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.
1100 * Open the external journal device
1102 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1104 struct block_device *bdev;
1106 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1112 ext4_msg(sb, KERN_ERR,
1113 "failed to open journal device unknown-block(%u,%u) %ld",
1114 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1119 * Release the journal device
1121 static void ext4_blkdev_put(struct block_device *bdev)
1123 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1126 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1128 struct block_device *bdev;
1129 bdev = sbi->s_journal_bdev;
1131 ext4_blkdev_put(bdev);
1132 sbi->s_journal_bdev = NULL;
1136 static inline struct inode *orphan_list_entry(struct list_head *l)
1138 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1141 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1143 struct list_head *l;
1145 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1146 le32_to_cpu(sbi->s_es->s_last_orphan));
1148 printk(KERN_ERR "sb_info orphan list:\n");
1149 list_for_each(l, &sbi->s_orphan) {
1150 struct inode *inode = orphan_list_entry(l);
1152 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1153 inode->i_sb->s_id, inode->i_ino, inode,
1154 inode->i_mode, inode->i_nlink,
1155 NEXT_ORPHAN(inode));
1160 static int ext4_quota_off(struct super_block *sb, int type);
1162 static inline void ext4_quota_off_umount(struct super_block *sb)
1166 /* Use our quota_off function to clear inode flags etc. */
1167 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1168 ext4_quota_off(sb, type);
1172 * This is a helper function which is used in the mount/remount
1173 * codepaths (which holds s_umount) to fetch the quota file name.
1175 static inline char *get_qf_name(struct super_block *sb,
1176 struct ext4_sb_info *sbi,
1179 return rcu_dereference_protected(sbi->s_qf_names[type],
1180 lockdep_is_held(&sb->s_umount));
1183 static inline void ext4_quota_off_umount(struct super_block *sb)
1188 static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1193 block = ext4_count_free_clusters(sbi->s_sb);
1194 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1195 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1198 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1199 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1200 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1204 err = percpu_counter_init(&sbi->s_dirs_counter,
1205 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1207 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1210 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1213 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1216 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1221 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1223 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1224 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1225 percpu_counter_destroy(&sbi->s_dirs_counter);
1226 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1227 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1228 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1231 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1233 struct buffer_head **group_desc;
1237 group_desc = rcu_dereference(sbi->s_group_desc);
1238 for (i = 0; i < sbi->s_gdb_count; i++)
1239 brelse(group_desc[i]);
1244 static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1246 struct flex_groups **flex_groups;
1250 flex_groups = rcu_dereference(sbi->s_flex_groups);
1252 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1253 kvfree(flex_groups[i]);
1254 kvfree(flex_groups);
1259 static void ext4_put_super(struct super_block *sb)
1261 struct ext4_sb_info *sbi = EXT4_SB(sb);
1262 struct ext4_super_block *es = sbi->s_es;
1267 * Unregister sysfs before destroying jbd2 journal.
1268 * Since we could still access attr_journal_task attribute via sysfs
1269 * path which could have sbi->s_journal->j_task as NULL
1270 * Unregister sysfs before flush sbi->s_error_work.
1271 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1272 * read metadata verify failed then will queue error work.
1273 * flush_stashed_error_work will call start_this_handle may trigger
1276 ext4_unregister_sysfs(sb);
1278 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1279 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1282 ext4_unregister_li_request(sb);
1283 ext4_quota_off_umount(sb);
1285 flush_work(&sbi->s_error_work);
1286 destroy_workqueue(sbi->rsv_conversion_wq);
1287 ext4_release_orphan_info(sb);
1289 if (sbi->s_journal) {
1290 aborted = is_journal_aborted(sbi->s_journal);
1291 err = jbd2_journal_destroy(sbi->s_journal);
1292 sbi->s_journal = NULL;
1293 if ((err < 0) && !aborted) {
1294 ext4_abort(sb, -err, "Couldn't clean up the journal");
1298 ext4_es_unregister_shrinker(sbi);
1299 timer_shutdown_sync(&sbi->s_err_report);
1300 ext4_release_system_zone(sb);
1301 ext4_mb_release(sb);
1302 ext4_ext_release(sb);
1304 if (!sb_rdonly(sb) && !aborted) {
1305 ext4_clear_feature_journal_needs_recovery(sb);
1306 ext4_clear_feature_orphan_present(sb);
1307 es->s_state = cpu_to_le16(sbi->s_mount_state);
1310 ext4_commit_super(sb);
1312 ext4_group_desc_free(sbi);
1313 ext4_flex_groups_free(sbi);
1314 ext4_percpu_param_destroy(sbi);
1316 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1317 kfree(get_qf_name(sb, sbi, i));
1320 /* Debugging code just in case the in-memory inode orphan list
1321 * isn't empty. The on-disk one can be non-empty if we've
1322 * detected an error and taken the fs readonly, but the
1323 * in-memory list had better be clean by this point. */
1324 if (!list_empty(&sbi->s_orphan))
1325 dump_orphan_list(sb, sbi);
1326 ASSERT(list_empty(&sbi->s_orphan));
1328 sync_blockdev(sb->s_bdev);
1329 invalidate_bdev(sb->s_bdev);
1330 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1332 * Invalidate the journal device's buffers. We don't want them
1333 * floating about in memory - the physical journal device may
1334 * hotswapped, and it breaks the `ro-after' testing code.
1336 sync_blockdev(sbi->s_journal_bdev);
1337 invalidate_bdev(sbi->s_journal_bdev);
1338 ext4_blkdev_remove(sbi);
1341 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1342 sbi->s_ea_inode_cache = NULL;
1344 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1345 sbi->s_ea_block_cache = NULL;
1347 ext4_stop_mmpd(sbi);
1350 sb->s_fs_info = NULL;
1352 * Now that we are completely done shutting down the
1353 * superblock, we need to actually destroy the kobject.
1355 kobject_put(&sbi->s_kobj);
1356 wait_for_completion(&sbi->s_kobj_unregister);
1357 if (sbi->s_chksum_driver)
1358 crypto_free_shash(sbi->s_chksum_driver);
1359 kfree(sbi->s_blockgroup_lock);
1360 fs_put_dax(sbi->s_daxdev, NULL);
1361 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1362 #if IS_ENABLED(CONFIG_UNICODE)
1363 utf8_unload(sb->s_encoding);
1368 static struct kmem_cache *ext4_inode_cachep;
1371 * Called inside transaction, so use GFP_NOFS
1373 static struct inode *ext4_alloc_inode(struct super_block *sb)
1375 struct ext4_inode_info *ei;
1377 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1381 inode_set_iversion(&ei->vfs_inode, 1);
1383 spin_lock_init(&ei->i_raw_lock);
1384 ei->i_prealloc_node = RB_ROOT;
1385 atomic_set(&ei->i_prealloc_active, 0);
1386 rwlock_init(&ei->i_prealloc_lock);
1387 ext4_es_init_tree(&ei->i_es_tree);
1388 rwlock_init(&ei->i_es_lock);
1389 INIT_LIST_HEAD(&ei->i_es_list);
1390 ei->i_es_all_nr = 0;
1391 ei->i_es_shk_nr = 0;
1392 ei->i_es_shrink_lblk = 0;
1393 ei->i_reserved_data_blocks = 0;
1394 spin_lock_init(&(ei->i_block_reservation_lock));
1395 ext4_init_pending_tree(&ei->i_pending_tree);
1397 ei->i_reserved_quota = 0;
1398 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1401 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1402 spin_lock_init(&ei->i_completed_io_lock);
1404 ei->i_datasync_tid = 0;
1405 atomic_set(&ei->i_unwritten, 0);
1406 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1407 ext4_fc_init_inode(&ei->vfs_inode);
1408 mutex_init(&ei->i_fc_lock);
1409 return &ei->vfs_inode;
1412 static int ext4_drop_inode(struct inode *inode)
1414 int drop = generic_drop_inode(inode);
1417 drop = fscrypt_drop_inode(inode);
1419 trace_ext4_drop_inode(inode, drop);
1423 static void ext4_free_in_core_inode(struct inode *inode)
1425 fscrypt_free_inode(inode);
1426 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1427 pr_warn("%s: inode %ld still in fc list",
1428 __func__, inode->i_ino);
1430 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1433 static void ext4_destroy_inode(struct inode *inode)
1435 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1436 ext4_msg(inode->i_sb, KERN_ERR,
1437 "Inode %lu (%p): orphan list check failed!",
1438 inode->i_ino, EXT4_I(inode));
1439 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1440 EXT4_I(inode), sizeof(struct ext4_inode_info),
1445 if (EXT4_I(inode)->i_reserved_data_blocks)
1446 ext4_msg(inode->i_sb, KERN_ERR,
1447 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1448 inode->i_ino, EXT4_I(inode),
1449 EXT4_I(inode)->i_reserved_data_blocks);
1452 static void init_once(void *foo)
1454 struct ext4_inode_info *ei = foo;
1456 INIT_LIST_HEAD(&ei->i_orphan);
1457 init_rwsem(&ei->xattr_sem);
1458 init_rwsem(&ei->i_data_sem);
1459 inode_init_once(&ei->vfs_inode);
1460 ext4_fc_init_inode(&ei->vfs_inode);
1463 static int __init init_inodecache(void)
1465 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1466 sizeof(struct ext4_inode_info), 0,
1467 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1469 offsetof(struct ext4_inode_info, i_data),
1470 sizeof_field(struct ext4_inode_info, i_data),
1472 if (ext4_inode_cachep == NULL)
1477 static void destroy_inodecache(void)
1480 * Make sure all delayed rcu free inodes are flushed before we
1484 kmem_cache_destroy(ext4_inode_cachep);
1487 void ext4_clear_inode(struct inode *inode)
1490 invalidate_inode_buffers(inode);
1492 ext4_discard_preallocations(inode, 0);
1493 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1495 if (EXT4_I(inode)->jinode) {
1496 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1497 EXT4_I(inode)->jinode);
1498 jbd2_free_inode(EXT4_I(inode)->jinode);
1499 EXT4_I(inode)->jinode = NULL;
1501 fscrypt_put_encryption_info(inode);
1502 fsverity_cleanup_inode(inode);
1505 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1506 u64 ino, u32 generation)
1508 struct inode *inode;
1511 * Currently we don't know the generation for parent directory, so
1512 * a generation of 0 means "accept any"
1514 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1516 return ERR_CAST(inode);
1517 if (generation && inode->i_generation != generation) {
1519 return ERR_PTR(-ESTALE);
1525 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1526 int fh_len, int fh_type)
1528 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1529 ext4_nfs_get_inode);
1532 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1533 int fh_len, int fh_type)
1535 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1536 ext4_nfs_get_inode);
1539 static int ext4_nfs_commit_metadata(struct inode *inode)
1541 struct writeback_control wbc = {
1542 .sync_mode = WB_SYNC_ALL
1545 trace_ext4_nfs_commit_metadata(inode);
1546 return ext4_write_inode(inode, &wbc);
1550 static const char * const quotatypes[] = INITQFNAMES;
1551 #define QTYPE2NAME(t) (quotatypes[t])
1553 static int ext4_write_dquot(struct dquot *dquot);
1554 static int ext4_acquire_dquot(struct dquot *dquot);
1555 static int ext4_release_dquot(struct dquot *dquot);
1556 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1557 static int ext4_write_info(struct super_block *sb, int type);
1558 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1559 const struct path *path);
1560 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1561 size_t len, loff_t off);
1562 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1563 const char *data, size_t len, loff_t off);
1564 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1565 unsigned int flags);
1567 static struct dquot **ext4_get_dquots(struct inode *inode)
1569 return EXT4_I(inode)->i_dquot;
1572 static const struct dquot_operations ext4_quota_operations = {
1573 .get_reserved_space = ext4_get_reserved_space,
1574 .write_dquot = ext4_write_dquot,
1575 .acquire_dquot = ext4_acquire_dquot,
1576 .release_dquot = ext4_release_dquot,
1577 .mark_dirty = ext4_mark_dquot_dirty,
1578 .write_info = ext4_write_info,
1579 .alloc_dquot = dquot_alloc,
1580 .destroy_dquot = dquot_destroy,
1581 .get_projid = ext4_get_projid,
1582 .get_inode_usage = ext4_get_inode_usage,
1583 .get_next_id = dquot_get_next_id,
1586 static const struct quotactl_ops ext4_qctl_operations = {
1587 .quota_on = ext4_quota_on,
1588 .quota_off = ext4_quota_off,
1589 .quota_sync = dquot_quota_sync,
1590 .get_state = dquot_get_state,
1591 .set_info = dquot_set_dqinfo,
1592 .get_dqblk = dquot_get_dqblk,
1593 .set_dqblk = dquot_set_dqblk,
1594 .get_nextdqblk = dquot_get_next_dqblk,
1598 static const struct super_operations ext4_sops = {
1599 .alloc_inode = ext4_alloc_inode,
1600 .free_inode = ext4_free_in_core_inode,
1601 .destroy_inode = ext4_destroy_inode,
1602 .write_inode = ext4_write_inode,
1603 .dirty_inode = ext4_dirty_inode,
1604 .drop_inode = ext4_drop_inode,
1605 .evict_inode = ext4_evict_inode,
1606 .put_super = ext4_put_super,
1607 .sync_fs = ext4_sync_fs,
1608 .freeze_fs = ext4_freeze,
1609 .unfreeze_fs = ext4_unfreeze,
1610 .statfs = ext4_statfs,
1611 .show_options = ext4_show_options,
1613 .quota_read = ext4_quota_read,
1614 .quota_write = ext4_quota_write,
1615 .get_dquots = ext4_get_dquots,
1619 static const struct export_operations ext4_export_ops = {
1620 .fh_to_dentry = ext4_fh_to_dentry,
1621 .fh_to_parent = ext4_fh_to_parent,
1622 .get_parent = ext4_get_parent,
1623 .commit_metadata = ext4_nfs_commit_metadata,
1627 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1628 Opt_resgid, Opt_resuid, Opt_sb,
1629 Opt_nouid32, Opt_debug, Opt_removed,
1630 Opt_user_xattr, Opt_acl,
1631 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1632 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1633 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1634 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1635 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1637 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1638 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1639 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1640 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1641 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1642 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1643 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1644 Opt_inode_readahead_blks, Opt_journal_ioprio,
1645 Opt_dioread_nolock, Opt_dioread_lock,
1646 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1647 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1648 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1649 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1650 #ifdef CONFIG_EXT4_DEBUG
1651 Opt_fc_debug_max_replay, Opt_fc_debug_force
1655 static const struct constant_table ext4_param_errors[] = {
1656 {"continue", EXT4_MOUNT_ERRORS_CONT},
1657 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1658 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1662 static const struct constant_table ext4_param_data[] = {
1663 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1664 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1665 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1669 static const struct constant_table ext4_param_data_err[] = {
1670 {"abort", Opt_data_err_abort},
1671 {"ignore", Opt_data_err_ignore},
1675 static const struct constant_table ext4_param_jqfmt[] = {
1676 {"vfsold", QFMT_VFS_OLD},
1677 {"vfsv0", QFMT_VFS_V0},
1678 {"vfsv1", QFMT_VFS_V1},
1682 static const struct constant_table ext4_param_dax[] = {
1683 {"always", Opt_dax_always},
1684 {"inode", Opt_dax_inode},
1685 {"never", Opt_dax_never},
1689 /* String parameter that allows empty argument */
1690 #define fsparam_string_empty(NAME, OPT) \
1691 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1694 * Mount option specification
1695 * We don't use fsparam_flag_no because of the way we set the
1696 * options and the way we show them in _ext4_show_options(). To
1697 * keep the changes to a minimum, let's keep the negative options
1700 static const struct fs_parameter_spec ext4_param_specs[] = {
1701 fsparam_flag ("bsddf", Opt_bsd_df),
1702 fsparam_flag ("minixdf", Opt_minix_df),
1703 fsparam_flag ("grpid", Opt_grpid),
1704 fsparam_flag ("bsdgroups", Opt_grpid),
1705 fsparam_flag ("nogrpid", Opt_nogrpid),
1706 fsparam_flag ("sysvgroups", Opt_nogrpid),
1707 fsparam_u32 ("resgid", Opt_resgid),
1708 fsparam_u32 ("resuid", Opt_resuid),
1709 fsparam_u32 ("sb", Opt_sb),
1710 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1711 fsparam_flag ("nouid32", Opt_nouid32),
1712 fsparam_flag ("debug", Opt_debug),
1713 fsparam_flag ("oldalloc", Opt_removed),
1714 fsparam_flag ("orlov", Opt_removed),
1715 fsparam_flag ("user_xattr", Opt_user_xattr),
1716 fsparam_flag ("acl", Opt_acl),
1717 fsparam_flag ("norecovery", Opt_noload),
1718 fsparam_flag ("noload", Opt_noload),
1719 fsparam_flag ("bh", Opt_removed),
1720 fsparam_flag ("nobh", Opt_removed),
1721 fsparam_u32 ("commit", Opt_commit),
1722 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1723 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1724 fsparam_u32 ("journal_dev", Opt_journal_dev),
1725 fsparam_bdev ("journal_path", Opt_journal_path),
1726 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1727 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1728 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1729 fsparam_flag ("abort", Opt_abort),
1730 fsparam_enum ("data", Opt_data, ext4_param_data),
1731 fsparam_enum ("data_err", Opt_data_err,
1732 ext4_param_data_err),
1733 fsparam_string_empty
1734 ("usrjquota", Opt_usrjquota),
1735 fsparam_string_empty
1736 ("grpjquota", Opt_grpjquota),
1737 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1738 fsparam_flag ("grpquota", Opt_grpquota),
1739 fsparam_flag ("quota", Opt_quota),
1740 fsparam_flag ("noquota", Opt_noquota),
1741 fsparam_flag ("usrquota", Opt_usrquota),
1742 fsparam_flag ("prjquota", Opt_prjquota),
1743 fsparam_flag ("barrier", Opt_barrier),
1744 fsparam_u32 ("barrier", Opt_barrier),
1745 fsparam_flag ("nobarrier", Opt_nobarrier),
1746 fsparam_flag ("i_version", Opt_removed),
1747 fsparam_flag ("dax", Opt_dax),
1748 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1749 fsparam_u32 ("stripe", Opt_stripe),
1750 fsparam_flag ("delalloc", Opt_delalloc),
1751 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1752 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1753 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1754 fsparam_u32 ("debug_want_extra_isize",
1755 Opt_debug_want_extra_isize),
1756 fsparam_flag ("mblk_io_submit", Opt_removed),
1757 fsparam_flag ("nomblk_io_submit", Opt_removed),
1758 fsparam_flag ("block_validity", Opt_block_validity),
1759 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1760 fsparam_u32 ("inode_readahead_blks",
1761 Opt_inode_readahead_blks),
1762 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1763 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1764 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1765 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1766 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1767 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1768 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1769 fsparam_flag ("discard", Opt_discard),
