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 folio *folio,
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 = folio_buffers(folio);
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 folio_redirty_for_writepage(wbc, folio);
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);
1053 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1054 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1057 percpu_counter_sub(&sbi->s_freeclusters_counter,
1061 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1062 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1067 count = ext4_free_inodes_count(sb, gdp);
1068 percpu_counter_sub(&sbi->s_freeinodes_counter,
1074 void ext4_update_dynamic_rev(struct super_block *sb)
1076 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1078 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1082 "updating to rev %d because of new feature flag, "
1083 "running e2fsck is recommended",
1086 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1087 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1088 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1089 /* leave es->s_feature_*compat flags alone */
1090 /* es->s_uuid will be set by e2fsck if empty */
1093 * The rest of the superblock fields should be zero, and if not it
1094 * means they are likely already in use, so leave them alone. We
1095 * can leave it up to e2fsck to clean up any inconsistencies there.
1099 static void ext4_bdev_mark_dead(struct block_device *bdev)
1101 ext4_force_shutdown(bdev->bd_holder, EXT4_GOING_FLAGS_NOLOGFLUSH);
1104 static const struct blk_holder_ops ext4_holder_ops = {
1105 .mark_dead = ext4_bdev_mark_dead,
1109 * Open the external journal device
1111 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1113 struct block_device *bdev;
1115 bdev = blkdev_get_by_dev(dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb,
1122 ext4_msg(sb, KERN_ERR,
1123 "failed to open journal device unknown-block(%u,%u) %ld",
1124 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1129 * Release the journal device
1131 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1133 struct block_device *bdev;
1134 bdev = sbi->s_journal_bdev;
1137 * Invalidate the journal device's buffers. We don't want them
1138 * floating about in memory - the physical journal device may
1139 * hotswapped, and it breaks the `ro-after' testing code.
1141 invalidate_bdev(bdev);
1142 blkdev_put(bdev, sbi->s_sb);
1143 sbi->s_journal_bdev = NULL;
1147 static inline struct inode *orphan_list_entry(struct list_head *l)
1149 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1152 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1154 struct list_head *l;
1156 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1157 le32_to_cpu(sbi->s_es->s_last_orphan));
1159 printk(KERN_ERR "sb_info orphan list:\n");
1160 list_for_each(l, &sbi->s_orphan) {
1161 struct inode *inode = orphan_list_entry(l);
1163 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1164 inode->i_sb->s_id, inode->i_ino, inode,
1165 inode->i_mode, inode->i_nlink,
1166 NEXT_ORPHAN(inode));
1171 static int ext4_quota_off(struct super_block *sb, int type);
1173 static inline void ext4_quotas_off(struct super_block *sb, int type)
1175 BUG_ON(type > EXT4_MAXQUOTAS);
1177 /* Use our quota_off function to clear inode flags etc. */
1178 for (type--; type >= 0; type--)
1179 ext4_quota_off(sb, type);
1183 * This is a helper function which is used in the mount/remount
1184 * codepaths (which holds s_umount) to fetch the quota file name.
1186 static inline char *get_qf_name(struct super_block *sb,
1187 struct ext4_sb_info *sbi,
1190 return rcu_dereference_protected(sbi->s_qf_names[type],
1191 lockdep_is_held(&sb->s_umount));
1194 static inline void ext4_quotas_off(struct super_block *sb, int type)
1199 static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1204 block = ext4_count_free_clusters(sbi->s_sb);
1205 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1206 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1209 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1210 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1211 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1215 err = percpu_counter_init(&sbi->s_dirs_counter,
1216 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1218 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1221 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1224 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1227 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1232 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1234 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1235 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1236 percpu_counter_destroy(&sbi->s_dirs_counter);
1237 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1238 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1239 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1242 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1244 struct buffer_head **group_desc;
1248 group_desc = rcu_dereference(sbi->s_group_desc);
1249 for (i = 0; i < sbi->s_gdb_count; i++)
1250 brelse(group_desc[i]);
1255 static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1257 struct flex_groups **flex_groups;
1261 flex_groups = rcu_dereference(sbi->s_flex_groups);
1263 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1264 kvfree(flex_groups[i]);
1265 kvfree(flex_groups);
1270 static void ext4_put_super(struct super_block *sb)
1272 struct ext4_sb_info *sbi = EXT4_SB(sb);
1273 struct ext4_super_block *es = sbi->s_es;
1278 * Unregister sysfs before destroying jbd2 journal.
1279 * Since we could still access attr_journal_task attribute via sysfs
1280 * path which could have sbi->s_journal->j_task as NULL
1281 * Unregister sysfs before flush sbi->s_error_work.
1282 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1283 * read metadata verify failed then will queue error work.
1284 * flush_stashed_error_work will call start_this_handle may trigger
1287 ext4_unregister_sysfs(sb);
1289 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1290 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1293 ext4_unregister_li_request(sb);
1294 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1296 flush_work(&sbi->s_error_work);
1297 destroy_workqueue(sbi->rsv_conversion_wq);
1298 ext4_release_orphan_info(sb);
1300 if (sbi->s_journal) {
1301 aborted = is_journal_aborted(sbi->s_journal);
1302 err = jbd2_journal_destroy(sbi->s_journal);
1303 sbi->s_journal = NULL;
1304 if ((err < 0) && !aborted) {
1305 ext4_abort(sb, -err, "Couldn't clean up the journal");
1309 ext4_es_unregister_shrinker(sbi);
1310 timer_shutdown_sync(&sbi->s_err_report);
1311 ext4_release_system_zone(sb);
1312 ext4_mb_release(sb);
1313 ext4_ext_release(sb);
1315 if (!sb_rdonly(sb) && !aborted) {
1316 ext4_clear_feature_journal_needs_recovery(sb);
1317 ext4_clear_feature_orphan_present(sb);
1318 es->s_state = cpu_to_le16(sbi->s_mount_state);
1321 ext4_commit_super(sb);
1323 ext4_group_desc_free(sbi);
1324 ext4_flex_groups_free(sbi);
1325 ext4_percpu_param_destroy(sbi);
1327 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1328 kfree(get_qf_name(sb, sbi, i));
1331 /* Debugging code just in case the in-memory inode orphan list
1332 * isn't empty. The on-disk one can be non-empty if we've
1333 * detected an error and taken the fs readonly, but the
1334 * in-memory list had better be clean by this point. */
1335 if (!list_empty(&sbi->s_orphan))
1336 dump_orphan_list(sb, sbi);
1337 ASSERT(list_empty(&sbi->s_orphan));
1339 sync_blockdev(sb->s_bdev);
1340 invalidate_bdev(sb->s_bdev);
1341 if (sbi->s_journal_bdev) {
1342 sync_blockdev(sbi->s_journal_bdev);
1343 ext4_blkdev_remove(sbi);
1346 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1347 sbi->s_ea_inode_cache = NULL;
1349 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1350 sbi->s_ea_block_cache = NULL;
1352 ext4_stop_mmpd(sbi);
1355 sb->s_fs_info = NULL;
1357 * Now that we are completely done shutting down the
1358 * superblock, we need to actually destroy the kobject.
1360 kobject_put(&sbi->s_kobj);
1361 wait_for_completion(&sbi->s_kobj_unregister);
1362 if (sbi->s_chksum_driver)
1363 crypto_free_shash(sbi->s_chksum_driver);
1364 kfree(sbi->s_blockgroup_lock);
1365 fs_put_dax(sbi->s_daxdev, NULL);
1366 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1367 #if IS_ENABLED(CONFIG_UNICODE)
1368 utf8_unload(sb->s_encoding);
1373 static struct kmem_cache *ext4_inode_cachep;
1376 * Called inside transaction, so use GFP_NOFS
1378 static struct inode *ext4_alloc_inode(struct super_block *sb)
1380 struct ext4_inode_info *ei;
1382 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1386 inode_set_iversion(&ei->vfs_inode, 1);
1388 spin_lock_init(&ei->i_raw_lock);
1389 ei->i_prealloc_node = RB_ROOT;
1390 atomic_set(&ei->i_prealloc_active, 0);
1391 rwlock_init(&ei->i_prealloc_lock);
1392 ext4_es_init_tree(&ei->i_es_tree);
1393 rwlock_init(&ei->i_es_lock);
1394 INIT_LIST_HEAD(&ei->i_es_list);
1395 ei->i_es_all_nr = 0;
1396 ei->i_es_shk_nr = 0;
1397 ei->i_es_shrink_lblk = 0;
1398 ei->i_reserved_data_blocks = 0;
1399 spin_lock_init(&(ei->i_block_reservation_lock));
1400 ext4_init_pending_tree(&ei->i_pending_tree);
1402 ei->i_reserved_quota = 0;
1403 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1406 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1407 spin_lock_init(&ei->i_completed_io_lock);
1409 ei->i_datasync_tid = 0;
1410 atomic_set(&ei->i_unwritten, 0);
1411 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1412 ext4_fc_init_inode(&ei->vfs_inode);
1413 mutex_init(&ei->i_fc_lock);
1414 return &ei->vfs_inode;
1417 static int ext4_drop_inode(struct inode *inode)
1419 int drop = generic_drop_inode(inode);
1422 drop = fscrypt_drop_inode(inode);
1424 trace_ext4_drop_inode(inode, drop);
1428 static void ext4_free_in_core_inode(struct inode *inode)
1430 fscrypt_free_inode(inode);
1431 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1432 pr_warn("%s: inode %ld still in fc list",
1433 __func__, inode->i_ino);
1435 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1438 static void ext4_destroy_inode(struct inode *inode)
1440 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1441 ext4_msg(inode->i_sb, KERN_ERR,
1442 "Inode %lu (%p): orphan list check failed!",
1443 inode->i_ino, EXT4_I(inode));
1444 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1445 EXT4_I(inode), sizeof(struct ext4_inode_info),
1450 if (EXT4_I(inode)->i_reserved_data_blocks)
1451 ext4_msg(inode->i_sb, KERN_ERR,
1452 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1453 inode->i_ino, EXT4_I(inode),
1454 EXT4_I(inode)->i_reserved_data_blocks);
1457 static void ext4_shutdown(struct super_block *sb)
1459 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1462 static void init_once(void *foo)
1464 struct ext4_inode_info *ei = foo;
1466 INIT_LIST_HEAD(&ei->i_orphan);
1467 init_rwsem(&ei->xattr_sem);
1468 init_rwsem(&ei->i_data_sem);
1469 inode_init_once(&ei->vfs_inode);
1470 ext4_fc_init_inode(&ei->vfs_inode);
1473 static int __init init_inodecache(void)
1475 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1476 sizeof(struct ext4_inode_info), 0,
1477 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1479 offsetof(struct ext4_inode_info, i_data),
1480 sizeof_field(struct ext4_inode_info, i_data),
1482 if (ext4_inode_cachep == NULL)
1487 static void destroy_inodecache(void)
1490 * Make sure all delayed rcu free inodes are flushed before we
1494 kmem_cache_destroy(ext4_inode_cachep);
1497 void ext4_clear_inode(struct inode *inode)
1500 invalidate_inode_buffers(inode);
1502 ext4_discard_preallocations(inode, 0);
1503 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1505 if (EXT4_I(inode)->jinode) {
1506 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1507 EXT4_I(inode)->jinode);
1508 jbd2_free_inode(EXT4_I(inode)->jinode);
1509 EXT4_I(inode)->jinode = NULL;
1511 fscrypt_put_encryption_info(inode);
1512 fsverity_cleanup_inode(inode);
1515 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1516 u64 ino, u32 generation)
1518 struct inode *inode;
1521 * Currently we don't know the generation for parent directory, so
1522 * a generation of 0 means "accept any"
1524 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1526 return ERR_CAST(inode);
1527 if (generation && inode->i_generation != generation) {
1529 return ERR_PTR(-ESTALE);
1535 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1536 int fh_len, int fh_type)
1538 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1539 ext4_nfs_get_inode);
1542 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1543 int fh_len, int fh_type)
1545 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1546 ext4_nfs_get_inode);
1549 static int ext4_nfs_commit_metadata(struct inode *inode)
1551 struct writeback_control wbc = {
1552 .sync_mode = WB_SYNC_ALL
1555 trace_ext4_nfs_commit_metadata(inode);
1556 return ext4_write_inode(inode, &wbc);
1560 static const char * const quotatypes[] = INITQFNAMES;
1561 #define QTYPE2NAME(t) (quotatypes[t])
1563 static int ext4_write_dquot(struct dquot *dquot);
1564 static int ext4_acquire_dquot(struct dquot *dquot);
1565 static int ext4_release_dquot(struct dquot *dquot);
1566 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1567 static int ext4_write_info(struct super_block *sb, int type);
1568 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1569 const struct path *path);
1570 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1571 size_t len, loff_t off);
1572 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1573 const char *data, size_t len, loff_t off);
1574 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1575 unsigned int flags);
1577 static struct dquot **ext4_get_dquots(struct inode *inode)
1579 return EXT4_I(inode)->i_dquot;
1582 static const struct dquot_operations ext4_quota_operations = {
1583 .get_reserved_space = ext4_get_reserved_space,
1584 .write_dquot = ext4_write_dquot,
1585 .acquire_dquot = ext4_acquire_dquot,
1586 .release_dquot = ext4_release_dquot,
1587 .mark_dirty = ext4_mark_dquot_dirty,
1588 .write_info = ext4_write_info,
1589 .alloc_dquot = dquot_alloc,
1590 .destroy_dquot = dquot_destroy,
1591 .get_projid = ext4_get_projid,
1592 .get_inode_usage = ext4_get_inode_usage,
1593 .get_next_id = dquot_get_next_id,
1596 static const struct quotactl_ops ext4_qctl_operations = {
1597 .quota_on = ext4_quota_on,
1598 .quota_off = ext4_quota_off,
1599 .quota_sync = dquot_quota_sync,
1600 .get_state = dquot_get_state,
1601 .set_info = dquot_set_dqinfo,
1602 .get_dqblk = dquot_get_dqblk,
1603 .set_dqblk = dquot_set_dqblk,
1604 .get_nextdqblk = dquot_get_next_dqblk,
1608 static const struct super_operations ext4_sops = {
1609 .alloc_inode = ext4_alloc_inode,
1610 .free_inode = ext4_free_in_core_inode,
1611 .destroy_inode = ext4_destroy_inode,
1612 .write_inode = ext4_write_inode,
1613 .dirty_inode = ext4_dirty_inode,
1614 .drop_inode = ext4_drop_inode,
1615 .evict_inode = ext4_evict_inode,
1616 .put_super = ext4_put_super,
1617 .sync_fs = ext4_sync_fs,
1618 .freeze_fs = ext4_freeze,
1619 .unfreeze_fs = ext4_unfreeze,
1620 .statfs = ext4_statfs,
1621 .show_options = ext4_show_options,
1622 .shutdown = ext4_shutdown,
1624 .quota_read = ext4_quota_read,
1625 .quota_write = ext4_quota_write,
1626 .get_dquots = ext4_get_dquots,
1630 static const struct export_operations ext4_export_ops = {
1631 .fh_to_dentry = ext4_fh_to_dentry,
1632 .fh_to_parent = ext4_fh_to_parent,
1633 .get_parent = ext4_get_parent,
1634 .commit_metadata = ext4_nfs_commit_metadata,
1638 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1639 Opt_resgid, Opt_resuid, Opt_sb,
1640 Opt_nouid32, Opt_debug, Opt_removed,
1641 Opt_user_xattr, Opt_acl,
1642 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1643 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1644 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1645 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1646 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1648 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1649 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1650 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1651 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1652 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1653 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1654 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1655 Opt_inode_readahead_blks, Opt_journal_ioprio,
1656 Opt_dioread_nolock, Opt_dioread_lock,
1657 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1658 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1659 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1660 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1661 #ifdef CONFIG_EXT4_DEBUG
1662 Opt_fc_debug_max_replay, Opt_fc_debug_force
1666 static const struct constant_table ext4_param_errors[] = {
1667 {"continue", EXT4_MOUNT_ERRORS_CONT},
1668 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1669 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1673 static const struct constant_table ext4_param_data[] = {
1674 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1675 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1676 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1680 static const struct constant_table ext4_param_data_err[] = {
1681 {"abort", Opt_data_err_abort},
1682 {"ignore", Opt_data_err_ignore},
1686 static const struct constant_table ext4_param_jqfmt[] = {
1687 {"vfsold", QFMT_VFS_OLD},
1688 {"vfsv0", QFMT_VFS_V0},
1689 {"vfsv1", QFMT_VFS_V1},
1693 static const struct constant_table ext4_param_dax[] = {
1694 {"always", Opt_dax_always},
1695 {"inode", Opt_dax_inode},
1696 {"never", Opt_dax_never},
1700 /* String parameter that allows empty argument */
1701 #define fsparam_string_empty(NAME, OPT) \
1702 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1705 * Mount option specification
1706 * We don't use fsparam_flag_no because of the way we set the
1707 * options and the way we show them in _ext4_show_options(). To
1708 * keep the changes to a minimum, let's keep the negative options
1711 static const struct fs_parameter_spec ext4_param_specs[] = {
1712 fsparam_flag ("bsddf", Opt_bsd_df),
1713 fsparam_flag ("minixdf", Opt_minix_df),
1714 fsparam_flag ("grpid", Opt_grpid),
1715 fsparam_flag ("bsdgroups", Opt_grpid),
1716 fsparam_flag ("nogrpid", Opt_nogrpid),
1717 fsparam_flag ("sysvgroups", Opt_nogrpid),
1718 fsparam_u32 ("resgid", Opt_resgid),
1719 fsparam_u32 ("resuid", Opt_resuid),
1720 fsparam_u32 ("sb", Opt_sb),
1721 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1722 fsparam_flag ("nouid32", Opt_nouid32),
1723 fsparam_flag ("debug", Opt_debug),
1724 fsparam_flag ("oldalloc", Opt_removed),
1725 fsparam_flag ("orlov", Opt_removed),
1726 fsparam_flag ("user_xattr", Opt_user_xattr),
1727 fsparam_flag ("acl", Opt_acl),
1728 fsparam_flag ("norecovery", Opt_noload),
1729 fsparam_flag ("noload", Opt_noload),
1730 fsparam_flag ("bh", Opt_removed),
1731 fsparam_flag ("nobh", Opt_removed),
1732 fsparam_u32 ("commit", Opt_commit),
1733 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1734 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1735 fsparam_u32 ("journal_dev", Opt_journal_dev),
1736 fsparam_bdev ("journal_path", Opt_journal_path),
1737 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1738 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1739 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1740 fsparam_flag ("abort", Opt_abort),
1741 fsparam_enum ("data", Opt_data, ext4_param_data),
1742 fsparam_enum ("data_err", Opt_data_err,
1743 ext4_param_data_err),
1744 fsparam_string_empty
1745 ("usrjquota", Opt_usrjquota),
1746 fsparam_string_empty
1747 ("grpjquota", Opt_grpjquota),
1748 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1749 fsparam_flag ("grpquota", Opt_grpquota),
1750 fsparam_flag ("quota", Opt_quota),
1751 fsparam_flag ("noquota", Opt_noquota),
1752 fsparam_flag ("usrquota", Opt_usrquota),
1753 fsparam_flag ("prjquota", Opt_prjquota),
1754 fsparam_flag ("barrier", Opt_barrier),
1755 fsparam_u32 ("barrier", Opt_barrier),
1756 fsparam_flag ("nobarrier", Opt_nobarrier),
1757 fsparam_flag ("i_version", Opt_removed),
1758 fsparam_flag ("dax", Opt_dax),
1759 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1760 fsparam_u32 ("stripe", Opt_stripe),
1761 fsparam_flag ("delalloc", Opt_delalloc),
1762 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1763 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1764 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1765 fsparam_u32 ("debug_want_extra_isize",
