {
4452 ext4_msg(sb, KERN_ERR,
4453 "unsupported inode size: %d",
4455 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4459 * i_atime_extra is the last extra field available for
4460 * [acm]times in struct ext4_inode. Checking for that
4461 * field should suffice to ensure we have extra space
4464 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4465 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4466 sb->s_time_gran = 1;
4467 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4469 sb->s_time_gran = NSEC_PER_SEC;
4470 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4472 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4475 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4476 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4477 EXT4_GOOD_OLD_INODE_SIZE;
4478 if (ext4_has_feature_extra_isize(sb)) {
4479 unsigned v, max = (sbi->s_inode_size -
4480 EXT4_GOOD_OLD_INODE_SIZE);
4482 v = le16_to_cpu(es->s_want_extra_isize);
4484 ext4_msg(sb, KERN_ERR,
4485 "bad s_want_extra_isize: %d", v);
4488 if (sbi->s_want_extra_isize < v)
4489 sbi->s_want_extra_isize = v;
4491 v = le16_to_cpu(es->s_min_extra_isize);
4493 ext4_msg(sb, KERN_ERR,
4494 "bad s_min_extra_isize: %d", v);
4497 if (sbi->s_want_extra_isize < v)
4498 sbi->s_want_extra_isize = v;
4505 #if IS_ENABLED(CONFIG_UNICODE)
4506 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4508 const struct ext4_sb_encodings *encoding_info;
4509 struct unicode_map *encoding;
4510 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4512 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4515 encoding_info = ext4_sb_read_encoding(es);
4516 if (!encoding_info) {
4517 ext4_msg(sb, KERN_ERR,
4518 "Encoding requested by superblock is unknown");
4522 encoding = utf8_load(encoding_info->version);
4523 if (IS_ERR(encoding)) {
4524 ext4_msg(sb, KERN_ERR,
4525 "can't mount with superblock charset: %s-%u.%u.%u "
4526 "not supported by the kernel. flags: 0x%x.",
4527 encoding_info->name,
4528 unicode_major(encoding_info->version),
4529 unicode_minor(encoding_info->version),
4530 unicode_rev(encoding_info->version),
4534 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4535 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4536 unicode_major(encoding_info->version),
4537 unicode_minor(encoding_info->version),
4538 unicode_rev(encoding_info->version),
4541 sb->s_encoding = encoding;
4542 sb->s_encoding_flags = encoding_flags;
4547 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4553 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4555 struct ext4_sb_info *sbi = EXT4_SB(sb);
4557 /* Warn if metadata_csum and gdt_csum are both set. */
4558 if (ext4_has_feature_metadata_csum(sb) &&
4559 ext4_has_feature_gdt_csum(sb))
4560 ext4_warning(sb, "metadata_csum and uninit_bg are "
4561 "redundant flags; please run fsck.");
4563 /* Check for a known checksum algorithm */
4564 if (!ext4_verify_csum_type(sb, es)) {
4565 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4566 "unknown checksum algorithm.");
4569 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4570 ext4_orphan_file_block_trigger);
4572 /* Load the checksum driver */
4573 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4574 if (IS_ERR(sbi->s_chksum_driver)) {
4575 int ret = PTR_ERR(sbi->s_chksum_driver);
4576 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4577 sbi->s_chksum_driver = NULL;
4581 /* Check superblock checksum */
4582 if (!ext4_superblock_csum_verify(sb, es)) {
4583 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4584 "invalid superblock checksum. Run e2fsck?");
4588 /* Precompute checksum seed for all metadata */
4589 if (ext4_has_feature_csum_seed(sb))
4590 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4591 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4592 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4593 sizeof(es->s_uuid));
4597 static int ext4_check_feature_compatibility(struct super_block *sb,
4598 struct ext4_super_block *es,
4601 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4602 (ext4_has_compat_features(sb) ||
4603 ext4_has_ro_compat_features(sb) ||
4604 ext4_has_incompat_features(sb)))
4605 ext4_msg(sb, KERN_WARNING,
4606 "feature flags set on rev 0 fs, "
4607 "running e2fsck is recommended");
4609 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4610 set_opt2(sb, HURD_COMPAT);
4611 if (ext4_has_feature_64bit(sb)) {
4612 ext4_msg(sb, KERN_ERR,
4613 "The Hurd can't support 64-bit file systems");
4618 * ea_inode feature uses l_i_version field which is not
4619 * available in HURD_COMPAT mode.
4621 if (ext4_has_feature_ea_inode(sb)) {
4622 ext4_msg(sb, KERN_ERR,
4623 "ea_inode feature is not supported for Hurd");
4628 if (IS_EXT2_SB(sb)) {
4629 if (ext2_feature_set_ok(sb))
4630 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4631 "using the ext4 subsystem");
4634 * If we're probing be silent, if this looks like
4635 * it's actually an ext[34] filesystem.
4637 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4639 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4640 "to feature incompatibilities");
4645 if (IS_EXT3_SB(sb)) {
4646 if (ext3_feature_set_ok(sb))
4647 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4648 "using the ext4 subsystem");
4651 * If we're probing be silent, if this looks like
4652 * it's actually an ext4 filesystem.
4654 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4656 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4657 "to feature incompatibilities");
4663 * Check feature flags regardless of the revision level, since we
4664 * previously didn't change the revision level when setting the flags,
4665 * so there is a chance incompat flags are set on a rev 0 filesystem.
4667 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4673 static int ext4_geometry_check(struct super_block *sb,
4674 struct ext4_super_block *es)
4676 struct ext4_sb_info *sbi = EXT4_SB(sb);
4679 /* check blocks count against device size */
4680 blocks_count = sb_bdev_nr_blocks(sb);
4681 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4682 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4683 "exceeds size of device (%llu blocks)",
4684 ext4_blocks_count(es), blocks_count);
4689 * It makes no sense for the first data block to be beyond the end
4690 * of the filesystem.