1770 fsparam_flag ("nodiscard", Opt_nodiscard),
1771 fsparam_u32 ("init_itable", Opt_init_itable),
1772 fsparam_flag ("init_itable", Opt_init_itable),
1773 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1774 #ifdef CONFIG_EXT4_DEBUG
1775 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1776 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1778 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1779 fsparam_flag ("test_dummy_encryption",
1780 Opt_test_dummy_encryption),
1781 fsparam_string ("test_dummy_encryption",
1782 Opt_test_dummy_encryption),
1783 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1784 fsparam_flag ("nombcache", Opt_nombcache),
1785 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1786 fsparam_flag ("prefetch_block_bitmaps",
1788 fsparam_flag ("no_prefetch_block_bitmaps",
1789 Opt_no_prefetch_block_bitmaps),
1790 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1791 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1792 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1793 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1794 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1795 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1799 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1801 #define MOPT_SET 0x0001
1802 #define MOPT_CLEAR 0x0002
1803 #define MOPT_NOSUPPORT 0x0004
1804 #define MOPT_EXPLICIT 0x0008
1807 #define MOPT_QFMT 0x0010
1809 #define MOPT_Q MOPT_NOSUPPORT
1810 #define MOPT_QFMT MOPT_NOSUPPORT
1812 #define MOPT_NO_EXT2 0x0020
1813 #define MOPT_NO_EXT3 0x0040
1814 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1815 #define MOPT_SKIP 0x0080
1816 #define MOPT_2 0x0100
1818 static const struct mount_opts {
1822 } ext4_mount_opts[] = {
1823 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1824 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1825 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1826 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1827 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1828 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1829 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1830 MOPT_EXT4_ONLY | MOPT_SET},
1831 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1832 MOPT_EXT4_ONLY | MOPT_CLEAR},
1833 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1834 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1835 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1836 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1837 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1838 MOPT_EXT4_ONLY | MOPT_CLEAR},
1839 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1840 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1841 {Opt_commit, 0, MOPT_NO_EXT2},
1842 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1843 MOPT_EXT4_ONLY | MOPT_CLEAR},
1844 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1845 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1846 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1847 EXT4_MOUNT_JOURNAL_CHECKSUM),
1848 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1849 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1850 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1851 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1852 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1853 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1854 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1855 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1856 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1857 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1858 {Opt_journal_path, 0, MOPT_NO_EXT2},
1859 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1860 {Opt_data, 0, MOPT_NO_EXT2},
1861 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1862 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1863 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1865 {Opt_acl, 0, MOPT_NOSUPPORT},
1867 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1868 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1869 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1870 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1872 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1874 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1876 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1877 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1878 MOPT_CLEAR | MOPT_Q},
1879 {Opt_usrjquota, 0, MOPT_Q},
1880 {Opt_grpjquota, 0, MOPT_Q},
1881 {Opt_jqfmt, 0, MOPT_QFMT},
1882 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1883 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1885 #ifdef CONFIG_EXT4_DEBUG
1886 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1887 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1892 #if IS_ENABLED(CONFIG_UNICODE)
1893 static const struct ext4_sb_encodings {
1896 unsigned int version;
1897 } ext4_sb_encoding_map[] = {
1898 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1901 static const struct ext4_sb_encodings *
1902 ext4_sb_read_encoding(const struct ext4_super_block *es)
1904 __u16 magic = le16_to_cpu(es->s_encoding);
1907 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1908 if (magic == ext4_sb_encoding_map[i].magic)
1909 return &ext4_sb_encoding_map[i];
1915 #define EXT4_SPEC_JQUOTA (1 << 0)
1916 #define EXT4_SPEC_JQFMT (1 << 1)
1917 #define EXT4_SPEC_DATAJ (1 << 2)
1918 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1919 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1920 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1921 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1922 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1923 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1924 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1925 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1926 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1927 #define EXT4_SPEC_s_stripe (1 << 13)
1928 #define EXT4_SPEC_s_resuid (1 << 14)
1929 #define EXT4_SPEC_s_resgid (1 << 15)
1930 #define EXT4_SPEC_s_commit_interval (1 << 16)
1931 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1932 #define EXT4_SPEC_s_sb_block (1 << 18)
1933 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1935 struct ext4_fs_context {
1936 char *s_qf_names[EXT4_MAXQUOTAS];
1937 struct fscrypt_dummy_policy dummy_enc_policy;
1938 int s_jquota_fmt; /* Format of quota to use */
1939 #ifdef CONFIG_EXT4_DEBUG
1940 int s_fc_debug_max_replay;
1942 unsigned short qname_spec;
1943 unsigned long vals_s_flags; /* Bits to set in s_flags */
1944 unsigned long mask_s_flags; /* Bits changed in s_flags */
1945 unsigned long journal_devnum;
1946 unsigned long s_commit_interval;
1947 unsigned long s_stripe;
1948 unsigned int s_inode_readahead_blks;
1949 unsigned int s_want_extra_isize;
1950 unsigned int s_li_wait_mult;
1951 unsigned int s_max_dir_size_kb;
1952 unsigned int journal_ioprio;
1953 unsigned int vals_s_mount_opt;
1954 unsigned int mask_s_mount_opt;
1955 unsigned int vals_s_mount_opt2;
1956 unsigned int mask_s_mount_opt2;
1957 unsigned long vals_s_mount_flags;
1958 unsigned long mask_s_mount_flags;
1959 unsigned int opt_flags; /* MOPT flags */
1961 u32 s_max_batch_time;
1962 u32 s_min_batch_time;
1965 ext4_fsblk_t s_sb_block;
1968 static void ext4_fc_free(struct fs_context *fc)
1970 struct ext4_fs_context *ctx = fc->fs_private;
1976 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1977 kfree(ctx->s_qf_names[i]);
1979 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1983 int ext4_init_fs_context(struct fs_context *fc)
1985 struct ext4_fs_context *ctx;
1987 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1991 fc->fs_private = ctx;
1992 fc->ops = &ext4_context_ops;
1999 * Note the name of the specified quota file.
2001 static int note_qf_name(struct fs_context *fc, int qtype,
2002 struct fs_parameter *param)
2004 struct ext4_fs_context *ctx = fc->fs_private;
2007 if (param->size < 1) {
2008 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2011 if (strchr(param->string, '/')) {
2012 ext4_msg(NULL, KERN_ERR,
2013 "quotafile must be on filesystem root");
2016 if (ctx->s_qf_names[qtype]) {
2017 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2018 ext4_msg(NULL, KERN_ERR,
2019 "%s quota file already specified",
2026 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2028 ext4_msg(NULL, KERN_ERR,
2029 "Not enough memory for storing quotafile name");
2032 ctx->s_qf_names[qtype] = qname;
2033 ctx->qname_spec |= 1 << qtype;
2034 ctx->spec |= EXT4_SPEC_JQUOTA;
2039 * Clear the name of the specified quota file.
2041 static int unnote_qf_name(struct fs_context *fc, int qtype)
2043 struct ext4_fs_context *ctx = fc->fs_private;
2045 if (ctx->s_qf_names[qtype])
2046 kfree(ctx->s_qf_names[qtype]);
2048 ctx->s_qf_names[qtype] = NULL;
2049 ctx->qname_spec |= 1 << qtype;
2050 ctx->spec |= EXT4_SPEC_JQUOTA;
2055 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2056 struct ext4_fs_context *ctx)
2060 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2061 ext4_msg(NULL, KERN_WARNING,
2062 "test_dummy_encryption option not supported");
2065 err = fscrypt_parse_test_dummy_encryption(param,
2066 &ctx->dummy_enc_policy);
2067 if (err == -EINVAL) {
2068 ext4_msg(NULL, KERN_WARNING,
2069 "Value of option \"%s\" is unrecognized", param->key);
2070 } else if (err == -EEXIST) {
2071 ext4_msg(NULL, KERN_WARNING,
2072 "Conflicting test_dummy_encryption options");
2078 #define EXT4_SET_CTX(name) \
2079 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2080 unsigned long flag) \
2082 ctx->mask_s_##name |= flag; \
2083 ctx->vals_s_##name |= flag; \
2086 #define EXT4_CLEAR_CTX(name) \
2087 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2088 unsigned long flag) \
2090 ctx->mask_s_##name |= flag; \
2091 ctx->vals_s_##name &= ~flag; \
2094 #define EXT4_TEST_CTX(name) \
2095 static inline unsigned long \
2096 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2098 return (ctx->vals_s_##name & flag); \
2101 EXT4_SET_CTX(flags); /* set only */
2102 EXT4_SET_CTX(mount_opt);
2103 EXT4_CLEAR_CTX(mount_opt);
2104 EXT4_TEST_CTX(mount_opt);
2105 EXT4_SET_CTX(mount_opt2);
2106 EXT4_CLEAR_CTX(mount_opt2);
2107 EXT4_TEST_CTX(mount_opt2);
2109 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2111 set_bit(bit, &ctx->mask_s_mount_flags);
2112 set_bit(bit, &ctx->vals_s_mount_flags);
2115 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2117 struct ext4_fs_context *ctx = fc->fs_private;
2118 struct fs_parse_result result;
2119 const struct mount_opts *m;
2125 token = fs_parse(fc, ext4_param_specs, param, &result);
2128 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2130 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2131 if (token == m->token)
2134 ctx->opt_flags |= m->flags;
2136 if (m->flags & MOPT_EXPLICIT) {
2137 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2138 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2139 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2140 ctx_set_mount_opt2(ctx,
2141 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2146 if (m->flags & MOPT_NOSUPPORT) {
2147 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2155 if (!*param->string)
2156 return unnote_qf_name(fc, USRQUOTA);
2158 return note_qf_name(fc, USRQUOTA, param);
2160 if (!*param->string)
2161 return unnote_qf_name(fc, GRPQUOTA);
2163 return note_qf_name(fc, GRPQUOTA, param);
2166 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2167 ext4_msg(NULL, KERN_WARNING,
2168 "Ignoring %s option on remount", param->key);
2170 ctx->s_sb_block = result.uint_32;
2171 ctx->spec |= EXT4_SPEC_s_sb_block;
2175 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2179 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2181 case Opt_inlinecrypt:
2182 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2183 ctx_set_flags(ctx, SB_INLINECRYPT);
2185 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2189 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2190 ctx_set_mount_opt(ctx, result.uint_32);
2194 ctx->s_jquota_fmt = result.uint_32;
2195 ctx->spec |= EXT4_SPEC_JQFMT;
2199 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2200 ctx_set_mount_opt(ctx, result.uint_32);
2201 ctx->spec |= EXT4_SPEC_DATAJ;
2204 if (result.uint_32 == 0)
2205 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2206 else if (result.uint_32 > INT_MAX / HZ) {
2207 ext4_msg(NULL, KERN_ERR,
2208 "Invalid commit interval %d, "
2209 "must be smaller than %d",
2210 result.uint_32, INT_MAX / HZ);
2213 ctx->s_commit_interval = HZ * result.uint_32;
2214 ctx->spec |= EXT4_SPEC_s_commit_interval;
2216 case Opt_debug_want_extra_isize:
2217 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2218 ext4_msg(NULL, KERN_ERR,
2219 "Invalid want_extra_isize %d", result.uint_32);
2222 ctx->s_want_extra_isize = result.uint_32;
2223 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2225 case Opt_max_batch_time:
2226 ctx->s_max_batch_time = result.uint_32;
2227 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2229 case Opt_min_batch_time:
2230 ctx->s_min_batch_time = result.uint_32;
2231 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2233 case Opt_inode_readahead_blks:
2234 if (result.uint_32 &&
2235 (result.uint_32 > (1 << 30) ||
2236 !is_power_of_2(result.uint_32))) {
2237 ext4_msg(NULL, KERN_ERR,
2238 "EXT4-fs: inode_readahead_blks must be "
2239 "0 or a power of 2 smaller than 2^31");
2242 ctx->s_inode_readahead_blks = result.uint_32;
2243 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2245 case Opt_init_itable:
2246 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2247 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2248 if (param->type == fs_value_is_string)
2249 ctx->s_li_wait_mult = result.uint_32;
2250 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2252 case Opt_max_dir_size_kb:
2253 ctx->s_max_dir_size_kb = result.uint_32;
2254 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2256 #ifdef CONFIG_EXT4_DEBUG
2257 case Opt_fc_debug_max_replay:
2258 ctx->s_fc_debug_max_replay = result.uint_32;
2259 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2263 ctx->s_stripe = result.uint_32;
2264 ctx->spec |= EXT4_SPEC_s_stripe;
2267 uid = make_kuid(current_user_ns(), result.uint_32);
2268 if (!uid_valid(uid)) {
2269 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2273 ctx->s_resuid = uid;
2274 ctx->spec |= EXT4_SPEC_s_resuid;
2277 gid = make_kgid(current_user_ns(), result.uint_32);
2278 if (!gid_valid(gid)) {
2279 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2283 ctx->s_resgid = gid;
2284 ctx->spec |= EXT4_SPEC_s_resgid;
2286 case Opt_journal_dev:
2288 ext4_msg(NULL, KERN_ERR,
2289 "Cannot specify journal on remount");
2292 ctx->journal_devnum = result.uint_32;
2293 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2295 case Opt_journal_path:
2297 struct inode *journal_inode;
2302 ext4_msg(NULL, KERN_ERR,
2303 "Cannot specify journal on remount");
2307 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2309 ext4_msg(NULL, KERN_ERR, "error: could not find "
2310 "journal device path");
2314 journal_inode = d_inode(path.dentry);
2315 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2316 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2320 case Opt_journal_ioprio:
2321 if (result.uint_32 > 7) {
2322 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2326 ctx->journal_ioprio =
2327 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2328 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2330 case Opt_test_dummy_encryption:
2331 return ext4_parse_test_dummy_encryption(param, ctx);
2334 #ifdef CONFIG_FS_DAX
2336 int type = (token == Opt_dax) ?
2337 Opt_dax : result.uint_32;
2341 case Opt_dax_always:
2342 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2343 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2346 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2347 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2350 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2351 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2352 /* Strictly for printing options */
2353 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2359 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2363 if (result.uint_32 == Opt_data_err_abort)
2364 ctx_set_mount_opt(ctx, m->mount_opt);
2365 else if (result.uint_32 == Opt_data_err_ignore)
2366 ctx_clear_mount_opt(ctx, m->mount_opt);
2368 case Opt_mb_optimize_scan:
2369 if (result.int_32 == 1) {
2370 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2371 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2372 } else if (result.int_32 == 0) {
2373 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2374 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2376 ext4_msg(NULL, KERN_WARNING,
2377 "mb_optimize_scan should be set to 0 or 1.");
2384 * At this point we should only be getting options requiring MOPT_SET,
2385 * or MOPT_CLEAR. Anything else is a bug
2387 if (m->token == Opt_err) {
2388 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2395 unsigned int set = 0;
2397 if ((param->type == fs_value_is_flag) ||
2401 if (m->flags & MOPT_CLEAR)
2403 else if (unlikely(!(m->flags & MOPT_SET))) {
2404 ext4_msg(NULL, KERN_WARNING,
2405 "buggy handling of option %s",
2410 if (m->flags & MOPT_2) {
2412 ctx_set_mount_opt2(ctx, m->mount_opt);
2414 ctx_clear_mount_opt2(ctx, m->mount_opt);
2417 ctx_set_mount_opt(ctx, m->mount_opt);
2419 ctx_clear_mount_opt(ctx, m->mount_opt);
2426 static int parse_options(struct fs_context *fc, char *options)
2428 struct fs_parameter param;
2435 while ((key = strsep(&options, ",")) != NULL) {
2438 char *value = strchr(key, '=');
2440 param.type = fs_value_is_flag;
2441 param.string = NULL;
2448 v_len = strlen(value);
2449 param.string = kmemdup_nul(value, v_len,
2453 param.type = fs_value_is_string;
2459 ret = ext4_parse_param(fc, ¶m);
2461 kfree(param.string);
2467 ret = ext4_validate_options(fc);
2474 static int parse_apply_sb_mount_options(struct super_block *sb,
2475 struct ext4_fs_context *m_ctx)
2477 struct ext4_sb_info *sbi = EXT4_SB(sb);
2478 char *s_mount_opts = NULL;
2479 struct ext4_fs_context *s_ctx = NULL;
2480 struct fs_context *fc = NULL;
2483 if (!sbi->s_es->s_mount_opts[0])
2486 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2487 sizeof(sbi->s_es->s_mount_opts),
2492 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2496 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2500 fc->fs_private = s_ctx;
2501 fc->s_fs_info = sbi;
2503 ret = parse_options(fc, s_mount_opts);
2507 ret = ext4_check_opt_consistency(fc, sb);
2510 ext4_msg(sb, KERN_WARNING,
2511 "failed to parse options in superblock: %s",
2517 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2518 m_ctx->journal_devnum = s_ctx->journal_devnum;
2519 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2520 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2522 ext4_apply_options(fc, sb);
2530 kfree(s_mount_opts);
2534 static void ext4_apply_quota_options(struct fs_context *fc,
2535 struct super_block *sb)
2538 bool quota_feature = ext4_has_feature_quota(sb);
2539 struct ext4_fs_context *ctx = fc->fs_private;
2540 struct ext4_sb_info *sbi = EXT4_SB(sb);
2547 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2548 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2549 if (!(ctx->qname_spec & (1 << i)))
2552 qname = ctx->s_qf_names[i]; /* May be NULL */
2555 ctx->s_qf_names[i] = NULL;
2556 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2557 lockdep_is_held(&sb->s_umount));
2559 kfree_rcu_mightsleep(qname);
2563 if (ctx->spec & EXT4_SPEC_JQFMT)
2564 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2569 * Check quota settings consistency.