1766 Opt_debug_want_extra_isize),
1767 fsparam_flag ("mblk_io_submit", Opt_removed),
1768 fsparam_flag ("nomblk_io_submit", Opt_removed),
1769 fsparam_flag ("block_validity", Opt_block_validity),
1770 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1771 fsparam_u32 ("inode_readahead_blks",
1772 Opt_inode_readahead_blks),
1773 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1774 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1775 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1776 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1777 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1778 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1779 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1780 fsparam_flag ("discard", Opt_discard),
1781 fsparam_flag ("nodiscard", Opt_nodiscard),
1782 fsparam_u32 ("init_itable", Opt_init_itable),
1783 fsparam_flag ("init_itable", Opt_init_itable),
1784 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1785 #ifdef CONFIG_EXT4_DEBUG
1786 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1787 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1789 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1790 fsparam_flag ("test_dummy_encryption",
1791 Opt_test_dummy_encryption),
1792 fsparam_string ("test_dummy_encryption",
1793 Opt_test_dummy_encryption),
1794 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1795 fsparam_flag ("nombcache", Opt_nombcache),
1796 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1797 fsparam_flag ("prefetch_block_bitmaps",
1799 fsparam_flag ("no_prefetch_block_bitmaps",
1800 Opt_no_prefetch_block_bitmaps),
1801 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1802 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1803 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1804 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1805 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1806 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1810 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1812 #define MOPT_SET 0x0001
1813 #define MOPT_CLEAR 0x0002
1814 #define MOPT_NOSUPPORT 0x0004
1815 #define MOPT_EXPLICIT 0x0008
1818 #define MOPT_QFMT 0x0010
1820 #define MOPT_Q MOPT_NOSUPPORT
1821 #define MOPT_QFMT MOPT_NOSUPPORT
1823 #define MOPT_NO_EXT2 0x0020
1824 #define MOPT_NO_EXT3 0x0040
1825 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1826 #define MOPT_SKIP 0x0080
1827 #define MOPT_2 0x0100
1829 static const struct mount_opts {
1833 } ext4_mount_opts[] = {
1834 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1835 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1836 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1837 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1838 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1839 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1840 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1841 MOPT_EXT4_ONLY | MOPT_SET},
1842 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1843 MOPT_EXT4_ONLY | MOPT_CLEAR},
1844 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1845 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1846 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1847 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1848 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1849 MOPT_EXT4_ONLY | MOPT_CLEAR},
1850 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1851 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1852 {Opt_commit, 0, MOPT_NO_EXT2},
1853 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1854 MOPT_EXT4_ONLY | MOPT_CLEAR},
1855 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1856 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1857 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1858 EXT4_MOUNT_JOURNAL_CHECKSUM),
1859 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1860 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1861 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1862 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1863 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1864 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1865 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1866 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1867 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1868 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1869 {Opt_journal_path, 0, MOPT_NO_EXT2},
1870 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1871 {Opt_data, 0, MOPT_NO_EXT2},
1872 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1873 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1874 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1876 {Opt_acl, 0, MOPT_NOSUPPORT},
1878 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1879 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1880 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1881 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1883 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1885 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1887 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1888 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1889 MOPT_CLEAR | MOPT_Q},
1890 {Opt_usrjquota, 0, MOPT_Q},
1891 {Opt_grpjquota, 0, MOPT_Q},
1892 {Opt_jqfmt, 0, MOPT_QFMT},
1893 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1894 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1896 #ifdef CONFIG_EXT4_DEBUG
1897 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1898 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1903 #if IS_ENABLED(CONFIG_UNICODE)
1904 static const struct ext4_sb_encodings {
1907 unsigned int version;
1908 } ext4_sb_encoding_map[] = {
1909 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1912 static const struct ext4_sb_encodings *
1913 ext4_sb_read_encoding(const struct ext4_super_block *es)
1915 __u16 magic = le16_to_cpu(es->s_encoding);
1918 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1919 if (magic == ext4_sb_encoding_map[i].magic)
1920 return &ext4_sb_encoding_map[i];
1926 #define EXT4_SPEC_JQUOTA (1 << 0)
1927 #define EXT4_SPEC_JQFMT (1 << 1)
1928 #define EXT4_SPEC_DATAJ (1 << 2)
1929 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1930 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1931 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1932 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1933 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1934 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1935 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1936 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1937 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1938 #define EXT4_SPEC_s_stripe (1 << 13)
1939 #define EXT4_SPEC_s_resuid (1 << 14)
1940 #define EXT4_SPEC_s_resgid (1 << 15)
1941 #define EXT4_SPEC_s_commit_interval (1 << 16)
1942 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1943 #define EXT4_SPEC_s_sb_block (1 << 18)
1944 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1946 struct ext4_fs_context {
1947 char *s_qf_names[EXT4_MAXQUOTAS];
1948 struct fscrypt_dummy_policy dummy_enc_policy;
1949 int s_jquota_fmt; /* Format of quota to use */
1950 #ifdef CONFIG_EXT4_DEBUG
1951 int s_fc_debug_max_replay;
1953 unsigned short qname_spec;
1954 unsigned long vals_s_flags; /* Bits to set in s_flags */
1955 unsigned long mask_s_flags; /* Bits changed in s_flags */
1956 unsigned long journal_devnum;
1957 unsigned long s_commit_interval;
1958 unsigned long s_stripe;
1959 unsigned int s_inode_readahead_blks;
1960 unsigned int s_want_extra_isize;
1961 unsigned int s_li_wait_mult;
1962 unsigned int s_max_dir_size_kb;
1963 unsigned int journal_ioprio;
1964 unsigned int vals_s_mount_opt;
1965 unsigned int mask_s_mount_opt;
1966 unsigned int vals_s_mount_opt2;
1967 unsigned int mask_s_mount_opt2;
1968 unsigned long vals_s_mount_flags;
1969 unsigned long mask_s_mount_flags;
1970 unsigned int opt_flags; /* MOPT flags */
1972 u32 s_max_batch_time;
1973 u32 s_min_batch_time;
1976 ext4_fsblk_t s_sb_block;
1979 static void ext4_fc_free(struct fs_context *fc)
1981 struct ext4_fs_context *ctx = fc->fs_private;
1987 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1988 kfree(ctx->s_qf_names[i]);
1990 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1994 int ext4_init_fs_context(struct fs_context *fc)
1996 struct ext4_fs_context *ctx;
1998 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2002 fc->fs_private = ctx;
2003 fc->ops = &ext4_context_ops;
2010 * Note the name of the specified quota file.
2012 static int note_qf_name(struct fs_context *fc, int qtype,
2013 struct fs_parameter *param)
2015 struct ext4_fs_context *ctx = fc->fs_private;
2018 if (param->size < 1) {
2019 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2022 if (strchr(param->string, '/')) {
2023 ext4_msg(NULL, KERN_ERR,
2024 "quotafile must be on filesystem root");
2027 if (ctx->s_qf_names[qtype]) {
2028 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2029 ext4_msg(NULL, KERN_ERR,
2030 "%s quota file already specified",
2037 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2039 ext4_msg(NULL, KERN_ERR,
2040 "Not enough memory for storing quotafile name");
2043 ctx->s_qf_names[qtype] = qname;
2044 ctx->qname_spec |= 1 << qtype;
2045 ctx->spec |= EXT4_SPEC_JQUOTA;
2050 * Clear the name of the specified quota file.
2052 static int unnote_qf_name(struct fs_context *fc, int qtype)
2054 struct ext4_fs_context *ctx = fc->fs_private;
2056 if (ctx->s_qf_names[qtype])
2057 kfree(ctx->s_qf_names[qtype]);
2059 ctx->s_qf_names[qtype] = NULL;
2060 ctx->qname_spec |= 1 << qtype;
2061 ctx->spec |= EXT4_SPEC_JQUOTA;
2066 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2067 struct ext4_fs_context *ctx)
2071 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2072 ext4_msg(NULL, KERN_WARNING,
2073 "test_dummy_encryption option not supported");
2076 err = fscrypt_parse_test_dummy_encryption(param,
2077 &ctx->dummy_enc_policy);
2078 if (err == -EINVAL) {
2079 ext4_msg(NULL, KERN_WARNING,
2080 "Value of option \"%s\" is unrecognized", param->key);
2081 } else if (err == -EEXIST) {
2082 ext4_msg(NULL, KERN_WARNING,
2083 "Conflicting test_dummy_encryption options");
2089 #define EXT4_SET_CTX(name) \
2090 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2091 unsigned long flag) \
2093 ctx->mask_s_##name |= flag; \
2094 ctx->vals_s_##name |= flag; \
2097 #define EXT4_CLEAR_CTX(name) \
2098 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2099 unsigned long flag) \
2101 ctx->mask_s_##name |= flag; \
2102 ctx->vals_s_##name &= ~flag; \
2105 #define EXT4_TEST_CTX(name) \
2106 static inline unsigned long \
2107 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2109 return (ctx->vals_s_##name & flag); \
2112 EXT4_SET_CTX(flags); /* set only */
2113 EXT4_SET_CTX(mount_opt);
2114 EXT4_CLEAR_CTX(mount_opt);
2115 EXT4_TEST_CTX(mount_opt);
2116 EXT4_SET_CTX(mount_opt2);
2117 EXT4_CLEAR_CTX(mount_opt2);
2118 EXT4_TEST_CTX(mount_opt2);
2120 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2122 set_bit(bit, &ctx->mask_s_mount_flags);
2123 set_bit(bit, &ctx->vals_s_mount_flags);
2126 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2128 struct ext4_fs_context *ctx = fc->fs_private;
2129 struct fs_parse_result result;
2130 const struct mount_opts *m;
2136 token = fs_parse(fc, ext4_param_specs, param, &result);
2139 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2141 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2142 if (token == m->token)
2145 ctx->opt_flags |= m->flags;
2147 if (m->flags & MOPT_EXPLICIT) {
2148 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2149 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2150 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2151 ctx_set_mount_opt2(ctx,
2152 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2157 if (m->flags & MOPT_NOSUPPORT) {
2158 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2166 if (!*param->string)
2167 return unnote_qf_name(fc, USRQUOTA);
2169 return note_qf_name(fc, USRQUOTA, param);
2171 if (!*param->string)
2172 return unnote_qf_name(fc, GRPQUOTA);
2174 return note_qf_name(fc, GRPQUOTA, param);
2177 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2178 ext4_msg(NULL, KERN_WARNING,
2179 "Ignoring %s option on remount", param->key);
2181 ctx->s_sb_block = result.uint_32;
2182 ctx->spec |= EXT4_SPEC_s_sb_block;
2186 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2190 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2192 case Opt_inlinecrypt:
2193 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2194 ctx_set_flags(ctx, SB_INLINECRYPT);
2196 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2200 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2201 ctx_set_mount_opt(ctx, result.uint_32);
2205 ctx->s_jquota_fmt = result.uint_32;
2206 ctx->spec |= EXT4_SPEC_JQFMT;
2210 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2211 ctx_set_mount_opt(ctx, result.uint_32);
2212 ctx->spec |= EXT4_SPEC_DATAJ;
2215 if (result.uint_32 == 0)
2216 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2217 else if (result.uint_32 > INT_MAX / HZ) {
2218 ext4_msg(NULL, KERN_ERR,
2219 "Invalid commit interval %d, "
2220 "must be smaller than %d",
2221 result.uint_32, INT_MAX / HZ);
2224 ctx->s_commit_interval = HZ * result.uint_32;
2225 ctx->spec |= EXT4_SPEC_s_commit_interval;
2227 case Opt_debug_want_extra_isize:
2228 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2229 ext4_msg(NULL, KERN_ERR,
2230 "Invalid want_extra_isize %d", result.uint_32);
2233 ctx->s_want_extra_isize = result.uint_32;
2234 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2236 case Opt_max_batch_time:
2237 ctx->s_max_batch_time = result.uint_32;
2238 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2240 case Opt_min_batch_time:
2241 ctx->s_min_batch_time = result.uint_32;
2242 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2244 case Opt_inode_readahead_blks:
2245 if (result.uint_32 &&
2246 (result.uint_32 > (1 << 30) ||
2247 !is_power_of_2(result.uint_32))) {
2248 ext4_msg(NULL, KERN_ERR,
2249 "EXT4-fs: inode_readahead_blks must be "
2250 "0 or a power of 2 smaller than 2^31");
2253 ctx->s_inode_readahead_blks = result.uint_32;
2254 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2256 case Opt_init_itable:
2257 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2258 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2259 if (param->type == fs_value_is_string)
2260 ctx->s_li_wait_mult = result.uint_32;
2261 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2263 case Opt_max_dir_size_kb:
2264 ctx->s_max_dir_size_kb = result.uint_32;
2265 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2267 #ifdef CONFIG_EXT4_DEBUG
2268 case Opt_fc_debug_max_replay:
2269 ctx->s_fc_debug_max_replay = result.uint_32;
2270 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2274 ctx->s_stripe = result.uint_32;
2275 ctx->spec |= EXT4_SPEC_s_stripe;
2278 uid = make_kuid(current_user_ns(), result.uint_32);
2279 if (!uid_valid(uid)) {
2280 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2284 ctx->s_resuid = uid;
2285 ctx->spec |= EXT4_SPEC_s_resuid;
2288 gid = make_kgid(current_user_ns(), result.uint_32);
2289 if (!gid_valid(gid)) {
2290 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2294 ctx->s_resgid = gid;
2295 ctx->spec |= EXT4_SPEC_s_resgid;
2297 case Opt_journal_dev:
2299 ext4_msg(NULL, KERN_ERR,
2300 "Cannot specify journal on remount");
2303 ctx->journal_devnum = result.uint_32;
2304 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2306 case Opt_journal_path:
2308 struct inode *journal_inode;
2313 ext4_msg(NULL, KERN_ERR,
2314 "Cannot specify journal on remount");
2318 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2320 ext4_msg(NULL, KERN_ERR, "error: could not find "
2321 "journal device path");
2325 journal_inode = d_inode(path.dentry);
2326 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2327 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2331 case Opt_journal_ioprio:
2332 if (result.uint_32 > 7) {
2333 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2337 ctx->journal_ioprio =
2338 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2339 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2341 case Opt_test_dummy_encryption:
2342 return ext4_parse_test_dummy_encryption(param, ctx);
2345 #ifdef CONFIG_FS_DAX
2347 int type = (token == Opt_dax) ?
2348 Opt_dax : result.uint_32;
2352 case Opt_dax_always:
2353 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2354 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2357 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2358 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2361 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2362 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2363 /* Strictly for printing options */
2364 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2370 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2374 if (result.uint_32 == Opt_data_err_abort)
2375 ctx_set_mount_opt(ctx, m->mount_opt);
2376 else if (result.uint_32 == Opt_data_err_ignore)
2377 ctx_clear_mount_opt(ctx, m->mount_opt);
2379 case Opt_mb_optimize_scan:
2380 if (result.int_32 == 1) {
2381 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2382 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2383 } else if (result.int_32 == 0) {
2384 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2385 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2387 ext4_msg(NULL, KERN_WARNING,
2388 "mb_optimize_scan should be set to 0 or 1.");
2395 * At this point we should only be getting options requiring MOPT_SET,
2396 * or MOPT_CLEAR. Anything else is a bug
2398 if (m->token == Opt_err) {
2399 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2406 unsigned int set = 0;
2408 if ((param->type == fs_value_is_flag) ||
2412 if (m->flags & MOPT_CLEAR)
2414 else if (unlikely(!(m->flags & MOPT_SET))) {
2415 ext4_msg(NULL, KERN_WARNING,
2416 "buggy handling of option %s",
2421 if (m->flags & MOPT_2) {
2423 ctx_set_mount_opt2(ctx, m->mount_opt);
2425 ctx_clear_mount_opt2(ctx, m->mount_opt);
2428 ctx_set_mount_opt(ctx, m->mount_opt);
2430 ctx_clear_mount_opt(ctx, m->mount_opt);
2437 static int parse_options(struct fs_context *fc, char *options)
2439 struct fs_parameter param;
2446 while ((key = strsep(&options, ",")) != NULL) {
2449 char *value = strchr(key, '=');
2451 param.type = fs_value_is_flag;
2452 param.string = NULL;
2459 v_len = strlen(value);
2460 param.string = kmemdup_nul(value, v_len,
2464 param.type = fs_value_is_string;
2470 ret = ext4_parse_param(fc, ¶m);
2472 kfree(param.string);
2478 ret = ext4_validate_options(fc);
2485 static int parse_apply_sb_mount_options(struct super_block *sb,
2486 struct ext4_fs_context *m_ctx)
2488 struct ext4_sb_info *sbi = EXT4_SB(sb);
2489 char *s_mount_opts = NULL;
2490 struct ext4_fs_context *s_ctx = NULL;
2491 struct fs_context *fc = NULL;
2494 if (!sbi->s_es->s_mount_opts[0])
2497 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2498 sizeof(sbi->s_es->s_mount_opts),
2503 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2507 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2511 fc->fs_private = s_ctx;
2512 fc->s_fs_info = sbi;
2514 ret = parse_options(fc, s_mount_opts);
2518 ret = ext4_check_opt_consistency(fc, sb);
2521 ext4_msg(sb, KERN_WARNING,
2522 "failed to parse options in superblock: %s",
2528 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2529 m_ctx->journal_devnum = s_ctx->journal_devnum;
2530 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2531 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2533 ext4_apply_options(fc, sb);
2541 kfree(s_mount_opts);
2545 static void ext4_apply_quota_options(struct fs_context *fc,
2546 struct super_block *sb)
2549 bool quota_feature = ext4_has_feature_quota(sb);
2550 struct ext4_fs_context *ctx = fc->fs_private;
2551 struct ext4_sb_info *sbi = EXT4_SB(sb);
2558 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2559 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2560 if (!(ctx->qname_spec & (1 << i)))
2563 qname = ctx->s_qf_names[i]; /* May be NULL */
2566 ctx->s_qf_names[i] = NULL;
2567 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2568 lockdep_is_held(&sb->s_umount));
2570 kfree_rcu_mightsleep(qname);
2574 if (ctx->spec & EXT4_SPEC_JQFMT)
2575 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2580 * Check quota settings consistency.