4692 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4693 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4694 "block %u is beyond end of filesystem (%llu)",
4695 le32_to_cpu(es->s_first_data_block),
4696 ext4_blocks_count(es));
4699 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4700 (sbi->s_cluster_ratio == 1)) {
4701 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4702 "block is 0 with a 1k block and cluster size");
4706 blocks_count = (ext4_blocks_count(es) -
4707 le32_to_cpu(es->s_first_data_block) +
4708 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4709 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4710 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4711 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4712 "(block count %llu, first data block %u, "
4713 "blocks per group %lu)", blocks_count,
4714 ext4_blocks_count(es),
4715 le32_to_cpu(es->s_first_data_block),
4716 EXT4_BLOCKS_PER_GROUP(sb));
4719 sbi->s_groups_count = blocks_count;
4720 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4721 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4722 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4723 le32_to_cpu(es->s_inodes_count)) {
4724 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4725 le32_to_cpu(es->s_inodes_count),
4726 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4733 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
4735 struct buffer_head **group_desc;
4739 group_desc = rcu_dereference(sbi->s_group_desc);
4740 for (i = 0; i < sbi->s_gdb_count; i++)
4741 brelse(group_desc[i]);
4746 static int ext4_group_desc_init(struct super_block *sb,
4747 struct ext4_super_block *es,
4748 ext4_fsblk_t logical_sb_block,
4749 ext4_group_t *first_not_zeroed)
4751 struct ext4_sb_info *sbi = EXT4_SB(sb);
4752 unsigned int db_count;
4756 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4757 EXT4_DESC_PER_BLOCK(sb);
4758 if (ext4_has_feature_meta_bg(sb)) {
4759 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4760 ext4_msg(sb, KERN_WARNING,
4761 "first meta block group too large: %u "
4762 "(group descriptor block count %u)",
4763 le32_to_cpu(es->s_first_meta_bg), db_count);
4767 rcu_assign_pointer(sbi->s_group_desc,
4768 kvmalloc_array(db_count,
4769 sizeof(struct buffer_head *),
4771 if (sbi->s_group_desc == NULL) {
4772 ext4_msg(sb, KERN_ERR, "not enough memory");
4776 bgl_lock_init(sbi->s_blockgroup_lock);
4778 /* Pre-read the descriptors into the buffer cache */
4779 for (i = 0; i < db_count; i++) {
4780 block = descriptor_loc(sb, logical_sb_block, i);
4781 ext4_sb_breadahead_unmovable(sb, block);
4784 for (i = 0; i < db_count; i++) {
4785 struct buffer_head *bh;
4787 block = descriptor_loc(sb, logical_sb_block, i);
4788 bh = ext4_sb_bread_unmovable(sb, block);
4790 ext4_msg(sb, KERN_ERR,
4791 "can't read group descriptor %d", i);
4792 sbi->s_gdb_count = i;
4796 rcu_dereference(sbi->s_group_desc)[i] = bh;
4799 sbi->s_gdb_count = db_count;
4800 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4801 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4802 return -EFSCORRUPTED;
4808 static int ext4_load_and_init_journal(struct super_block *sb,
4809 struct ext4_super_block *es,
4810 struct ext4_fs_context *ctx)
4812 struct ext4_sb_info *sbi = EXT4_SB(sb);
4815 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4819 if (ext4_has_feature_64bit(sb) &&
4820 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4821 JBD2_FEATURE_INCOMPAT_64BIT)) {
4822 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4826 if (!set_journal_csum_feature_set(sb)) {
4827 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4832 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4833 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4834 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4835 ext4_msg(sb, KERN_ERR,
4836 "Failed to set fast commit journal feature");
4840 /* We have now updated the journal if required, so we can
4841 * validate the data journaling mode. */
4842 switch (test_opt(sb, DATA_FLAGS)) {
4844 /* No mode set, assume a default based on the journal
4845 * capabilities: ORDERED_DATA if the journal can
4846 * cope, else JOURNAL_DATA
4848 if (jbd2_journal_check_available_features
4849 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4850 set_opt(sb, ORDERED_DATA);
4851 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4853 set_opt(sb, JOURNAL_DATA);
4854 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4858 case EXT4_MOUNT_ORDERED_DATA:
4859 case EXT4_MOUNT_WRITEBACK_DATA:
4860 if (!jbd2_journal_check_available_features
4861 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4862 ext4_msg(sb, KERN_ERR, "Journal does not support "
4863 "requested data journaling mode");
4871 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4872 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4873 ext4_msg(sb, KERN_ERR, "can't mount with "
4874 "journal_async_commit in data=ordered mode");
4878 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4880 sbi->s_journal->j_submit_inode_data_buffers =
4881 ext4_journal_submit_inode_data_buffers;
4882 sbi->s_journal->j_finish_inode_data_buffers =
4883 ext4_journal_finish_inode_data_buffers;
4888 /* flush s_error_work before journal destroy. */
4889 flush_work(&sbi->s_error_work);
4890 jbd2_journal_destroy(sbi->s_journal);
4891 sbi->s_journal = NULL;
4895 static int ext4_journal_data_mode_check(struct super_block *sb)
4897 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4898 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4899 "data=journal disables delayed allocation, "
4900 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4901 /* can't mount with both data=journal and dioread_nolock. */
4902 clear_opt(sb, DIOREAD_NOLOCK);
4903 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4904 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4905 ext4_msg(sb, KERN_ERR, "can't mount with "
4906 "both data=journal and delalloc");
4909 if (test_opt(sb, DAX_ALWAYS)) {
4910 ext4_msg(sb, KERN_ERR, "can't mount with "
4911 "both data=journal and dax");
4914 if (ext4_has_feature_encrypt(sb)) {
4915 ext4_msg(sb, KERN_WARNING,
4916 "encrypted files will use data=ordered "
4917 "instead of data journaling mode");
4919 if (test_opt(sb, DELALLOC))
4920 clear_opt(sb, DELALLOC);
4922 sb->s_iflags |= SB_I_CGROUPWB;
4928 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
4931 struct ext4_sb_info *sbi = EXT4_SB(sb);
4932 struct ext4_super_block *es;
4933 ext4_fsblk_t logical_sb_block;
4934 unsigned long offset = 0;
4935 struct buffer_head *bh;
4939 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4941 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4946 * The ext4 superblock will not be buffer aligned for other than 1kB
4947 * block sizes. We need to calculate the offset from buffer start.
4949 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4950 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4951 offset = do_div(logical_sb_block, blocksize);
4953 logical_sb_block = sbi->s_sb_block;
4956 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4958 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4962 * Note: s_es must be initialized as soon as possible because
4963 * some ext4 macro-instructions depend on its value
4965 es = (struct ext4_super_block *) (bh->b_data + offset);
4967 sb->s_magic = le16_to_cpu(es->s_magic);
4968 if (sb->s_magic != EXT4_SUPER_MAGIC) {
4970 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4974 if (le32_to_cpu(es->s_log_block_size) >
4975 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4976 ext4_msg(sb, KERN_ERR,
4977 "Invalid log block size: %u",
4978 le32_to_cpu(es->s_log_block_size));
4981 if (le32_to_cpu(es->s_log_cluster_size) >
4982 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4983 ext4_msg(sb, KERN_ERR,
4984 "Invalid log cluster size: %u",
4985 le32_to_cpu(es->s_log_cluster_size));
4989 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4992 * If the default block size is not the same as the real block size,
4993 * we need to reload it.
4995 if (sb->s_blocksize == blocksize) {
4996 *lsb = logical_sb_block;
5002 * bh must be released before kill_bdev(), otherwise
5003 * it won't be freed and its page also. kill_bdev()
5004 * is called by sb_set_blocksize().
5007 /* Validate the filesystem blocksize */
5008 if (!sb_set_blocksize(sb, blocksize)) {
5009 ext4_msg(sb, KERN_ERR, "bad block size %d",
5015 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5016 offset = do_div(logical_sb_block, blocksize);
5017 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5019 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5024 es = (struct ext4_super_block *)(bh->b_data + offset);
5026 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5027 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5030 *lsb = logical_sb_block;
5038 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5040 struct ext4_super_block *es = NULL;
5041 struct ext4_sb_info *sbi = EXT4_SB(sb);
5042 struct flex_groups **flex_groups;
5044 ext4_fsblk_t logical_sb_block;
5048 int needs_recovery, has_huge_files;
5050 ext4_group_t first_not_zeroed;
5051 struct ext4_fs_context *ctx = fc->fs_private;
5052 int silent = fc->sb_flags & SB_SILENT;
5054 /* Set defaults for the variables that will be set during parsing */
5055 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5056 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5058 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5059 sbi->s_sectors_written_start =
5060 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5062 /* -EINVAL is default */
5064 err = ext4_load_super(sb, &logical_sb_block, silent);
5069 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5071 err = ext4_init_metadata_csum(sb, es);
5075 ext4_set_def_opts(sb, es);
5077 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5078 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5079 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5080 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5081 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5084 * set default s_li_wait_mult for lazyinit, for the case there is
5085 * no mount option specified.
5087 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5089 if (ext4_inode_info_init(sb, es))
5092 err = parse_apply_sb_mount_options(sb, ctx);
5096 sbi->s_def_mount_opt = sbi->s_mount_opt;
5097 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5099 err = ext4_check_opt_consistency(fc, sb);
5103 ext4_apply_options(fc, sb);
5105 if (ext4_encoding_init(sb, es))
5108 if (ext4_journal_data_mode_check(sb))
5111 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5112 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5114 /* i_version is always enabled now */
5115 sb->s_flags |= SB_I_VERSION;
5117 if (ext4_check_feature_compatibility(sb, es, silent))
5120 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
5121 ext4_msg(sb, KERN_ERR,
5122 "Number of reserved GDT blocks insanely large: %d",
5123 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
5127 if (sbi->s_daxdev) {
5128 if (sb->s_blocksize == PAGE_SIZE)
5129 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
5131 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
5134 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
5135 if (ext4_has_feature_inline_data(sb)) {
5136 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
5137 " that may contain inline data");
5140 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
5141 ext4_msg(sb, KERN_ERR,
5142 "DAX unsupported by block device.");
5147 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
5148 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
5149 es->s_encryption_level);
5153 has_huge_files = ext4_has_feature_huge_file(sb);
5154 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5156 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5158 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5159 if (ext4_has_feature_64bit(sb)) {
5160 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5161 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5162 !is_power_of_2(sbi->s_desc_size)) {
5163 ext4_msg(sb, KERN_ERR,
5164 "unsupported descriptor size %lu",
5169 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5171 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5172 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5174 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5175 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5177 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5180 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5181 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5182 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5183 sbi->s_inodes_per_group);
5186 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5187 sbi->s_inodes_per_block;
5188 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5189 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5190 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5191 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5193 for (i = 0; i < 4; i++)
5194 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5195 sbi->s_def_hash_version = es->s_def_hash_version;
5196 if (ext4_has_feature_dir_index(sb)) {
5197 i = le32_to_cpu(es->s_flags);
5198 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5199 sbi->s_hash_unsigned = 3;
5200 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5201 #ifdef __CHAR_UNSIGNED__
5204 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5205 sbi->s_hash_unsigned = 3;
5209 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5214 if (ext4_handle_clustersize(sb))
5218 * Test whether we have more sectors than will fit in sector_t,
5219 * and whether the max offset is addressable by the page cache.