2571 static int ext4_check_quota_consistency(struct fs_context *fc,
2572 struct super_block *sb)
2575 struct ext4_fs_context *ctx = fc->fs_private;
2576 struct ext4_sb_info *sbi = EXT4_SB(sb);
2577 bool quota_feature = ext4_has_feature_quota(sb);
2578 bool quota_loaded = sb_any_quota_loaded(sb);
2579 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2583 * We do the test below only for project quotas. 'usrquota' and
2584 * 'grpquota' mount options are allowed even without quota feature
2585 * to support legacy quotas in quota files.
2587 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2588 !ext4_has_feature_project(sb)) {
2589 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2590 "Cannot enable project quota enforcement.");
2594 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2595 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2597 ctx->mask_s_mount_opt & quota_flags &&
2598 !ctx_test_mount_opt(ctx, quota_flags))
2599 goto err_quota_change;
2601 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2603 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2604 if (!(ctx->qname_spec & (1 << i)))
2608 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2609 goto err_jquota_change;
2611 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2612 strcmp(get_qf_name(sb, sbi, i),
2613 ctx->s_qf_names[i]) != 0)
2614 goto err_jquota_specified;
2617 if (quota_feature) {
2618 ext4_msg(NULL, KERN_INFO,
2619 "Journaled quota options ignored when "
2620 "QUOTA feature is enabled");
2625 if (ctx->spec & EXT4_SPEC_JQFMT) {
2626 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2627 goto err_jquota_change;
2628 if (quota_feature) {
2629 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2630 "ignored when QUOTA feature is enabled");
2635 /* Make sure we don't mix old and new quota format */
2636 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2637 ctx->s_qf_names[USRQUOTA]);
2638 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2639 ctx->s_qf_names[GRPQUOTA]);
2641 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2642 test_opt(sb, USRQUOTA));
2644 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2645 test_opt(sb, GRPQUOTA));
2648 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2652 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2656 if (usr_qf_name || grp_qf_name) {
2657 if (usrquota || grpquota) {
2658 ext4_msg(NULL, KERN_ERR, "old and new quota "
2663 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2664 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2673 ext4_msg(NULL, KERN_ERR,
2674 "Cannot change quota options when quota turned on");
2677 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2678 "options when quota turned on");
2680 err_jquota_specified:
2681 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2689 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2690 struct super_block *sb)
2692 const struct ext4_fs_context *ctx = fc->fs_private;
2693 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2695 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2698 if (!ext4_has_feature_encrypt(sb)) {
2699 ext4_msg(NULL, KERN_WARNING,
2700 "test_dummy_encryption requires encrypt feature");
2704 * This mount option is just for testing, and it's not worthwhile to
2705 * implement the extra complexity (e.g. RCU protection) that would be
2706 * needed to allow it to be set or changed during remount. We do allow
2707 * it to be specified during remount, but only if there is no change.
2709 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2710 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2711 &ctx->dummy_enc_policy))
2713 ext4_msg(NULL, KERN_WARNING,
2714 "Can't set or change test_dummy_encryption on remount");
2717 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2718 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2719 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2720 &ctx->dummy_enc_policy))
2722 ext4_msg(NULL, KERN_WARNING,
2723 "Conflicting test_dummy_encryption options");
2729 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2730 struct super_block *sb)
2732 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2733 /* if already set, it was already verified to be the same */
2734 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2736 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2737 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2738 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2741 static int ext4_check_opt_consistency(struct fs_context *fc,
2742 struct super_block *sb)
2744 struct ext4_fs_context *ctx = fc->fs_private;
2745 struct ext4_sb_info *sbi = fc->s_fs_info;
2746 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2749 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2750 ext4_msg(NULL, KERN_ERR,
2751 "Mount option(s) incompatible with ext2");
2754 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2755 ext4_msg(NULL, KERN_ERR,
2756 "Mount option(s) incompatible with ext3");
2760 if (ctx->s_want_extra_isize >
2761 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2762 ext4_msg(NULL, KERN_ERR,
2763 "Invalid want_extra_isize %d",
2764 ctx->s_want_extra_isize);
2768 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2770 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2771 if (blocksize < PAGE_SIZE)
2772 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2773 "experimental mount option 'dioread_nolock' "
2774 "for blocksize < PAGE_SIZE");
2777 err = ext4_check_test_dummy_encryption(fc, sb);
2781 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2782 if (!sbi->s_journal) {
2783 ext4_msg(NULL, KERN_WARNING,
2784 "Remounting file system with no journal "
2785 "so ignoring journalled data option");
2786 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2787 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2788 test_opt(sb, DATA_FLAGS)) {
2789 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2796 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2797 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2798 ext4_msg(NULL, KERN_ERR, "can't mount with "
2799 "both data=journal and dax");
2803 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2804 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2805 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2806 fail_dax_change_remount:
2807 ext4_msg(NULL, KERN_ERR, "can't change "
2808 "dax mount option while remounting");
2810 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2811 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2812 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2813 goto fail_dax_change_remount;
2814 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2815 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2816 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2817 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2818 goto fail_dax_change_remount;
2822 return ext4_check_quota_consistency(fc, sb);
2825 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2827 struct ext4_fs_context *ctx = fc->fs_private;
2828 struct ext4_sb_info *sbi = fc->s_fs_info;
2830 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2831 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2832 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2833 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2834 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2835 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2836 sb->s_flags &= ~ctx->mask_s_flags;
2837 sb->s_flags |= ctx->vals_s_flags;
2839 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2840 APPLY(s_commit_interval);
2842 APPLY(s_max_batch_time);
2843 APPLY(s_min_batch_time);
2844 APPLY(s_want_extra_isize);
2845 APPLY(s_inode_readahead_blks);
2846 APPLY(s_max_dir_size_kb);
2847 APPLY(s_li_wait_mult);
2851 #ifdef CONFIG_EXT4_DEBUG
2852 APPLY(s_fc_debug_max_replay);
2855 ext4_apply_quota_options(fc, sb);
2856 ext4_apply_test_dummy_encryption(ctx, sb);
2860 static int ext4_validate_options(struct fs_context *fc)
2863 struct ext4_fs_context *ctx = fc->fs_private;
2864 char *usr_qf_name, *grp_qf_name;
2866 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2867 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2869 if (usr_qf_name || grp_qf_name) {
2870 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2871 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2873 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2874 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2876 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2877 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2878 ext4_msg(NULL, KERN_ERR, "old and new quota "
2887 static inline void ext4_show_quota_options(struct seq_file *seq,
2888 struct super_block *sb)
2890 #if defined(CONFIG_QUOTA)
2891 struct ext4_sb_info *sbi = EXT4_SB(sb);
2892 char *usr_qf_name, *grp_qf_name;
2894 if (sbi->s_jquota_fmt) {
2897 switch (sbi->s_jquota_fmt) {
2908 seq_printf(seq, ",jqfmt=%s", fmtname);
2912 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2913 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2915 seq_show_option(seq, "usrjquota", usr_qf_name);
2917 seq_show_option(seq, "grpjquota", grp_qf_name);
2922 static const char *token2str(int token)
2924 const struct fs_parameter_spec *spec;
2926 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2927 if (spec->opt == token && !spec->type)
2934 * - it's set to a non-default value OR
2935 * - if the per-sb default is different from the global default
2937 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2940 struct ext4_sb_info *sbi = EXT4_SB(sb);
2941 struct ext4_super_block *es = sbi->s_es;
2943 const struct mount_opts *m;
2944 char sep = nodefs ? '\n' : ',';
2946 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2947 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2949 if (sbi->s_sb_block != 1)
2950 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2952 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2953 int want_set = m->flags & MOPT_SET;
2954 int opt_2 = m->flags & MOPT_2;
2955 unsigned int mount_opt, def_mount_opt;
2957 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2958 m->flags & MOPT_SKIP)
2962 mount_opt = sbi->s_mount_opt2;
2963 def_mount_opt = sbi->s_def_mount_opt2;
2965 mount_opt = sbi->s_mount_opt;
2966 def_mount_opt = sbi->s_def_mount_opt;
2968 /* skip if same as the default */
2969 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2971 /* select Opt_noFoo vs Opt_Foo */
2973 (mount_opt & m->mount_opt) != m->mount_opt) ||
2974 (!want_set && (mount_opt & m->mount_opt)))
2976 SEQ_OPTS_PRINT("%s", token2str(m->token));
2979 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2980 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2981 SEQ_OPTS_PRINT("resuid=%u",
2982 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2983 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2984 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2985 SEQ_OPTS_PRINT("resgid=%u",
2986 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2987 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2988 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2989 SEQ_OPTS_PUTS("errors=remount-ro");
2990 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2991 SEQ_OPTS_PUTS("errors=continue");
2992 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2993 SEQ_OPTS_PUTS("errors=panic");
2994 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2995 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2996 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2997 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2998 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2999 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3000 if (nodefs || sbi->s_stripe)
3001 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3002 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3003 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3004 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3005 SEQ_OPTS_PUTS("data=journal");
3006 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3007 SEQ_OPTS_PUTS("data=ordered");
3008 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3009 SEQ_OPTS_PUTS("data=writeback");
3012 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3013 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3014 sbi->s_inode_readahead_blks);
3016 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3017 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3018 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3019 if (nodefs || sbi->s_max_dir_size_kb)
3020 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3021 if (test_opt(sb, DATA_ERR_ABORT))
3022 SEQ_OPTS_PUTS("data_err=abort");
3024 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3026 if (sb->s_flags & SB_INLINECRYPT)
3027 SEQ_OPTS_PUTS("inlinecrypt");
3029 if (test_opt(sb, DAX_ALWAYS)) {
3031 SEQ_OPTS_PUTS("dax");
3033 SEQ_OPTS_PUTS("dax=always");
3034 } else if (test_opt2(sb, DAX_NEVER)) {
3035 SEQ_OPTS_PUTS("dax=never");
3036 } else if (test_opt2(sb, DAX_INODE)) {
3037 SEQ_OPTS_PUTS("dax=inode");
3040 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3041 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3042 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3043 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3044 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3045 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3048 ext4_show_quota_options(seq, sb);
3052 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3054 return _ext4_show_options(seq, root->d_sb, 0);
3057 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3059 struct super_block *sb = seq->private;
3062 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3063 rc = _ext4_show_options(seq, sb, 1);
3064 seq_puts(seq, "\n");
3068 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3071 struct ext4_sb_info *sbi = EXT4_SB(sb);
3074 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3075 ext4_msg(sb, KERN_ERR, "revision level too high, "
3076 "forcing read-only mode");
3082 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3083 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3084 "running e2fsck is recommended");
3085 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3086 ext4_msg(sb, KERN_WARNING,
3087 "warning: mounting fs with errors, "
3088 "running e2fsck is recommended");
3089 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3090 le16_to_cpu(es->s_mnt_count) >=
3091 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3092 ext4_msg(sb, KERN_WARNING,
3093 "warning: maximal mount count reached, "
3094 "running e2fsck is recommended");
3095 else if (le32_to_cpu(es->s_checkinterval) &&
3096 (ext4_get_tstamp(es, s_lastcheck) +
3097 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3098 ext4_msg(sb, KERN_WARNING,
3099 "warning: checktime reached, "
3100 "running e2fsck is recommended");
3101 if (!sbi->s_journal)
3102 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3103 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3104 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3105 le16_add_cpu(&es->s_mnt_count, 1);
3106 ext4_update_tstamp(es, s_mtime);
3107 if (sbi->s_journal) {
3108 ext4_set_feature_journal_needs_recovery(sb);
3109 if (ext4_has_feature_orphan_file(sb))
3110 ext4_set_feature_orphan_present(sb);
3113 err = ext4_commit_super(sb);
3115 if (test_opt(sb, DEBUG))
3116 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3117 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3119 sbi->s_groups_count,
3120 EXT4_BLOCKS_PER_GROUP(sb),
3121 EXT4_INODES_PER_GROUP(sb),
3122 sbi->s_mount_opt, sbi->s_mount_opt2);
3126 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3128 struct ext4_sb_info *sbi = EXT4_SB(sb);
3129 struct flex_groups **old_groups, **new_groups;
3132 if (!sbi->s_log_groups_per_flex)
3135 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3136 if (size <= sbi->s_flex_groups_allocated)
3139 new_groups = kvzalloc(roundup_pow_of_two(size *
3140 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3142 ext4_msg(sb, KERN_ERR,
3143 "not enough memory for %d flex group pointers", size);
3146 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3147 new_groups[i] = kvzalloc(roundup_pow_of_two(
3148 sizeof(struct flex_groups)),
3150 if (!new_groups[i]) {
3151 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3152 kvfree(new_groups[j]);
3154 ext4_msg(sb, KERN_ERR,
3155 "not enough memory for %d flex groups", size);
3160 old_groups = rcu_dereference(sbi->s_flex_groups);
3162 memcpy(new_groups, old_groups,
3163 (sbi->s_flex_groups_allocated *
3164 sizeof(struct flex_groups *)));
3166 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3167 sbi->s_flex_groups_allocated = size;
3169 ext4_kvfree_array_rcu(old_groups);
3173 static int ext4_fill_flex_info(struct super_block *sb)
3175 struct ext4_sb_info *sbi = EXT4_SB(sb);
3176 struct ext4_group_desc *gdp = NULL;
3177 struct flex_groups *fg;
3178 ext4_group_t flex_group;
3181 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3182 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3183 sbi->s_log_groups_per_flex = 0;
3187 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3191 for (i = 0; i < sbi->s_groups_count; i++) {
3192 gdp = ext4_get_group_desc(sb, i, NULL);
3194 flex_group = ext4_flex_group(sbi, i);
3195 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3196 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3197 atomic64_add(ext4_free_group_clusters(sb, gdp),
3198 &fg->free_clusters);
3199 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3207 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3208 struct ext4_group_desc *gdp)
3210 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3212 __le32 le_group = cpu_to_le32(block_group);
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3215 if (ext4_has_metadata_csum(sbi->s_sb)) {
3216 /* Use new metadata_csum algorithm */
3218 __u16 dummy_csum = 0;
3220 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3222 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3223 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3224 sizeof(dummy_csum));
3225 offset += sizeof(dummy_csum);
3226 if (offset < sbi->s_desc_size)
3227 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3228 sbi->s_desc_size - offset);
3230 crc = csum32 & 0xFFFF;
3234 /* old crc16 code */
3235 if (!ext4_has_feature_gdt_csum(sb))
3238 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3239 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3240 crc = crc16(crc, (__u8 *)gdp, offset);
3241 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3242 /* for checksum of struct ext4_group_desc do the rest...*/
3243 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3244 crc = crc16(crc, (__u8 *)gdp + offset,
3245 sbi->s_desc_size - offset);
3248 return cpu_to_le16(crc);
3251 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3252 struct ext4_group_desc *gdp)
3254 if (ext4_has_group_desc_csum(sb) &&
3255 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3261 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3262 struct ext4_group_desc *gdp)
3264 if (!ext4_has_group_desc_csum(sb))
3266 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3269 /* Called at mount-time, super-block is locked */
3270 static int ext4_check_descriptors(struct super_block *sb,
3271 ext4_fsblk_t sb_block,
3272 ext4_group_t *first_not_zeroed)
3274 struct ext4_sb_info *sbi = EXT4_SB(sb);
3275 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3276 ext4_fsblk_t last_block;
3277 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3278 ext4_fsblk_t block_bitmap;
3279 ext4_fsblk_t inode_bitmap;
3280 ext4_fsblk_t inode_table;
3281 int flexbg_flag = 0;
3282 ext4_group_t i, grp = sbi->s_groups_count;
3284 if (ext4_has_feature_flex_bg(sb))
3287 ext4_debug("Checking group descriptors");
3289 for (i = 0; i < sbi->s_groups_count; i++) {
3290 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3292 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3293 last_block = ext4_blocks_count(sbi->s_es) - 1;
3295 last_block = first_block +
3296 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3298 if ((grp == sbi->s_groups_count) &&
3299 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3302 block_bitmap = ext4_block_bitmap(sb, gdp);
3303 if (block_bitmap == sb_block) {
3304 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3305 "Block bitmap for group %u overlaps "
3310 if (block_bitmap >= sb_block + 1 &&
3311 block_bitmap <= last_bg_block) {
3312 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3313 "Block bitmap for group %u overlaps "
3314 "block group descriptors", i);
3318 if (block_bitmap < first_block || block_bitmap > last_block) {
3319 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3320 "Block bitmap for group %u not in group "
3321 "(block %llu)!", i, block_bitmap);
3324 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3325 if (inode_bitmap == sb_block) {
3326 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3327 "Inode bitmap for group %u overlaps "
3332 if (inode_bitmap >= sb_block + 1 &&
3333 inode_bitmap <= last_bg_block) {
3334 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3335 "Inode bitmap for group %u overlaps "
3336 "block group descriptors", i);
3340 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3342 "Inode bitmap for group %u not in group "
3343 "(block %llu)!", i, inode_bitmap);
3346 inode_table = ext4_inode_table(sb, gdp);
3347 if (inode_table == sb_block) {
3348 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3349 "Inode table for group %u overlaps "
3354 if (inode_table >= sb_block + 1 &&
3355 inode_table <= last_bg_block) {
3356 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3357 "Inode table for group %u overlaps "
3358 "block group descriptors", i);
3362 if (inode_table < first_block ||
3363 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3364 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3365 "Inode table for group %u not in group "
3366 "(block %llu)!", i, inode_table);
3369 ext4_lock_group(sb, i);
3370 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3371 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3372 "Checksum for group %u failed (%u!=%u)",
3373 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3374 gdp)), le16_to_cpu(gdp->bg_checksum));
3375 if (!sb_rdonly(sb)) {
3376 ext4_unlock_group(sb, i);
3380 ext4_unlock_group(sb, i);
3382 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3384 if (NULL != first_not_zeroed)
3385 *first_not_zeroed = grp;
3390 * Maximal extent format file size.