2582 static int ext4_check_quota_consistency(struct fs_context *fc,
2583 struct super_block *sb)
2586 struct ext4_fs_context *ctx = fc->fs_private;
2587 struct ext4_sb_info *sbi = EXT4_SB(sb);
2588 bool quota_feature = ext4_has_feature_quota(sb);
2589 bool quota_loaded = sb_any_quota_loaded(sb);
2590 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2594 * We do the test below only for project quotas. 'usrquota' and
2595 * 'grpquota' mount options are allowed even without quota feature
2596 * to support legacy quotas in quota files.
2598 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2599 !ext4_has_feature_project(sb)) {
2600 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2601 "Cannot enable project quota enforcement.");
2605 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2606 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2608 ctx->mask_s_mount_opt & quota_flags &&
2609 !ctx_test_mount_opt(ctx, quota_flags))
2610 goto err_quota_change;
2612 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2614 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2615 if (!(ctx->qname_spec & (1 << i)))
2619 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2620 goto err_jquota_change;
2622 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2623 strcmp(get_qf_name(sb, sbi, i),
2624 ctx->s_qf_names[i]) != 0)
2625 goto err_jquota_specified;
2628 if (quota_feature) {
2629 ext4_msg(NULL, KERN_INFO,
2630 "Journaled quota options ignored when "
2631 "QUOTA feature is enabled");
2636 if (ctx->spec & EXT4_SPEC_JQFMT) {
2637 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2638 goto err_jquota_change;
2639 if (quota_feature) {
2640 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2641 "ignored when QUOTA feature is enabled");
2646 /* Make sure we don't mix old and new quota format */
2647 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2648 ctx->s_qf_names[USRQUOTA]);
2649 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2650 ctx->s_qf_names[GRPQUOTA]);
2652 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2653 test_opt(sb, USRQUOTA));
2655 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2656 test_opt(sb, GRPQUOTA));
2659 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2663 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2667 if (usr_qf_name || grp_qf_name) {
2668 if (usrquota || grpquota) {
2669 ext4_msg(NULL, KERN_ERR, "old and new quota "
2674 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2675 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2684 ext4_msg(NULL, KERN_ERR,
2685 "Cannot change quota options when quota turned on");
2688 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2689 "options when quota turned on");
2691 err_jquota_specified:
2692 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2700 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2701 struct super_block *sb)
2703 const struct ext4_fs_context *ctx = fc->fs_private;
2704 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2706 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2709 if (!ext4_has_feature_encrypt(sb)) {
2710 ext4_msg(NULL, KERN_WARNING,
2711 "test_dummy_encryption requires encrypt feature");
2715 * This mount option is just for testing, and it's not worthwhile to
2716 * implement the extra complexity (e.g. RCU protection) that would be
2717 * needed to allow it to be set or changed during remount. We do allow
2718 * it to be specified during remount, but only if there is no change.
2720 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2721 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2722 &ctx->dummy_enc_policy))
2724 ext4_msg(NULL, KERN_WARNING,
2725 "Can't set or change test_dummy_encryption on remount");
2728 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2729 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2730 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2731 &ctx->dummy_enc_policy))
2733 ext4_msg(NULL, KERN_WARNING,
2734 "Conflicting test_dummy_encryption options");
2740 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2741 struct super_block *sb)
2743 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2744 /* if already set, it was already verified to be the same */
2745 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2747 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2748 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2749 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2752 static int ext4_check_opt_consistency(struct fs_context *fc,
2753 struct super_block *sb)
2755 struct ext4_fs_context *ctx = fc->fs_private;
2756 struct ext4_sb_info *sbi = fc->s_fs_info;
2757 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2760 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2761 ext4_msg(NULL, KERN_ERR,
2762 "Mount option(s) incompatible with ext2");
2765 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2766 ext4_msg(NULL, KERN_ERR,
2767 "Mount option(s) incompatible with ext3");
2771 if (ctx->s_want_extra_isize >
2772 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2773 ext4_msg(NULL, KERN_ERR,
2774 "Invalid want_extra_isize %d",
2775 ctx->s_want_extra_isize);
2779 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2781 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2782 if (blocksize < PAGE_SIZE)
2783 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2784 "experimental mount option 'dioread_nolock' "
2785 "for blocksize < PAGE_SIZE");
2788 err = ext4_check_test_dummy_encryption(fc, sb);
2792 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2793 if (!sbi->s_journal) {
2794 ext4_msg(NULL, KERN_WARNING,
2795 "Remounting file system with no journal "
2796 "so ignoring journalled data option");
2797 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2798 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2799 test_opt(sb, DATA_FLAGS)) {
2800 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2807 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2808 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2809 ext4_msg(NULL, KERN_ERR, "can't mount with "
2810 "both data=journal and dax");
2814 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2815 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2816 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2817 fail_dax_change_remount:
2818 ext4_msg(NULL, KERN_ERR, "can't change "
2819 "dax mount option while remounting");
2821 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2822 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2823 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2824 goto fail_dax_change_remount;
2825 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2826 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2827 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2828 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2829 goto fail_dax_change_remount;
2833 return ext4_check_quota_consistency(fc, sb);
2836 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2838 struct ext4_fs_context *ctx = fc->fs_private;
2839 struct ext4_sb_info *sbi = fc->s_fs_info;
2841 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2842 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2843 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2844 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2845 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2846 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2847 sb->s_flags &= ~ctx->mask_s_flags;
2848 sb->s_flags |= ctx->vals_s_flags;
2850 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2851 APPLY(s_commit_interval);
2853 APPLY(s_max_batch_time);
2854 APPLY(s_min_batch_time);
2855 APPLY(s_want_extra_isize);
2856 APPLY(s_inode_readahead_blks);
2857 APPLY(s_max_dir_size_kb);
2858 APPLY(s_li_wait_mult);
2862 #ifdef CONFIG_EXT4_DEBUG
2863 APPLY(s_fc_debug_max_replay);
2866 ext4_apply_quota_options(fc, sb);
2867 ext4_apply_test_dummy_encryption(ctx, sb);
2871 static int ext4_validate_options(struct fs_context *fc)
2874 struct ext4_fs_context *ctx = fc->fs_private;
2875 char *usr_qf_name, *grp_qf_name;
2877 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2878 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2880 if (usr_qf_name || grp_qf_name) {
2881 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2882 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2884 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2885 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2887 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2888 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2889 ext4_msg(NULL, KERN_ERR, "old and new quota "
2898 static inline void ext4_show_quota_options(struct seq_file *seq,
2899 struct super_block *sb)
2901 #if defined(CONFIG_QUOTA)
2902 struct ext4_sb_info *sbi = EXT4_SB(sb);
2903 char *usr_qf_name, *grp_qf_name;
2905 if (sbi->s_jquota_fmt) {
2908 switch (sbi->s_jquota_fmt) {
2919 seq_printf(seq, ",jqfmt=%s", fmtname);
2923 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2924 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2926 seq_show_option(seq, "usrjquota", usr_qf_name);
2928 seq_show_option(seq, "grpjquota", grp_qf_name);
2933 static const char *token2str(int token)
2935 const struct fs_parameter_spec *spec;
2937 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2938 if (spec->opt == token && !spec->type)
2945 * - it's set to a non-default value OR
2946 * - if the per-sb default is different from the global default
2948 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2951 struct ext4_sb_info *sbi = EXT4_SB(sb);
2952 struct ext4_super_block *es = sbi->s_es;
2954 const struct mount_opts *m;
2955 char sep = nodefs ? '\n' : ',';
2957 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2958 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2960 if (sbi->s_sb_block != 1)
2961 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2963 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2964 int want_set = m->flags & MOPT_SET;
2965 int opt_2 = m->flags & MOPT_2;
2966 unsigned int mount_opt, def_mount_opt;
2968 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2969 m->flags & MOPT_SKIP)
2973 mount_opt = sbi->s_mount_opt2;
2974 def_mount_opt = sbi->s_def_mount_opt2;
2976 mount_opt = sbi->s_mount_opt;
2977 def_mount_opt = sbi->s_def_mount_opt;
2979 /* skip if same as the default */
2980 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2982 /* select Opt_noFoo vs Opt_Foo */
2984 (mount_opt & m->mount_opt) != m->mount_opt) ||
2985 (!want_set && (mount_opt & m->mount_opt)))
2987 SEQ_OPTS_PRINT("%s", token2str(m->token));
2990 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2991 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2992 SEQ_OPTS_PRINT("resuid=%u",
2993 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2994 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2995 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2996 SEQ_OPTS_PRINT("resgid=%u",
2997 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2998 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2999 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3000 SEQ_OPTS_PUTS("errors=remount-ro");
3001 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
3002 SEQ_OPTS_PUTS("errors=continue");
3003 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
3004 SEQ_OPTS_PUTS("errors=panic");
3005 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
3006 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
3007 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
3008 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
3009 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
3010 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3011 if (nodefs || sbi->s_stripe)
3012 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3013 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3014 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3015 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3016 SEQ_OPTS_PUTS("data=journal");
3017 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3018 SEQ_OPTS_PUTS("data=ordered");
3019 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3020 SEQ_OPTS_PUTS("data=writeback");
3023 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3024 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3025 sbi->s_inode_readahead_blks);
3027 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3028 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3029 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3030 if (nodefs || sbi->s_max_dir_size_kb)
3031 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3032 if (test_opt(sb, DATA_ERR_ABORT))
3033 SEQ_OPTS_PUTS("data_err=abort");
3035 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3037 if (sb->s_flags & SB_INLINECRYPT)
3038 SEQ_OPTS_PUTS("inlinecrypt");
3040 if (test_opt(sb, DAX_ALWAYS)) {
3042 SEQ_OPTS_PUTS("dax");
3044 SEQ_OPTS_PUTS("dax=always");
3045 } else if (test_opt2(sb, DAX_NEVER)) {
3046 SEQ_OPTS_PUTS("dax=never");
3047 } else if (test_opt2(sb, DAX_INODE)) {
3048 SEQ_OPTS_PUTS("dax=inode");
3051 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3052 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3053 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3054 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3055 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3056 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3059 ext4_show_quota_options(seq, sb);
3063 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3065 return _ext4_show_options(seq, root->d_sb, 0);
3068 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3070 struct super_block *sb = seq->private;
3073 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3074 rc = _ext4_show_options(seq, sb, 1);
3075 seq_puts(seq, "\n");
3079 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3082 struct ext4_sb_info *sbi = EXT4_SB(sb);
3085 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3086 ext4_msg(sb, KERN_ERR, "revision level too high, "
3087 "forcing read-only mode");
3093 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3094 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3095 "running e2fsck is recommended");
3096 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3097 ext4_msg(sb, KERN_WARNING,
3098 "warning: mounting fs with errors, "
3099 "running e2fsck is recommended");
3100 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3101 le16_to_cpu(es->s_mnt_count) >=
3102 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3103 ext4_msg(sb, KERN_WARNING,
3104 "warning: maximal mount count reached, "
3105 "running e2fsck is recommended");
3106 else if (le32_to_cpu(es->s_checkinterval) &&
3107 (ext4_get_tstamp(es, s_lastcheck) +
3108 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3109 ext4_msg(sb, KERN_WARNING,
3110 "warning: checktime reached, "
3111 "running e2fsck is recommended");
3112 if (!sbi->s_journal)
3113 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3114 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3115 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3116 le16_add_cpu(&es->s_mnt_count, 1);
3117 ext4_update_tstamp(es, s_mtime);
3118 if (sbi->s_journal) {
3119 ext4_set_feature_journal_needs_recovery(sb);
3120 if (ext4_has_feature_orphan_file(sb))
3121 ext4_set_feature_orphan_present(sb);
3124 err = ext4_commit_super(sb);
3126 if (test_opt(sb, DEBUG))
3127 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3128 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3130 sbi->s_groups_count,
3131 EXT4_BLOCKS_PER_GROUP(sb),
3132 EXT4_INODES_PER_GROUP(sb),
3133 sbi->s_mount_opt, sbi->s_mount_opt2);
3137 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
3140 struct flex_groups **old_groups, **new_groups;
3143 if (!sbi->s_log_groups_per_flex)
3146 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3147 if (size <= sbi->s_flex_groups_allocated)
3150 new_groups = kvzalloc(roundup_pow_of_two(size *
3151 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3153 ext4_msg(sb, KERN_ERR,
3154 "not enough memory for %d flex group pointers", size);
3157 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3158 new_groups[i] = kvzalloc(roundup_pow_of_two(
3159 sizeof(struct flex_groups)),
3161 if (!new_groups[i]) {
3162 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3163 kvfree(new_groups[j]);
3165 ext4_msg(sb, KERN_ERR,
3166 "not enough memory for %d flex groups", size);
3171 old_groups = rcu_dereference(sbi->s_flex_groups);
3173 memcpy(new_groups, old_groups,
3174 (sbi->s_flex_groups_allocated *
3175 sizeof(struct flex_groups *)));
3177 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3178 sbi->s_flex_groups_allocated = size;
3180 ext4_kvfree_array_rcu(old_groups);
3184 static int ext4_fill_flex_info(struct super_block *sb)
3186 struct ext4_sb_info *sbi = EXT4_SB(sb);
3187 struct ext4_group_desc *gdp = NULL;
3188 struct flex_groups *fg;
3189 ext4_group_t flex_group;
3192 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3193 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3194 sbi->s_log_groups_per_flex = 0;
3198 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3202 for (i = 0; i < sbi->s_groups_count; i++) {
3203 gdp = ext4_get_group_desc(sb, i, NULL);
3205 flex_group = ext4_flex_group(sbi, i);
3206 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3207 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3208 atomic64_add(ext4_free_group_clusters(sb, gdp),
3209 &fg->free_clusters);
3210 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3218 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3219 struct ext4_group_desc *gdp)
3221 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3223 __le32 le_group = cpu_to_le32(block_group);
3224 struct ext4_sb_info *sbi = EXT4_SB(sb);
3226 if (ext4_has_metadata_csum(sbi->s_sb)) {
3227 /* Use new metadata_csum algorithm */
3229 __u16 dummy_csum = 0;
3231 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3233 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3234 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3235 sizeof(dummy_csum));
3236 offset += sizeof(dummy_csum);
3237 if (offset < sbi->s_desc_size)
3238 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3239 sbi->s_desc_size - offset);
3241 crc = csum32 & 0xFFFF;
3245 /* old crc16 code */
3246 if (!ext4_has_feature_gdt_csum(sb))
3249 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3250 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3251 crc = crc16(crc, (__u8 *)gdp, offset);
3252 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3253 /* for checksum of struct ext4_group_desc do the rest...*/
3254 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3255 crc = crc16(crc, (__u8 *)gdp + offset,
3256 sbi->s_desc_size - offset);
3259 return cpu_to_le16(crc);
3262 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3263 struct ext4_group_desc *gdp)
3265 if (ext4_has_group_desc_csum(sb) &&
3266 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3272 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3273 struct ext4_group_desc *gdp)
3275 if (!ext4_has_group_desc_csum(sb))
3277 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3280 /* Called at mount-time, super-block is locked */
3281 static int ext4_check_descriptors(struct super_block *sb,
3282 ext4_fsblk_t sb_block,
3283 ext4_group_t *first_not_zeroed)
3285 struct ext4_sb_info *sbi = EXT4_SB(sb);
3286 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3287 ext4_fsblk_t last_block;
3288 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3289 ext4_fsblk_t block_bitmap;
3290 ext4_fsblk_t inode_bitmap;
3291 ext4_fsblk_t inode_table;
3292 int flexbg_flag = 0;
3293 ext4_group_t i, grp = sbi->s_groups_count;
3295 if (ext4_has_feature_flex_bg(sb))
3298 ext4_debug("Checking group descriptors");
3300 for (i = 0; i < sbi->s_groups_count; i++) {
3301 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3303 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3304 last_block = ext4_blocks_count(sbi->s_es) - 1;
3306 last_block = first_block +
3307 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3309 if ((grp == sbi->s_groups_count) &&
3310 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3313 block_bitmap = ext4_block_bitmap(sb, gdp);
3314 if (block_bitmap == sb_block) {
3315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3316 "Block bitmap for group %u overlaps "
3321 if (block_bitmap >= sb_block + 1 &&
3322 block_bitmap <= last_bg_block) {
3323 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3324 "Block bitmap for group %u overlaps "
3325 "block group descriptors", i);
3329 if (block_bitmap < first_block || block_bitmap > last_block) {
3330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3331 "Block bitmap for group %u not in group "
3332 "(block %llu)!", i, block_bitmap);
3335 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3336 if (inode_bitmap == sb_block) {
3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3338 "Inode bitmap for group %u overlaps "
3343 if (inode_bitmap >= sb_block + 1 &&
3344 inode_bitmap <= last_bg_block) {
3345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3346 "Inode bitmap for group %u overlaps "
3347 "block group descriptors", i);
3351 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3352 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3353 "Inode bitmap for group %u not in group "
3354 "(block %llu)!", i, inode_bitmap);
3357 inode_table = ext4_inode_table(sb, gdp);
3358 if (inode_table == sb_block) {
3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3360 "Inode table for group %u overlaps "
3365 if (inode_table >= sb_block + 1 &&
3366 inode_table <= last_bg_block) {
3367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3368 "Inode table for group %u overlaps "
3369 "block group descriptors", i);
3373 if (inode_table < first_block ||
3374 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3375 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3376 "Inode table for group %u not in group "
3377 "(block %llu)!", i, inode_table);
3380 ext4_lock_group(sb, i);
3381 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3382 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3383 "Checksum for group %u failed (%u!=%u)",
3384 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3385 gdp)), le16_to_cpu(gdp->bg_checksum));
3386 if (!sb_rdonly(sb)) {
3387 ext4_unlock_group(sb, i);
3391 ext4_unlock_group(sb, i);
3393 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3395 if (NULL != first_not_zeroed)
3396 *first_not_zeroed = grp;
3401 * Maximal extent format file size.
