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/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
49 #include <linux/fsnotify.h>
50 #include <linux/fs_context.h>
51 #include <linux/fs_parser.h>
54 #include "ext4_extents.h" /* Needed for trace points definition */
55 #include "ext4_jbd2.h"
61 #define CREATE_TRACE_POINTS
62 #include <trace/events/ext4.h>
64 static struct ext4_lazy_init *ext4_li_info;
65 static DEFINE_MUTEX(ext4_li_mtx);
66 static struct ratelimit_state ext4_mount_msg_ratelimit;
68 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
69 unsigned long journal_devnum);
70 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
71 static void ext4_update_super(struct super_block *sb);
72 static int ext4_commit_super(struct super_block *sb);
73 static int ext4_mark_recovery_complete(struct super_block *sb,
74 struct ext4_super_block *es);
75 static int ext4_clear_journal_err(struct super_block *sb,
76 struct ext4_super_block *es);
77 static int ext4_sync_fs(struct super_block *sb, int wait);
78 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
79 static int ext4_unfreeze(struct super_block *sb);
80 static int ext4_freeze(struct super_block *sb);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
88 static int ext4_validate_options(struct fs_context *fc);
89 static int ext4_check_opt_consistency(struct fs_context *fc,
90 struct super_block *sb);
91 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb);
92 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
93 static int ext4_get_tree(struct fs_context *fc);
94 static int ext4_reconfigure(struct fs_context *fc);
95 static void ext4_fc_free(struct fs_context *fc);
96 static int ext4_init_fs_context(struct fs_context *fc);
97 static const struct fs_parameter_spec ext4_param_specs[];
103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
104 * -> page lock -> i_data_sem (rw)
106 * buffered write path:
107 * sb_start_write -> i_mutex -> mmap_lock
108 * sb_start_write -> i_mutex -> transaction start -> page lock ->
112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
118 * sb_start_write -> i_mutex -> mmap_lock
119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
122 * transaction start -> page lock(s) -> i_data_sem (rw)
125 static const struct fs_context_operations ext4_context_ops = {
126 .parse_param = ext4_parse_param,
127 .get_tree = ext4_get_tree,
128 .reconfigure = ext4_reconfigure,
129 .free = ext4_fc_free,
133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
134 static struct file_system_type ext2_fs_type = {
135 .owner = THIS_MODULE,
137 .init_fs_context = ext4_init_fs_context,
138 .parameters = ext4_param_specs,
139 .kill_sb = kill_block_super,
140 .fs_flags = FS_REQUIRES_DEV,
142 MODULE_ALIAS_FS("ext2");
143 MODULE_ALIAS("ext2");
144 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
146 #define IS_EXT2_SB(sb) (0)
150 static struct file_system_type ext3_fs_type = {
151 .owner = THIS_MODULE,
153 .init_fs_context = ext4_init_fs_context,
154 .parameters = ext4_param_specs,
155 .kill_sb = kill_block_super,
156 .fs_flags = FS_REQUIRES_DEV,
158 MODULE_ALIAS_FS("ext3");
159 MODULE_ALIAS("ext3");
160 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
163 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
167 * buffer's verified bit is no longer valid after reading from
168 * disk again due to write out error, clear it to make sure we
169 * recheck the buffer contents.