3391 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3392 * extent format containers, within a sector_t, and within i_blocks
3393 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3394 * so that won't be a limiting factor.
3396 * However there is other limiting factor. We do store extents in the form
3397 * of starting block and length, hence the resulting length of the extent
3398 * covering maximum file size must fit into on-disk format containers as
3399 * well. Given that length is always by 1 unit bigger than max unit (because
3400 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3402 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3404 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3407 loff_t upper_limit = MAX_LFS_FILESIZE;
3409 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3411 if (!has_huge_files) {
3412 upper_limit = (1LL << 32) - 1;
3414 /* total blocks in file system block size */
3415 upper_limit >>= (blkbits - 9);
3416 upper_limit <<= blkbits;
3420 * 32-bit extent-start container, ee_block. We lower the maxbytes
3421 * by one fs block, so ee_len can cover the extent of maximum file
3424 res = (1LL << 32) - 1;
3427 /* Sanity check against vm- & vfs- imposed limits */
3428 if (res > upper_limit)
3435 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3436 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3437 * We need to be 1 filesystem block less than the 2^48 sector limit.
3439 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3441 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3443 unsigned int ppb = 1 << (bits - 2);
3446 * This is calculated to be the largest file size for a dense, block
3447 * mapped file such that the file's total number of 512-byte sectors,
3448 * including data and all indirect blocks, does not exceed (2^48 - 1).
3450 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3451 * number of 512-byte sectors of the file.
3453 if (!has_huge_files) {
3455 * !has_huge_files or implies that the inode i_block field
3456 * represents total file blocks in 2^32 512-byte sectors ==
3457 * size of vfs inode i_blocks * 8
3459 upper_limit = (1LL << 32) - 1;
3461 /* total blocks in file system block size */
3462 upper_limit >>= (bits - 9);
3466 * We use 48 bit ext4_inode i_blocks
3467 * With EXT4_HUGE_FILE_FL set the i_blocks
3468 * represent total number of blocks in
3469 * file system block size
3471 upper_limit = (1LL << 48) - 1;
3475 /* Compute how many blocks we can address by block tree */
3478 res += ((loff_t)ppb) * ppb * ppb;
3479 /* Compute how many metadata blocks are needed */
3481 meta_blocks += 1 + ppb;
3482 meta_blocks += 1 + ppb + ppb * ppb;
3483 /* Does block tree limit file size? */
3484 if (res + meta_blocks <= upper_limit)
3488 /* How many metadata blocks are needed for addressing upper_limit? */
3489 upper_limit -= EXT4_NDIR_BLOCKS;
3490 /* indirect blocks */
3493 /* double indirect blocks */
3494 if (upper_limit < ppb * ppb) {
3495 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3499 meta_blocks += 1 + ppb;
3500 upper_limit -= ppb * ppb;
3501 /* tripple indirect blocks for the rest */
3502 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3503 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3507 if (res > MAX_LFS_FILESIZE)
3508 res = MAX_LFS_FILESIZE;
3513 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3514 ext4_fsblk_t logical_sb_block, int nr)
3516 struct ext4_sb_info *sbi = EXT4_SB(sb);
3517 ext4_group_t bg, first_meta_bg;
3520 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3522 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3523 return logical_sb_block + nr + 1;
3524 bg = sbi->s_desc_per_block * nr;
3525 if (ext4_bg_has_super(sb, bg))
3529 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3530 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3531 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3534 if (sb->s_blocksize == 1024 && nr == 0 &&
3535 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3538 return (has_super + ext4_group_first_block_no(sb, bg));
3542 * ext4_get_stripe_size: Get the stripe size.
3543 * @sbi: In memory super block info
3545 * If we have specified it via mount option, then
3546 * use the mount option value. If the value specified at mount time is
3547 * greater than the blocks per group use the super block value.
3548 * If the super block value is greater than blocks per group return 0.
3549 * Allocator needs it be less than blocks per group.
3552 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3554 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3555 unsigned long stripe_width =
3556 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3559 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3560 ret = sbi->s_stripe;
3561 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3563 else if (stride && stride <= sbi->s_blocks_per_group)
3569 * If the stripe width is 1, this makes no sense and
3570 * we set it to 0 to turn off stripe handling code.
3579 * Check whether this filesystem can be mounted based on
3580 * the features present and the RDONLY/RDWR mount requested.
3581 * Returns 1 if this filesystem can be mounted as requested,
3582 * 0 if it cannot be.
3584 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3586 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3587 ext4_msg(sb, KERN_ERR,
3588 "Couldn't mount because of "
3589 "unsupported optional features (%x)",
3590 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3591 ~EXT4_FEATURE_INCOMPAT_SUPP));
3595 #if !IS_ENABLED(CONFIG_UNICODE)
3596 if (ext4_has_feature_casefold(sb)) {
3597 ext4_msg(sb, KERN_ERR,
3598 "Filesystem with casefold feature cannot be "
3599 "mounted without CONFIG_UNICODE");
3607 if (ext4_has_feature_readonly(sb)) {
3608 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3609 sb->s_flags |= SB_RDONLY;
3613 /* Check that feature set is OK for a read-write mount */
3614 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3615 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3616 "unsupported optional features (%x)",
3617 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3618 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3621 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3622 ext4_msg(sb, KERN_ERR,
3623 "Can't support bigalloc feature without "
3624 "extents feature\n");
3628 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3629 if (!readonly && (ext4_has_feature_quota(sb) ||
3630 ext4_has_feature_project(sb))) {
3631 ext4_msg(sb, KERN_ERR,
3632 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3635 #endif /* CONFIG_QUOTA */
3640 * This function is called once a day if we have errors logged
3641 * on the file system
3643 static void print_daily_error_info(struct timer_list *t)
3645 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3646 struct super_block *sb = sbi->s_sb;
3647 struct ext4_super_block *es = sbi->s_es;
3649 if (es->s_error_count)
3650 /* fsck newer than v1.41.13 is needed to clean this condition. */
3651 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3652 le32_to_cpu(es->s_error_count));
3653 if (es->s_first_error_time) {
3654 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3656 ext4_get_tstamp(es, s_first_error_time),
3657 (int) sizeof(es->s_first_error_func),
3658 es->s_first_error_func,
3659 le32_to_cpu(es->s_first_error_line));
3660 if (es->s_first_error_ino)
3661 printk(KERN_CONT ": inode %u",
3662 le32_to_cpu(es->s_first_error_ino));
3663 if (es->s_first_error_block)
3664 printk(KERN_CONT ": block %llu", (unsigned long long)
3665 le64_to_cpu(es->s_first_error_block));
3666 printk(KERN_CONT "\n");
3668 if (es->s_last_error_time) {
3669 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3671 ext4_get_tstamp(es, s_last_error_time),
3672 (int) sizeof(es->s_last_error_func),
3673 es->s_last_error_func,
3674 le32_to_cpu(es->s_last_error_line));
3675 if (es->s_last_error_ino)
3676 printk(KERN_CONT ": inode %u",
3677 le32_to_cpu(es->s_last_error_ino));
3678 if (es->s_last_error_block)
3679 printk(KERN_CONT ": block %llu", (unsigned long long)
3680 le64_to_cpu(es->s_last_error_block));
3681 printk(KERN_CONT "\n");
3683 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3686 /* Find next suitable group and run ext4_init_inode_table */
3687 static int ext4_run_li_request(struct ext4_li_request *elr)
3689 struct ext4_group_desc *gdp = NULL;
3690 struct super_block *sb = elr->lr_super;
3691 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3692 ext4_group_t group = elr->lr_next_group;
3693 unsigned int prefetch_ios = 0;
3697 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3698 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3699 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3701 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3703 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3705 if (group >= elr->lr_next_group) {
3707 if (elr->lr_first_not_zeroed != ngroups &&
3708 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3709 elr->lr_next_group = elr->lr_first_not_zeroed;
3710 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3717 for (; group < ngroups; group++) {
3718 gdp = ext4_get_group_desc(sb, group, NULL);
3724 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3728 if (group >= ngroups)
3732 start_time = ktime_get_real_ns();
3733 ret = ext4_init_inode_table(sb, group,
3734 elr->lr_timeout ? 0 : 1);
3735 trace_ext4_lazy_itable_init(sb, group);
3736 if (elr->lr_timeout == 0) {
3737 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3738 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3740 elr->lr_next_sched = jiffies + elr->lr_timeout;
3741 elr->lr_next_group = group + 1;
3747 * Remove lr_request from the list_request and free the
3748 * request structure. Should be called with li_list_mtx held
3750 static void ext4_remove_li_request(struct ext4_li_request *elr)
3755 list_del(&elr->lr_request);
3756 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3760 static void ext4_unregister_li_request(struct super_block *sb)
3762 mutex_lock(&ext4_li_mtx);
3763 if (!ext4_li_info) {
3764 mutex_unlock(&ext4_li_mtx);
3768 mutex_lock(&ext4_li_info->li_list_mtx);
3769 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3770 mutex_unlock(&ext4_li_info->li_list_mtx);
3771 mutex_unlock(&ext4_li_mtx);
3774 static struct task_struct *ext4_lazyinit_task;
3777 * This is the function where ext4lazyinit thread lives. It walks
3778 * through the request list searching for next scheduled filesystem.
3779 * When such a fs is found, run the lazy initialization request
3780 * (ext4_rn_li_request) and keep track of the time spend in this
3781 * function. Based on that time we compute next schedule time of
3782 * the request. When walking through the list is complete, compute
3783 * next waking time and put itself into sleep.
3785 static int ext4_lazyinit_thread(void *arg)
3787 struct ext4_lazy_init *eli = arg;
3788 struct list_head *pos, *n;
3789 struct ext4_li_request *elr;
3790 unsigned long next_wakeup, cur;
3792 BUG_ON(NULL == eli);
3797 next_wakeup = MAX_JIFFY_OFFSET;
3799 mutex_lock(&eli->li_list_mtx);
3800 if (list_empty(&eli->li_request_list)) {
3801 mutex_unlock(&eli->li_list_mtx);
3804 list_for_each_safe(pos, n, &eli->li_request_list) {
3807 elr = list_entry(pos, struct ext4_li_request,
3810 if (time_before(jiffies, elr->lr_next_sched)) {
3811 if (time_before(elr->lr_next_sched, next_wakeup))
3812 next_wakeup = elr->lr_next_sched;
3815 if (down_read_trylock(&elr->lr_super->s_umount)) {
3816 if (sb_start_write_trylock(elr->lr_super)) {
3819 * We hold sb->s_umount, sb can not
3820 * be removed from the list, it is
3821 * now safe to drop li_list_mtx
3823 mutex_unlock(&eli->li_list_mtx);
3824 err = ext4_run_li_request(elr);
3825 sb_end_write(elr->lr_super);
3826 mutex_lock(&eli->li_list_mtx);
3829 up_read((&elr->lr_super->s_umount));
3831 /* error, remove the lazy_init job */
3833 ext4_remove_li_request(elr);
3837 elr->lr_next_sched = jiffies +
3838 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3840 if (time_before(elr->lr_next_sched, next_wakeup))
3841 next_wakeup = elr->lr_next_sched;
3843 mutex_unlock(&eli->li_list_mtx);
3848 if ((time_after_eq(cur, next_wakeup)) ||
3849 (MAX_JIFFY_OFFSET == next_wakeup)) {
3854 schedule_timeout_interruptible(next_wakeup - cur);
3856 if (kthread_should_stop()) {
3857 ext4_clear_request_list();
3864 * It looks like the request list is empty, but we need
3865 * to check it under the li_list_mtx lock, to prevent any
3866 * additions into it, and of course we should lock ext4_li_mtx
3867 * to atomically free the list and ext4_li_info, because at
3868 * this point another ext4 filesystem could be registering
3871 mutex_lock(&ext4_li_mtx);
3872 mutex_lock(&eli->li_list_mtx);
3873 if (!list_empty(&eli->li_request_list)) {
3874 mutex_unlock(&eli->li_list_mtx);
3875 mutex_unlock(&ext4_li_mtx);
3878 mutex_unlock(&eli->li_list_mtx);
3879 kfree(ext4_li_info);
3880 ext4_li_info = NULL;
3881 mutex_unlock(&ext4_li_mtx);
3886 static void ext4_clear_request_list(void)
3888 struct list_head *pos, *n;
3889 struct ext4_li_request *elr;
3891 mutex_lock(&ext4_li_info->li_list_mtx);
3892 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3893 elr = list_entry(pos, struct ext4_li_request,
3895 ext4_remove_li_request(elr);
3897 mutex_unlock(&ext4_li_info->li_list_mtx);
3900 static int ext4_run_lazyinit_thread(void)
3902 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3903 ext4_li_info, "ext4lazyinit");
3904 if (IS_ERR(ext4_lazyinit_task)) {
3905 int err = PTR_ERR(ext4_lazyinit_task);
3906 ext4_clear_request_list();
3907 kfree(ext4_li_info);
3908 ext4_li_info = NULL;
3909 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3910 "initialization thread\n",
3914 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3919 * Check whether it make sense to run itable init. thread or not.
3920 * If there is at least one uninitialized inode table, return
3921 * corresponding group number, else the loop goes through all
3922 * groups and return total number of groups.
3924 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3926 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3927 struct ext4_group_desc *gdp = NULL;
3929 if (!ext4_has_group_desc_csum(sb))
3932 for (group = 0; group < ngroups; group++) {
3933 gdp = ext4_get_group_desc(sb, group, NULL);
3937 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3944 static int ext4_li_info_new(void)
3946 struct ext4_lazy_init *eli = NULL;
3948 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3952 INIT_LIST_HEAD(&eli->li_request_list);
3953 mutex_init(&eli->li_list_mtx);
3955 eli->li_state |= EXT4_LAZYINIT_QUIT;
3962 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3965 struct ext4_li_request *elr;
3967 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3972 elr->lr_first_not_zeroed = start;
3973 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3974 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3975 elr->lr_next_group = start;
3977 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3981 * Randomize first schedule time of the request to
3982 * spread the inode table initialization requests
3985 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3989 int ext4_register_li_request(struct super_block *sb,
3990 ext4_group_t first_not_zeroed)
3992 struct ext4_sb_info *sbi = EXT4_SB(sb);
3993 struct ext4_li_request *elr = NULL;
3994 ext4_group_t ngroups = sbi->s_groups_count;
3997 mutex_lock(&ext4_li_mtx);
3998 if (sbi->s_li_request != NULL) {
4000 * Reset timeout so it can be computed again, because
4001 * s_li_wait_mult might have changed.
4003 sbi->s_li_request->lr_timeout = 0;
4007 if (sb_rdonly(sb) ||
4008 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4009 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4012 elr = ext4_li_request_new(sb, first_not_zeroed);
4018 if (NULL == ext4_li_info) {
4019 ret = ext4_li_info_new();
4024 mutex_lock(&ext4_li_info->li_list_mtx);
4025 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4026 mutex_unlock(&ext4_li_info->li_list_mtx);
4028 sbi->s_li_request = elr;
4030 * set elr to NULL here since it has been inserted to
4031 * the request_list and the removal and free of it is
4032 * handled by ext4_clear_request_list from now on.
4036 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4037 ret = ext4_run_lazyinit_thread();
4042 mutex_unlock(&ext4_li_mtx);
4049 * We do not need to lock anything since this is called on
4052 static void ext4_destroy_lazyinit_thread(void)
4055 * If thread exited earlier
4056 * there's nothing to be done.
4058 if (!ext4_li_info || !ext4_lazyinit_task)
4061 kthread_stop(ext4_lazyinit_task);
4064 static int set_journal_csum_feature_set(struct super_block *sb)
4067 int compat, incompat;
4068 struct ext4_sb_info *sbi = EXT4_SB(sb);
4070 if (ext4_has_metadata_csum(sb)) {
4071 /* journal checksum v3 */
4073 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4075 /* journal checksum v1 */
4076 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4080 jbd2_journal_clear_features(sbi->s_journal,
4081 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4082 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4083 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4084 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4085 ret = jbd2_journal_set_features(sbi->s_journal,
4087 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4089 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4090 ret = jbd2_journal_set_features(sbi->s_journal,
4093 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4094 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4096 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4097 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4104 * Note: calculating the overhead so we can be compatible with
4105 * historical BSD practice is quite difficult in the face of
4106 * clusters/bigalloc. This is because multiple metadata blocks from
4107 * different block group can end up in the same allocation cluster.
4108 * Calculating the exact overhead in the face of clustered allocation
4109 * requires either O(all block bitmaps) in memory or O(number of block
4110 * groups**2) in time. We will still calculate the superblock for
4111 * older file systems --- and if we come across with a bigalloc file
4112 * system with zero in s_overhead_clusters the estimate will be close to
4113 * correct especially for very large cluster sizes --- but for newer
4114 * file systems, it's better to calculate this figure once at mkfs
4115 * time, and store it in the superblock. If the superblock value is
4116 * present (even for non-bigalloc file systems), we will use it.