3402 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3403 * extent format containers, within a sector_t, and within i_blocks
3404 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3405 * so that won't be a limiting factor.
3407 * However there is other limiting factor. We do store extents in the form
3408 * of starting block and length, hence the resulting length of the extent
3409 * covering maximum file size must fit into on-disk format containers as
3410 * well. Given that length is always by 1 unit bigger than max unit (because
3411 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3413 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3415 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3418 loff_t upper_limit = MAX_LFS_FILESIZE;
3420 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3422 if (!has_huge_files) {
3423 upper_limit = (1LL << 32) - 1;
3425 /* total blocks in file system block size */
3426 upper_limit >>= (blkbits - 9);
3427 upper_limit <<= blkbits;
3431 * 32-bit extent-start container, ee_block. We lower the maxbytes
3432 * by one fs block, so ee_len can cover the extent of maximum file
3435 res = (1LL << 32) - 1;
3438 /* Sanity check against vm- & vfs- imposed limits */
3439 if (res > upper_limit)
3446 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3447 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3448 * We need to be 1 filesystem block less than the 2^48 sector limit.
3450 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3452 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3454 unsigned int ppb = 1 << (bits - 2);
3457 * This is calculated to be the largest file size for a dense, block
3458 * mapped file such that the file's total number of 512-byte sectors,
3459 * including data and all indirect blocks, does not exceed (2^48 - 1).
3461 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3462 * number of 512-byte sectors of the file.
3464 if (!has_huge_files) {
3466 * !has_huge_files or implies that the inode i_block field
3467 * represents total file blocks in 2^32 512-byte sectors ==
3468 * size of vfs inode i_blocks * 8
3470 upper_limit = (1LL << 32) - 1;
3472 /* total blocks in file system block size */
3473 upper_limit >>= (bits - 9);
3477 * We use 48 bit ext4_inode i_blocks
3478 * With EXT4_HUGE_FILE_FL set the i_blocks
3479 * represent total number of blocks in
3480 * file system block size
3482 upper_limit = (1LL << 48) - 1;
3486 /* Compute how many blocks we can address by block tree */
3489 res += ((loff_t)ppb) * ppb * ppb;
3490 /* Compute how many metadata blocks are needed */
3492 meta_blocks += 1 + ppb;
3493 meta_blocks += 1 + ppb + ppb * ppb;
3494 /* Does block tree limit file size? */
3495 if (res + meta_blocks <= upper_limit)
3499 /* How many metadata blocks are needed for addressing upper_limit? */
3500 upper_limit -= EXT4_NDIR_BLOCKS;
3501 /* indirect blocks */
3504 /* double indirect blocks */
3505 if (upper_limit < ppb * ppb) {
3506 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3510 meta_blocks += 1 + ppb;
3511 upper_limit -= ppb * ppb;
3512 /* tripple indirect blocks for the rest */
3513 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3514 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3518 if (res > MAX_LFS_FILESIZE)
3519 res = MAX_LFS_FILESIZE;
3524 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3525 ext4_fsblk_t logical_sb_block, int nr)
3527 struct ext4_sb_info *sbi = EXT4_SB(sb);
3528 ext4_group_t bg, first_meta_bg;
3531 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3533 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3534 return logical_sb_block + nr + 1;
3535 bg = sbi->s_desc_per_block * nr;
3536 if (ext4_bg_has_super(sb, bg))
3540 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3541 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3542 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3545 if (sb->s_blocksize == 1024 && nr == 0 &&
3546 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3549 return (has_super + ext4_group_first_block_no(sb, bg));
3553 * ext4_get_stripe_size: Get the stripe size.
3554 * @sbi: In memory super block info
3556 * If we have specified it via mount option, then
3557 * use the mount option value. If the value specified at mount time is
3558 * greater than the blocks per group use the super block value.
3559 * If the super block value is greater than blocks per group return 0.
3560 * Allocator needs it be less than blocks per group.
3563 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3565 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3566 unsigned long stripe_width =
3567 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3570 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3571 ret = sbi->s_stripe;
3572 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3574 else if (stride && stride <= sbi->s_blocks_per_group)
3580 * If the stripe width is 1, this makes no sense and
3581 * we set it to 0 to turn off stripe handling code.
3590 * Check whether this filesystem can be mounted based on
3591 * the features present and the RDONLY/RDWR mount requested.
3592 * Returns 1 if this filesystem can be mounted as requested,
3593 * 0 if it cannot be.
3595 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3597 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3598 ext4_msg(sb, KERN_ERR,
3599 "Couldn't mount because of "
3600 "unsupported optional features (%x)",
3601 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3602 ~EXT4_FEATURE_INCOMPAT_SUPP));
3606 #if !IS_ENABLED(CONFIG_UNICODE)
3607 if (ext4_has_feature_casefold(sb)) {
3608 ext4_msg(sb, KERN_ERR,
3609 "Filesystem with casefold feature cannot be "
3610 "mounted without CONFIG_UNICODE");
3618 if (ext4_has_feature_readonly(sb)) {
3619 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3620 sb->s_flags |= SB_RDONLY;
3624 /* Check that feature set is OK for a read-write mount */
3625 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3626 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3627 "unsupported optional features (%x)",
3628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3629 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3632 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3633 ext4_msg(sb, KERN_ERR,
3634 "Can't support bigalloc feature without "
3635 "extents feature\n");
3639 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3640 if (!readonly && (ext4_has_feature_quota(sb) ||
3641 ext4_has_feature_project(sb))) {
3642 ext4_msg(sb, KERN_ERR,
3643 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3646 #endif /* CONFIG_QUOTA */
3651 * This function is called once a day if we have errors logged
3652 * on the file system
3654 static void print_daily_error_info(struct timer_list *t)
3656 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3657 struct super_block *sb = sbi->s_sb;
3658 struct ext4_super_block *es = sbi->s_es;
3660 if (es->s_error_count)
3661 /* fsck newer than v1.41.13 is needed to clean this condition. */
3662 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3663 le32_to_cpu(es->s_error_count));
3664 if (es->s_first_error_time) {
3665 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3667 ext4_get_tstamp(es, s_first_error_time),
3668 (int) sizeof(es->s_first_error_func),
3669 es->s_first_error_func,
3670 le32_to_cpu(es->s_first_error_line));
3671 if (es->s_first_error_ino)
3672 printk(KERN_CONT ": inode %u",
3673 le32_to_cpu(es->s_first_error_ino));
3674 if (es->s_first_error_block)
3675 printk(KERN_CONT ": block %llu", (unsigned long long)
3676 le64_to_cpu(es->s_first_error_block));
3677 printk(KERN_CONT "\n");
3679 if (es->s_last_error_time) {
3680 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3682 ext4_get_tstamp(es, s_last_error_time),
3683 (int) sizeof(es->s_last_error_func),
3684 es->s_last_error_func,
3685 le32_to_cpu(es->s_last_error_line));
3686 if (es->s_last_error_ino)
3687 printk(KERN_CONT ": inode %u",
3688 le32_to_cpu(es->s_last_error_ino));
3689 if (es->s_last_error_block)
3690 printk(KERN_CONT ": block %llu", (unsigned long long)
3691 le64_to_cpu(es->s_last_error_block));
3692 printk(KERN_CONT "\n");
3694 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3697 /* Find next suitable group and run ext4_init_inode_table */
3698 static int ext4_run_li_request(struct ext4_li_request *elr)
3700 struct ext4_group_desc *gdp = NULL;
3701 struct super_block *sb = elr->lr_super;
3702 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3703 ext4_group_t group = elr->lr_next_group;
3704 unsigned int prefetch_ios = 0;
3706 int nr = EXT4_SB(sb)->s_mb_prefetch;
3709 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3710 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3711 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3712 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3713 if (group >= elr->lr_next_group) {
3715 if (elr->lr_first_not_zeroed != ngroups &&
3716 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3717 elr->lr_next_group = elr->lr_first_not_zeroed;
3718 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3725 for (; group < ngroups; group++) {
3726 gdp = ext4_get_group_desc(sb, group, NULL);
3732 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3736 if (group >= ngroups)
3740 start_time = ktime_get_real_ns();
3741 ret = ext4_init_inode_table(sb, group,
3742 elr->lr_timeout ? 0 : 1);
3743 trace_ext4_lazy_itable_init(sb, group);
3744 if (elr->lr_timeout == 0) {
3745 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3746 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3748 elr->lr_next_sched = jiffies + elr->lr_timeout;
3749 elr->lr_next_group = group + 1;
3755 * Remove lr_request from the list_request and free the
3756 * request structure. Should be called with li_list_mtx held
3758 static void ext4_remove_li_request(struct ext4_li_request *elr)
3763 list_del(&elr->lr_request);
3764 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3768 static void ext4_unregister_li_request(struct super_block *sb)
3770 mutex_lock(&ext4_li_mtx);
3771 if (!ext4_li_info) {
3772 mutex_unlock(&ext4_li_mtx);
3776 mutex_lock(&ext4_li_info->li_list_mtx);
3777 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3778 mutex_unlock(&ext4_li_info->li_list_mtx);
3779 mutex_unlock(&ext4_li_mtx);
3782 static struct task_struct *ext4_lazyinit_task;
3785 * This is the function where ext4lazyinit thread lives. It walks
3786 * through the request list searching for next scheduled filesystem.
3787 * When such a fs is found, run the lazy initialization request
3788 * (ext4_rn_li_request) and keep track of the time spend in this
3789 * function. Based on that time we compute next schedule time of
3790 * the request. When walking through the list is complete, compute
3791 * next waking time and put itself into sleep.
3793 static int ext4_lazyinit_thread(void *arg)
3795 struct ext4_lazy_init *eli = arg;
3796 struct list_head *pos, *n;
3797 struct ext4_li_request *elr;
3798 unsigned long next_wakeup, cur;
3800 BUG_ON(NULL == eli);
3805 next_wakeup = MAX_JIFFY_OFFSET;
3807 mutex_lock(&eli->li_list_mtx);
3808 if (list_empty(&eli->li_request_list)) {
3809 mutex_unlock(&eli->li_list_mtx);
3812 list_for_each_safe(pos, n, &eli->li_request_list) {
3815 elr = list_entry(pos, struct ext4_li_request,
3818 if (time_before(jiffies, elr->lr_next_sched)) {
3819 if (time_before(elr->lr_next_sched, next_wakeup))
3820 next_wakeup = elr->lr_next_sched;
3823 if (down_read_trylock(&elr->lr_super->s_umount)) {
3824 if (sb_start_write_trylock(elr->lr_super)) {
3827 * We hold sb->s_umount, sb can not
3828 * be removed from the list, it is
3829 * now safe to drop li_list_mtx
3831 mutex_unlock(&eli->li_list_mtx);
3832 err = ext4_run_li_request(elr);
3833 sb_end_write(elr->lr_super);
3834 mutex_lock(&eli->li_list_mtx);
3837 up_read((&elr->lr_super->s_umount));
3839 /* error, remove the lazy_init job */
3841 ext4_remove_li_request(elr);
3845 elr->lr_next_sched = jiffies +
3846 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3848 if (time_before(elr->lr_next_sched, next_wakeup))
3849 next_wakeup = elr->lr_next_sched;
3851 mutex_unlock(&eli->li_list_mtx);
3856 if ((time_after_eq(cur, next_wakeup)) ||
3857 (MAX_JIFFY_OFFSET == next_wakeup)) {
3862 schedule_timeout_interruptible(next_wakeup - cur);
3864 if (kthread_should_stop()) {
3865 ext4_clear_request_list();
3872 * It looks like the request list is empty, but we need
3873 * to check it under the li_list_mtx lock, to prevent any
3874 * additions into it, and of course we should lock ext4_li_mtx
3875 * to atomically free the list and ext4_li_info, because at
3876 * this point another ext4 filesystem could be registering
3879 mutex_lock(&ext4_li_mtx);
3880 mutex_lock(&eli->li_list_mtx);
3881 if (!list_empty(&eli->li_request_list)) {
3882 mutex_unlock(&eli->li_list_mtx);
3883 mutex_unlock(&ext4_li_mtx);
3886 mutex_unlock(&eli->li_list_mtx);
3887 kfree(ext4_li_info);
3888 ext4_li_info = NULL;
3889 mutex_unlock(&ext4_li_mtx);
3894 static void ext4_clear_request_list(void)
3896 struct list_head *pos, *n;
3897 struct ext4_li_request *elr;
3899 mutex_lock(&ext4_li_info->li_list_mtx);
3900 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3901 elr = list_entry(pos, struct ext4_li_request,
3903 ext4_remove_li_request(elr);
3905 mutex_unlock(&ext4_li_info->li_list_mtx);
3908 static int ext4_run_lazyinit_thread(void)
3910 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3911 ext4_li_info, "ext4lazyinit");
3912 if (IS_ERR(ext4_lazyinit_task)) {
3913 int err = PTR_ERR(ext4_lazyinit_task);
3914 ext4_clear_request_list();
3915 kfree(ext4_li_info);
3916 ext4_li_info = NULL;
3917 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3918 "initialization thread\n",
3922 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3927 * Check whether it make sense to run itable init. thread or not.
3928 * If there is at least one uninitialized inode table, return
3929 * corresponding group number, else the loop goes through all
3930 * groups and return total number of groups.
3932 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3934 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3935 struct ext4_group_desc *gdp = NULL;
3937 if (!ext4_has_group_desc_csum(sb))
3940 for (group = 0; group < ngroups; group++) {
3941 gdp = ext4_get_group_desc(sb, group, NULL);
3945 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3952 static int ext4_li_info_new(void)
3954 struct ext4_lazy_init *eli = NULL;
3956 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3960 INIT_LIST_HEAD(&eli->li_request_list);
3961 mutex_init(&eli->li_list_mtx);
3963 eli->li_state |= EXT4_LAZYINIT_QUIT;
3970 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3973 struct ext4_li_request *elr;
3975 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3980 elr->lr_first_not_zeroed = start;
3981 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3982 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3983 elr->lr_next_group = start;
3985 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3989 * Randomize first schedule time of the request to
3990 * spread the inode table initialization requests
3993 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3997 int ext4_register_li_request(struct super_block *sb,
3998 ext4_group_t first_not_zeroed)
4000 struct ext4_sb_info *sbi = EXT4_SB(sb);
4001 struct ext4_li_request *elr = NULL;
4002 ext4_group_t ngroups = sbi->s_groups_count;
4005 mutex_lock(&ext4_li_mtx);
4006 if (sbi->s_li_request != NULL) {
4008 * Reset timeout so it can be computed again, because
4009 * s_li_wait_mult might have changed.
4011 sbi->s_li_request->lr_timeout = 0;
4015 if (sb_rdonly(sb) ||
4016 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4017 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4020 elr = ext4_li_request_new(sb, first_not_zeroed);
4026 if (NULL == ext4_li_info) {
4027 ret = ext4_li_info_new();
4032 mutex_lock(&ext4_li_info->li_list_mtx);
4033 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4034 mutex_unlock(&ext4_li_info->li_list_mtx);
4036 sbi->s_li_request = elr;
4038 * set elr to NULL here since it has been inserted to
4039 * the request_list and the removal and free of it is
4040 * handled by ext4_clear_request_list from now on.
4044 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4045 ret = ext4_run_lazyinit_thread();
4050 mutex_unlock(&ext4_li_mtx);
4057 * We do not need to lock anything since this is called on
4060 static void ext4_destroy_lazyinit_thread(void)
4063 * If thread exited earlier
4064 * there's nothing to be done.
4066 if (!ext4_li_info || !ext4_lazyinit_task)
4069 kthread_stop(ext4_lazyinit_task);
4072 static int set_journal_csum_feature_set(struct super_block *sb)
4075 int compat, incompat;
4076 struct ext4_sb_info *sbi = EXT4_SB(sb);
4078 if (ext4_has_metadata_csum(sb)) {
4079 /* journal checksum v3 */
4081 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4083 /* journal checksum v1 */
4084 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4088 jbd2_journal_clear_features(sbi->s_journal,
4089 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4090 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4091 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4092 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4093 ret = jbd2_journal_set_features(sbi->s_journal,
4095 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4097 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4098 ret = jbd2_journal_set_features(sbi->s_journal,
4101 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4102 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4104 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4105 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4112 * Note: calculating the overhead so we can be compatible with
4113 * historical BSD practice is quite difficult in the face of
4114 * clusters/bigalloc. This is because multiple metadata blocks from
4115 * different block group can end up in the same allocation cluster.
4116 * Calculating the exact overhead in the face of clustered allocation
4117 * requires either O(all block bitmaps) in memory or O(number of block
4118 * groups**2) in time. We will still calculate the superblock for
4119 * older file systems --- and if we come across with a bigalloc file
4120 * system with zero in s_overhead_clusters the estimate will be close to
4121 * correct especially for very large cluster sizes --- but for newer
4122 * file systems, it's better to calculate this figure once at mkfs
4123 * time, and store it in the superblock. If the superblock value is
4124 * present (even for non-bigalloc file systems), we will use it.