5221 err = generic_check_addressable(sb->s_blocksize_bits,
5222 ext4_blocks_count(es));
5224 ext4_msg(sb, KERN_ERR, "filesystem"
5225 " too large to mount safely on this system");
5229 if (ext4_geometry_check(sb, es))
5232 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5233 spin_lock_init(&sbi->s_error_lock);
5234 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5236 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5240 /* Register extent status tree shrinker */
5241 if (ext4_es_register_shrinker(sbi))
5244 sbi->s_stripe = ext4_get_stripe_size(sbi);
5245 sbi->s_extent_max_zeroout_kb = 32;
5248 * set up enough so that it can read an inode
5250 sb->s_op = &ext4_sops;
5251 sb->s_export_op = &ext4_export_ops;
5252 sb->s_xattr = ext4_xattr_handlers;
5253 #ifdef CONFIG_FS_ENCRYPTION
5254 sb->s_cop = &ext4_cryptops;
5256 #ifdef CONFIG_FS_VERITY
5257 sb->s_vop = &ext4_verityops;
5260 sb->dq_op = &ext4_quota_operations;
5261 if (ext4_has_feature_quota(sb))
5262 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5264 sb->s_qcop = &ext4_qctl_operations;
5265 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5267 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5269 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5270 mutex_init(&sbi->s_orphan_lock);
5272 ext4_fast_commit_init(sb);
5276 needs_recovery = (es->s_last_orphan != 0 ||
5277 ext4_has_feature_orphan_present(sb) ||
5278 ext4_has_feature_journal_needs_recovery(sb));
5280 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5281 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5282 goto failed_mount3a;
5285 * The first inode we look at is the journal inode. Don't try
5286 * root first: it may be modified in the journal!
5288 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5289 err = ext4_load_and_init_journal(sb, es, ctx);
5291 goto failed_mount3a;
5292 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5293 ext4_has_feature_journal_needs_recovery(sb)) {
5294 ext4_msg(sb, KERN_ERR, "required journal recovery "
5295 "suppressed and not mounted read-only");
5296 goto failed_mount3a;
5298 /* Nojournal mode, all journal mount options are illegal */
5299 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5300 ext4_msg(sb, KERN_ERR, "can't mount with "
5301 "journal_async_commit, fs mounted w/o journal");
5302 goto failed_mount3a;
5305 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5306 ext4_msg(sb, KERN_ERR, "can't mount with "
5307 "journal_checksum, fs mounted w/o journal");
5308 goto failed_mount3a;
5310 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5311 ext4_msg(sb, KERN_ERR, "can't mount with "
5312 "commit=%lu, fs mounted w/o journal",
5313 sbi->s_commit_interval / HZ);
5314 goto failed_mount3a;
5316 if (EXT4_MOUNT_DATA_FLAGS &
5317 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5318 ext4_msg(sb, KERN_ERR, "can't mount with "
5319 "data=, fs mounted w/o journal");
5320 goto failed_mount3a;
5322 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5323 clear_opt(sb, JOURNAL_CHECKSUM);
5324 clear_opt(sb, DATA_FLAGS);
5325 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5326 sbi->s_journal = NULL;
5330 if (!test_opt(sb, NO_MBCACHE)) {
5331 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5332 if (!sbi->s_ea_block_cache) {
5333 ext4_msg(sb, KERN_ERR,
5334 "Failed to create ea_block_cache");
5335 goto failed_mount_wq;
5338 if (ext4_has_feature_ea_inode(sb)) {
5339 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5340 if (!sbi->s_ea_inode_cache) {
5341 ext4_msg(sb, KERN_ERR,
5342 "Failed to create ea_inode_cache");
5343 goto failed_mount_wq;
5348 if (ext4_has_feature_verity(sb) && sb->s_blocksize != PAGE_SIZE) {
5349 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5350 goto failed_mount_wq;
5354 * Get the # of file system overhead blocks from the
5355 * superblock if present.
5357 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5358 /* ignore the precalculated value if it is ridiculous */
5359 if (sbi->s_overhead > ext4_blocks_count(es))
5360 sbi->s_overhead = 0;
5362 * If the bigalloc feature is not enabled recalculating the
5363 * overhead doesn't take long, so we might as well just redo
5364 * it to make sure we are using the correct value.
5366 if (!ext4_has_feature_bigalloc(sb))
5367 sbi->s_overhead = 0;
5368 if (sbi->s_overhead == 0) {
5369 err = ext4_calculate_overhead(sb);
5371 goto failed_mount_wq;
5375 * The maximum number of concurrent works can be high and
5376 * concurrency isn't really necessary. Limit it to 1.
5378 EXT4_SB(sb)->rsv_conversion_wq =
5379 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5380 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5381 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5387 * The jbd2_journal_load will have done any necessary log recovery,
5388 * so we can safely mount the rest of the filesystem now.
5391 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5393 ext4_msg(sb, KERN_ERR, "get root inode failed");
5394 ret = PTR_ERR(root);
5398 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5399 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5404 sb->s_root = d_make_root(root);
5406 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5411 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5412 if (ret == -EROFS) {
5413 sb->s_flags |= SB_RDONLY;
5416 goto failed_mount4a;
5418 ext4_set_resv_clusters(sb);
5420 if (test_opt(sb, BLOCK_VALIDITY)) {
5421 err = ext4_setup_system_zone(sb);
5423 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5425 goto failed_mount4a;
5428 ext4_fc_replay_cleanup(sb);
5433 * Enable optimize_scan if number of groups is > threshold. This can be
5434 * turned off by passing "mb_optimize_scan=0". This can also be
5435 * turned on forcefully by passing "mb_optimize_scan=1".
5437 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5438 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5439 set_opt2(sb, MB_OPTIMIZE_SCAN);
5441 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5444 err = ext4_mb_init(sb);
5446 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5452 * We can only set up the journal commit callback once
5453 * mballoc is initialized
5456 sbi->s_journal->j_commit_callback =
5457 ext4_journal_commit_callback;
5459 block = ext4_count_free_clusters(sb);
5460 ext4_free_blocks_count_set(sbi->s_es,
5461 EXT4_C2B(sbi, block));
5462 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5465 unsigned long freei = ext4_count_free_inodes(sb);
5466 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5467 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5471 err = percpu_counter_init(&sbi->s_dirs_counter,
5472 ext4_count_dirs(sb), GFP_KERNEL);
5474 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5477 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5480 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5483 ext4_msg(sb, KERN_ERR, "insufficient memory");
5487 if (ext4_has_feature_flex_bg(sb))
5488 if (!ext4_fill_flex_info(sb)) {
5489 ext4_msg(sb, KERN_ERR,
5490 "unable to initialize "
5491 "flex_bg meta info!");
5496 err = ext4_register_li_request(sb, first_not_zeroed);
5500 err = ext4_register_sysfs(sb);
5504 err = ext4_init_orphan_info(sb);
5508 /* Enable quota usage during mount. */
5509 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5510 err = ext4_enable_quotas(sb);
5514 #endif /* CONFIG_QUOTA */
5517 * Save the original bdev mapping's wb_err value which could be
5518 * used to detect the metadata async write error.