4118 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4121 struct ext4_sb_info *sbi = EXT4_SB(sb);
4122 struct ext4_group_desc *gdp;
4123 ext4_fsblk_t first_block, last_block, b;
4124 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4125 int s, j, count = 0;
4126 int has_super = ext4_bg_has_super(sb, grp);
4128 if (!ext4_has_feature_bigalloc(sb))
4129 return (has_super + ext4_bg_num_gdb(sb, grp) +
4130 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4131 sbi->s_itb_per_group + 2);
4133 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4134 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4135 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4136 for (i = 0; i < ngroups; i++) {
4137 gdp = ext4_get_group_desc(sb, i, NULL);
4138 b = ext4_block_bitmap(sb, gdp);
4139 if (b >= first_block && b <= last_block) {
4140 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4143 b = ext4_inode_bitmap(sb, gdp);
4144 if (b >= first_block && b <= last_block) {
4145 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4148 b = ext4_inode_table(sb, gdp);
4149 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4150 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4151 int c = EXT4_B2C(sbi, b - first_block);
4152 ext4_set_bit(c, buf);
4158 if (ext4_bg_has_super(sb, grp)) {
4159 ext4_set_bit(s++, buf);
4162 j = ext4_bg_num_gdb(sb, grp);
4163 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4164 ext4_error(sb, "Invalid number of block group "
4165 "descriptor blocks: %d", j);
4166 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4170 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4174 return EXT4_CLUSTERS_PER_GROUP(sb) -
4175 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4179 * Compute the overhead and stash it in sbi->s_overhead
4181 int ext4_calculate_overhead(struct super_block *sb)
4183 struct ext4_sb_info *sbi = EXT4_SB(sb);
4184 struct ext4_super_block *es = sbi->s_es;
4185 struct inode *j_inode;
4186 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4187 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4188 ext4_fsblk_t overhead = 0;
4189 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4195 * Compute the overhead (FS structures). This is constant
4196 * for a given filesystem unless the number of block groups
4197 * changes so we cache the previous value until it does.
4201 * All of the blocks before first_data_block are overhead
4203 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4206 * Add the overhead found in each block group
4208 for (i = 0; i < ngroups; i++) {
4211 blks = count_overhead(sb, i, buf);
4214 memset(buf, 0, PAGE_SIZE);
4219 * Add the internal journal blocks whether the journal has been
4222 if (sbi->s_journal && !sbi->s_journal_bdev)
4223 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4224 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4225 /* j_inum for internal journal is non-zero */
4226 j_inode = ext4_get_journal_inode(sb, j_inum);
4228 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4229 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4232 ext4_msg(sb, KERN_ERR, "can't get journal size");
4235 sbi->s_overhead = overhead;
4237 free_page((unsigned long) buf);
4241 static void ext4_set_resv_clusters(struct super_block *sb)
4243 ext4_fsblk_t resv_clusters;
4244 struct ext4_sb_info *sbi = EXT4_SB(sb);
4247 * There's no need to reserve anything when we aren't using extents.
4248 * The space estimates are exact, there are no unwritten extents,
4249 * hole punching doesn't need new metadata... This is needed especially
4250 * to keep ext2/3 backward compatibility.
4252 if (!ext4_has_feature_extents(sb))
4255 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4256 * This should cover the situations where we can not afford to run
4257 * out of space like for example punch hole, or converting
4258 * unwritten extents in delalloc path. In most cases such
4259 * allocation would require 1, or 2 blocks, higher numbers are
4262 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4263 sbi->s_cluster_bits);
4265 do_div(resv_clusters, 50);
4266 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4268 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4271 static const char *ext4_quota_mode(struct super_block *sb)
4274 if (!ext4_quota_capable(sb))
4277 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4278 return "journalled";
4286 static void ext4_setup_csum_trigger(struct super_block *sb,
4287 enum ext4_journal_trigger_type type,
4289 struct jbd2_buffer_trigger_type *type,
4290 struct buffer_head *bh,
4294 struct ext4_sb_info *sbi = EXT4_SB(sb);
4296 sbi->s_journal_triggers[type].sb = sb;
4297 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4300 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4305 kfree(sbi->s_blockgroup_lock);
4306 fs_put_dax(sbi->s_daxdev, NULL);
4310 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4312 struct ext4_sb_info *sbi;
4314 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4318 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4321 sbi->s_blockgroup_lock =
4322 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4324 if (!sbi->s_blockgroup_lock)
4327 sb->s_fs_info = sbi;
4331 fs_put_dax(sbi->s_daxdev, NULL);
4336 static void ext4_set_def_opts(struct super_block *sb,
4337 struct ext4_super_block *es)
4339 unsigned long def_mount_opts;
4341 /* Set defaults before we parse the mount options */
4342 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4343 set_opt(sb, INIT_INODE_TABLE);
4344 if (def_mount_opts & EXT4_DEFM_DEBUG)
4346 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4348 if (def_mount_opts & EXT4_DEFM_UID16)
4349 set_opt(sb, NO_UID32);
4350 /* xattr user namespace & acls are now defaulted on */
4351 set_opt(sb, XATTR_USER);
4352 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4353 set_opt(sb, POSIX_ACL);
4355 if (ext4_has_feature_fast_commit(sb))
4356 set_opt2(sb, JOURNAL_FAST_COMMIT);
4357 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4358 if (ext4_has_metadata_csum(sb))
4359 set_opt(sb, JOURNAL_CHECKSUM);
4361 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4362 set_opt(sb, JOURNAL_DATA);
4363 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4364 set_opt(sb, ORDERED_DATA);
4365 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4366 set_opt(sb, WRITEBACK_DATA);
4368 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4369 set_opt(sb, ERRORS_PANIC);
4370 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4371 set_opt(sb, ERRORS_CONT);
4373 set_opt(sb, ERRORS_RO);
4374 /* block_validity enabled by default; disable with noblock_validity */
4375 set_opt(sb, BLOCK_VALIDITY);
4376 if (def_mount_opts & EXT4_DEFM_DISCARD)
4377 set_opt(sb, DISCARD);
4379 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4380 set_opt(sb, BARRIER);
4383 * enable delayed allocation by default
4384 * Use -o nodelalloc to turn it off
4386 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4387 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4388 set_opt(sb, DELALLOC);
4390 if (sb->s_blocksize == PAGE_SIZE)
4391 set_opt(sb, DIOREAD_NOLOCK);
4394 static int ext4_handle_clustersize(struct super_block *sb)
4396 struct ext4_sb_info *sbi = EXT4_SB(sb);
4397 struct ext4_super_block *es = sbi->s_es;
4400 /* Handle clustersize */
4401 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4402 if (ext4_has_feature_bigalloc(sb)) {
4403 if (clustersize < sb->s_blocksize) {
4404 ext4_msg(sb, KERN_ERR,
4405 "cluster size (%d) smaller than "
4406 "block size (%lu)", clustersize, sb->s_blocksize);
4409 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4410 le32_to_cpu(es->s_log_block_size);
4411 sbi->s_clusters_per_group =
4412 le32_to_cpu(es->s_clusters_per_group);
4413 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4414 ext4_msg(sb, KERN_ERR,
4415 "#clusters per group too big: %lu",
4416 sbi->s_clusters_per_group);
4419 if (sbi->s_blocks_per_group !=
4420 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4421 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4422 "clusters per group (%lu) inconsistent",
4423 sbi->s_blocks_per_group,
4424 sbi->s_clusters_per_group);
4428 if (clustersize != sb->s_blocksize) {
4429 ext4_msg(sb, KERN_ERR,
4430 "fragment/cluster size (%d) != "
4431 "block size (%lu)", clustersize, sb->s_blocksize);
4434 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4435 ext4_msg(sb, KERN_ERR,
4436 "#blocks per group too big: %lu",
4437 sbi->s_blocks_per_group);
4440 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4441 sbi->s_cluster_bits = 0;
4443 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4445 /* Do we have standard group size of clustersize * 8 blocks ? */
4446 if (sbi->s_blocks_per_group == clustersize << 3)
4447 set_opt2(sb, STD_GROUP_SIZE);
4452 static void ext4_fast_commit_init(struct super_block *sb)
4454 struct ext4_sb_info *sbi = EXT4_SB(sb);
4456 /* Initialize fast commit stuff */
4457 atomic_set(&sbi->s_fc_subtid, 0);
4458 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4459 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4460 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4461 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4462 sbi->s_fc_bytes = 0;
4463 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4464 sbi->s_fc_ineligible_tid = 0;
4465 spin_lock_init(&sbi->s_fc_lock);
4466 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4467 sbi->s_fc_replay_state.fc_regions = NULL;
4468 sbi->s_fc_replay_state.fc_regions_size = 0;
4469 sbi->s_fc_replay_state.fc_regions_used = 0;
4470 sbi->s_fc_replay_state.fc_regions_valid = 0;
4471 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4472 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4473 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4476 static int ext4_inode_info_init(struct super_block *sb,
4477 struct ext4_super_block *es)
4479 struct ext4_sb_info *sbi = EXT4_SB(sb);
4481 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4482 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4483 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4485 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4486 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4487 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4488 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4492 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4493 (!is_power_of_2(sbi->s_inode_size)) ||
4494 (sbi->s_inode_size > sb->s_blocksize)) {
4495 ext4_msg(sb, KERN_ERR,
4496 "unsupported inode size: %d",
4498 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4502 * i_atime_extra is the last extra field available for
4503 * [acm]times in struct ext4_inode. Checking for that
4504 * field should suffice to ensure we have extra space
4507 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4508 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4509 sb->s_time_gran = 1;
4510 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4512 sb->s_time_gran = NSEC_PER_SEC;
4513 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4515 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4518 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4519 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4520 EXT4_GOOD_OLD_INODE_SIZE;
4521 if (ext4_has_feature_extra_isize(sb)) {
4522 unsigned v, max = (sbi->s_inode_size -
4523 EXT4_GOOD_OLD_INODE_SIZE);
4525 v = le16_to_cpu(es->s_want_extra_isize);
4527 ext4_msg(sb, KERN_ERR,
4528 "bad s_want_extra_isize: %d", v);
4531 if (sbi->s_want_extra_isize < v)
4532 sbi->s_want_extra_isize = v;
4534 v = le16_to_cpu(es->s_min_extra_isize);
4536 ext4_msg(sb, KERN_ERR,
4537 "bad s_min_extra_isize: %d", v);
4540 if (sbi->s_want_extra_isize < v)
4541 sbi->s_want_extra_isize = v;
4548 #if IS_ENABLED(CONFIG_UNICODE)
4549 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4551 const struct ext4_sb_encodings *encoding_info;
4552 struct unicode_map *encoding;
4553 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4555 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4558 encoding_info = ext4_sb_read_encoding(es);
4559 if (!encoding_info) {
4560 ext4_msg(sb, KERN_ERR,
4561 "Encoding requested by superblock is unknown");
4565 encoding = utf8_load(encoding_info->version);
4566 if (IS_ERR(encoding)) {
4567 ext4_msg(sb, KERN_ERR,
4568 "can't mount with superblock charset: %s-%u.%u.%u "
4569 "not supported by the kernel. flags: 0x%x.",
4570 encoding_info->name,
4571 unicode_major(encoding_info->version),
4572 unicode_minor(encoding_info->version),
4573 unicode_rev(encoding_info->version),
4577 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4578 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4579 unicode_major(encoding_info->version),
4580 unicode_minor(encoding_info->version),
4581 unicode_rev(encoding_info->version),
4584 sb->s_encoding = encoding;
4585 sb->s_encoding_flags = encoding_flags;
4590 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4596 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4598 struct ext4_sb_info *sbi = EXT4_SB(sb);
4600 /* Warn if metadata_csum and gdt_csum are both set. */
4601 if (ext4_has_feature_metadata_csum(sb) &&
4602 ext4_has_feature_gdt_csum(sb))
4603 ext4_warning(sb, "metadata_csum and uninit_bg are "
4604 "redundant flags; please run fsck.");
4606 /* Check for a known checksum algorithm */
4607 if (!ext4_verify_csum_type(sb, es)) {
4608 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4609 "unknown checksum algorithm.");
4612 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4613 ext4_orphan_file_block_trigger);
4615 /* Load the checksum driver */
4616 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4617 if (IS_ERR(sbi->s_chksum_driver)) {
4618 int ret = PTR_ERR(sbi->s_chksum_driver);
4619 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4620 sbi->s_chksum_driver = NULL;
4624 /* Check superblock checksum */
4625 if (!ext4_superblock_csum_verify(sb, es)) {
4626 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4627 "invalid superblock checksum. Run e2fsck?");
4631 /* Precompute checksum seed for all metadata */
4632 if (ext4_has_feature_csum_seed(sb))
4633 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4634 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4635 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4636 sizeof(es->s_uuid));
4640 static int ext4_check_feature_compatibility(struct super_block *sb,
4641 struct ext4_super_block *es,
4644 struct ext4_sb_info *sbi = EXT4_SB(sb);
4646 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4647 (ext4_has_compat_features(sb) ||
4648 ext4_has_ro_compat_features(sb) ||
4649 ext4_has_incompat_features(sb)))
4650 ext4_msg(sb, KERN_WARNING,
4651 "feature flags set on rev 0 fs, "
4652 "running e2fsck is recommended");
4654 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4655 set_opt2(sb, HURD_COMPAT);
4656 if (ext4_has_feature_64bit(sb)) {
4657 ext4_msg(sb, KERN_ERR,
4658 "The Hurd can't support 64-bit file systems");
4663 * ea_inode feature uses l_i_version field which is not
4664 * available in HURD_COMPAT mode.
4666 if (ext4_has_feature_ea_inode(sb)) {
4667 ext4_msg(sb, KERN_ERR,
4668 "ea_inode feature is not supported for Hurd");
4673 if (IS_EXT2_SB(sb)) {
4674 if (ext2_feature_set_ok(sb))
4675 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4676 "using the ext4 subsystem");
4679 * If we're probing be silent, if this looks like
4680 * it's actually an ext[34] filesystem.
4682 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4684 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4685 "to feature incompatibilities");
4690 if (IS_EXT3_SB(sb)) {
4691 if (ext3_feature_set_ok(sb))
4692 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4693 "using the ext4 subsystem");
4696 * If we're probing be silent, if this looks like
4697 * it's actually an ext4 filesystem.
4699 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4701 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4702 "to feature incompatibilities");
4708 * Check feature flags regardless of the revision level, since we
4709 * previously didn't change the revision level when setting the flags,
4710 * so there is a chance incompat flags are set on a rev 0 filesystem.
4712 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4715 if (sbi->s_daxdev) {
4716 if (sb->s_blocksize == PAGE_SIZE)
4717 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4719 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4722 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4723 if (ext4_has_feature_inline_data(sb)) {
4724 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4725 " that may contain inline data");
4728 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4729 ext4_msg(sb, KERN_ERR,
4730 "DAX unsupported by block device.");
4735 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4736 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4737 es->s_encryption_level);
4744 static int ext4_check_geometry(struct super_block *sb,
4745 struct ext4_super_block *es)
4747 struct ext4_sb_info *sbi = EXT4_SB(sb);
4751 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4752 ext4_msg(sb, KERN_ERR,
4753 "Number of reserved GDT blocks insanely large: %d",
4754 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4758 * Test whether we have more sectors than will fit in sector_t,
4759 * and whether the max offset is addressable by the page cache.
4761 err = generic_check_addressable(sb->s_blocksize_bits,
4762 ext4_blocks_count(es));
4764 ext4_msg(sb, KERN_ERR, "filesystem"
4765 " too large to mount safely on this system");
4769 /* check blocks count against device size */
4770 blocks_count = sb_bdev_nr_blocks(sb);
4771 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4772 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4773 "exceeds size of device (%llu blocks)",
4774 ext4_blocks_count(es), blocks_count);
4779 * It makes no sense for the first data block to be beyond the end
4780 * of the filesystem.