4126 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4129 struct ext4_sb_info *sbi = EXT4_SB(sb);
4130 struct ext4_group_desc *gdp;
4131 ext4_fsblk_t first_block, last_block, b;
4132 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4133 int s, j, count = 0;
4134 int has_super = ext4_bg_has_super(sb, grp);
4136 if (!ext4_has_feature_bigalloc(sb))
4137 return (has_super + ext4_bg_num_gdb(sb, grp) +
4138 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4139 sbi->s_itb_per_group + 2);
4141 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4142 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4143 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4144 for (i = 0; i < ngroups; i++) {
4145 gdp = ext4_get_group_desc(sb, i, NULL);
4146 b = ext4_block_bitmap(sb, gdp);
4147 if (b >= first_block && b <= last_block) {
4148 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4151 b = ext4_inode_bitmap(sb, gdp);
4152 if (b >= first_block && b <= last_block) {
4153 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4156 b = ext4_inode_table(sb, gdp);
4157 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4158 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4159 int c = EXT4_B2C(sbi, b - first_block);
4160 ext4_set_bit(c, buf);
4166 if (ext4_bg_has_super(sb, grp)) {
4167 ext4_set_bit(s++, buf);
4170 j = ext4_bg_num_gdb(sb, grp);
4171 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4172 ext4_error(sb, "Invalid number of block group "
4173 "descriptor blocks: %d", j);
4174 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4178 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4182 return EXT4_CLUSTERS_PER_GROUP(sb) -
4183 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4187 * Compute the overhead and stash it in sbi->s_overhead
4189 int ext4_calculate_overhead(struct super_block *sb)
4191 struct ext4_sb_info *sbi = EXT4_SB(sb);
4192 struct ext4_super_block *es = sbi->s_es;
4193 struct inode *j_inode;
4194 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4195 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4196 ext4_fsblk_t overhead = 0;
4197 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4203 * Compute the overhead (FS structures). This is constant
4204 * for a given filesystem unless the number of block groups
4205 * changes so we cache the previous value until it does.
4209 * All of the blocks before first_data_block are overhead
4211 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4214 * Add the overhead found in each block group
4216 for (i = 0; i < ngroups; i++) {
4219 blks = count_overhead(sb, i, buf);
4222 memset(buf, 0, PAGE_SIZE);
4227 * Add the internal journal blocks whether the journal has been
4230 if (sbi->s_journal && !sbi->s_journal_bdev)
4231 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4232 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4233 /* j_inum for internal journal is non-zero */
4234 j_inode = ext4_get_journal_inode(sb, j_inum);
4236 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4237 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4240 ext4_msg(sb, KERN_ERR, "can't get journal size");
4243 sbi->s_overhead = overhead;
4245 free_page((unsigned long) buf);
4249 static void ext4_set_resv_clusters(struct super_block *sb)
4251 ext4_fsblk_t resv_clusters;
4252 struct ext4_sb_info *sbi = EXT4_SB(sb);
4255 * There's no need to reserve anything when we aren't using extents.
4256 * The space estimates are exact, there are no unwritten extents,
4257 * hole punching doesn't need new metadata... This is needed especially
4258 * to keep ext2/3 backward compatibility.
4260 if (!ext4_has_feature_extents(sb))
4263 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4264 * This should cover the situations where we can not afford to run
4265 * out of space like for example punch hole, or converting
4266 * unwritten extents in delalloc path. In most cases such
4267 * allocation would require 1, or 2 blocks, higher numbers are
4270 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4271 sbi->s_cluster_bits);
4273 do_div(resv_clusters, 50);
4274 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4276 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4279 static const char *ext4_quota_mode(struct super_block *sb)
4282 if (!ext4_quota_capable(sb))
4285 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4286 return "journalled";
4294 static void ext4_setup_csum_trigger(struct super_block *sb,
4295 enum ext4_journal_trigger_type type,
4297 struct jbd2_buffer_trigger_type *type,
4298 struct buffer_head *bh,
4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
4304 sbi->s_journal_triggers[type].sb = sb;
4305 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4308 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4313 kfree(sbi->s_blockgroup_lock);
4314 fs_put_dax(sbi->s_daxdev, NULL);
4318 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4320 struct ext4_sb_info *sbi;
4322 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4326 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4329 sbi->s_blockgroup_lock =
4330 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4332 if (!sbi->s_blockgroup_lock)
4335 sb->s_fs_info = sbi;
4339 fs_put_dax(sbi->s_daxdev, NULL);
4344 static void ext4_set_def_opts(struct super_block *sb,
4345 struct ext4_super_block *es)
4347 unsigned long def_mount_opts;
4349 /* Set defaults before we parse the mount options */
4350 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4351 set_opt(sb, INIT_INODE_TABLE);
4352 if (def_mount_opts & EXT4_DEFM_DEBUG)
4354 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4356 if (def_mount_opts & EXT4_DEFM_UID16)
4357 set_opt(sb, NO_UID32);
4358 /* xattr user namespace & acls are now defaulted on */
4359 set_opt(sb, XATTR_USER);
4360 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4361 set_opt(sb, POSIX_ACL);
4363 if (ext4_has_feature_fast_commit(sb))
4364 set_opt2(sb, JOURNAL_FAST_COMMIT);
4365 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4366 if (ext4_has_metadata_csum(sb))
4367 set_opt(sb, JOURNAL_CHECKSUM);
4369 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4370 set_opt(sb, JOURNAL_DATA);
4371 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4372 set_opt(sb, ORDERED_DATA);
4373 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4374 set_opt(sb, WRITEBACK_DATA);
4376 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4377 set_opt(sb, ERRORS_PANIC);
4378 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4379 set_opt(sb, ERRORS_CONT);
4381 set_opt(sb, ERRORS_RO);
4382 /* block_validity enabled by default; disable with noblock_validity */
4383 set_opt(sb, BLOCK_VALIDITY);
4384 if (def_mount_opts & EXT4_DEFM_DISCARD)
4385 set_opt(sb, DISCARD);
4387 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4388 set_opt(sb, BARRIER);
4391 * enable delayed allocation by default
4392 * Use -o nodelalloc to turn it off
4394 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4395 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4396 set_opt(sb, DELALLOC);
4398 if (sb->s_blocksize == PAGE_SIZE)
4399 set_opt(sb, DIOREAD_NOLOCK);
4402 static int ext4_handle_clustersize(struct super_block *sb)
4404 struct ext4_sb_info *sbi = EXT4_SB(sb);
4405 struct ext4_super_block *es = sbi->s_es;
4408 /* Handle clustersize */
4409 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4410 if (ext4_has_feature_bigalloc(sb)) {
4411 if (clustersize < sb->s_blocksize) {
4412 ext4_msg(sb, KERN_ERR,
4413 "cluster size (%d) smaller than "
4414 "block size (%lu)", clustersize, sb->s_blocksize);
4417 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4418 le32_to_cpu(es->s_log_block_size);
4419 sbi->s_clusters_per_group =
4420 le32_to_cpu(es->s_clusters_per_group);
4421 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4422 ext4_msg(sb, KERN_ERR,
4423 "#clusters per group too big: %lu",
4424 sbi->s_clusters_per_group);
4427 if (sbi->s_blocks_per_group !=
4428 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4429 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4430 "clusters per group (%lu) inconsistent",
4431 sbi->s_blocks_per_group,
4432 sbi->s_clusters_per_group);
4436 if (clustersize != sb->s_blocksize) {
4437 ext4_msg(sb, KERN_ERR,
4438 "fragment/cluster size (%d) != "
4439 "block size (%lu)", clustersize, sb->s_blocksize);
4442 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4443 ext4_msg(sb, KERN_ERR,
4444 "#blocks per group too big: %lu",
4445 sbi->s_blocks_per_group);
4448 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4449 sbi->s_cluster_bits = 0;
4451 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4453 /* Do we have standard group size of clustersize * 8 blocks ? */
4454 if (sbi->s_blocks_per_group == clustersize << 3)
4455 set_opt2(sb, STD_GROUP_SIZE);
4460 static void ext4_fast_commit_init(struct super_block *sb)
4462 struct ext4_sb_info *sbi = EXT4_SB(sb);
4464 /* Initialize fast commit stuff */
4465 atomic_set(&sbi->s_fc_subtid, 0);
4466 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4467 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4468 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4469 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4470 sbi->s_fc_bytes = 0;
4471 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4472 sbi->s_fc_ineligible_tid = 0;
4473 spin_lock_init(&sbi->s_fc_lock);
4474 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4475 sbi->s_fc_replay_state.fc_regions = NULL;
4476 sbi->s_fc_replay_state.fc_regions_size = 0;
4477 sbi->s_fc_replay_state.fc_regions_used = 0;
4478 sbi->s_fc_replay_state.fc_regions_valid = 0;
4479 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4480 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4481 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4484 static int ext4_inode_info_init(struct super_block *sb,
4485 struct ext4_super_block *es)
4487 struct ext4_sb_info *sbi = EXT4_SB(sb);
4489 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4490 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4491 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4493 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4494 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4495 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4496 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4500 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4501 (!is_power_of_2(sbi->s_inode_size)) ||
4502 (sbi->s_inode_size > sb->s_blocksize)) {
4503 ext4_msg(sb, KERN_ERR,
4504 "unsupported inode size: %d",
4506 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4510 * i_atime_extra is the last extra field available for
4511 * [acm]times in struct ext4_inode. Checking for that
4512 * field should suffice to ensure we have extra space
4515 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4516 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4517 sb->s_time_gran = 1;
4518 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4520 sb->s_time_gran = NSEC_PER_SEC;
4521 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4523 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4526 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4527 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4528 EXT4_GOOD_OLD_INODE_SIZE;
4529 if (ext4_has_feature_extra_isize(sb)) {
4530 unsigned v, max = (sbi->s_inode_size -
4531 EXT4_GOOD_OLD_INODE_SIZE);
4533 v = le16_to_cpu(es->s_want_extra_isize);
4535 ext4_msg(sb, KERN_ERR,
4536 "bad s_want_extra_isize: %d", v);
4539 if (sbi->s_want_extra_isize < v)
4540 sbi->s_want_extra_isize = v;
4542 v = le16_to_cpu(es->s_min_extra_isize);
4544 ext4_msg(sb, KERN_ERR,
4545 "bad s_min_extra_isize: %d", v);
4548 if (sbi->s_want_extra_isize < v)
4549 sbi->s_want_extra_isize = v;
4556 #if IS_ENABLED(CONFIG_UNICODE)
4557 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4559 const struct ext4_sb_encodings *encoding_info;
4560 struct unicode_map *encoding;
4561 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4563 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4566 encoding_info = ext4_sb_read_encoding(es);
4567 if (!encoding_info) {
4568 ext4_msg(sb, KERN_ERR,
4569 "Encoding requested by superblock is unknown");
4573 encoding = utf8_load(encoding_info->version);
4574 if (IS_ERR(encoding)) {
4575 ext4_msg(sb, KERN_ERR,
4576 "can't mount with superblock charset: %s-%u.%u.%u "
4577 "not supported by the kernel. flags: 0x%x.",
4578 encoding_info->name,
4579 unicode_major(encoding_info->version),
4580 unicode_minor(encoding_info->version),
4581 unicode_rev(encoding_info->version),
4585 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4586 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4587 unicode_major(encoding_info->version),
4588 unicode_minor(encoding_info->version),
4589 unicode_rev(encoding_info->version),
4592 sb->s_encoding = encoding;
4593 sb->s_encoding_flags = encoding_flags;
4598 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4604 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4606 struct ext4_sb_info *sbi = EXT4_SB(sb);
4608 /* Warn if metadata_csum and gdt_csum are both set. */
4609 if (ext4_has_feature_metadata_csum(sb) &&
4610 ext4_has_feature_gdt_csum(sb))
4611 ext4_warning(sb, "metadata_csum and uninit_bg are "
4612 "redundant flags; please run fsck.");
4614 /* Check for a known checksum algorithm */
4615 if (!ext4_verify_csum_type(sb, es)) {
4616 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4617 "unknown checksum algorithm.");
4620 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4621 ext4_orphan_file_block_trigger);
4623 /* Load the checksum driver */
4624 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4625 if (IS_ERR(sbi->s_chksum_driver)) {
4626 int ret = PTR_ERR(sbi->s_chksum_driver);
4627 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4628 sbi->s_chksum_driver = NULL;
4632 /* Check superblock checksum */
4633 if (!ext4_superblock_csum_verify(sb, es)) {
4634 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4635 "invalid superblock checksum. Run e2fsck?");
4639 /* Precompute checksum seed for all metadata */
4640 if (ext4_has_feature_csum_seed(sb))
4641 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4642 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4643 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4644 sizeof(es->s_uuid));
4648 static int ext4_check_feature_compatibility(struct super_block *sb,
4649 struct ext4_super_block *es,
4652 struct ext4_sb_info *sbi = EXT4_SB(sb);
4654 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4655 (ext4_has_compat_features(sb) ||
4656 ext4_has_ro_compat_features(sb) ||
4657 ext4_has_incompat_features(sb)))
4658 ext4_msg(sb, KERN_WARNING,
4659 "feature flags set on rev 0 fs, "
4660 "running e2fsck is recommended");
4662 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4663 set_opt2(sb, HURD_COMPAT);
4664 if (ext4_has_feature_64bit(sb)) {
4665 ext4_msg(sb, KERN_ERR,
4666 "The Hurd can't support 64-bit file systems");
4671 * ea_inode feature uses l_i_version field which is not
4672 * available in HURD_COMPAT mode.
4674 if (ext4_has_feature_ea_inode(sb)) {
4675 ext4_msg(sb, KERN_ERR,
4676 "ea_inode feature is not supported for Hurd");
4681 if (IS_EXT2_SB(sb)) {
4682 if (ext2_feature_set_ok(sb))
4683 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4684 "using the ext4 subsystem");
4687 * If we're probing be silent, if this looks like
4688 * it's actually an ext[34] filesystem.
4690 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4692 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4693 "to feature incompatibilities");
4698 if (IS_EXT3_SB(sb)) {
4699 if (ext3_feature_set_ok(sb))
4700 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4701 "using the ext4 subsystem");
4704 * If we're probing be silent, if this looks like
4705 * it's actually an ext4 filesystem.
4707 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4709 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4710 "to feature incompatibilities");
4716 * Check feature flags regardless of the revision level, since we
4717 * previously didn't change the revision level when setting the flags,
4718 * so there is a chance incompat flags are set on a rev 0 filesystem.
4720 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4723 if (sbi->s_daxdev) {
4724 if (sb->s_blocksize == PAGE_SIZE)
4725 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4727 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4730 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4731 if (ext4_has_feature_inline_data(sb)) {
4732 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4733 " that may contain inline data");
4736 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4737 ext4_msg(sb, KERN_ERR,
4738 "DAX unsupported by block device.");
4743 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4744 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4745 es->s_encryption_level);
4752 static int ext4_check_geometry(struct super_block *sb,
4753 struct ext4_super_block *es)
4755 struct ext4_sb_info *sbi = EXT4_SB(sb);
4759 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4760 ext4_msg(sb, KERN_ERR,
4761 "Number of reserved GDT blocks insanely large: %d",
4762 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4766 * Test whether we have more sectors than will fit in sector_t,
4767 * and whether the max offset is addressable by the page cache.
4769 err = generic_check_addressable(sb->s_blocksize_bits,
4770 ext4_blocks_count(es));
4772 ext4_msg(sb, KERN_ERR, "filesystem"
4773 " too large to mount safely on this system");
4777 /* check blocks count against device size */
4778 blocks_count = sb_bdev_nr_blocks(sb);
4779 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4780 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4781 "exceeds size of device (%llu blocks)",
4782 ext4_blocks_count(es), blocks_count);
4787 * It makes no sense for the first data block to be beyond the end
4788 * of the filesystem.