5520 spin_lock_init(&sbi->s_bdev_wb_lock);
5521 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5522 &sbi->s_bdev_wb_err);
5523 sb->s_bdev->bd_super = sb;
5524 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5525 ext4_orphan_cleanup(sb, es);
5526 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5528 * Update the checksum after updating free space/inode counters and
5529 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5530 * checksum in the buffer cache until it is written out and
5531 * e2fsprogs programs trying to open a file system immediately
5532 * after it is mounted can fail.
5534 ext4_superblock_csum_set(sb);
5535 if (needs_recovery) {
5536 ext4_msg(sb, KERN_INFO, "recovery complete");
5537 err = ext4_mark_recovery_complete(sb, es);
5542 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5543 ext4_msg(sb, KERN_WARNING,
5544 "mounting with \"discard\" option, but the device does not support discard");
5546 if (es->s_error_count)
5547 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5549 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5550 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5551 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5552 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5553 atomic_set(&sbi->s_warning_count, 0);
5554 atomic_set(&sbi->s_msg_count, 0);
5559 ext4_release_orphan_info(sb);
5561 ext4_unregister_sysfs(sb);
5562 kobject_put(&sbi->s_kobj);
5564 ext4_unregister_li_request(sb);
5566 ext4_mb_release(sb);
5568 flex_groups = rcu_dereference(sbi->s_flex_groups);
5570 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5571 kvfree(flex_groups[i]);
5572 kvfree(flex_groups);
5575 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5576 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5577 percpu_counter_destroy(&sbi->s_dirs_counter);
5578 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5579 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5580 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5582 ext4_ext_release(sb);
5583 ext4_release_system_zone(sb);
5588 ext4_msg(sb, KERN_ERR, "mount failed");
5589 if (EXT4_SB(sb)->rsv_conversion_wq)
5590 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5592 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5593 sbi->s_ea_inode_cache = NULL;
5595 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5596 sbi->s_ea_block_cache = NULL;
5598 if (sbi->s_journal) {
5599 /* flush s_error_work before journal destroy. */
5600 flush_work(&sbi->s_error_work);
5601 jbd2_journal_destroy(sbi->s_journal);
5602 sbi->s_journal = NULL;
5605 ext4_es_unregister_shrinker(sbi);
5607 /* flush s_error_work before sbi destroy */
5608 flush_work(&sbi->s_error_work);
5609 del_timer_sync(&sbi->s_err_report);
5610 ext4_stop_mmpd(sbi);
5611 ext4_group_desc_free(sbi);
5613 if (sbi->s_chksum_driver)
5614 crypto_free_shash(sbi->s_chksum_driver);
5616 #if IS_ENABLED(CONFIG_UNICODE)
5617 utf8_unload(sb->s_encoding);
5621 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5622 kfree(get_qf_name(sb, sbi, i));
5624 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5625 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5627 ext4_blkdev_remove(sbi);
5629 sb->s_fs_info = NULL;
5630 return err ? err : ret;
5633 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5635 struct ext4_fs_context *ctx = fc->fs_private;
5636 struct ext4_sb_info *sbi;
5640 sbi = ext4_alloc_sbi(sb);
5644 fc->s_fs_info = sbi;
5646 /* Cleanup superblock name */
5647 strreplace(sb->s_id, '/', '!');
5649 sbi->s_sb_block = 1; /* Default super block location */
5650 if (ctx->spec & EXT4_SPEC_s_sb_block)
5651 sbi->s_sb_block = ctx->s_sb_block;
5653 ret = __ext4_fill_super(fc, sb);
5657 if (sbi->s_journal) {
5658 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5659 descr = " journalled data mode";
5660 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5661 descr = " ordered data mode";
5663 descr = " writeback data mode";
5665 descr = "out journal";
5667 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5668 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU with%s. "
5669 "Quota mode: %s.", &sb->s_uuid, descr,
5670 ext4_quota_mode(sb));
5672 /* Update the s_overhead_clusters if necessary */
5673 ext4_update_overhead(sb, false);
5678 fc->s_fs_info = NULL;
5682 static int ext4_get_tree(struct fs_context *fc)
5684 return get_tree_bdev(fc, ext4_fill_super);
5688 * Setup any per-fs journal parameters now. We'll do this both on
5689 * initial mount, once the journal has been initialised but before we've
5690 * done any recovery; and again on any subsequent remount.
5692 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5694 struct ext4_sb_info *sbi = EXT4_SB(sb);
5696 journal->j_commit_interval = sbi->s_commit_interval;
5697 journal->j_min_batch_time = sbi->s_min_batch_time;
5698 journal->j_max_batch_time = sbi->s_max_batch_time;
5699 ext4_fc_init(sb, journal);
5701 write_lock(&journal->j_state_lock);
5702 if (test_opt(sb, BARRIER))
5703 journal->j_flags |= JBD2_BARRIER;
5705 journal->j_flags &= ~JBD2_BARRIER;
5706 if (test_opt(sb, DATA_ERR_ABORT))
5707 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5709 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5710 write_unlock(&journal->j_state_lock);
5713 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5714 unsigned int journal_inum)
5716 struct inode *journal_inode;
5719 * Test for the existence of a valid inode on disk. Bad things
5720 * happen if we iget() an unused inode, as the subsequent iput()
5721 * will try to delete it.
5723 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5724 if (IS_ERR(journal_inode)) {
5725 ext4_msg(sb, KERN_ERR, "no journal found");
5728 if (!journal_inode->i_nlink) {
5729 make_bad_inode(journal_inode);
5730 iput(journal_inode);
5731 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5735 ext4_debug("Journal inode found at %p: %lld bytes\n",
5736 journal_inode, journal_inode->i_size);
5737 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5738 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5739 iput(journal_inode);
5742 return journal_inode;
5745 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5747 struct ext4_map_blocks map;
5750 if (journal->j_inode == NULL)
5753 map.m_lblk = *block;
5755 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5757 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5758 "journal bmap failed: block %llu ret %d\n",
5760 jbd2_journal_abort(journal, ret ? ret : -EIO);
5763 *block = map.m_pblk;
5767 static journal_t *ext4_get_journal(struct super_block *sb,
5768 unsigned int journal_inum)
5770 struct inode *journal_inode;
5773 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5776 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5780 journal = jbd2_journal_init_inode(journal_inode);
5782 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5783 iput(journal_inode);
5786 journal->j_private = sb;
5787 journal->j_bmap = ext4_journal_bmap;
5788 ext4_init_journal_params(sb, journal);
5792 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5795 struct buffer_head *bh;
5799 int hblock, blocksize;
5800 ext4_fsblk_t sb_block;
5801 unsigned long offset;
5802 struct ext4_super_block *es;
5803 struct block_device *bdev;
5805 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5808 bdev = ext4_blkdev_get(j_dev, sb);
5812 blocksize = sb->s_blocksize;
5813 hblock = bdev_logical_block_size(bdev);
5814 if (blocksize < hblock) {
5815 ext4_msg(sb, KERN_ERR,
5816 "blocksize too small for journal device");
5820 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5821 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5822 set_blocksize(bdev, blocksize);
5823 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5824 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5825 "external journal");
5829 es = (struct ext4_super_block *) (bh->b_data + offset);
5830 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5831 !(le32_to_cpu(es->s_feature_incompat) &
5832 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5833 ext4_msg(sb, KERN_ERR, "external journal has "
5839 if ((le32_to_cpu(es->s_feature_ro_compat) &
5840 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5841 es->s_checksum != ext4_superblock_csum(sb, es)) {
5842 ext4_msg(sb, KERN_ERR, "external journal has "
5843 "corrupt superblock");
5848 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5849 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5854 len = ext4_blocks_count(es);
5855 start = sb_block + 1;
5856 brelse(bh); /* we're done with the superblock */
5858 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5859 start, len, blocksize);
5861 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5864 journal->j_private = sb;
5865 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5866 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5869 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5870 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5871 "user (unsupported) - %d",
5872 be32_to_cpu(journal->j_superblock->s_nr_users));
5875 EXT4_SB(sb)->s_journal_bdev = bdev;
5876 ext4_init_journal_params(sb, journal);
5880 jbd2_journal_destroy(journal);
5882 ext4_blkdev_put(bdev);
5886 static int ext4_load_journal(struct super_block *sb,
5887 struct ext4_super_block *es,
5888 unsigned long journal_devnum)
5891 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5894 int really_read_only;
5897 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5898 return -EFSCORRUPTED;
5900 if (journal_devnum &&
5901 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5902 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5903 "numbers have changed");
5904 journal_dev = new_decode_dev(journal_devnum);
5906 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5908 if (journal_inum && journal_dev) {
5909 ext4_msg(sb, KERN_ERR,
5910 "filesystem has both journal inode and journal device!");
5915 journal = ext4_get_journal(sb, journal_inum);
5919 journal = ext4_get_dev_journal(sb, journal_dev);
5924 journal_dev_ro = bdev_read_only(journal->j_dev);
5925 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5927 if (journal_dev_ro && !sb_rdonly(sb)) {
5928 ext4_msg(sb, KERN_ERR,
5929 "journal device read-only, try mounting with '-o ro'");
5935 * Are we loading a blank journal or performing recovery after a
5936 * crash? For recovery, we need to check in advance whether we
5937 * can get read-write access to the device.