4782 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4783 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4784 "block %u is beyond end of filesystem (%llu)",
4785 le32_to_cpu(es->s_first_data_block),
4786 ext4_blocks_count(es));
4789 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4790 (sbi->s_cluster_ratio == 1)) {
4791 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4792 "block is 0 with a 1k block and cluster size");
4796 blocks_count = (ext4_blocks_count(es) -
4797 le32_to_cpu(es->s_first_data_block) +
4798 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4799 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4800 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4801 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4802 "(block count %llu, first data block %u, "
4803 "blocks per group %lu)", blocks_count,
4804 ext4_blocks_count(es),
4805 le32_to_cpu(es->s_first_data_block),
4806 EXT4_BLOCKS_PER_GROUP(sb));
4809 sbi->s_groups_count = blocks_count;
4810 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4811 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4812 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4813 le32_to_cpu(es->s_inodes_count)) {
4814 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4815 le32_to_cpu(es->s_inodes_count),
4816 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4823 static int ext4_group_desc_init(struct super_block *sb,
4824 struct ext4_super_block *es,
4825 ext4_fsblk_t logical_sb_block,
4826 ext4_group_t *first_not_zeroed)
4828 struct ext4_sb_info *sbi = EXT4_SB(sb);
4829 unsigned int db_count;
4833 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4834 EXT4_DESC_PER_BLOCK(sb);
4835 if (ext4_has_feature_meta_bg(sb)) {
4836 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4837 ext4_msg(sb, KERN_WARNING,
4838 "first meta block group too large: %u "
4839 "(group descriptor block count %u)",
4840 le32_to_cpu(es->s_first_meta_bg), db_count);
4844 rcu_assign_pointer(sbi->s_group_desc,
4845 kvmalloc_array(db_count,
4846 sizeof(struct buffer_head *),
4848 if (sbi->s_group_desc == NULL) {
4849 ext4_msg(sb, KERN_ERR, "not enough memory");
4853 bgl_lock_init(sbi->s_blockgroup_lock);
4855 /* Pre-read the descriptors into the buffer cache */
4856 for (i = 0; i < db_count; i++) {
4857 block = descriptor_loc(sb, logical_sb_block, i);
4858 ext4_sb_breadahead_unmovable(sb, block);
4861 for (i = 0; i < db_count; i++) {
4862 struct buffer_head *bh;
4864 block = descriptor_loc(sb, logical_sb_block, i);
4865 bh = ext4_sb_bread_unmovable(sb, block);
4867 ext4_msg(sb, KERN_ERR,
4868 "can't read group descriptor %d", i);
4869 sbi->s_gdb_count = i;
4873 rcu_dereference(sbi->s_group_desc)[i] = bh;
4876 sbi->s_gdb_count = db_count;
4877 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4878 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4879 return -EFSCORRUPTED;
4885 static int ext4_load_and_init_journal(struct super_block *sb,
4886 struct ext4_super_block *es,
4887 struct ext4_fs_context *ctx)
4889 struct ext4_sb_info *sbi = EXT4_SB(sb);
4892 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4896 if (ext4_has_feature_64bit(sb) &&
4897 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4898 JBD2_FEATURE_INCOMPAT_64BIT)) {
4899 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4903 if (!set_journal_csum_feature_set(sb)) {
4904 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4909 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4910 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4911 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4912 ext4_msg(sb, KERN_ERR,
4913 "Failed to set fast commit journal feature");
4917 /* We have now updated the journal if required, so we can
4918 * validate the data journaling mode. */
4919 switch (test_opt(sb, DATA_FLAGS)) {
4921 /* No mode set, assume a default based on the journal
4922 * capabilities: ORDERED_DATA if the journal can
4923 * cope, else JOURNAL_DATA
4925 if (jbd2_journal_check_available_features
4926 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4927 set_opt(sb, ORDERED_DATA);
4928 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4930 set_opt(sb, JOURNAL_DATA);
4931 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4935 case EXT4_MOUNT_ORDERED_DATA:
4936 case EXT4_MOUNT_WRITEBACK_DATA:
4937 if (!jbd2_journal_check_available_features
4938 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4939 ext4_msg(sb, KERN_ERR, "Journal does not support "
4940 "requested data journaling mode");
4948 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4949 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4950 ext4_msg(sb, KERN_ERR, "can't mount with "
4951 "journal_async_commit in data=ordered mode");
4955 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4957 sbi->s_journal->j_submit_inode_data_buffers =
4958 ext4_journal_submit_inode_data_buffers;
4959 sbi->s_journal->j_finish_inode_data_buffers =
4960 ext4_journal_finish_inode_data_buffers;
4965 /* flush s_error_work before journal destroy. */
4966 flush_work(&sbi->s_error_work);
4967 jbd2_journal_destroy(sbi->s_journal);
4968 sbi->s_journal = NULL;
4972 static int ext4_check_journal_data_mode(struct super_block *sb)
4974 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4975 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4976 "data=journal disables delayed allocation, "
4977 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4978 /* can't mount with both data=journal and dioread_nolock. */
4979 clear_opt(sb, DIOREAD_NOLOCK);
4980 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4981 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4982 ext4_msg(sb, KERN_ERR, "can't mount with "
4983 "both data=journal and delalloc");
4986 if (test_opt(sb, DAX_ALWAYS)) {
4987 ext4_msg(sb, KERN_ERR, "can't mount with "
4988 "both data=journal and dax");
4991 if (ext4_has_feature_encrypt(sb)) {
4992 ext4_msg(sb, KERN_WARNING,
4993 "encrypted files will use data=ordered "
4994 "instead of data journaling mode");
4996 if (test_opt(sb, DELALLOC))
4997 clear_opt(sb, DELALLOC);
4999 sb->s_iflags |= SB_I_CGROUPWB;
5005 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5008 struct ext4_sb_info *sbi = EXT4_SB(sb);
5009 struct ext4_super_block *es;
5010 ext4_fsblk_t logical_sb_block;
5011 unsigned long offset = 0;
5012 struct buffer_head *bh;
5016 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5018 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5023 * The ext4 superblock will not be buffer aligned for other than 1kB
5024 * block sizes. We need to calculate the offset from buffer start.
5026 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5027 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5028 offset = do_div(logical_sb_block, blocksize);
5030 logical_sb_block = sbi->s_sb_block;
5033 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5035 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5039 * Note: s_es must be initialized as soon as possible because
5040 * some ext4 macro-instructions depend on its value
5042 es = (struct ext4_super_block *) (bh->b_data + offset);
5044 sb->s_magic = le16_to_cpu(es->s_magic);
5045 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5047 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5051 if (le32_to_cpu(es->s_log_block_size) >
5052 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5053 ext4_msg(sb, KERN_ERR,
5054 "Invalid log block size: %u",
5055 le32_to_cpu(es->s_log_block_size));
5058 if (le32_to_cpu(es->s_log_cluster_size) >
5059 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5060 ext4_msg(sb, KERN_ERR,
5061 "Invalid log cluster size: %u",
5062 le32_to_cpu(es->s_log_cluster_size));
5066 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5069 * If the default block size is not the same as the real block size,
5070 * we need to reload it.
5072 if (sb->s_blocksize == blocksize) {
5073 *lsb = logical_sb_block;
5079 * bh must be released before kill_bdev(), otherwise
5080 * it won't be freed and its page also. kill_bdev()
5081 * is called by sb_set_blocksize().
5084 /* Validate the filesystem blocksize */
5085 if (!sb_set_blocksize(sb, blocksize)) {
5086 ext4_msg(sb, KERN_ERR, "bad block size %d",
5092 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5093 offset = do_div(logical_sb_block, blocksize);
5094 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5096 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5101 es = (struct ext4_super_block *)(bh->b_data + offset);
5103 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5104 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5107 *lsb = logical_sb_block;
5115 static void ext4_hash_info_init(struct super_block *sb)
5117 struct ext4_sb_info *sbi = EXT4_SB(sb);
5118 struct ext4_super_block *es = sbi->s_es;
5121 for (i = 0; i < 4; i++)
5122 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5124 sbi->s_def_hash_version = es->s_def_hash_version;
5125 if (ext4_has_feature_dir_index(sb)) {
5126 i = le32_to_cpu(es->s_flags);
5127 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5128 sbi->s_hash_unsigned = 3;
5129 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5130 #ifdef __CHAR_UNSIGNED__
5133 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5134 sbi->s_hash_unsigned = 3;
5138 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5144 static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5146 struct ext4_sb_info *sbi = EXT4_SB(sb);
5147 struct ext4_super_block *es = sbi->s_es;
5150 has_huge_files = ext4_has_feature_huge_file(sb);
5151 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5153 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5155 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5156 if (ext4_has_feature_64bit(sb)) {
5157 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5158 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5159 !is_power_of_2(sbi->s_desc_size)) {
5160 ext4_msg(sb, KERN_ERR,
5161 "unsupported descriptor size %lu",
5166 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5168 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5169 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5171 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5172 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5174 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5177 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5178 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5179 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5180 sbi->s_inodes_per_group);
5183 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5184 sbi->s_inodes_per_block;
5185 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5186 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5187 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5188 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5193 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5195 struct ext4_super_block *es = NULL;
5196 struct ext4_sb_info *sbi = EXT4_SB(sb);
5197 ext4_fsblk_t logical_sb_block;
5201 ext4_group_t first_not_zeroed;
5202 struct ext4_fs_context *ctx = fc->fs_private;
5203 int silent = fc->sb_flags & SB_SILENT;
5205 /* Set defaults for the variables that will be set during parsing */
5206 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5207 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5209 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5210 sbi->s_sectors_written_start =
5211 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5213 err = ext4_load_super(sb, &logical_sb_block, silent);
5218 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5220 err = ext4_init_metadata_csum(sb, es);
5224 ext4_set_def_opts(sb, es);
5226 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5227 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5228 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5229 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5230 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5233 * set default s_li_wait_mult for lazyinit, for the case there is
5234 * no mount option specified.
5236 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5238 err = ext4_inode_info_init(sb, es);
5242 err = parse_apply_sb_mount_options(sb, ctx);
5246 sbi->s_def_mount_opt = sbi->s_mount_opt;
5247 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5249 err = ext4_check_opt_consistency(fc, sb);
5253 ext4_apply_options(fc, sb);
5255 err = ext4_encoding_init(sb, es);
5259 err = ext4_check_journal_data_mode(sb);
5263 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5264 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5266 /* i_version is always enabled now */
5267 sb->s_flags |= SB_I_VERSION;
5269 err = ext4_check_feature_compatibility(sb, es, silent);
5273 err = ext4_block_group_meta_init(sb, silent);
5277 ext4_hash_info_init(sb);
5279 err = ext4_handle_clustersize(sb);
5283 err = ext4_check_geometry(sb, es);
5287 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5288 spin_lock_init(&sbi->s_error_lock);
5289 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5291 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5295 err = ext4_es_register_shrinker(sbi);
5299 sbi->s_stripe = ext4_get_stripe_size(sbi);
5300 sbi->s_extent_max_zeroout_kb = 32;
5303 * set up enough so that it can read an inode
5305 sb->s_op = &ext4_sops;
5306 sb->s_export_op = &ext4_export_ops;
5307 sb->s_xattr = ext4_xattr_handlers;
5308 #ifdef CONFIG_FS_ENCRYPTION
5309 sb->s_cop = &ext4_cryptops;
5311 #ifdef CONFIG_FS_VERITY
5312 sb->s_vop = &ext4_verityops;
5315 sb->dq_op = &ext4_quota_operations;
5316 if (ext4_has_feature_quota(sb))
5317 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5319 sb->s_qcop = &ext4_qctl_operations;
5320 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5322 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5324 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5325 mutex_init(&sbi->s_orphan_lock);
5327 ext4_fast_commit_init(sb);
5331 needs_recovery = (es->s_last_orphan != 0 ||
5332 ext4_has_feature_orphan_present(sb) ||
5333 ext4_has_feature_journal_needs_recovery(sb));
5335 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5336 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5338 goto failed_mount3a;
5343 * The first inode we look at is the journal inode. Don't try
5344 * root first: it may be modified in the journal!
5346 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5347 err = ext4_load_and_init_journal(sb, es, ctx);
5349 goto failed_mount3a;
5350 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5351 ext4_has_feature_journal_needs_recovery(sb)) {
5352 ext4_msg(sb, KERN_ERR, "required journal recovery "
5353 "suppressed and not mounted read-only");
5354 goto failed_mount3a;
5356 /* Nojournal mode, all journal mount options are illegal */
5357 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5358 ext4_msg(sb, KERN_ERR, "can't mount with "
5359 "journal_async_commit, fs mounted w/o journal");
5360 goto failed_mount3a;
5363 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5364 ext4_msg(sb, KERN_ERR, "can't mount with "
5365 "journal_checksum, fs mounted w/o journal");
5366 goto failed_mount3a;
5368 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5369 ext4_msg(sb, KERN_ERR, "can't mount with "
5370 "commit=%lu, fs mounted w/o journal",
5371 sbi->s_commit_interval / HZ);
5372 goto failed_mount3a;
5374 if (EXT4_MOUNT_DATA_FLAGS &
5375 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5376 ext4_msg(sb, KERN_ERR, "can't mount with "
5377 "data=, fs mounted w/o journal");
5378 goto failed_mount3a;
5380 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5381 clear_opt(sb, JOURNAL_CHECKSUM);
5382 clear_opt(sb, DATA_FLAGS);
5383 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5384 sbi->s_journal = NULL;
5388 if (!test_opt(sb, NO_MBCACHE)) {
5389 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5390 if (!sbi->s_ea_block_cache) {
5391 ext4_msg(sb, KERN_ERR,
5392 "Failed to create ea_block_cache");
5394 goto failed_mount_wq;
5397 if (ext4_has_feature_ea_inode(sb)) {
5398 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5399 if (!sbi->s_ea_inode_cache) {
5400 ext4_msg(sb, KERN_ERR,
5401 "Failed to create ea_inode_cache");
5403 goto failed_mount_wq;
5409 * Get the # of file system overhead blocks from the
5410 * superblock if present.
5412 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5413 /* ignore the precalculated value if it is ridiculous */
5414 if (sbi->s_overhead > ext4_blocks_count(es))
5415 sbi->s_overhead = 0;
5417 * If the bigalloc feature is not enabled recalculating the
5418 * overhead doesn't take long, so we might as well just redo
5419 * it to make sure we are using the correct value.
5421 if (!ext4_has_feature_bigalloc(sb))
5422 sbi->s_overhead = 0;
5423 if (sbi->s_overhead == 0) {
5424 err = ext4_calculate_overhead(sb);
5426 goto failed_mount_wq;
5430 * The maximum number of concurrent works can be high and
5431 * concurrency isn't really necessary. Limit it to 1.
5433 EXT4_SB(sb)->rsv_conversion_wq =
5434 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5435 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5436 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5442 * The jbd2_journal_load will have done any necessary log recovery,
5443 * so we can safely mount the rest of the filesystem now.
5446 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5448 ext4_msg(sb, KERN_ERR, "get root inode failed");
5449 err = PTR_ERR(root);
5453 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5454 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5456 err = -EFSCORRUPTED;
5460 sb->s_root = d_make_root(root);
5462 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5467 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5468 if (err == -EROFS) {
5469 sb->s_flags |= SB_RDONLY;
5471 goto failed_mount4a;
5473 ext4_set_resv_clusters(sb);
5475 if (test_opt(sb, BLOCK_VALIDITY)) {
5476 err = ext4_setup_system_zone(sb);
5478 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5480 goto failed_mount4a;
5483 ext4_fc_replay_cleanup(sb);
5488 * Enable optimize_scan if number of groups is > threshold. This can be
5489 * turned off by passing "mb_optimize_scan=0". This can also be
5490 * turned on forcefully by passing "mb_optimize_scan=1".
5492 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5493 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5494 set_opt2(sb, MB_OPTIMIZE_SCAN);
5496 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5499 err = ext4_mb_init(sb);
5501 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5507 * We can only set up the journal commit callback once
5508 * mballoc is initialized
5511 sbi->s_journal->j_commit_callback =
5512 ext4_journal_commit_callback;
5514 err = ext4_percpu_param_init(sbi);
5518 if (ext4_has_feature_flex_bg(sb))
5519 if (!ext4_fill_flex_info(sb)) {
5520 ext4_msg(sb, KERN_ERR,
5521 "unable to initialize "
5522 "flex_bg meta info!");
5527 err = ext4_register_li_request(sb, first_not_zeroed);
5531 err = ext4_register_sysfs(sb);
5535 err = ext4_init_orphan_info(sb);
5539 /* Enable quota usage during mount. */
5540 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5541 err = ext4_enable_quotas(sb);
5545 #endif /* CONFIG_QUOTA */
5548 * Save the original bdev mapping's wb_err value which could be
5549 * used to detect the metadata async write error.
5551 spin_lock_init(&sbi->s_bdev_wb_lock);
5552 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5553 &sbi->s_bdev_wb_err);
5554 sb->s_bdev->bd_super = sb;
5555 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5556 ext4_orphan_cleanup(sb, es);
5557 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5559 * Update the checksum after updating free space/inode counters and
5560 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5561 * checksum in the buffer cache until it is written out and
5562 * e2fsprogs programs trying to open a file system immediately
5563 * after it is mounted can fail.
5565 ext4_superblock_csum_set(sb);
5566 if (needs_recovery) {
5567 ext4_msg(sb, KERN_INFO, "recovery complete");
5568 err = ext4_mark_recovery_complete(sb, es);
5573 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5574 ext4_msg(sb, KERN_WARNING,
5575 "mounting with \"discard\" option, but the device does not support discard");
5577 if (es->s_error_count)
5578 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5580 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5581 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5582 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5583 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5584 atomic_set(&sbi->s_warning_count, 0);
5585 atomic_set(&sbi->s_msg_count, 0);
5590 ext4_release_orphan_info(sb);
5592 ext4_unregister_sysfs(sb);
5593 kobject_put(&sbi->s_kobj);
5595 ext4_unregister_li_request(sb);
5597 ext4_mb_release(sb);
5598 ext4_flex_groups_free(sbi);
5599 ext4_percpu_param_destroy(sbi);
5601 ext4_ext_release(sb);
5602 ext4_release_system_zone(sb);
5607 ext4_msg(sb, KERN_ERR, "mount failed");
5608 if (EXT4_SB(sb)->rsv_conversion_wq)
5609 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5611 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5612 sbi->s_ea_inode_cache = NULL;
5614 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5615 sbi->s_ea_block_cache = NULL;
5617 if (sbi->s_journal) {
5618 /* flush s_error_work before journal destroy. */
5619 flush_work(&sbi->s_error_work);
5620 jbd2_journal_destroy(sbi->s_journal);
5621 sbi->s_journal = NULL;
5624 ext4_es_unregister_shrinker(sbi);
5626 /* flush s_error_work before sbi destroy */
5627 flush_work(&sbi->s_error_work);
5628 del_timer_sync(&sbi->s_err_report);
5629 ext4_stop_mmpd(sbi);
5630 ext4_group_desc_free(sbi);
5632 if (sbi->s_chksum_driver)
5633 crypto_free_shash(sbi->s_chksum_driver);
5635 #if IS_ENABLED(CONFIG_UNICODE)
5636 utf8_unload(sb->s_encoding);
5640 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5641 kfree(get_qf_name(sb, sbi, i));
5643 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5644 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5646 ext4_blkdev_remove(sbi);
5648 sb->s_fs_info = NULL;
5652 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5654 struct ext4_fs_context *ctx = fc->fs_private;
5655 struct ext4_sb_info *sbi;
5659 sbi = ext4_alloc_sbi(sb);
5663 fc->s_fs_info = sbi;
5665 /* Cleanup superblock name */
5666 strreplace(sb->s_id, '/', '!');
5668 sbi->s_sb_block = 1; /* Default super block location */
5669 if (ctx->spec & EXT4_SPEC_s_sb_block)
5670 sbi->s_sb_block = ctx->s_sb_block;
5672 ret = __ext4_fill_super(fc, sb);
5676 if (sbi->s_journal) {
5677 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5678 descr = " journalled data mode";
5679 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5680 descr = " ordered data mode";
5682 descr = " writeback data mode";
5684 descr = "out journal";
5686 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5687 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU with%s. "
5688 "Quota mode: %s.", &sb->s_uuid, descr,
5689 ext4_quota_mode(sb));
5691 /* Update the s_overhead_clusters if necessary */
5692 ext4_update_overhead(sb, false);
5697 fc->s_fs_info = NULL;
5701 static int ext4_get_tree(struct fs_context *fc)
5703 return get_tree_bdev(fc, ext4_fill_super);
5707 * Setup any per-fs journal parameters now. We'll do this both on
5708 * initial mount, once the journal has been initialised but before we've
5709 * done any recovery; and again on any subsequent remount.