4790 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4791 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4792 "block %u is beyond end of filesystem (%llu)",
4793 le32_to_cpu(es->s_first_data_block),
4794 ext4_blocks_count(es));
4797 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4798 (sbi->s_cluster_ratio == 1)) {
4799 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4800 "block is 0 with a 1k block and cluster size");
4804 blocks_count = (ext4_blocks_count(es) -
4805 le32_to_cpu(es->s_first_data_block) +
4806 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4807 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4808 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4809 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4810 "(block count %llu, first data block %u, "
4811 "blocks per group %lu)", blocks_count,
4812 ext4_blocks_count(es),
4813 le32_to_cpu(es->s_first_data_block),
4814 EXT4_BLOCKS_PER_GROUP(sb));
4817 sbi->s_groups_count = blocks_count;
4818 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4819 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4820 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4821 le32_to_cpu(es->s_inodes_count)) {
4822 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4823 le32_to_cpu(es->s_inodes_count),
4824 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4831 static int ext4_group_desc_init(struct super_block *sb,
4832 struct ext4_super_block *es,
4833 ext4_fsblk_t logical_sb_block,
4834 ext4_group_t *first_not_zeroed)
4836 struct ext4_sb_info *sbi = EXT4_SB(sb);
4837 unsigned int db_count;
4841 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4842 EXT4_DESC_PER_BLOCK(sb);
4843 if (ext4_has_feature_meta_bg(sb)) {
4844 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4845 ext4_msg(sb, KERN_WARNING,
4846 "first meta block group too large: %u "
4847 "(group descriptor block count %u)",
4848 le32_to_cpu(es->s_first_meta_bg), db_count);
4852 rcu_assign_pointer(sbi->s_group_desc,
4853 kvmalloc_array(db_count,
4854 sizeof(struct buffer_head *),
4856 if (sbi->s_group_desc == NULL) {
4857 ext4_msg(sb, KERN_ERR, "not enough memory");
4861 bgl_lock_init(sbi->s_blockgroup_lock);
4863 /* Pre-read the descriptors into the buffer cache */
4864 for (i = 0; i < db_count; i++) {
4865 block = descriptor_loc(sb, logical_sb_block, i);
4866 ext4_sb_breadahead_unmovable(sb, block);
4869 for (i = 0; i < db_count; i++) {
4870 struct buffer_head *bh;
4872 block = descriptor_loc(sb, logical_sb_block, i);
4873 bh = ext4_sb_bread_unmovable(sb, block);
4875 ext4_msg(sb, KERN_ERR,
4876 "can't read group descriptor %d", i);
4877 sbi->s_gdb_count = i;
4881 rcu_dereference(sbi->s_group_desc)[i] = bh;
4884 sbi->s_gdb_count = db_count;
4885 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4886 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4887 return -EFSCORRUPTED;
4893 static int ext4_load_and_init_journal(struct super_block *sb,
4894 struct ext4_super_block *es,
4895 struct ext4_fs_context *ctx)
4897 struct ext4_sb_info *sbi = EXT4_SB(sb);
4900 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4904 if (ext4_has_feature_64bit(sb) &&
4905 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4906 JBD2_FEATURE_INCOMPAT_64BIT)) {
4907 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4911 if (!set_journal_csum_feature_set(sb)) {
4912 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4917 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4918 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4919 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4920 ext4_msg(sb, KERN_ERR,
4921 "Failed to set fast commit journal feature");
4925 /* We have now updated the journal if required, so we can
4926 * validate the data journaling mode. */
4927 switch (test_opt(sb, DATA_FLAGS)) {
4929 /* No mode set, assume a default based on the journal
4930 * capabilities: ORDERED_DATA if the journal can
4931 * cope, else JOURNAL_DATA
4933 if (jbd2_journal_check_available_features
4934 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4935 set_opt(sb, ORDERED_DATA);
4936 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4938 set_opt(sb, JOURNAL_DATA);
4939 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4943 case EXT4_MOUNT_ORDERED_DATA:
4944 case EXT4_MOUNT_WRITEBACK_DATA:
4945 if (!jbd2_journal_check_available_features
4946 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4947 ext4_msg(sb, KERN_ERR, "Journal does not support "
4948 "requested data journaling mode");
4956 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4957 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4958 ext4_msg(sb, KERN_ERR, "can't mount with "
4959 "journal_async_commit in data=ordered mode");
4963 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4965 sbi->s_journal->j_submit_inode_data_buffers =
4966 ext4_journal_submit_inode_data_buffers;
4967 sbi->s_journal->j_finish_inode_data_buffers =
4968 ext4_journal_finish_inode_data_buffers;
4973 /* flush s_error_work before journal destroy. */
4974 flush_work(&sbi->s_error_work);
4975 jbd2_journal_destroy(sbi->s_journal);
4976 sbi->s_journal = NULL;
4980 static int ext4_check_journal_data_mode(struct super_block *sb)
4982 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4983 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4984 "data=journal disables delayed allocation, "
4985 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4986 /* can't mount with both data=journal and dioread_nolock. */
4987 clear_opt(sb, DIOREAD_NOLOCK);
4988 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4989 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4990 ext4_msg(sb, KERN_ERR, "can't mount with "
4991 "both data=journal and delalloc");
4994 if (test_opt(sb, DAX_ALWAYS)) {
4995 ext4_msg(sb, KERN_ERR, "can't mount with "
4996 "both data=journal and dax");
4999 if (ext4_has_feature_encrypt(sb)) {
5000 ext4_msg(sb, KERN_WARNING,
5001 "encrypted files will use data=ordered "
5002 "instead of data journaling mode");
5004 if (test_opt(sb, DELALLOC))
5005 clear_opt(sb, DELALLOC);
5007 sb->s_iflags |= SB_I_CGROUPWB;
5013 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5016 struct ext4_sb_info *sbi = EXT4_SB(sb);
5017 struct ext4_super_block *es;
5018 ext4_fsblk_t logical_sb_block;
5019 unsigned long offset = 0;
5020 struct buffer_head *bh;
5024 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5026 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5031 * The ext4 superblock will not be buffer aligned for other than 1kB
5032 * block sizes. We need to calculate the offset from buffer start.
5034 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5035 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5036 offset = do_div(logical_sb_block, blocksize);
5038 logical_sb_block = sbi->s_sb_block;
5041 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5043 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5047 * Note: s_es must be initialized as soon as possible because
5048 * some ext4 macro-instructions depend on its value
5050 es = (struct ext4_super_block *) (bh->b_data + offset);
5052 sb->s_magic = le16_to_cpu(es->s_magic);
5053 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5055 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5059 if (le32_to_cpu(es->s_log_block_size) >
5060 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5061 ext4_msg(sb, KERN_ERR,
5062 "Invalid log block size: %u",
5063 le32_to_cpu(es->s_log_block_size));
5066 if (le32_to_cpu(es->s_log_cluster_size) >
5067 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5068 ext4_msg(sb, KERN_ERR,
5069 "Invalid log cluster size: %u",
5070 le32_to_cpu(es->s_log_cluster_size));
5074 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5077 * If the default block size is not the same as the real block size,
5078 * we need to reload it.
5080 if (sb->s_blocksize == blocksize) {
5081 *lsb = logical_sb_block;
5087 * bh must be released before kill_bdev(), otherwise
5088 * it won't be freed and its page also. kill_bdev()
5089 * is called by sb_set_blocksize().
5092 /* Validate the filesystem blocksize */
5093 if (!sb_set_blocksize(sb, blocksize)) {
5094 ext4_msg(sb, KERN_ERR, "bad block size %d",
5100 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5101 offset = do_div(logical_sb_block, blocksize);
5102 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5104 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5109 es = (struct ext4_super_block *)(bh->b_data + offset);
5111 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5112 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5115 *lsb = logical_sb_block;
5123 static void ext4_hash_info_init(struct super_block *sb)
5125 struct ext4_sb_info *sbi = EXT4_SB(sb);
5126 struct ext4_super_block *es = sbi->s_es;
5129 for (i = 0; i < 4; i++)
5130 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5132 sbi->s_def_hash_version = es->s_def_hash_version;
5133 if (ext4_has_feature_dir_index(sb)) {
5134 i = le32_to_cpu(es->s_flags);
5135 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5136 sbi->s_hash_unsigned = 3;
5137 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5138 #ifdef __CHAR_UNSIGNED__
5141 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5142 sbi->s_hash_unsigned = 3;
5146 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5152 static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5154 struct ext4_sb_info *sbi = EXT4_SB(sb);
5155 struct ext4_super_block *es = sbi->s_es;
5158 has_huge_files = ext4_has_feature_huge_file(sb);
5159 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5161 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5163 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5164 if (ext4_has_feature_64bit(sb)) {
5165 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5166 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5167 !is_power_of_2(sbi->s_desc_size)) {
5168 ext4_msg(sb, KERN_ERR,
5169 "unsupported descriptor size %lu",
5174 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5176 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5177 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5179 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5180 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5182 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5185 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5186 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5187 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5188 sbi->s_inodes_per_group);
5191 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5192 sbi->s_inodes_per_block;
5193 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5194 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5195 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5196 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5201 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5203 struct ext4_super_block *es = NULL;
5204 struct ext4_sb_info *sbi = EXT4_SB(sb);
5205 ext4_fsblk_t logical_sb_block;
5209 ext4_group_t first_not_zeroed;
5210 struct ext4_fs_context *ctx = fc->fs_private;
5211 int silent = fc->sb_flags & SB_SILENT;
5213 /* Set defaults for the variables that will be set during parsing */
5214 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5215 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5217 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5218 sbi->s_sectors_written_start =
5219 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5221 err = ext4_load_super(sb, &logical_sb_block, silent);
5226 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5228 err = ext4_init_metadata_csum(sb, es);
5232 ext4_set_def_opts(sb, es);
5234 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5235 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5236 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5237 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5238 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5241 * set default s_li_wait_mult for lazyinit, for the case there is
5242 * no mount option specified.
5244 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5246 err = ext4_inode_info_init(sb, es);
5250 err = parse_apply_sb_mount_options(sb, ctx);
5254 sbi->s_def_mount_opt = sbi->s_mount_opt;
5255 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5257 err = ext4_check_opt_consistency(fc, sb);
5261 ext4_apply_options(fc, sb);
5263 err = ext4_encoding_init(sb, es);
5267 err = ext4_check_journal_data_mode(sb);
5271 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5272 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5274 /* i_version is always enabled now */
5275 sb->s_flags |= SB_I_VERSION;
5277 err = ext4_check_feature_compatibility(sb, es, silent);
5281 err = ext4_block_group_meta_init(sb, silent);
5285 ext4_hash_info_init(sb);
5287 err = ext4_handle_clustersize(sb);
5291 err = ext4_check_geometry(sb, es);
5295 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5296 spin_lock_init(&sbi->s_error_lock);
5297 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5299 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5303 err = ext4_es_register_shrinker(sbi);
5307 sbi->s_stripe = ext4_get_stripe_size(sbi);
5309 * It's hard to get stripe aligned blocks if stripe is not aligned with
5310 * cluster, just disable stripe and alert user to simpfy code and avoid
5311 * stripe aligned allocation which will rarely successes.
5313 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5314 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5315 ext4_msg(sb, KERN_WARNING,
5316 "stripe (%lu) is not aligned with cluster size (%u), "
5317 "stripe is disabled",
5318 sbi->s_stripe, sbi->s_cluster_ratio);
5321 sbi->s_extent_max_zeroout_kb = 32;
5324 * set up enough so that it can read an inode
5326 sb->s_op = &ext4_sops;
5327 sb->s_export_op = &ext4_export_ops;
5328 sb->s_xattr = ext4_xattr_handlers;
5329 #ifdef CONFIG_FS_ENCRYPTION
5330 sb->s_cop = &ext4_cryptops;
5332 #ifdef CONFIG_FS_VERITY
5333 sb->s_vop = &ext4_verityops;
5336 sb->dq_op = &ext4_quota_operations;
5337 if (ext4_has_feature_quota(sb))
5338 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5340 sb->s_qcop = &ext4_qctl_operations;
5341 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5343 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5345 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5346 mutex_init(&sbi->s_orphan_lock);
5348 ext4_fast_commit_init(sb);
5352 needs_recovery = (es->s_last_orphan != 0 ||
5353 ext4_has_feature_orphan_present(sb) ||
5354 ext4_has_feature_journal_needs_recovery(sb));
5356 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5357 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5359 goto failed_mount3a;
5364 * The first inode we look at is the journal inode. Don't try
5365 * root first: it may be modified in the journal!
5367 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5368 err = ext4_load_and_init_journal(sb, es, ctx);
5370 goto failed_mount3a;
5371 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5372 ext4_has_feature_journal_needs_recovery(sb)) {
5373 ext4_msg(sb, KERN_ERR, "required journal recovery "
5374 "suppressed and not mounted read-only");
5375 goto failed_mount3a;
5377 /* Nojournal mode, all journal mount options are illegal */
5378 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5379 ext4_msg(sb, KERN_ERR, "can't mount with "
5380 "journal_async_commit, fs mounted w/o journal");
5381 goto failed_mount3a;
5384 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5385 ext4_msg(sb, KERN_ERR, "can't mount with "
5386 "journal_checksum, fs mounted w/o journal");
5387 goto failed_mount3a;
5389 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5390 ext4_msg(sb, KERN_ERR, "can't mount with "
5391 "commit=%lu, fs mounted w/o journal",
5392 sbi->s_commit_interval / HZ);
5393 goto failed_mount3a;
5395 if (EXT4_MOUNT_DATA_FLAGS &
5396 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5397 ext4_msg(sb, KERN_ERR, "can't mount with "
5398 "data=, fs mounted w/o journal");
5399 goto failed_mount3a;
5401 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5402 clear_opt(sb, JOURNAL_CHECKSUM);
5403 clear_opt(sb, DATA_FLAGS);
5404 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5405 sbi->s_journal = NULL;
5409 if (!test_opt(sb, NO_MBCACHE)) {
5410 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5411 if (!sbi->s_ea_block_cache) {
5412 ext4_msg(sb, KERN_ERR,
5413 "Failed to create ea_block_cache");
5415 goto failed_mount_wq;
5418 if (ext4_has_feature_ea_inode(sb)) {
5419 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5420 if (!sbi->s_ea_inode_cache) {
5421 ext4_msg(sb, KERN_ERR,
5422 "Failed to create ea_inode_cache");
5424 goto failed_mount_wq;
5430 * Get the # of file system overhead blocks from the
5431 * superblock if present.
5433 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5434 /* ignore the precalculated value if it is ridiculous */
5435 if (sbi->s_overhead > ext4_blocks_count(es))
5436 sbi->s_overhead = 0;
5438 * If the bigalloc feature is not enabled recalculating the
5439 * overhead doesn't take long, so we might as well just redo
5440 * it to make sure we are using the correct value.
5442 if (!ext4_has_feature_bigalloc(sb))
5443 sbi->s_overhead = 0;
5444 if (sbi->s_overhead == 0) {
5445 err = ext4_calculate_overhead(sb);
5447 goto failed_mount_wq;
5451 * The maximum number of concurrent works can be high and
5452 * concurrency isn't really necessary. Limit it to 1.
5454 EXT4_SB(sb)->rsv_conversion_wq =
5455 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5456 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5457 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5463 * The jbd2_journal_load will have done any necessary log recovery,
5464 * so we can safely mount the rest of the filesystem now.
5467 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5469 ext4_msg(sb, KERN_ERR, "get root inode failed");
5470 err = PTR_ERR(root);
5474 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5475 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5477 err = -EFSCORRUPTED;
5481 sb->s_root = d_make_root(root);
5483 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5488 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5489 if (err == -EROFS) {
5490 sb->s_flags |= SB_RDONLY;
5492 goto failed_mount4a;
5494 ext4_set_resv_clusters(sb);
5496 if (test_opt(sb, BLOCK_VALIDITY)) {
5497 err = ext4_setup_system_zone(sb);
5499 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5501 goto failed_mount4a;
5504 ext4_fc_replay_cleanup(sb);
5509 * Enable optimize_scan if number of groups is > threshold. This can be
5510 * turned off by passing "mb_optimize_scan=0". This can also be
5511 * turned on forcefully by passing "mb_optimize_scan=1".
5513 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5514 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5515 set_opt2(sb, MB_OPTIMIZE_SCAN);
5517 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5520 err = ext4_mb_init(sb);
5522 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5528 * We can only set up the journal commit callback once
5529 * mballoc is initialized
5532 sbi->s_journal->j_commit_callback =
5533 ext4_journal_commit_callback;
5535 err = ext4_percpu_param_init(sbi);
5539 if (ext4_has_feature_flex_bg(sb))
5540 if (!ext4_fill_flex_info(sb)) {
5541 ext4_msg(sb, KERN_ERR,
5542 "unable to initialize "
5543 "flex_bg meta info!");
5548 err = ext4_register_li_request(sb, first_not_zeroed);
5552 err = ext4_register_sysfs(sb);
5556 err = ext4_init_orphan_info(sb);
5560 /* Enable quota usage during mount. */
5561 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5562 err = ext4_enable_quotas(sb);
5566 #endif /* CONFIG_QUOTA */
5569 * Save the original bdev mapping's wb_err value which could be
5570 * used to detect the metadata async write error.
5572 spin_lock_init(&sbi->s_bdev_wb_lock);
5573 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5574 &sbi->s_bdev_wb_err);
5575 sb->s_bdev->bd_super = sb;
5576 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5577 ext4_orphan_cleanup(sb, es);
5578 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5580 * Update the checksum after updating free space/inode counters and
5581 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5582 * checksum in the buffer cache until it is written out and
5583 * e2fsprogs programs trying to open a file system immediately
5584 * after it is mounted can fail.
5586 ext4_superblock_csum_set(sb);
5587 if (needs_recovery) {
5588 ext4_msg(sb, KERN_INFO, "recovery complete");
5589 err = ext4_mark_recovery_complete(sb, es);
5591 goto failed_mount10;
5594 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5595 ext4_msg(sb, KERN_WARNING,
5596 "mounting with \"discard\" option, but the device does not support discard");
5598 if (es->s_error_count)
5599 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5601 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5602 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5603 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5604 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5605 atomic_set(&sbi->s_warning_count, 0);
5606 atomic_set(&sbi->s_msg_count, 0);
5611 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5612 failed_mount9: __maybe_unused
5613 ext4_release_orphan_info(sb);
5615 ext4_unregister_sysfs(sb);
5616 kobject_put(&sbi->s_kobj);
5618 ext4_unregister_li_request(sb);
5620 ext4_mb_release(sb);
5621 ext4_flex_groups_free(sbi);
5622 ext4_percpu_param_destroy(sbi);
5624 ext4_ext_release(sb);
5625 ext4_release_system_zone(sb);
5630 ext4_msg(sb, KERN_ERR, "mount failed");
5631 if (EXT4_SB(sb)->rsv_conversion_wq)
5632 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5634 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5635 sbi->s_ea_inode_cache = NULL;
5637 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5638 sbi->s_ea_block_cache = NULL;
5640 if (sbi->s_journal) {
5641 /* flush s_error_work before journal destroy. */
5642 flush_work(&sbi->s_error_work);
5643 jbd2_journal_destroy(sbi->s_journal);
5644 sbi->s_journal = NULL;
5647 ext4_es_unregister_shrinker(sbi);
5649 /* flush s_error_work before sbi destroy */
5650 flush_work(&sbi->s_error_work);
5651 del_timer_sync(&sbi->s_err_report);
5652 ext4_stop_mmpd(sbi);
5653 ext4_group_desc_free(sbi);
5655 if (sbi->s_chksum_driver)
5656 crypto_free_shash(sbi->s_chksum_driver);
5658 #if IS_ENABLED(CONFIG_UNICODE)
5659 utf8_unload(sb->s_encoding);
5663 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5664 kfree(get_qf_name(sb, sbi, i));
5666 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5667 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5669 ext4_blkdev_remove(sbi);
5671 invalidate_bdev(sb->s_bdev);
5672 sb->s_fs_info = NULL;
5676 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5678 struct ext4_fs_context *ctx = fc->fs_private;
5679 struct ext4_sb_info *sbi;
5683 sbi = ext4_alloc_sbi(sb);
5687 fc->s_fs_info = sbi;
5689 /* Cleanup superblock name */
5690 strreplace(sb->s_id, '/', '!');
5692 sbi->s_sb_block = 1; /* Default super block location */
5693 if (ctx->spec & EXT4_SPEC_s_sb_block)
5694 sbi->s_sb_block = ctx->s_sb_block;
5696 ret = __ext4_fill_super(fc, sb);
5700 if (sbi->s_journal) {
5701 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5702 descr = " journalled data mode";
5703 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5704 descr = " ordered data mode";
5706 descr = " writeback data mode";
5708 descr = "out journal";
5710 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5711 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5712 "Quota mode: %s.", &sb->s_uuid,
5713 sb_rdonly(sb) ? "ro" : "r/w", descr,
5714 ext4_quota_mode(sb));
5716 /* Update the s_overhead_clusters if necessary */
5717 ext4_update_overhead(sb, false);
5722 fc->s_fs_info = NULL;
5726 static int ext4_get_tree(struct fs_context *fc)
5728 return get_tree_bdev(fc, ext4_fill_super);
5732 * Setup any per-fs journal parameters now. We'll do this both on
5733 * initial mount, once the journal has been initialised but before we've
5734 * done any recovery; and again on any subsequent remount.