5939 if (ext4_has_feature_journal_needs_recovery(sb)) {
5940 if (sb_rdonly(sb)) {
5941 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5942 "required on readonly filesystem");
5943 if (really_read_only) {
5944 ext4_msg(sb, KERN_ERR, "write access "
5945 "unavailable, cannot proceed "
5946 "(try mounting with noload)");
5950 ext4_msg(sb, KERN_INFO, "write access will "
5951 "be enabled during recovery");
5955 if (!(journal->j_flags & JBD2_BARRIER))
5956 ext4_msg(sb, KERN_INFO, "barriers disabled");
5958 if (!ext4_has_feature_journal_needs_recovery(sb))
5959 err = jbd2_journal_wipe(journal, !really_read_only);
5961 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5963 memcpy(save, ((char *) es) +
5964 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5965 err = jbd2_journal_load(journal);
5967 memcpy(((char *) es) + EXT4_S_ERR_START,
5968 save, EXT4_S_ERR_LEN);
5973 ext4_msg(sb, KERN_ERR, "error loading journal");
5977 EXT4_SB(sb)->s_journal = journal;
5978 err = ext4_clear_journal_err(sb, es);
5980 EXT4_SB(sb)->s_journal = NULL;
5981 jbd2_journal_destroy(journal);
5985 if (!really_read_only && journal_devnum &&
5986 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5987 es->s_journal_dev = cpu_to_le32(journal_devnum);
5988 ext4_commit_super(sb);
5990 if (!really_read_only && journal_inum &&
5991 journal_inum != le32_to_cpu(es->s_journal_inum)) {
5992 es->s_journal_inum = cpu_to_le32(journal_inum);
5993 ext4_commit_super(sb);
5999 jbd2_journal_destroy(journal);
6003 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6004 static void ext4_update_super(struct super_block *sb)
6006 struct ext4_sb_info *sbi = EXT4_SB(sb);
6007 struct ext4_super_block *es = sbi->s_es;
6008 struct buffer_head *sbh = sbi->s_sbh;
6012 * If the file system is mounted read-only, don't update the
6013 * superblock write time. This avoids updating the superblock
6014 * write time when we are mounting the root file system
6015 * read/only but we need to replay the journal; at that point,
6016 * for people who are east of GMT and who make their clock
6017 * tick in localtime for Windows bug-for-bug compatibility,
6018 * the clock is set in the future, and this will cause e2fsck
6019 * to complain and force a full file system check.
6021 if (!(sb->s_flags & SB_RDONLY))
6022 ext4_update_tstamp(es, s_wtime);
6023 es->s_kbytes_written =
6024 cpu_to_le64(sbi->s_kbytes_written +
6025 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6026 sbi->s_sectors_written_start) >> 1));
6027 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6028 ext4_free_blocks_count_set(es,
6029 EXT4_C2B(sbi, percpu_counter_sum_positive(
6030 &sbi->s_freeclusters_counter)));
6031 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6032 es->s_free_inodes_count =
6033 cpu_to_le32(percpu_counter_sum_positive(
6034 &sbi->s_freeinodes_counter));
6035 /* Copy error information to the on-disk superblock */
6036 spin_lock(&sbi->s_error_lock);
6037 if (sbi->s_add_error_count > 0) {
6038 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6039 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6040 __ext4_update_tstamp(&es->s_first_error_time,
6041 &es->s_first_error_time_hi,
6042 sbi->s_first_error_time);
6043 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6044 sizeof(es->s_first_error_func));
6045 es->s_first_error_line =
6046 cpu_to_le32(sbi->s_first_error_line);
6047 es->s_first_error_ino =
6048 cpu_to_le32(sbi->s_first_error_ino);
6049 es->s_first_error_block =
6050 cpu_to_le64(sbi->s_first_error_block);
6051 es->s_first_error_errcode =
6052 ext4_errno_to_code(sbi->s_first_error_code);
6054 __ext4_update_tstamp(&es->s_last_error_time,
6055 &es->s_last_error_time_hi,
6056 sbi->s_last_error_time);
6057 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6058 sizeof(es->s_last_error_func));
6059 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6060 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6061 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6062 es->s_last_error_errcode =
6063 ext4_errno_to_code(sbi->s_last_error_code);
6065 * Start the daily error reporting function if it hasn't been
6068 if (!es->s_error_count)
6069 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6070 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6071 sbi->s_add_error_count = 0;
6073 spin_unlock(&sbi->s_error_lock);
6075 ext4_superblock_csum_set(sb);
6079 static int ext4_commit_super(struct super_block *sb)
6081 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6085 if (block_device_ejected(sb))
6088 ext4_update_super(sb);
6091 /* Buffer got discarded which means block device got invalidated */
6092 if (!buffer_mapped(sbh)) {
6097 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6099 * Oh, dear. A previous attempt to write the
6100 * superblock failed. This could happen because the
6101 * USB device was yanked out. Or it could happen to
6102 * be a transient write error and maybe the block will
6103 * be remapped. Nothing we can do but to retry the
6104 * write and hope for the best.
6106 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6107 "superblock detected");
6108 clear_buffer_write_io_error(sbh);
6109 set_buffer_uptodate(sbh);
6112 /* Clear potential dirty bit if it was journalled update */
6113 clear_buffer_dirty(sbh);
6114 sbh->b_end_io = end_buffer_write_sync;
6115 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6116 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6117 wait_on_buffer(sbh);
6118 if (buffer_write_io_error(sbh)) {
6119 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6121 clear_buffer_write_io_error(sbh);
6122 set_buffer_uptodate(sbh);
6129 * Have we just finished recovery? If so, and if we are mounting (or
6130 * remounting) the filesystem readonly, then we will end up with a
6131 * consistent fs on disk. Record that fact.
6133 static int ext4_mark_recovery_complete(struct super_block *sb,
6134 struct ext4_super_block *es)
6137 journal_t *journal = EXT4_SB(sb)->s_journal;
6139 if (!ext4_has_feature_journal(sb)) {
6140 if (journal != NULL) {
6141 ext4_error(sb, "Journal got removed while the fs was "
6143 return -EFSCORRUPTED;
6147 jbd2_journal_lock_updates(journal);
6148 err = jbd2_journal_flush(journal, 0);
6152 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6153 ext4_has_feature_orphan_present(sb))) {
6154 if (!ext4_orphan_file_empty(sb)) {
6155 ext4_error(sb, "Orphan file not empty on read-only fs.");
6156 err = -EFSCORRUPTED;
6159 ext4_clear_feature_journal_needs_recovery(sb);
6160 ext4_clear_feature_orphan_present(sb);
6161 ext4_commit_super(sb);
6164 jbd2_journal_unlock_updates(journal);
6169 * If we are mounting (or read-write remounting) a filesystem whose journal
6170 * has recorded an error from a previous lifetime, move that error to the
6171 * main filesystem now.