5711 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5713 struct ext4_sb_info *sbi = EXT4_SB(sb);
5715 journal->j_commit_interval = sbi->s_commit_interval;
5716 journal->j_min_batch_time = sbi->s_min_batch_time;
5717 journal->j_max_batch_time = sbi->s_max_batch_time;
5718 ext4_fc_init(sb, journal);
5720 write_lock(&journal->j_state_lock);
5721 if (test_opt(sb, BARRIER))
5722 journal->j_flags |= JBD2_BARRIER;
5724 journal->j_flags &= ~JBD2_BARRIER;
5725 if (test_opt(sb, DATA_ERR_ABORT))
5726 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5728 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5729 write_unlock(&journal->j_state_lock);
5732 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5733 unsigned int journal_inum)
5735 struct inode *journal_inode;
5738 * Test for the existence of a valid inode on disk. Bad things
5739 * happen if we iget() an unused inode, as the subsequent iput()
5740 * will try to delete it.
5742 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5743 if (IS_ERR(journal_inode)) {
5744 ext4_msg(sb, KERN_ERR, "no journal found");
5747 if (!journal_inode->i_nlink) {
5748 make_bad_inode(journal_inode);
5749 iput(journal_inode);
5750 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5754 ext4_debug("Journal inode found at %p: %lld bytes\n",
5755 journal_inode, journal_inode->i_size);
5756 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5757 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5758 iput(journal_inode);
5761 return journal_inode;
5764 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5766 struct ext4_map_blocks map;
5769 if (journal->j_inode == NULL)
5772 map.m_lblk = *block;
5774 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5776 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5777 "journal bmap failed: block %llu ret %d\n",
5779 jbd2_journal_abort(journal, ret ? ret : -EIO);
5782 *block = map.m_pblk;
5786 static journal_t *ext4_get_journal(struct super_block *sb,
5787 unsigned int journal_inum)
5789 struct inode *journal_inode;
5792 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5795 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5799 journal = jbd2_journal_init_inode(journal_inode);
5801 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5802 iput(journal_inode);
5805 journal->j_private = sb;
5806 journal->j_bmap = ext4_journal_bmap;
5807 ext4_init_journal_params(sb, journal);
5811 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5814 struct buffer_head *bh;
5818 int hblock, blocksize;
5819 ext4_fsblk_t sb_block;
5820 unsigned long offset;
5821 struct ext4_super_block *es;
5822 struct block_device *bdev;
5824 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5827 bdev = ext4_blkdev_get(j_dev, sb);
5831 blocksize = sb->s_blocksize;
5832 hblock = bdev_logical_block_size(bdev);
5833 if (blocksize < hblock) {
5834 ext4_msg(sb, KERN_ERR,
5835 "blocksize too small for journal device");
5839 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5840 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5841 set_blocksize(bdev, blocksize);
5842 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5843 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5844 "external journal");
5848 es = (struct ext4_super_block *) (bh->b_data + offset);
5849 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5850 !(le32_to_cpu(es->s_feature_incompat) &
5851 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5852 ext4_msg(sb, KERN_ERR, "external journal has "
5858 if ((le32_to_cpu(es->s_feature_ro_compat) &
5859 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5860 es->s_checksum != ext4_superblock_csum(sb, es)) {
5861 ext4_msg(sb, KERN_ERR, "external journal has "
5862 "corrupt superblock");
5867 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5868 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5873 len = ext4_blocks_count(es);
5874 start = sb_block + 1;
5875 brelse(bh); /* we're done with the superblock */
5877 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5878 start, len, blocksize);
5880 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5883 journal->j_private = sb;
5884 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5885 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5888 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5889 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5890 "user (unsupported) - %d",
5891 be32_to_cpu(journal->j_superblock->s_nr_users));
5894 EXT4_SB(sb)->s_journal_bdev = bdev;
5895 ext4_init_journal_params(sb, journal);
5899 jbd2_journal_destroy(journal);
5901 ext4_blkdev_put(bdev);
5905 static int ext4_load_journal(struct super_block *sb,
5906 struct ext4_super_block *es,
5907 unsigned long journal_devnum)
5910 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5913 int really_read_only;
5916 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5917 return -EFSCORRUPTED;
5919 if (journal_devnum &&
5920 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5921 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5922 "numbers have changed");
5923 journal_dev = new_decode_dev(journal_devnum);
5925 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5927 if (journal_inum && journal_dev) {
5928 ext4_msg(sb, KERN_ERR,
5929 "filesystem has both journal inode and journal device!");
5934 journal = ext4_get_journal(sb, journal_inum);
5938 journal = ext4_get_dev_journal(sb, journal_dev);
5943 journal_dev_ro = bdev_read_only(journal->j_dev);
5944 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5946 if (journal_dev_ro && !sb_rdonly(sb)) {
5947 ext4_msg(sb, KERN_ERR,
5948 "journal device read-only, try mounting with '-o ro'");
5954 * Are we loading a blank journal or performing recovery after a
5955 * crash? For recovery, we need to check in advance whether we
5956 * can get read-write access to the device.
5958 if (ext4_has_feature_journal_needs_recovery(sb)) {
5959 if (sb_rdonly(sb)) {
5960 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5961 "required on readonly filesystem");
5962 if (really_read_only) {
5963 ext4_msg(sb, KERN_ERR, "write access "
5964 "unavailable, cannot proceed "
5965 "(try mounting with noload)");
5969 ext4_msg(sb, KERN_INFO, "write access will "
5970 "be enabled during recovery");
5974 if (!(journal->j_flags & JBD2_BARRIER))
5975 ext4_msg(sb, KERN_INFO, "barriers disabled");
5977 if (!ext4_has_feature_journal_needs_recovery(sb))
5978 err = jbd2_journal_wipe(journal, !really_read_only);
5980 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5983 memcpy(save, ((char *) es) +
5984 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5985 err = jbd2_journal_load(journal);
5987 memcpy(((char *) es) + EXT4_S_ERR_START,
5988 save, EXT4_S_ERR_LEN);
5990 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
5992 /* Write out restored error information to the superblock */
5993 if (!bdev_read_only(sb->s_bdev)) {
5995 err2 = ext4_commit_super(sb);
6001 ext4_msg(sb, KERN_ERR, "error loading journal");
6005 EXT4_SB(sb)->s_journal = journal;
6006 err = ext4_clear_journal_err(sb, es);
6008 EXT4_SB(sb)->s_journal = NULL;
6009 jbd2_journal_destroy(journal);
6013 if (!really_read_only && journal_devnum &&
6014 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6015 es->s_journal_dev = cpu_to_le32(journal_devnum);
6016 ext4_commit_super(sb);
6018 if (!really_read_only && journal_inum &&
6019 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6020 es->s_journal_inum = cpu_to_le32(journal_inum);
6021 ext4_commit_super(sb);
6027 jbd2_journal_destroy(journal);
6031 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6032 static void ext4_update_super(struct super_block *sb)
6034 struct ext4_sb_info *sbi = EXT4_SB(sb);
6035 struct ext4_super_block *es = sbi->s_es;
6036 struct buffer_head *sbh = sbi->s_sbh;
6040 * If the file system is mounted read-only, don't update the
6041 * superblock write time. This avoids updating the superblock
6042 * write time when we are mounting the root file system
6043 * read/only but we need to replay the journal; at that point,
6044 * for people who are east of GMT and who make their clock
6045 * tick in localtime for Windows bug-for-bug compatibility,
6046 * the clock is set in the future, and this will cause e2fsck
6047 * to complain and force a full file system check.
6049 if (!(sb->s_flags & SB_RDONLY))
6050 ext4_update_tstamp(es, s_wtime);
6051 es->s_kbytes_written =
6052 cpu_to_le64(sbi->s_kbytes_written +
6053 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6054 sbi->s_sectors_written_start) >> 1));
6055 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6056 ext4_free_blocks_count_set(es,
6057 EXT4_C2B(sbi, percpu_counter_sum_positive(
6058 &sbi->s_freeclusters_counter)));
6059 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6060 es->s_free_inodes_count =
6061 cpu_to_le32(percpu_counter_sum_positive(
6062 &sbi->s_freeinodes_counter));
6063 /* Copy error information to the on-disk superblock */
6064 spin_lock(&sbi->s_error_lock);
6065 if (sbi->s_add_error_count > 0) {
6066 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6067 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6068 __ext4_update_tstamp(&es->s_first_error_time,
6069 &es->s_first_error_time_hi,
6070 sbi->s_first_error_time);
6071 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6072 sizeof(es->s_first_error_func));
6073 es->s_first_error_line =
6074 cpu_to_le32(sbi->s_first_error_line);
6075 es->s_first_error_ino =
6076 cpu_to_le32(sbi->s_first_error_ino);
6077 es->s_first_error_block =
6078 cpu_to_le64(sbi->s_first_error_block);
6079 es->s_first_error_errcode =
6080 ext4_errno_to_code(sbi->s_first_error_code);
6082 __ext4_update_tstamp(&es->s_last_error_time,
6083 &es->s_last_error_time_hi,
6084 sbi->s_last_error_time);
6085 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6086 sizeof(es->s_last_error_func));
6087 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6088 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6089 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6090 es->s_last_error_errcode =
6091 ext4_errno_to_code(sbi->s_last_error_code);
6093 * Start the daily error reporting function if it hasn't been
6096 if (!es->s_error_count)
6097 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6098 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6099 sbi->s_add_error_count = 0;
6101 spin_unlock(&sbi->s_error_lock);
6103 ext4_superblock_csum_set(sb);
6107 static int ext4_commit_super(struct super_block *sb)
6109 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6113 if (block_device_ejected(sb))
6116 ext4_update_super(sb);
6119 /* Buffer got discarded which means block device got invalidated */
6120 if (!buffer_mapped(sbh)) {
6125 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6127 * Oh, dear. A previous attempt to write the
6128 * superblock failed. This could happen because the
6129 * USB device was yanked out. Or it could happen to
6130 * be a transient write error and maybe the block will
6131 * be remapped. Nothing we can do but to retry the
6132 * write and hope for the best.
6134 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6135 "superblock detected");
6136 clear_buffer_write_io_error(sbh);
6137 set_buffer_uptodate(sbh);
6140 /* Clear potential dirty bit if it was journalled update */
6141 clear_buffer_dirty(sbh);
6142 sbh->b_end_io = end_buffer_write_sync;
6143 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6144 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6145 wait_on_buffer(sbh);
6146 if (buffer_write_io_error(sbh)) {
6147 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6149 clear_buffer_write_io_error(sbh);
6150 set_buffer_uptodate(sbh);
6157 * Have we just finished recovery? If so, and if we are mounting (or
6158 * remounting) the filesystem readonly, then we will end up with a
6159 * consistent fs on disk. Record that fact.
6161 static int ext4_mark_recovery_complete(struct super_block *sb,
6162 struct ext4_super_block *es)
6165 journal_t *journal = EXT4_SB(sb)->s_journal;
6167 if (!ext4_has_feature_journal(sb)) {
6168 if (journal != NULL) {
6169 ext4_error(sb, "Journal got removed while the fs was "
6171 return -EFSCORRUPTED;
6175 jbd2_journal_lock_updates(journal);
6176 err = jbd2_journal_flush(journal, 0);
6180 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6181 ext4_has_feature_orphan_present(sb))) {
6182 if (!ext4_orphan_file_empty(sb)) {
6183 ext4_error(sb, "Orphan file not empty on read-only fs.");
6184 err = -EFSCORRUPTED;
6187 ext4_clear_feature_journal_needs_recovery(sb);
6188 ext4_clear_feature_orphan_present(sb);
6189 ext4_commit_super(sb);
6192 jbd2_journal_unlock_updates(journal);
6197 * If we are mounting (or read-write remounting) a filesystem whose journal
6198 * has recorded an error from a previous lifetime, move that error to the
6199 * main filesystem now.
6201 static int ext4_clear_journal_err(struct super_block *sb,
6202 struct ext4_super_block *es)
6208 if (!ext4_has_feature_journal(sb)) {
6209 ext4_error(sb, "Journal got removed while the fs was mounted!");
6210 return -EFSCORRUPTED;
6213 journal = EXT4_SB(sb)->s_journal;
6216 * Now check for any error status which may have been recorded in the
6217 * journal by a prior ext4_error() or ext4_abort()
6220 j_errno = jbd2_journal_errno(journal);
6224 errstr = ext4_decode_error(sb, j_errno, nbuf);
6225 ext4_warning(sb, "Filesystem error recorded "
6226 "from previous mount: %s", errstr);
6228 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6229 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6230 j_errno = ext4_commit_super(sb);
6233 ext4_warning(sb, "Marked fs in need of filesystem check.");
6235 jbd2_journal_clear_err(journal);
6236 jbd2_journal_update_sb_errno(journal);
6242 * Force the running and committing transactions to commit,
6243 * and wait on the commit.
6245 int ext4_force_commit(struct super_block *sb)
6252 journal = EXT4_SB(sb)->s_journal;
6253 return ext4_journal_force_commit(journal);
6256 static int ext4_sync_fs(struct super_block *sb, int wait)
6260 bool needs_barrier = false;
6261 struct ext4_sb_info *sbi = EXT4_SB(sb);
6263 if (unlikely(ext4_forced_shutdown(sbi)))
6266 trace_ext4_sync_fs(sb, wait);
6267 flush_workqueue(sbi->rsv_conversion_wq);
6269 * Writeback quota in non-journalled quota case - journalled quota has
6272 dquot_writeback_dquots(sb, -1);
6274 * Data writeback is possible w/o journal transaction, so barrier must
6275 * being sent at the end of the function. But we can skip it if
6276 * transaction_commit will do it for us.
6278 if (sbi->s_journal) {
6279 target = jbd2_get_latest_transaction(sbi->s_journal);
6280 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6281 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6282 needs_barrier = true;
6284 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6286 ret = jbd2_log_wait_commit(sbi->s_journal,
6289 } else if (wait && test_opt(sb, BARRIER))
6290 needs_barrier = true;
6291 if (needs_barrier) {
6293 err = blkdev_issue_flush(sb->s_bdev);
6302 * LVM calls this function before a (read-only) snapshot is created. This
6303 * gives us a chance to flush the journal completely and mark the fs clean.
6305 * Note that only this function cannot bring a filesystem to be in a clean
6306 * state independently. It relies on upper layer to stop all data & metadata
6309 static int ext4_freeze(struct super_block *sb)
6317 journal = EXT4_SB(sb)->s_journal;
6320 /* Now we set up the journal barrier. */
6321 jbd2_journal_lock_updates(journal);
6324 * Don't clear the needs_recovery flag if we failed to
6325 * flush the journal.
6327 error = jbd2_journal_flush(journal, 0);
6331 /* Journal blocked and flushed, clear needs_recovery flag. */
6332 ext4_clear_feature_journal_needs_recovery(sb);
6333 if (ext4_orphan_file_empty(sb))
6334 ext4_clear_feature_orphan_present(sb);
6337 error = ext4_commit_super(sb);
6340 /* we rely on upper layer to stop further updates */
6341 jbd2_journal_unlock_updates(journal);
6346 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6347 * flag here, even though the filesystem is not technically dirty yet.
6349 static int ext4_unfreeze(struct super_block *sb)
6351 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6354 if (EXT4_SB(sb)->s_journal) {
6355 /* Reset the needs_recovery flag before the fs is unlocked. */
6356 ext4_set_feature_journal_needs_recovery(sb);
6357 if (ext4_has_feature_orphan_file(sb))
6358 ext4_set_feature_orphan_present(sb);
6361 ext4_commit_super(sb);
6366 * Structure to save mount options for ext4_remount's benefit
6368 struct ext4_mount_options {
6369 unsigned long s_mount_opt;
6370 unsigned long s_mount_opt2;
6373 unsigned long s_commit_interval;
6374 u32 s_min_batch_time, s_max_batch_time;
6377 char *s_qf_names[EXT4_MAXQUOTAS];
6381 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6383 struct ext4_fs_context *ctx = fc->fs_private;
6384 struct ext4_super_block *es;
6385 struct ext4_sb_info *sbi = EXT4_SB(sb);
6386 unsigned long old_sb_flags;
6387 struct ext4_mount_options old_opts;
6392 int enable_quota = 0;
6394 char *to_free[EXT4_MAXQUOTAS];
6398 /* Store the original options */
6399 old_sb_flags = sb->s_flags;
6400 old_opts.s_mount_opt = sbi->s_mount_opt;
6401 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6402 old_opts.s_resuid = sbi->s_resuid;
6403 old_opts.s_resgid = sbi->s_resgid;
6404 old_opts.s_commit_interval = sbi->s_commit_interval;
6405 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6406 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6408 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6409 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6410 if (sbi->s_qf_names[i]) {
6411 char *qf_name = get_qf_name(sb, sbi, i);
6413 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6414 if (!old_opts.s_qf_names[i]) {
6415 for (j = 0; j < i; j++)
6416 kfree(old_opts.s_qf_names[j]);
6420 old_opts.s_qf_names[i] = NULL;
6422 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6423 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6424 ctx->journal_ioprio =
6425 sbi->s_journal->j_task->io_context->ioprio;
6427 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6431 ext4_apply_options(fc, sb);
6433 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6434 test_opt(sb, JOURNAL_CHECKSUM)) {
6435 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6436 "during remount not supported; ignoring");
6437 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6440 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6441 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6442 ext4_msg(sb, KERN_ERR, "can't mount with "
6443 "both data=journal and delalloc");
6447 if (test_opt(sb, DIOREAD_NOLOCK)) {
6448 ext4_msg(sb, KERN_ERR, "can't mount with "
6449 "both data=journal and dioread_nolock");
6453 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6454 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6455 ext4_msg(sb, KERN_ERR, "can't mount with "
6456 "journal_async_commit in data=ordered mode");
6462 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6463 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6468 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6469 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6471 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6472 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6476 if (sbi->s_journal) {
6477 ext4_init_journal_params(sb, sbi->s_journal);
6478 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6481 /* Flush outstanding errors before changing fs state */
6482 flush_work(&sbi->s_error_work);
6484 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6485 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6490 if (fc->sb_flags & SB_RDONLY) {
6491 err = sync_filesystem(sb);
6494 err = dquot_suspend(sb, -1);
6499 * First of all, the unconditional stuff we have to do
6500 * to disable replay of the journal when we next remount
6502 sb->s_flags |= SB_RDONLY;
6505 * OK, test if we are remounting a valid rw partition
6506 * readonly, and if so set the rdonly flag and then
6507 * mark the partition as valid again.