5736 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5738 struct ext4_sb_info *sbi = EXT4_SB(sb);
5740 journal->j_commit_interval = sbi->s_commit_interval;
5741 journal->j_min_batch_time = sbi->s_min_batch_time;
5742 journal->j_max_batch_time = sbi->s_max_batch_time;
5743 ext4_fc_init(sb, journal);
5745 write_lock(&journal->j_state_lock);
5746 if (test_opt(sb, BARRIER))
5747 journal->j_flags |= JBD2_BARRIER;
5749 journal->j_flags &= ~JBD2_BARRIER;
5750 if (test_opt(sb, DATA_ERR_ABORT))
5751 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5753 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5755 * Always enable journal cycle record option, letting the journal
5756 * records log transactions continuously between each mount.
5758 journal->j_flags |= JBD2_CYCLE_RECORD;
5759 write_unlock(&journal->j_state_lock);
5762 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5763 unsigned int journal_inum)
5765 struct inode *journal_inode;
5768 * Test for the existence of a valid inode on disk. Bad things
5769 * happen if we iget() an unused inode, as the subsequent iput()
5770 * will try to delete it.
5772 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5773 if (IS_ERR(journal_inode)) {
5774 ext4_msg(sb, KERN_ERR, "no journal found");
5777 if (!journal_inode->i_nlink) {
5778 make_bad_inode(journal_inode);
5779 iput(journal_inode);
5780 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5784 ext4_debug("Journal inode found at %p: %lld bytes\n",
5785 journal_inode, journal_inode->i_size);
5786 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5787 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5788 iput(journal_inode);
5791 return journal_inode;
5794 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5796 struct ext4_map_blocks map;
5799 if (journal->j_inode == NULL)
5802 map.m_lblk = *block;
5804 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5806 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5807 "journal bmap failed: block %llu ret %d\n",
5809 jbd2_journal_abort(journal, ret ? ret : -EIO);
5812 *block = map.m_pblk;
5816 static journal_t *ext4_get_journal(struct super_block *sb,
5817 unsigned int journal_inum)
5819 struct inode *journal_inode;
5822 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5825 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5829 journal = jbd2_journal_init_inode(journal_inode);
5831 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5832 iput(journal_inode);
5835 journal->j_private = sb;
5836 journal->j_bmap = ext4_journal_bmap;
5837 ext4_init_journal_params(sb, journal);
5841 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5844 struct buffer_head *bh;
5848 int hblock, blocksize;
5849 ext4_fsblk_t sb_block;
5850 unsigned long offset;
5851 struct ext4_super_block *es;
5852 struct block_device *bdev;
5854 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5857 bdev = ext4_blkdev_get(j_dev, sb);
5861 blocksize = sb->s_blocksize;
5862 hblock = bdev_logical_block_size(bdev);
5863 if (blocksize < hblock) {
5864 ext4_msg(sb, KERN_ERR,
5865 "blocksize too small for journal device");
5869 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5870 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5871 set_blocksize(bdev, blocksize);
5872 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5873 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5874 "external journal");
5878 es = (struct ext4_super_block *) (bh->b_data + offset);
5879 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5880 !(le32_to_cpu(es->s_feature_incompat) &
5881 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5882 ext4_msg(sb, KERN_ERR, "external journal has "
5888 if ((le32_to_cpu(es->s_feature_ro_compat) &
5889 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5890 es->s_checksum != ext4_superblock_csum(sb, es)) {
5891 ext4_msg(sb, KERN_ERR, "external journal has "
5892 "corrupt superblock");
5897 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5898 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5903 len = ext4_blocks_count(es);
5904 start = sb_block + 1;
5905 brelse(bh); /* we're done with the superblock */
5907 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5908 start, len, blocksize);
5910 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5913 journal->j_private = sb;
5914 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5915 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5918 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5919 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5920 "user (unsupported) - %d",
5921 be32_to_cpu(journal->j_superblock->s_nr_users));
5924 EXT4_SB(sb)->s_journal_bdev = bdev;
5925 ext4_init_journal_params(sb, journal);
5929 jbd2_journal_destroy(journal);
5931 blkdev_put(bdev, sb);
5935 static int ext4_load_journal(struct super_block *sb,
5936 struct ext4_super_block *es,
5937 unsigned long journal_devnum)
5940 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5943 int really_read_only;
5946 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5947 return -EFSCORRUPTED;
5949 if (journal_devnum &&
5950 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5951 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5952 "numbers have changed");
5953 journal_dev = new_decode_dev(journal_devnum);
5955 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5957 if (journal_inum && journal_dev) {
5958 ext4_msg(sb, KERN_ERR,
5959 "filesystem has both journal inode and journal device!");
5964 journal = ext4_get_journal(sb, journal_inum);
5968 journal = ext4_get_dev_journal(sb, journal_dev);
5973 journal_dev_ro = bdev_read_only(journal->j_dev);
5974 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5976 if (journal_dev_ro && !sb_rdonly(sb)) {
5977 ext4_msg(sb, KERN_ERR,
5978 "journal device read-only, try mounting with '-o ro'");
5984 * Are we loading a blank journal or performing recovery after a
5985 * crash? For recovery, we need to check in advance whether we
5986 * can get read-write access to the device.
5988 if (ext4_has_feature_journal_needs_recovery(sb)) {
5989 if (sb_rdonly(sb)) {
5990 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5991 "required on readonly filesystem");
5992 if (really_read_only) {
5993 ext4_msg(sb, KERN_ERR, "write access "
5994 "unavailable, cannot proceed "
5995 "(try mounting with noload)");
5999 ext4_msg(sb, KERN_INFO, "write access will "
6000 "be enabled during recovery");
6004 if (!(journal->j_flags & JBD2_BARRIER))
6005 ext4_msg(sb, KERN_INFO, "barriers disabled");
6007 if (!ext4_has_feature_journal_needs_recovery(sb))
6008 err = jbd2_journal_wipe(journal, !really_read_only);
6010 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6012 bool changed = false;
6015 memcpy(save, ((char *) es) +
6016 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6017 err = jbd2_journal_load(journal);
6018 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6019 save, EXT4_S_ERR_LEN)) {
6020 memcpy(((char *) es) + EXT4_S_ERR_START,
6021 save, EXT4_S_ERR_LEN);
6025 orig_state = es->s_state;
6026 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6028 if (orig_state != es->s_state)
6030 /* Write out restored error information to the superblock */
6031 if (changed && !really_read_only) {
6033 err2 = ext4_commit_super(sb);
6039 ext4_msg(sb, KERN_ERR, "error loading journal");
6043 EXT4_SB(sb)->s_journal = journal;
6044 err = ext4_clear_journal_err(sb, es);
6046 EXT4_SB(sb)->s_journal = NULL;
6047 jbd2_journal_destroy(journal);
6051 if (!really_read_only && journal_devnum &&
6052 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6053 es->s_journal_dev = cpu_to_le32(journal_devnum);
6054 ext4_commit_super(sb);
6056 if (!really_read_only && journal_inum &&
6057 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6058 es->s_journal_inum = cpu_to_le32(journal_inum);
6059 ext4_commit_super(sb);
6065 jbd2_journal_destroy(journal);
6069 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6070 static void ext4_update_super(struct super_block *sb)
6072 struct ext4_sb_info *sbi = EXT4_SB(sb);
6073 struct ext4_super_block *es = sbi->s_es;
6074 struct buffer_head *sbh = sbi->s_sbh;
6078 * If the file system is mounted read-only, don't update the
6079 * superblock write time. This avoids updating the superblock
6080 * write time when we are mounting the root file system
6081 * read/only but we need to replay the journal; at that point,
6082 * for people who are east of GMT and who make their clock
6083 * tick in localtime for Windows bug-for-bug compatibility,
6084 * the clock is set in the future, and this will cause e2fsck
6085 * to complain and force a full file system check.
6087 if (!(sb->s_flags & SB_RDONLY))
6088 ext4_update_tstamp(es, s_wtime);
6089 es->s_kbytes_written =
6090 cpu_to_le64(sbi->s_kbytes_written +
6091 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6092 sbi->s_sectors_written_start) >> 1));
6093 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6094 ext4_free_blocks_count_set(es,
6095 EXT4_C2B(sbi, percpu_counter_sum_positive(
6096 &sbi->s_freeclusters_counter)));
6097 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6098 es->s_free_inodes_count =
6099 cpu_to_le32(percpu_counter_sum_positive(
6100 &sbi->s_freeinodes_counter));
6101 /* Copy error information to the on-disk superblock */
6102 spin_lock(&sbi->s_error_lock);
6103 if (sbi->s_add_error_count > 0) {
6104 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6105 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6106 __ext4_update_tstamp(&es->s_first_error_time,
6107 &es->s_first_error_time_hi,
6108 sbi->s_first_error_time);
6109 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6110 sizeof(es->s_first_error_func));
6111 es->s_first_error_line =
6112 cpu_to_le32(sbi->s_first_error_line);
6113 es->s_first_error_ino =
6114 cpu_to_le32(sbi->s_first_error_ino);
6115 es->s_first_error_block =
6116 cpu_to_le64(sbi->s_first_error_block);
6117 es->s_first_error_errcode =
6118 ext4_errno_to_code(sbi->s_first_error_code);
6120 __ext4_update_tstamp(&es->s_last_error_time,
6121 &es->s_last_error_time_hi,
6122 sbi->s_last_error_time);
6123 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6124 sizeof(es->s_last_error_func));
6125 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6126 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6127 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6128 es->s_last_error_errcode =
6129 ext4_errno_to_code(sbi->s_last_error_code);
6131 * Start the daily error reporting function if it hasn't been
6134 if (!es->s_error_count)
6135 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6136 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6137 sbi->s_add_error_count = 0;
6139 spin_unlock(&sbi->s_error_lock);
6141 ext4_superblock_csum_set(sb);
6145 static int ext4_commit_super(struct super_block *sb)
6147 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6151 if (block_device_ejected(sb))
6154 ext4_update_super(sb);
6157 /* Buffer got discarded which means block device got invalidated */
6158 if (!buffer_mapped(sbh)) {
6163 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6165 * Oh, dear. A previous attempt to write the
6166 * superblock failed. This could happen because the
6167 * USB device was yanked out. Or it could happen to
6168 * be a transient write error and maybe the block will
6169 * be remapped. Nothing we can do but to retry the
6170 * write and hope for the best.
6172 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6173 "superblock detected");
6174 clear_buffer_write_io_error(sbh);
6175 set_buffer_uptodate(sbh);
6178 /* Clear potential dirty bit if it was journalled update */
6179 clear_buffer_dirty(sbh);
6180 sbh->b_end_io = end_buffer_write_sync;
6181 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6182 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6183 wait_on_buffer(sbh);
6184 if (buffer_write_io_error(sbh)) {
6185 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6187 clear_buffer_write_io_error(sbh);
6188 set_buffer_uptodate(sbh);
6195 * Have we just finished recovery? If so, and if we are mounting (or
6196 * remounting) the filesystem readonly, then we will end up with a
6197 * consistent fs on disk. Record that fact.
6199 static int ext4_mark_recovery_complete(struct super_block *sb,
6200 struct ext4_super_block *es)
6203 journal_t *journal = EXT4_SB(sb)->s_journal;
6205 if (!ext4_has_feature_journal(sb)) {
6206 if (journal != NULL) {
6207 ext4_error(sb, "Journal got removed while the fs was "
6209 return -EFSCORRUPTED;
6213 jbd2_journal_lock_updates(journal);
6214 err = jbd2_journal_flush(journal, 0);
6218 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6219 ext4_has_feature_orphan_present(sb))) {
6220 if (!ext4_orphan_file_empty(sb)) {
6221 ext4_error(sb, "Orphan file not empty on read-only fs.");
6222 err = -EFSCORRUPTED;
6225 ext4_clear_feature_journal_needs_recovery(sb);
6226 ext4_clear_feature_orphan_present(sb);
6227 ext4_commit_super(sb);
6230 jbd2_journal_unlock_updates(journal);
6235 * If we are mounting (or read-write remounting) a filesystem whose journal
6236 * has recorded an error from a previous lifetime, move that error to the
6237 * main filesystem now.
6239 static int ext4_clear_journal_err(struct super_block *sb,
6240 struct ext4_super_block *es)
6246 if (!ext4_has_feature_journal(sb)) {
6247 ext4_error(sb, "Journal got removed while the fs was mounted!");
6248 return -EFSCORRUPTED;
6251 journal = EXT4_SB(sb)->s_journal;
6254 * Now check for any error status which may have been recorded in the
6255 * journal by a prior ext4_error() or ext4_abort()
6258 j_errno = jbd2_journal_errno(journal);
6262 errstr = ext4_decode_error(sb, j_errno, nbuf);
6263 ext4_warning(sb, "Filesystem error recorded "
6264 "from previous mount: %s", errstr);
6266 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6267 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6268 j_errno = ext4_commit_super(sb);
6271 ext4_warning(sb, "Marked fs in need of filesystem check.");
6273 jbd2_journal_clear_err(journal);
6274 jbd2_journal_update_sb_errno(journal);
6280 * Force the running and committing transactions to commit,
6281 * and wait on the commit.
6283 int ext4_force_commit(struct super_block *sb)
6290 journal = EXT4_SB(sb)->s_journal;
6291 return ext4_journal_force_commit(journal);
6294 static int ext4_sync_fs(struct super_block *sb, int wait)
6298 bool needs_barrier = false;
6299 struct ext4_sb_info *sbi = EXT4_SB(sb);
6301 if (unlikely(ext4_forced_shutdown(sbi)))
6304 trace_ext4_sync_fs(sb, wait);
6305 flush_workqueue(sbi->rsv_conversion_wq);
6307 * Writeback quota in non-journalled quota case - journalled quota has
6310 dquot_writeback_dquots(sb, -1);
6312 * Data writeback is possible w/o journal transaction, so barrier must
6313 * being sent at the end of the function. But we can skip it if
6314 * transaction_commit will do it for us.
6316 if (sbi->s_journal) {
6317 target = jbd2_get_latest_transaction(sbi->s_journal);
6318 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6319 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6320 needs_barrier = true;
6322 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6324 ret = jbd2_log_wait_commit(sbi->s_journal,
6327 } else if (wait && test_opt(sb, BARRIER))
6328 needs_barrier = true;
6329 if (needs_barrier) {
6331 err = blkdev_issue_flush(sb->s_bdev);
6340 * LVM calls this function before a (read-only) snapshot is created. This
6341 * gives us a chance to flush the journal completely and mark the fs clean.
6343 * Note that only this function cannot bring a filesystem to be in a clean
6344 * state independently. It relies on upper layer to stop all data & metadata
6347 static int ext4_freeze(struct super_block *sb)
6355 journal = EXT4_SB(sb)->s_journal;
6358 /* Now we set up the journal barrier. */
6359 jbd2_journal_lock_updates(journal);
6362 * Don't clear the needs_recovery flag if we failed to
6363 * flush the journal.
6365 error = jbd2_journal_flush(journal, 0);
6369 /* Journal blocked and flushed, clear needs_recovery flag. */
6370 ext4_clear_feature_journal_needs_recovery(sb);
6371 if (ext4_orphan_file_empty(sb))
6372 ext4_clear_feature_orphan_present(sb);
6375 error = ext4_commit_super(sb);
6378 /* we rely on upper layer to stop further updates */
6379 jbd2_journal_unlock_updates(journal);
6384 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6385 * flag here, even though the filesystem is not technically dirty yet.
6387 static int ext4_unfreeze(struct super_block *sb)
6389 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6392 if (EXT4_SB(sb)->s_journal) {
6393 /* Reset the needs_recovery flag before the fs is unlocked. */
6394 ext4_set_feature_journal_needs_recovery(sb);
6395 if (ext4_has_feature_orphan_file(sb))
6396 ext4_set_feature_orphan_present(sb);
6399 ext4_commit_super(sb);
6404 * Structure to save mount options for ext4_remount's benefit
6406 struct ext4_mount_options {
6407 unsigned long s_mount_opt;
6408 unsigned long s_mount_opt2;
6411 unsigned long s_commit_interval;
6412 u32 s_min_batch_time, s_max_batch_time;
6415 char *s_qf_names[EXT4_MAXQUOTAS];
6419 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6421 struct ext4_fs_context *ctx = fc->fs_private;
6422 struct ext4_super_block *es;
6423 struct ext4_sb_info *sbi = EXT4_SB(sb);
6424 unsigned long old_sb_flags;
6425 struct ext4_mount_options old_opts;
6429 int enable_quota = 0;
6431 char *to_free[EXT4_MAXQUOTAS];
6435 /* Store the original options */
6436 old_sb_flags = sb->s_flags;
6437 old_opts.s_mount_opt = sbi->s_mount_opt;
6438 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6439 old_opts.s_resuid = sbi->s_resuid;
6440 old_opts.s_resgid = sbi->s_resgid;
6441 old_opts.s_commit_interval = sbi->s_commit_interval;
6442 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6443 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6445 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6446 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6447 if (sbi->s_qf_names[i]) {
6448 char *qf_name = get_qf_name(sb, sbi, i);
6450 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6451 if (!old_opts.s_qf_names[i]) {
6452 for (j = 0; j < i; j++)
6453 kfree(old_opts.s_qf_names[j]);
6457 old_opts.s_qf_names[i] = NULL;
6459 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6460 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6461 ctx->journal_ioprio =
6462 sbi->s_journal->j_task->io_context->ioprio;
6464 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6468 ext4_apply_options(fc, sb);
6470 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6471 test_opt(sb, JOURNAL_CHECKSUM)) {
6472 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6473 "during remount not supported; ignoring");
6474 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6477 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6478 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6479 ext4_msg(sb, KERN_ERR, "can't mount with "
6480 "both data=journal and delalloc");
6484 if (test_opt(sb, DIOREAD_NOLOCK)) {
6485 ext4_msg(sb, KERN_ERR, "can't mount with "
6486 "both data=journal and dioread_nolock");
6490 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6491 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6492 ext4_msg(sb, KERN_ERR, "can't mount with "
6493 "journal_async_commit in data=ordered mode");
6499 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6500 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6505 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6506 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6508 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6509 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6513 if (sbi->s_journal) {
6514 ext4_init_journal_params(sb, sbi->s_journal);
6515 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6518 /* Flush outstanding errors before changing fs state */
6519 flush_work(&sbi->s_error_work);
6521 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6522 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6527 if (fc->sb_flags & SB_RDONLY) {
6528 err = sync_filesystem(sb);
6531 err = dquot_suspend(sb, -1);
6536 * First of all, the unconditional stuff we have to do
6537 * to disable replay of the journal when we next remount
6539 sb->s_flags |= SB_RDONLY;
6542 * OK, test if we are remounting a valid rw partition
6543 * readonly, and if so set the rdonly flag and then
6544 * mark the partition as valid again.