6173 static int ext4_clear_journal_err(struct super_block *sb,
6174 struct ext4_super_block *es)
6180 if (!ext4_has_feature_journal(sb)) {
6181 ext4_error(sb, "Journal got removed while the fs was mounted!");
6182 return -EFSCORRUPTED;
6185 journal = EXT4_SB(sb)->s_journal;
6188 * Now check for any error status which may have been recorded in the
6189 * journal by a prior ext4_error() or ext4_abort()
6192 j_errno = jbd2_journal_errno(journal);
6196 errstr = ext4_decode_error(sb, j_errno, nbuf);
6197 ext4_warning(sb, "Filesystem error recorded "
6198 "from previous mount: %s", errstr);
6199 ext4_warning(sb, "Marking fs in need of filesystem check.");
6201 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6202 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6203 ext4_commit_super(sb);
6205 jbd2_journal_clear_err(journal);
6206 jbd2_journal_update_sb_errno(journal);
6212 * Force the running and committing transactions to commit,
6213 * and wait on the commit.
6215 int ext4_force_commit(struct super_block *sb)
6222 journal = EXT4_SB(sb)->s_journal;
6223 return ext4_journal_force_commit(journal);
6226 static int ext4_sync_fs(struct super_block *sb, int wait)
6230 bool needs_barrier = false;
6231 struct ext4_sb_info *sbi = EXT4_SB(sb);
6233 if (unlikely(ext4_forced_shutdown(sbi)))
6236 trace_ext4_sync_fs(sb, wait);
6237 flush_workqueue(sbi->rsv_conversion_wq);
6239 * Writeback quota in non-journalled quota case - journalled quota has
6242 dquot_writeback_dquots(sb, -1);
6244 * Data writeback is possible w/o journal transaction, so barrier must
6245 * being sent at the end of the function. But we can skip it if
6246 * transaction_commit will do it for us.
6248 if (sbi->s_journal) {
6249 target = jbd2_get_latest_transaction(sbi->s_journal);
6250 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6251 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6252 needs_barrier = true;
6254 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6256 ret = jbd2_log_wait_commit(sbi->s_journal,
6259 } else if (wait && test_opt(sb, BARRIER))
6260 needs_barrier = true;
6261 if (needs_barrier) {
6263 err = blkdev_issue_flush(sb->s_bdev);
6272 * LVM calls this function before a (read-only) snapshot is created. This
6273 * gives us a chance to flush the journal completely and mark the fs clean.
6275 * Note that only this function cannot bring a filesystem to be in a clean
6276 * state independently. It relies on upper layer to stop all data & metadata
6279 static int ext4_freeze(struct super_block *sb)
6287 journal = EXT4_SB(sb)->s_journal;
6290 /* Now we set up the journal barrier. */
6291 jbd2_journal_lock_updates(journal);
6294 * Don't clear the needs_recovery flag if we failed to
6295 * flush the journal.
6297 error = jbd2_journal_flush(journal, 0);
6301 /* Journal blocked and flushed, clear needs_recovery flag. */
6302 ext4_clear_feature_journal_needs_recovery(sb);
6303 if (ext4_orphan_file_empty(sb))
6304 ext4_clear_feature_orphan_present(sb);
6307 error = ext4_commit_super(sb);
6310 /* we rely on upper layer to stop further updates */
6311 jbd2_journal_unlock_updates(journal);
6316 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6317 * flag here, even though the filesystem is not technically dirty yet.
6319 static int ext4_unfreeze(struct super_block *sb)
6321 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6324 if (EXT4_SB(sb)->s_journal) {
6325 /* Reset the needs_recovery flag before the fs is unlocked. */
6326 ext4_set_feature_journal_needs_recovery(sb);
6327 if (ext4_has_feature_orphan_file(sb))
6328 ext4_set_feature_orphan_present(sb);
6331 ext4_commit_super(sb);
6336 * Structure to save mount options for ext4_remount's benefit
6338 struct ext4_mount_options {
6339 unsigned long s_mount_opt;
6340 unsigned long s_mount_opt2;
6343 unsigned long s_commit_interval;
6344 u32 s_min_batch_time, s_max_batch_time;
6347 char *s_qf_names[EXT4_MAXQUOTAS];
6351 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6353 struct ext4_fs_context *ctx = fc->fs_private;
6354 struct ext4_super_block *es;
6355 struct ext4_sb_info *sbi = EXT4_SB(sb);
6356 unsigned long old_sb_flags;
6357 struct ext4_mount_options old_opts;
6361 int enable_quota = 0;
6363 char *to_free[EXT4_MAXQUOTAS];
6367 /* Store the original options */
6368 old_sb_flags = sb->s_flags;
6369 old_opts.s_mount_opt = sbi->s_mount_opt;
6370 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6371 old_opts.s_resuid = sbi->s_resuid;
6372 old_opts.s_resgid = sbi->s_resgid;
6373 old_opts.s_commit_interval = sbi->s_commit_interval;
6374 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6375 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6377 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6378 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6379 if (sbi->s_qf_names[i]) {
6380 char *qf_name = get_qf_name(sb, sbi, i);
6382 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6383 if (!old_opts.s_qf_names[i]) {
6384 for (j = 0; j < i; j++)
6385 kfree(old_opts.s_qf_names[j]);
6389 old_opts.s_qf_names[i] = NULL;
6391 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6392 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6393 ctx->journal_ioprio =
6394 sbi->s_journal->j_task->io_context->ioprio;
6396 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6400 ext4_apply_options(fc, sb);
6402 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6403 test_opt(sb, JOURNAL_CHECKSUM)) {
6404 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6405 "during remount not supported; ignoring");
6406 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6409 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6410 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6411 ext4_msg(sb, KERN_ERR, "can't mount with "
6412 "both data=journal and delalloc");
6416 if (test_opt(sb, DIOREAD_NOLOCK)) {
6417 ext4_msg(sb, KERN_ERR, "can't mount with "
6418 "both data=journal and dioread_nolock");
6422 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6423 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6424 ext4_msg(sb, KERN_ERR, "can't mount with "
6425 "journal_async_commit in data=ordered mode");
6431 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6432 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6437 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6438 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6440 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6441 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6445 if (sbi->s_journal) {
6446 ext4_init_journal_params(sb, sbi->s_journal);
6447 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6450 /* Flush outstanding errors before changing fs state */
6451 flush_work(&sbi->s_error_work);
6453 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6454 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6459 if (fc->sb_flags & SB_RDONLY) {
6460 err = sync_filesystem(sb);
6463 err = dquot_suspend(sb, -1);
6468 * First of all, the unconditional stuff we have to do
6469 * to disable replay of the journal when we next remount
6471 sb->s_flags |= SB_RDONLY;
6474 * OK, test if we are remounting a valid rw partition
6475 * readonly, and if so set the rdonly flag and then
6476 * mark the partition as valid again.
6478 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6479 (sbi->s_mount_state & EXT4_VALID_FS))
6480 es->s_state = cpu_to_le16(sbi->s_mount_state);
6482 if (sbi->s_journal) {
6484 * We let remount-ro finish even if marking fs
6485 * as clean failed...
6487 ext4_mark_recovery_complete(sb, es);
6490 /* Make sure we can mount this feature set readwrite */
6491 if (ext4_has_feature_readonly(sb) ||
6492 !ext4_feature_set_ok(sb, 0)) {
6497 * Make sure the group descriptor checksums
6498 * are sane. If they aren't, refuse to remount r/w.
6500 for (g = 0; g < sbi->s_groups_count; g++) {
6501 struct ext4_group_desc *gdp =
6502 ext4_get_group_desc(sb, g, NULL);
6504 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6505 ext4_msg(sb, KERN_ERR,
6506 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6507 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6508 le16_to_cpu(gdp->bg_checksum));
6515 * If we have an unprocessed orphan list hanging
6516 * around from a previously readonly bdev mount,
6517 * require a full umount/remount for now.