6509 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6510 (sbi->s_mount_state & EXT4_VALID_FS))
6511 es->s_state = cpu_to_le16(sbi->s_mount_state);
6513 if (sbi->s_journal) {
6515 * We let remount-ro finish even if marking fs
6516 * as clean failed...
6518 ext4_mark_recovery_complete(sb, es);
6521 /* Make sure we can mount this feature set readwrite */
6522 if (ext4_has_feature_readonly(sb) ||
6523 !ext4_feature_set_ok(sb, 0)) {
6528 * Make sure the group descriptor checksums
6529 * are sane. If they aren't, refuse to remount r/w.
6531 for (g = 0; g < sbi->s_groups_count; g++) {
6532 struct ext4_group_desc *gdp =
6533 ext4_get_group_desc(sb, g, NULL);
6535 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6536 ext4_msg(sb, KERN_ERR,
6537 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6538 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6539 le16_to_cpu(gdp->bg_checksum));
6546 * If we have an unprocessed orphan list hanging
6547 * around from a previously readonly bdev mount,
6548 * require a full umount/remount for now.
6550 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6551 ext4_msg(sb, KERN_WARNING, "Couldn't "
6552 "remount RDWR because of unprocessed "
6553 "orphan inode list. Please "
6554 "umount/remount instead");
6560 * Mounting a RDONLY partition read-write, so reread
6561 * and store the current valid flag. (It may have
6562 * been changed by e2fsck since we originally mounted
6565 if (sbi->s_journal) {
6566 err = ext4_clear_journal_err(sb, es);
6570 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6573 err = ext4_setup_super(sb, es, 0);
6578 if (ext4_has_feature_mmp(sb)) {
6579 err = ext4_multi_mount_protect(sb,
6580 le64_to_cpu(es->s_mmp_block));
6591 * Reinitialize lazy itable initialization thread based on
6594 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6595 ext4_unregister_li_request(sb);
6597 ext4_group_t first_not_zeroed;
6598 first_not_zeroed = ext4_has_uninit_itable(sb);
6599 ext4_register_li_request(sb, first_not_zeroed);
6603 * Handle creation of system zone data early because it can fail.
6604 * Releasing of existing data is done when we are sure remount will
6607 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6608 err = ext4_setup_system_zone(sb);
6613 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6614 err = ext4_commit_super(sb);
6621 if (sb_any_quota_suspended(sb))
6622 dquot_resume(sb, -1);
6623 else if (ext4_has_feature_quota(sb)) {
6624 err = ext4_enable_quotas(sb);
6629 /* Release old quota file names */
6630 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6631 kfree(old_opts.s_qf_names[i]);
6633 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6634 ext4_release_system_zone(sb);
6637 sb->s_flags &= ~SB_RDONLY;
6639 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6640 ext4_stop_mmpd(sbi);
6646 * If there was a failing r/w to ro transition, we may need to
6649 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6650 sb_any_quota_suspended(sb))
6651 dquot_resume(sb, -1);
6652 sb->s_flags = old_sb_flags;
6653 sbi->s_mount_opt = old_opts.s_mount_opt;
6654 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6655 sbi->s_resuid = old_opts.s_resuid;
6656 sbi->s_resgid = old_opts.s_resgid;
6657 sbi->s_commit_interval = old_opts.s_commit_interval;
6658 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6659 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6660 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6661 ext4_release_system_zone(sb);
6663 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6664 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6665 to_free[i] = get_qf_name(sb, sbi, i);
6666 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6669 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6672 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6673 ext4_stop_mmpd(sbi);
6677 static int ext4_reconfigure(struct fs_context *fc)
6679 struct super_block *sb = fc->root->d_sb;
6682 fc->s_fs_info = EXT4_SB(sb);
6684 ret = ext4_check_opt_consistency(fc, sb);
6688 ret = __ext4_remount(fc, sb);
6692 ext4_msg(sb, KERN_INFO, "re-mounted %pU. Quota mode: %s.",
6693 &sb->s_uuid, ext4_quota_mode(sb));
6699 static int ext4_statfs_project(struct super_block *sb,
6700 kprojid_t projid, struct kstatfs *buf)
6703 struct dquot *dquot;
6707 qid = make_kqid_projid(projid);
6708 dquot = dqget(sb, qid);
6710 return PTR_ERR(dquot);
6711 spin_lock(&dquot->dq_dqb_lock);
6713 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6714 dquot->dq_dqb.dqb_bhardlimit);
6715 limit >>= sb->s_blocksize_bits;
6717 if (limit && buf->f_blocks > limit) {
6718 curblock = (dquot->dq_dqb.dqb_curspace +
6719 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6720 buf->f_blocks = limit;
6721 buf->f_bfree = buf->f_bavail =
6722 (buf->f_blocks > curblock) ?
6723 (buf->f_blocks - curblock) : 0;
6726 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6727 dquot->dq_dqb.dqb_ihardlimit);
6728 if (limit && buf->f_files > limit) {
6729 buf->f_files = limit;
6731 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6732 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6735 spin_unlock(&dquot->dq_dqb_lock);
6741 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6743 struct super_block *sb = dentry->d_sb;
6744 struct ext4_sb_info *sbi = EXT4_SB(sb);
6745 struct ext4_super_block *es = sbi->s_es;
6746 ext4_fsblk_t overhead = 0, resv_blocks;
6748 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6750 if (!test_opt(sb, MINIX_DF))
6751 overhead = sbi->s_overhead;
6753 buf->f_type = EXT4_SUPER_MAGIC;
6754 buf->f_bsize = sb->s_blocksize;
6755 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6756 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6757 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6758 /* prevent underflow in case that few free space is available */
6759 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6760 buf->f_bavail = buf->f_bfree -
6761 (ext4_r_blocks_count(es) + resv_blocks);
6762 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6764 buf->f_files = le32_to_cpu(es->s_inodes_count);
6765 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6766 buf->f_namelen = EXT4_NAME_LEN;
6767 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6770 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6771 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6772 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6781 * Helper functions so that transaction is started before we acquire dqio_sem
6782 * to keep correct lock ordering of transaction > dqio_sem
6784 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6786 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6789 static int ext4_write_dquot(struct dquot *dquot)
6793 struct inode *inode;
6795 inode = dquot_to_inode(dquot);
6796 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6797 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6799 return PTR_ERR(handle);
6800 ret = dquot_commit(dquot);
6801 err = ext4_journal_stop(handle);
6807 static int ext4_acquire_dquot(struct dquot *dquot)
6812 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6813 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6815 return PTR_ERR(handle);
6816 ret = dquot_acquire(dquot);
6817 err = ext4_journal_stop(handle);
6823 static int ext4_release_dquot(struct dquot *dquot)
6828 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6829 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6830 if (IS_ERR(handle)) {
6831 /* Release dquot anyway to avoid endless cycle in dqput() */
6832 dquot_release(dquot);
6833 return PTR_ERR(handle);
6835 ret = dquot_release(dquot);
6836 err = ext4_journal_stop(handle);
6842 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6844 struct super_block *sb = dquot->dq_sb;
6846 if (ext4_is_quota_journalled(sb)) {
6847 dquot_mark_dquot_dirty(dquot);
6848 return ext4_write_dquot(dquot);
6850 return dquot_mark_dquot_dirty(dquot);
6854 static int ext4_write_info(struct super_block *sb, int type)
6859 /* Data block + inode block */
6860 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6862 return PTR_ERR(handle);
6863 ret = dquot_commit_info(sb, type);
6864 err = ext4_journal_stop(handle);
6870 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6872 struct ext4_inode_info *ei = EXT4_I(inode);
6874 /* The first argument of lockdep_set_subclass has to be
6875 * *exactly* the same as the argument to init_rwsem() --- in
6876 * this case, in init_once() --- or lockdep gets unhappy
6877 * because the name of the lock is set using the
6878 * stringification of the argument to init_rwsem().
6880 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6881 lockdep_set_subclass(&ei->i_data_sem, subclass);
6885 * Standard function to be called on quota_on
6887 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6888 const struct path *path)
6892 if (!test_opt(sb, QUOTA))
6895 /* Quotafile not on the same filesystem? */
6896 if (path->dentry->d_sb != sb)
6899 /* Quota already enabled for this file? */
6900 if (IS_NOQUOTA(d_inode(path->dentry)))
6903 /* Journaling quota? */
6904 if (EXT4_SB(sb)->s_qf_names[type]) {
6905 /* Quotafile not in fs root? */
6906 if (path->dentry->d_parent != sb->s_root)
6907 ext4_msg(sb, KERN_WARNING,
6908 "Quota file not on filesystem root. "
6909 "Journaled quota will not work");
6910 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6913 * Clear the flag just in case mount options changed since
6916 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6919 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6920 err = dquot_quota_on(sb, type, format_id, path);
6922 struct inode *inode = d_inode(path->dentry);
6926 * Set inode flags to prevent userspace from messing with quota
6927 * files. If this fails, we return success anyway since quotas
6928 * are already enabled and this is not a hard failure.
6931 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6934 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6935 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6936 S_NOATIME | S_IMMUTABLE);
6937 err = ext4_mark_inode_dirty(handle, inode);
6938 ext4_journal_stop(handle);
6940 inode_unlock(inode);
6942 dquot_quota_off(sb, type);
6945 lockdep_set_quota_inode(path->dentry->d_inode,
6950 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6954 return qf_inum == EXT4_USR_QUOTA_INO;
6956 return qf_inum == EXT4_GRP_QUOTA_INO;
6958 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6964 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6968 struct inode *qf_inode;
6969 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6970 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6971 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6972 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6975 BUG_ON(!ext4_has_feature_quota(sb));
6977 if (!qf_inums[type])
6980 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6981 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6982 qf_inums[type], type);
6986 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6987 if (IS_ERR(qf_inode)) {
6988 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6989 qf_inums[type], type);
6990 return PTR_ERR(qf_inode);
6993 /* Don't account quota for quota files to avoid recursion */
6994 qf_inode->i_flags |= S_NOQUOTA;
6995 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6996 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6998 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7004 /* Enable usage tracking for all quota types. */
7005 int ext4_enable_quotas(struct super_block *sb)
7008 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7009 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7010 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7011 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7013 bool quota_mopt[EXT4_MAXQUOTAS] = {
7014 test_opt(sb, USRQUOTA),
7015 test_opt(sb, GRPQUOTA),
7016 test_opt(sb, PRJQUOTA),
7019 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7020 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7021 if (qf_inums[type]) {
7022 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7023 DQUOT_USAGE_ENABLED |
7024 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7027 "Failed to enable quota tracking "
7028 "(type=%d, err=%d, ino=%lu). "
7029 "Please run e2fsck to fix.", type,
7030 err, qf_inums[type]);
7031 for (type--; type >= 0; type--) {
7032 struct inode *inode;
7034 inode = sb_dqopt(sb)->files[type];
7036 inode = igrab(inode);
7037 dquot_quota_off(sb, type);
7039 lockdep_set_quota_inode(inode,
7052 static int ext4_quota_off(struct super_block *sb, int type)
7054 struct inode *inode = sb_dqopt(sb)->files[type];
7058 /* Force all delayed allocation blocks to be allocated.
7059 * Caller already holds s_umount sem */
7060 if (test_opt(sb, DELALLOC))
7061 sync_filesystem(sb);
7063 if (!inode || !igrab(inode))
7066 err = dquot_quota_off(sb, type);
7067 if (err || ext4_has_feature_quota(sb))
7072 * Update modification times of quota files when userspace can
7073 * start looking at them. If we fail, we return success anyway since
7074 * this is not a hard failure and quotas are already disabled.
7076 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7077 if (IS_ERR(handle)) {
7078 err = PTR_ERR(handle);
7081 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7082 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7083 inode->i_mtime = inode->i_ctime = current_time(inode);
7084 err = ext4_mark_inode_dirty(handle, inode);
7085 ext4_journal_stop(handle);
7087 inode_unlock(inode);
7089 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7093 return dquot_quota_off(sb, type);
7096 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7097 * acquiring the locks... As quota files are never truncated and quota code
7098 * itself serializes the operations (and no one else should touch the files)
7099 * we don't have to be afraid of races */
7100 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7101 size_t len, loff_t off)
7103 struct inode *inode = sb_dqopt(sb)->files[type];
7104 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7105 int offset = off & (sb->s_blocksize - 1);
7108 struct buffer_head *bh;
7109 loff_t i_size = i_size_read(inode);
7113 if (off+len > i_size)
7116 while (toread > 0) {
7117 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7118 bh = ext4_bread(NULL, inode, blk, 0);
7121 if (!bh) /* A hole? */
7122 memset(data, 0, tocopy);
7124 memcpy(data, bh->b_data+offset, tocopy);
7134 /* Write to quotafile (we know the transaction is already started and has
7135 * enough credits) */
7136 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7137 const char *data, size_t len, loff_t off)
7139 struct inode *inode = sb_dqopt(sb)->files[type];
7140 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7141 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7143 struct buffer_head *bh;
7144 handle_t *handle = journal_current_handle();
7147 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7148 " cancelled because transaction is not started",
7149 (unsigned long long)off, (unsigned long long)len);
7153 * Since we account only one data block in transaction credits,
7154 * then it is impossible to cross a block boundary.
7156 if (sb->s_blocksize - offset < len) {
7157 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7158 " cancelled because not block aligned",
7159 (unsigned long long)off, (unsigned long long)len);
7164 bh = ext4_bread(handle, inode, blk,
7165 EXT4_GET_BLOCKS_CREATE |
7166 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7167 } while (PTR_ERR(bh) == -ENOSPC &&
7168 ext4_should_retry_alloc(inode->i_sb, &retries));
7173 BUFFER_TRACE(bh, "get write access");
7174 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7180 memcpy(bh->b_data+offset, data, len);
7181 flush_dcache_page(bh->b_page);
7183 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7186 if (inode->i_size < off + len) {
7187 i_size_write(inode, off + len);
7188 EXT4_I(inode)->i_disksize = inode->i_size;
7189 err2 = ext4_mark_inode_dirty(handle, inode);
7190 if (unlikely(err2 && !err))
7193 return err ? err : len;
7197 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7198 static inline void register_as_ext2(void)
7200 int err = register_filesystem(&ext2_fs_type);
7203 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7206 static inline void unregister_as_ext2(void)
7208 unregister_filesystem(&ext2_fs_type);
7211 static inline int ext2_feature_set_ok(struct super_block *sb)
7213 if (ext4_has_unknown_ext2_incompat_features(sb))
7217 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7222 static inline void register_as_ext2(void) { }
7223 static inline void unregister_as_ext2(void) { }
7224 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7227 static inline void register_as_ext3(void)
7229 int err = register_filesystem(&ext3_fs_type);
7232 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7235 static inline void unregister_as_ext3(void)
7237 unregister_filesystem(&ext3_fs_type);
7240 static inline int ext3_feature_set_ok(struct super_block *sb)
7242 if (ext4_has_unknown_ext3_incompat_features(sb))
7244 if (!ext4_has_feature_journal(sb))
7248 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7253 static struct file_system_type ext4_fs_type = {
7254 .owner = THIS_MODULE,
7256 .init_fs_context = ext4_init_fs_context,
7257 .parameters = ext4_param_specs,
7258 .kill_sb = kill_block_super,
7259 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7261 MODULE_ALIAS_FS("ext4");
7263 /* Shared across all ext4 file systems */
7264 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7266 static int __init ext4_init_fs(void)
7270 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7271 ext4_li_info = NULL;
7273 /* Build-time check for flags consistency */
7274 ext4_check_flag_values();
7276 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7277 init_waitqueue_head(&ext4__ioend_wq[i]);
7279 err = ext4_init_es();
7283 err = ext4_init_pending();
7287 err = ext4_init_post_read_processing();
7291 err = ext4_init_pageio();
7295 err = ext4_init_system_zone();
7299 err = ext4_init_sysfs();
7303 err = ext4_init_mballoc();
7306 err = init_inodecache();
7310 err = ext4_fc_init_dentry_cache();
7316 err = register_filesystem(&ext4_fs_type);
7322 unregister_as_ext2();
7323 unregister_as_ext3();
7324 ext4_fc_destroy_dentry_cache();
7326 destroy_inodecache();
7328 ext4_exit_mballoc();
7332 ext4_exit_system_zone();
7336 ext4_exit_post_read_processing();
7338 ext4_exit_pending();
7345 static void __exit ext4_exit_fs(void)
7347 ext4_destroy_lazyinit_thread();
7348 unregister_as_ext2();
7349 unregister_as_ext3();
7350 unregister_filesystem(&ext4_fs_type);
7351 ext4_fc_destroy_dentry_cache();
7352 destroy_inodecache();
7353 ext4_exit_mballoc();
7355 ext4_exit_system_zone();
7357 ext4_exit_post_read_processing();
7359 ext4_exit_pending();
7362 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7363 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7364 MODULE_LICENSE("GPL");
7365 MODULE_SOFTDEP("pre: crc32c");
7366 module_init(ext4_init_fs)
7367 module_exit(ext4_exit_fs)