6546 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6547 (sbi->s_mount_state & EXT4_VALID_FS))
6548 es->s_state = cpu_to_le16(sbi->s_mount_state);
6550 if (sbi->s_journal) {
6552 * We let remount-ro finish even if marking fs
6553 * as clean failed...
6555 ext4_mark_recovery_complete(sb, es);
6558 /* Make sure we can mount this feature set readwrite */
6559 if (ext4_has_feature_readonly(sb) ||
6560 !ext4_feature_set_ok(sb, 0)) {
6565 * Make sure the group descriptor checksums
6566 * are sane. If they aren't, refuse to remount r/w.
6568 for (g = 0; g < sbi->s_groups_count; g++) {
6569 struct ext4_group_desc *gdp =
6570 ext4_get_group_desc(sb, g, NULL);
6572 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6573 ext4_msg(sb, KERN_ERR,
6574 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6575 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6576 le16_to_cpu(gdp->bg_checksum));
6583 * If we have an unprocessed orphan list hanging
6584 * around from a previously readonly bdev mount,
6585 * require a full umount/remount for now.
6587 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6588 ext4_msg(sb, KERN_WARNING, "Couldn't "
6589 "remount RDWR because of unprocessed "
6590 "orphan inode list. Please "
6591 "umount/remount instead");
6597 * Mounting a RDONLY partition read-write, so reread
6598 * and store the current valid flag. (It may have
6599 * been changed by e2fsck since we originally mounted
6602 if (sbi->s_journal) {
6603 err = ext4_clear_journal_err(sb, es);
6607 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6610 err = ext4_setup_super(sb, es, 0);
6614 sb->s_flags &= ~SB_RDONLY;
6615 if (ext4_has_feature_mmp(sb)) {
6616 err = ext4_multi_mount_protect(sb,
6617 le64_to_cpu(es->s_mmp_block));
6628 * Handle creation of system zone data early because it can fail.
6629 * Releasing of existing data is done when we are sure remount will
6632 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6633 err = ext4_setup_system_zone(sb);
6638 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6639 err = ext4_commit_super(sb);
6646 if (sb_any_quota_suspended(sb))
6647 dquot_resume(sb, -1);
6648 else if (ext4_has_feature_quota(sb)) {
6649 err = ext4_enable_quotas(sb);
6654 /* Release old quota file names */
6655 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6656 kfree(old_opts.s_qf_names[i]);
6658 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6659 ext4_release_system_zone(sb);
6662 * Reinitialize lazy itable initialization thread based on
6665 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6666 ext4_unregister_li_request(sb);
6668 ext4_group_t first_not_zeroed;
6669 first_not_zeroed = ext4_has_uninit_itable(sb);
6670 ext4_register_li_request(sb, first_not_zeroed);
6673 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6674 ext4_stop_mmpd(sbi);
6680 * If there was a failing r/w to ro transition, we may need to
6683 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6684 sb_any_quota_suspended(sb))
6685 dquot_resume(sb, -1);
6686 sb->s_flags = old_sb_flags;
6687 sbi->s_mount_opt = old_opts.s_mount_opt;
6688 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6689 sbi->s_resuid = old_opts.s_resuid;
6690 sbi->s_resgid = old_opts.s_resgid;
6691 sbi->s_commit_interval = old_opts.s_commit_interval;
6692 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6693 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6694 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6695 ext4_release_system_zone(sb);
6697 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6698 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6699 to_free[i] = get_qf_name(sb, sbi, i);
6700 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6703 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6706 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6707 ext4_stop_mmpd(sbi);
6711 static int ext4_reconfigure(struct fs_context *fc)
6713 struct super_block *sb = fc->root->d_sb;
6716 fc->s_fs_info = EXT4_SB(sb);
6718 ret = ext4_check_opt_consistency(fc, sb);
6722 ret = __ext4_remount(fc, sb);
6726 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6727 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6728 ext4_quota_mode(sb));
6734 static int ext4_statfs_project(struct super_block *sb,
6735 kprojid_t projid, struct kstatfs *buf)
6738 struct dquot *dquot;
6742 qid = make_kqid_projid(projid);
6743 dquot = dqget(sb, qid);
6745 return PTR_ERR(dquot);
6746 spin_lock(&dquot->dq_dqb_lock);
6748 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6749 dquot->dq_dqb.dqb_bhardlimit);
6750 limit >>= sb->s_blocksize_bits;
6752 if (limit && buf->f_blocks > limit) {
6753 curblock = (dquot->dq_dqb.dqb_curspace +
6754 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6755 buf->f_blocks = limit;
6756 buf->f_bfree = buf->f_bavail =
6757 (buf->f_blocks > curblock) ?
6758 (buf->f_blocks - curblock) : 0;
6761 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6762 dquot->dq_dqb.dqb_ihardlimit);
6763 if (limit && buf->f_files > limit) {
6764 buf->f_files = limit;
6766 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6767 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6770 spin_unlock(&dquot->dq_dqb_lock);
6776 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6778 struct super_block *sb = dentry->d_sb;
6779 struct ext4_sb_info *sbi = EXT4_SB(sb);
6780 struct ext4_super_block *es = sbi->s_es;
6781 ext4_fsblk_t overhead = 0, resv_blocks;
6783 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6785 if (!test_opt(sb, MINIX_DF))
6786 overhead = sbi->s_overhead;
6788 buf->f_type = EXT4_SUPER_MAGIC;
6789 buf->f_bsize = sb->s_blocksize;
6790 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6791 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6792 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6793 /* prevent underflow in case that few free space is available */
6794 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6795 buf->f_bavail = buf->f_bfree -
6796 (ext4_r_blocks_count(es) + resv_blocks);
6797 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6799 buf->f_files = le32_to_cpu(es->s_inodes_count);
6800 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6801 buf->f_namelen = EXT4_NAME_LEN;
6802 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6805 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6806 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6807 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6816 * Helper functions so that transaction is started before we acquire dqio_sem
6817 * to keep correct lock ordering of transaction > dqio_sem
6819 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6821 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6824 static int ext4_write_dquot(struct dquot *dquot)
6828 struct inode *inode;
6830 inode = dquot_to_inode(dquot);
6831 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6832 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6834 return PTR_ERR(handle);
6835 ret = dquot_commit(dquot);
6836 err = ext4_journal_stop(handle);
6842 static int ext4_acquire_dquot(struct dquot *dquot)
6847 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6848 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6850 return PTR_ERR(handle);
6851 ret = dquot_acquire(dquot);
6852 err = ext4_journal_stop(handle);
6858 static int ext4_release_dquot(struct dquot *dquot)
6863 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6864 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6865 if (IS_ERR(handle)) {
6866 /* Release dquot anyway to avoid endless cycle in dqput() */
6867 dquot_release(dquot);
6868 return PTR_ERR(handle);
6870 ret = dquot_release(dquot);
6871 err = ext4_journal_stop(handle);
6877 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6879 struct super_block *sb = dquot->dq_sb;
6881 if (ext4_is_quota_journalled(sb)) {
6882 dquot_mark_dquot_dirty(dquot);
6883 return ext4_write_dquot(dquot);
6885 return dquot_mark_dquot_dirty(dquot);
6889 static int ext4_write_info(struct super_block *sb, int type)
6894 /* Data block + inode block */
6895 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6897 return PTR_ERR(handle);
6898 ret = dquot_commit_info(sb, type);
6899 err = ext4_journal_stop(handle);
6905 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6907 struct ext4_inode_info *ei = EXT4_I(inode);
6909 /* The first argument of lockdep_set_subclass has to be
6910 * *exactly* the same as the argument to init_rwsem() --- in
6911 * this case, in init_once() --- or lockdep gets unhappy
6912 * because the name of the lock is set using the
6913 * stringification of the argument to init_rwsem().
6915 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6916 lockdep_set_subclass(&ei->i_data_sem, subclass);
6920 * Standard function to be called on quota_on
6922 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6923 const struct path *path)
6927 if (!test_opt(sb, QUOTA))
6930 /* Quotafile not on the same filesystem? */
6931 if (path->dentry->d_sb != sb)
6934 /* Quota already enabled for this file? */
6935 if (IS_NOQUOTA(d_inode(path->dentry)))
6938 /* Journaling quota? */
6939 if (EXT4_SB(sb)->s_qf_names[type]) {
6940 /* Quotafile not in fs root? */
6941 if (path->dentry->d_parent != sb->s_root)
6942 ext4_msg(sb, KERN_WARNING,
6943 "Quota file not on filesystem root. "
6944 "Journaled quota will not work");
6945 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6948 * Clear the flag just in case mount options changed since
6951 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6954 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6955 err = dquot_quota_on(sb, type, format_id, path);
6957 struct inode *inode = d_inode(path->dentry);
6961 * Set inode flags to prevent userspace from messing with quota
6962 * files. If this fails, we return success anyway since quotas
6963 * are already enabled and this is not a hard failure.
6966 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6969 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6970 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6971 S_NOATIME | S_IMMUTABLE);
6972 err = ext4_mark_inode_dirty(handle, inode);
6973 ext4_journal_stop(handle);
6975 inode_unlock(inode);
6977 dquot_quota_off(sb, type);
6980 lockdep_set_quota_inode(path->dentry->d_inode,
6985 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6989 return qf_inum == EXT4_USR_QUOTA_INO;
6991 return qf_inum == EXT4_GRP_QUOTA_INO;
6993 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6999 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7003 struct inode *qf_inode;
7004 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7005 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7006 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7007 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7010 BUG_ON(!ext4_has_feature_quota(sb));
7012 if (!qf_inums[type])
7015 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7016 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7017 qf_inums[type], type);
7021 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7022 if (IS_ERR(qf_inode)) {
7023 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7024 qf_inums[type], type);
7025 return PTR_ERR(qf_inode);
7028 /* Don't account quota for quota files to avoid recursion */
7029 qf_inode->i_flags |= S_NOQUOTA;
7030 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7031 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
7033 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7039 /* Enable usage tracking for all quota types. */
7040 int ext4_enable_quotas(struct super_block *sb)
7043 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7044 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7045 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7046 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7048 bool quota_mopt[EXT4_MAXQUOTAS] = {
7049 test_opt(sb, USRQUOTA),
7050 test_opt(sb, GRPQUOTA),
7051 test_opt(sb, PRJQUOTA),
7054 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7055 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7056 if (qf_inums[type]) {
7057 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7058 DQUOT_USAGE_ENABLED |
7059 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7062 "Failed to enable quota tracking "
7063 "(type=%d, err=%d, ino=%lu). "
7064 "Please run e2fsck to fix.", type,
7065 err, qf_inums[type]);
7067 ext4_quotas_off(sb, type);
7075 static int ext4_quota_off(struct super_block *sb, int type)
7077 struct inode *inode = sb_dqopt(sb)->files[type];
7081 /* Force all delayed allocation blocks to be allocated.
7082 * Caller already holds s_umount sem */
7083 if (test_opt(sb, DELALLOC))
7084 sync_filesystem(sb);
7086 if (!inode || !igrab(inode))
7089 err = dquot_quota_off(sb, type);
7090 if (err || ext4_has_feature_quota(sb))
7095 * Update modification times of quota files when userspace can
7096 * start looking at them. If we fail, we return success anyway since
7097 * this is not a hard failure and quotas are already disabled.
7099 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7100 if (IS_ERR(handle)) {
7101 err = PTR_ERR(handle);
7104 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7105 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7106 inode->i_mtime = inode->i_ctime = current_time(inode);
7107 err = ext4_mark_inode_dirty(handle, inode);
7108 ext4_journal_stop(handle);
7110 inode_unlock(inode);
7112 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7116 return dquot_quota_off(sb, type);
7119 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7120 * acquiring the locks... As quota files are never truncated and quota code
7121 * itself serializes the operations (and no one else should touch the files)
7122 * we don't have to be afraid of races */
7123 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7124 size_t len, loff_t off)
7126 struct inode *inode = sb_dqopt(sb)->files[type];
7127 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7128 int offset = off & (sb->s_blocksize - 1);
7131 struct buffer_head *bh;
7132 loff_t i_size = i_size_read(inode);
7136 if (off+len > i_size)
7139 while (toread > 0) {
7140 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7141 bh = ext4_bread(NULL, inode, blk, 0);
7144 if (!bh) /* A hole? */
7145 memset(data, 0, tocopy);
7147 memcpy(data, bh->b_data+offset, tocopy);
7157 /* Write to quotafile (we know the transaction is already started and has
7158 * enough credits) */
7159 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7160 const char *data, size_t len, loff_t off)
7162 struct inode *inode = sb_dqopt(sb)->files[type];
7163 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7164 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7166 struct buffer_head *bh;
7167 handle_t *handle = journal_current_handle();
7170 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7171 " cancelled because transaction is not started",
7172 (unsigned long long)off, (unsigned long long)len);
7176 * Since we account only one data block in transaction credits,
7177 * then it is impossible to cross a block boundary.
7179 if (sb->s_blocksize - offset < len) {
7180 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7181 " cancelled because not block aligned",
7182 (unsigned long long)off, (unsigned long long)len);
7187 bh = ext4_bread(handle, inode, blk,
7188 EXT4_GET_BLOCKS_CREATE |
7189 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7190 } while (PTR_ERR(bh) == -ENOSPC &&
7191 ext4_should_retry_alloc(inode->i_sb, &retries));
7196 BUFFER_TRACE(bh, "get write access");
7197 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7203 memcpy(bh->b_data+offset, data, len);
7204 flush_dcache_page(bh->b_page);
7206 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7209 if (inode->i_size < off + len) {
7210 i_size_write(inode, off + len);
7211 EXT4_I(inode)->i_disksize = inode->i_size;
7212 err2 = ext4_mark_inode_dirty(handle, inode);
7213 if (unlikely(err2 && !err))
7216 return err ? err : len;
7220 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7221 static inline void register_as_ext2(void)
7223 int err = register_filesystem(&ext2_fs_type);
7226 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7229 static inline void unregister_as_ext2(void)
7231 unregister_filesystem(&ext2_fs_type);
7234 static inline int ext2_feature_set_ok(struct super_block *sb)
7236 if (ext4_has_unknown_ext2_incompat_features(sb))
7240 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7245 static inline void register_as_ext2(void) { }
7246 static inline void unregister_as_ext2(void) { }
7247 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7250 static inline void register_as_ext3(void)
7252 int err = register_filesystem(&ext3_fs_type);
7255 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7258 static inline void unregister_as_ext3(void)
7260 unregister_filesystem(&ext3_fs_type);
7263 static inline int ext3_feature_set_ok(struct super_block *sb)
7265 if (ext4_has_unknown_ext3_incompat_features(sb))
7267 if (!ext4_has_feature_journal(sb))
7271 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7276 static struct file_system_type ext4_fs_type = {
7277 .owner = THIS_MODULE,
7279 .init_fs_context = ext4_init_fs_context,
7280 .parameters = ext4_param_specs,
7281 .kill_sb = kill_block_super,
7282 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7284 MODULE_ALIAS_FS("ext4");
7286 /* Shared across all ext4 file systems */
7287 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7289 static int __init ext4_init_fs(void)
7293 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7294 ext4_li_info = NULL;
7296 /* Build-time check for flags consistency */
7297 ext4_check_flag_values();
7299 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7300 init_waitqueue_head(&ext4__ioend_wq[i]);
7302 err = ext4_init_es();
7306 err = ext4_init_pending();
7310 err = ext4_init_post_read_processing();
7314 err = ext4_init_pageio();
7318 err = ext4_init_system_zone();
7322 err = ext4_init_sysfs();
7326 err = ext4_init_mballoc();
7329 err = init_inodecache();
7333 err = ext4_fc_init_dentry_cache();
7339 err = register_filesystem(&ext4_fs_type);
7345 unregister_as_ext2();
7346 unregister_as_ext3();
7347 ext4_fc_destroy_dentry_cache();
7349 destroy_inodecache();
7351 ext4_exit_mballoc();
7355 ext4_exit_system_zone();
7359 ext4_exit_post_read_processing();
7361 ext4_exit_pending();
7368 static void __exit ext4_exit_fs(void)
7370 ext4_destroy_lazyinit_thread();
7371 unregister_as_ext2();
7372 unregister_as_ext3();
7373 unregister_filesystem(&ext4_fs_type);
7374 ext4_fc_destroy_dentry_cache();
7375 destroy_inodecache();
7376 ext4_exit_mballoc();
7378 ext4_exit_system_zone();
7380 ext4_exit_post_read_processing();
7382 ext4_exit_pending();
7385 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7386 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7387 MODULE_LICENSE("GPL");
7388 MODULE_SOFTDEP("pre: crc32c");
7389 module_init(ext4_init_fs)
7390 module_exit(ext4_exit_fs)