6519 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6520 ext4_msg(sb, KERN_WARNING, "Couldn't "
6521 "remount RDWR because of unprocessed "
6522 "orphan inode list. Please "
6523 "umount/remount instead");
6529 * Mounting a RDONLY partition read-write, so reread
6530 * and store the current valid flag. (It may have
6531 * been changed by e2fsck since we originally mounted
6534 if (sbi->s_journal) {
6535 err = ext4_clear_journal_err(sb, es);
6539 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6542 err = ext4_setup_super(sb, es, 0);
6546 sb->s_flags &= ~SB_RDONLY;
6547 if (ext4_has_feature_mmp(sb))
6548 if (ext4_multi_mount_protect(sb,
6549 le64_to_cpu(es->s_mmp_block))) {
6560 * Reinitialize lazy itable initialization thread based on
6563 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6564 ext4_unregister_li_request(sb);
6566 ext4_group_t first_not_zeroed;
6567 first_not_zeroed = ext4_has_uninit_itable(sb);
6568 ext4_register_li_request(sb, first_not_zeroed);
6572 * Handle creation of system zone data early because it can fail.
6573 * Releasing of existing data is done when we are sure remount will
6576 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6577 err = ext4_setup_system_zone(sb);
6582 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6583 err = ext4_commit_super(sb);
6589 /* Release old quota file names */
6590 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6591 kfree(old_opts.s_qf_names[i]);
6593 if (sb_any_quota_suspended(sb))
6594 dquot_resume(sb, -1);
6595 else if (ext4_has_feature_quota(sb)) {
6596 err = ext4_enable_quotas(sb);
6602 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6603 ext4_release_system_zone(sb);
6605 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6606 ext4_stop_mmpd(sbi);
6611 sb->s_flags = old_sb_flags;
6612 sbi->s_mount_opt = old_opts.s_mount_opt;
6613 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6614 sbi->s_resuid = old_opts.s_resuid;
6615 sbi->s_resgid = old_opts.s_resgid;
6616 sbi->s_commit_interval = old_opts.s_commit_interval;
6617 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6618 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6619 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6620 ext4_release_system_zone(sb);
6622 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6623 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6624 to_free[i] = get_qf_name(sb, sbi, i);
6625 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6628 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6631 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6632 ext4_stop_mmpd(sbi);
6636 static int ext4_reconfigure(struct fs_context *fc)
6638 struct super_block *sb = fc->root->d_sb;
6641 fc->s_fs_info = EXT4_SB(sb);
6643 ret = ext4_check_opt_consistency(fc, sb);
6647 ret = __ext4_remount(fc, sb);
6651 ext4_msg(sb, KERN_INFO, "re-mounted %pU. Quota mode: %s.",
6652 &sb->s_uuid, ext4_quota_mode(sb));
6658 static int ext4_statfs_project(struct super_block *sb,
6659 kprojid_t projid, struct kstatfs *buf)
6662 struct dquot *dquot;
6666 qid = make_kqid_projid(projid);
6667 dquot = dqget(sb, qid);
6669 return PTR_ERR(dquot);
6670 spin_lock(&dquot->dq_dqb_lock);
6672 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6673 dquot->dq_dqb.dqb_bhardlimit);
6674 limit >>= sb->s_blocksize_bits;
6676 if (limit && buf->f_blocks > limit) {
6677 curblock = (dquot->dq_dqb.dqb_curspace +
6678 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6679 buf->f_blocks = limit;
6680 buf->f_bfree = buf->f_bavail =
6681 (buf->f_blocks > curblock) ?
6682 (buf->f_blocks - curblock) : 0;
6685 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6686 dquot->dq_dqb.dqb_ihardlimit);
6687 if (limit && buf->f_files > limit) {
6688 buf->f_files = limit;
6690 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6691 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6694 spin_unlock(&dquot->dq_dqb_lock);
6700 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6702 struct super_block *sb = dentry->d_sb;
6703 struct ext4_sb_info *sbi = EXT4_SB(sb);
6704 struct ext4_super_block *es = sbi->s_es;
6705 ext4_fsblk_t overhead = 0, resv_blocks;
6707 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6709 if (!test_opt(sb, MINIX_DF))
6710 overhead = sbi->s_overhead;
6712 buf->f_type = EXT4_SUPER_MAGIC;
6713 buf->f_bsize = sb->s_blocksize;
6714 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6715 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6716 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6717 /* prevent underflow in case that few free space is available */
6718 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6719 buf->f_bavail = buf->f_bfree -
6720 (ext4_r_blocks_count(es) + resv_blocks);
6721 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6723 buf->f_files = le32_to_cpu(es->s_inodes_count);
6724 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6725 buf->f_namelen = EXT4_NAME_LEN;
6726 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6729 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6730 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6731 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6740 * Helper functions so that transaction is started before we acquire dqio_sem
6741 * to keep correct lock ordering of transaction > dqio_sem
6743 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6745 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6748 static int ext4_write_dquot(struct dquot *dquot)
6752 struct inode *inode;
6754 inode = dquot_to_inode(dquot);
6755 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6756 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6758 return PTR_ERR(handle);
6759 ret = dquot_commit(dquot);
6760 err = ext4_journal_stop(handle);
6766 static int ext4_acquire_dquot(struct dquot *dquot)
6771 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6772 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6774 return PTR_ERR(handle);
6775 ret = dquot_acquire(dquot);
6776 err = ext4_journal_stop(handle);
6782 static int ext4_release_dquot(struct dquot *dquot)
6787 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6788 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6789 if (IS_ERR(handle)) {
6790 /* Release dquot anyway to avoid endless cycle in dqput() */
6791 dquot_release(dquot);
6792 return PTR_ERR(handle);
6794 ret = dquot_release(dquot);
6795 err = ext4_journal_stop(handle);
6801 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6803 struct super_block *sb = dquot->dq_sb;
6805 if (ext4_is_quota_journalled(sb)) {
6806 dquot_mark_dquot_dirty(dquot);
6807 return ext4_write_dquot(dquot);
6809 return dquot_mark_dquot_dirty(dquot);
6813 static int ext4_write_info(struct super_block *sb, int type)
6818 /* Data block + inode block */
6819 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6821 return PTR_ERR(handle);
6822 ret = dquot_commit_info(sb, type);
6823 err = ext4_journal_stop(handle);
6829 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6831 struct ext4_inode_info *ei = EXT4_I(inode);
6833 /* The first argument of lockdep_set_subclass has to be
6834 * *exactly* the same as the argument to init_rwsem() --- in
6835 * this case, in init_once() --- or lockdep gets unhappy
6836 * because the name of the lock is set using the
6837 * stringification of the argument to init_rwsem().
6839 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6840 lockdep_set_subclass(&ei->i_data_sem, subclass);
6844 * Standard function to be called on quota_on
6846 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6847 const struct path *path)
6851 if (!test_opt(sb, QUOTA))
6854 /* Quotafile not on the same filesystem? */
6855 if (path->dentry->d_sb != sb)
6858 /* Quota already enabled for this file? */
6859 if (IS_NOQUOTA(d_inode(path->dentry)))
6862 /* Journaling quota? */
6863 if (EXT4_SB(sb)->s_qf_names[type]) {
6864 /* Quotafile not in fs root? */
6865 if (path->dentry->d_parent != sb->s_root)
6866 ext4_msg(sb, KERN_WARNING,
6867 "Quota file not on filesystem root. "
6868 "Journaled quota will not work");
6869 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6872 * Clear the flag just in case mount options changed since
6875 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6879 * When we journal data on quota file, we have to flush journal to see
6880 * all updates to the file when we bypass pagecache...
6882 if (EXT4_SB(sb)->s_journal &&
6883 ext4_should_journal_data(d_inode(path->dentry))) {
6885 * We don't need to lock updates but journal_flush() could
6886 * otherwise be livelocked...
6888 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6889 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6890 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6895 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6896 err = dquot_quota_on(sb, type, format_id, path);
6898 struct inode *inode = d_inode(path->dentry);
6902 * Set inode flags to prevent userspace from messing with quota
6903 * files. If this fails, we return success anyway since quotas
6904 * are already enabled and this is not a hard failure.
6907 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6910 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6911 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6912 S_NOATIME | S_IMMUTABLE);
6913 err = ext4_mark_inode_dirty(handle, inode);
6914 ext4_journal_stop(handle);
6916 inode_unlock(inode);
6918 dquot_quota_off(sb, type);
6921 lockdep_set_quota_inode(path->dentry->d_inode,
6926 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6930 return qf_inum == EXT4_USR_QUOTA_INO;
6932 return qf_inum == EXT4_GRP_QUOTA_INO;
6934 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6940 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6944 struct inode *qf_inode;
6945 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6946 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6947 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6948 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6951 BUG_ON(!ext4_has_feature_quota(sb));
6953 if (!qf_inums[type])
6956 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6957 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6958 qf_inums[type], type);
6962 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6963 if (IS_ERR(qf_inode)) {
6964 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6965 qf_inums[type], type);
6966 return PTR_ERR(qf_inode);
6969 /* Don't account quota for quota files to avoid recursion */
6970 qf_inode->i_flags |= S_NOQUOTA;
6971 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6972 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6974 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6980 /* Enable usage tracking for all quota types. */
6981 int ext4_enable_quotas(struct super_block *sb)
6984 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6985 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6986 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6987 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6989 bool quota_mopt[EXT4_MAXQUOTAS] = {
6990 test_opt(sb, USRQUOTA),
6991 test_opt(sb, GRPQUOTA),
6992 test_opt(sb, PRJQUOTA),
6995 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6996 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6997 if (qf_inums[type]) {
6998 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6999 DQUOT_USAGE_ENABLED |
7000 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7003 "Failed to enable quota tracking "
7004 "(type=%d, err=%d, ino=%lu). "
7005 "Please run e2fsck to fix.", type,
7006 err, qf_inums[type]);
7007 for (type--; type >= 0; type--) {
7008 struct inode *inode;
7010 inode = sb_dqopt(sb)->files[type];
7012 inode = igrab(inode);
7013 dquot_quota_off(sb, type);
7015 lockdep_set_quota_inode(inode,
7028 static int ext4_quota_off(struct super_block *sb, int type)
7030 struct inode *inode = sb_dqopt(sb)->files[type];
7034 /* Force all delayed allocation blocks to be allocated.
7035 * Caller already holds s_umount sem */
7036 if (test_opt(sb, DELALLOC))
7037 sync_filesystem(sb);
7039 if (!inode || !igrab(inode))
7042 err = dquot_quota_off(sb, type);
7043 if (err || ext4_has_feature_quota(sb))
7048 * Update modification times of quota files when userspace can
7049 * start looking at them. If we fail, we return success anyway since
7050 * this is not a hard failure and quotas are already disabled.
7052 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7053 if (IS_ERR(handle)) {
7054 err = PTR_ERR(handle);
7057 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7058 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7059 inode->i_mtime = inode->i_ctime = current_time(inode);
7060 err = ext4_mark_inode_dirty(handle, inode);
7061 ext4_journal_stop(handle);
7063 inode_unlock(inode);
7065 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7069 return dquot_quota_off(sb, type);
7072 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7073 * acquiring the locks... As quota files are never truncated and quota code
7074 * itself serializes the operations (and no one else should touch the files)
7075 * we don't have to be afraid of races */
7076 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7077 size_t len, loff_t off)
7079 struct inode *inode = sb_dqopt(sb)->files[type];
7080 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7081 int offset = off & (sb->s_blocksize - 1);
7084 struct buffer_head *bh;
7085 loff_t i_size = i_size_read(inode);
7089 if (off+len > i_size)
7092 while (toread > 0) {
7093 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7094 bh = ext4_bread(NULL, inode, blk, 0);
7097 if (!bh) /* A hole? */
7098 memset(data, 0, tocopy);
7100 memcpy(data, bh->b_data+offset, tocopy);
7110 /* Write to quotafile (we know the transaction is already started and has
7111 * enough credits) */
7112 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7113 const char *data, size_t len, loff_t off)
7115 struct inode *inode = sb_dqopt(sb)->files[type];
7116 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7117 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7119 struct buffer_head *bh;
7120 handle_t *handle = journal_current_handle();
7123 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7124 " cancelled because transaction is not started",
7125 (unsigned long long)off, (unsigned long long)len);
7129 * Since we account only one data block in transaction credits,
7130 * then it is impossible to cross a block boundary.
7132 if (sb->s_blocksize - offset < len) {
7133 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7134 " cancelled because not block aligned",
7135 (unsigned long long)off, (unsigned long long)len);
7140 bh = ext4_bread(handle, inode, blk,
7141 EXT4_GET_BLOCKS_CREATE |
7142 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7143 } while (PTR_ERR(bh) == -ENOSPC &&
7144 ext4_should_retry_alloc(inode->i_sb, &retries));
7149 BUFFER_TRACE(bh, "get write access");
7150 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7156 memcpy(bh->b_data+offset, data, len);
7157 flush_dcache_page(bh->b_page);
7159 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7162 if (inode->i_size < off + len) {
7163 i_size_write(inode, off + len);
7164 EXT4_I(inode)->i_disksize = inode->i_size;
7165 err2 = ext4_mark_inode_dirty(handle, inode);
7166 if (unlikely(err2 && !err))
7169 return err ? err : len;
7173 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7174 static inline void register_as_ext2(void)
7176 int err = register_filesystem(&ext2_fs_type);
7179 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7182 static inline void unregister_as_ext2(void)
7184 unregister_filesystem(&ext2_fs_type);
7187 static inline int ext2_feature_set_ok(struct super_block *sb)
7189 if (ext4_has_unknown_ext2_incompat_features(sb))
7193 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7198 static inline void register_as_ext2(void) { }
7199 static inline void unregister_as_ext2(void) { }
7200 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7203 static inline void register_as_ext3(void)
7205 int err = register_filesystem(&ext3_fs_type);
7208 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7211 static inline void unregister_as_ext3(void)
7213 unregister_filesystem(&ext3_fs_type);
7216 static inline int ext3_feature_set_ok(struct super_block *sb)
7218 if (ext4_has_unknown_ext3_incompat_features(sb))
7220 if (!ext4_has_feature_journal(sb))
7224 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7229 static struct file_system_type ext4_fs_type = {
7230 .owner = THIS_MODULE,
7232 .init_fs_context = ext4_init_fs_context,
7233 .parameters = ext4_param_specs,
7234 .kill_sb = kill_block_super,
7235 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7237 MODULE_ALIAS_FS("ext4");
7239 /* Shared across all ext4 file systems */
7240 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7242 static int __init ext4_init_fs(void)
7246 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7247 ext4_li_info = NULL;
7249 /* Build-time check for flags consistency */
7250 ext4_check_flag_values();
7252 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7253 init_waitqueue_head(&ext4__ioend_wq[i]);
7255 err = ext4_init_es();
7259 err = ext4_init_pending();
7263 err = ext4_init_post_read_processing();
7267 err = ext4_init_pageio();
7271 err = ext4_init_system_zone();
7275 err = ext4_init_sysfs();
7279 err = ext4_init_mballoc();
7282 err = init_inodecache();
7286 err = ext4_fc_init_dentry_cache();
7292 err = register_filesystem(&ext4_fs_type);
7298 unregister_as_ext2();
7299 unregister_as_ext3();
7300 ext4_fc_destroy_dentry_cache();
7302 destroy_inodecache();
7304 ext4_exit_mballoc();
7308 ext4_exit_system_zone();
7312 ext4_exit_post_read_processing();
7314 ext4_exit_pending();
7321 static void __exit ext4_exit_fs(void)
7323 ext4_destroy_lazyinit_thread();
7324 unregister_as_ext2();
7325 unregister_as_ext3();
7326 unregister_filesystem(&ext4_fs_type);
7327 ext4_fc_destroy_dentry_cache();
7328 destroy_inodecache();
7329 ext4_exit_mballoc();
7331 ext4_exit_system_zone();
7333 ext4_exit_post_read_processing();
7335 ext4_exit_pending();
7338 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7339 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7340 MODULE_LICENSE("GPL");
7341 MODULE_SOFTDEP("pre: crc32c");
7342 module_init(ext4_init_fs)
7343 module_exit(ext4_exit_fs)