171 clear_buffer_verified(bh);
173 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
175 submit_bh(REQ_OP_READ, op_flags, bh);
178 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
181 BUG_ON(!buffer_locked(bh));
183 if (ext4_buffer_uptodate(bh)) {
187 __ext4_read_bh(bh, op_flags, end_io);
190 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
192 BUG_ON(!buffer_locked(bh));
194 if (ext4_buffer_uptodate(bh)) {
199 __ext4_read_bh(bh, op_flags, end_io);
202 if (buffer_uptodate(bh))
207 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
209 if (trylock_buffer(bh)) {
211 return ext4_read_bh(bh, op_flags, NULL);
212 ext4_read_bh_nowait(bh, op_flags, NULL);
217 if (buffer_uptodate(bh))
225 * This works like __bread_gfp() except it uses ERR_PTR for error
226 * returns. Currently with sb_bread it's impossible to distinguish
227 * between ENOMEM and EIO situations (since both result in a NULL
230 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
231 sector_t block, int op_flags,
234 struct buffer_head *bh;
237 bh = sb_getblk_gfp(sb, block, gfp);
239 return ERR_PTR(-ENOMEM);
240 if (ext4_buffer_uptodate(bh))
243 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
251 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
254 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
257 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
260 return __ext4_sb_bread_gfp(sb, block, 0, 0);
263 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
265 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
268 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
273 static int ext4_verify_csum_type(struct super_block *sb,
274 struct ext4_super_block *es)
276 if (!ext4_has_feature_metadata_csum(sb))
279 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
282 static __le32 ext4_superblock_csum(struct super_block *sb,
283 struct ext4_super_block *es)
285 struct ext4_sb_info *sbi = EXT4_SB(sb);
286 int offset = offsetof(struct ext4_super_block, s_checksum);
289 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
291 return cpu_to_le32(csum);
294 static int ext4_superblock_csum_verify(struct super_block *sb,
295 struct ext4_super_block *es)
297 if (!ext4_has_metadata_csum(sb))
300 return es->s_checksum == ext4_superblock_csum(sb, es);
303 void ext4_superblock_csum_set(struct super_block *sb)
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307 if (!ext4_has_metadata_csum(sb))
310 es->s_checksum = ext4_superblock_csum(sb, es);
313 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
314 struct ext4_group_desc *bg)
316 return le32_to_cpu(bg->bg_block_bitmap_lo) |
317 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
318 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
321 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
322 struct ext4_group_desc *bg)
324 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
325 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
326 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
329 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
330 struct ext4_group_desc *bg)
332 return le32_to_cpu(bg->bg_inode_table_lo) |
333 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
334 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
337 __u32 ext4_free_group_clusters(struct super_block *sb,
338 struct ext4_group_desc *bg)
340 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
341 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
342 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
345 __u32 ext4_free_inodes_count(struct super_block *sb,
346 struct ext4_group_desc *bg)
348 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
349 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
350 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
353 __u32 ext4_used_dirs_count(struct super_block *sb,
354 struct ext4_group_desc *bg)
356 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
357 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
358 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
361 __u32 ext4_itable_unused_count(struct super_block *sb,
362 struct ext4_group_desc *bg)
364 return le16_to_cpu(bg->bg_itable_unused_lo) |
365 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
366 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
369 void ext4_block_bitmap_set(struct super_block *sb,
370 struct ext4_group_desc *bg, ext4_fsblk_t blk)
372 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
373 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
374 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
377 void ext4_inode_bitmap_set(struct super_block *sb,
378 struct ext4_group_desc *bg, ext4_fsblk_t blk)
380 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
381 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
382 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
385 void ext4_inode_table_set(struct super_block *sb,
386 struct ext4_group_desc *bg, ext4_fsblk_t blk)
388 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
389 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
390 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
393 void ext4_free_group_clusters_set(struct super_block *sb,
394 struct ext4_group_desc *bg, __u32 count)
396 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
397 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
398 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
401 void ext4_free_inodes_set(struct super_block *sb,
402 struct ext4_group_desc *bg, __u32 count)
404 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
405 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
406 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
409 void ext4_used_dirs_set(struct super_block *sb,
410 struct ext4_group_desc *bg, __u32 count)
412 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
413 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
414 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
417 void ext4_itable_unused_set(struct super_block *sb,
418 struct ext4_group_desc *bg, __u32 count)
420 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
421 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
422 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
425 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
427 now = clamp_val(now, 0, (1ull << 40) - 1);
429 *lo = cpu_to_le32(lower_32_bits(now));
430 *hi = upper_32_bits(now);
433 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
435 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
437 #define ext4_update_tstamp(es, tstamp) \
438 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
439 ktime_get_real_seconds())
440 #define ext4_get_tstamp(es, tstamp) \
441 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
444 * The del_gendisk() function uninitializes the disk-specific data
445 * structures, including the bdi structure, without telling anyone
446 * else. Once this happens, any attempt to call mark_buffer_dirty()
447 * (for example, by ext4_commit_super), will cause a kernel OOPS.
448 * This is a kludge to prevent these oops until we can put in a proper
449 * hook in del_gendisk() to inform the VFS and file system layers.
451 static int block_device_ejected(struct super_block *sb)
453 struct inode *bd_inode = sb->s_bdev->bd_inode;
454 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
456 return bdi->dev == NULL;
459 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
461 struct super_block *sb = journal->j_private;
462 struct ext4_sb_info *sbi = EXT4_SB(sb);
463 int error = is_journal_aborted(journal);
464 struct ext4_journal_cb_entry *jce;
466 BUG_ON(txn->t_state == T_FINISHED);
468 ext4_process_freed_data(sb, txn->t_tid);
470 spin_lock(&sbi->s_md_lock);
471 while (!list_empty(&txn->t_private_list)) {
472 jce = list_entry(txn->t_private_list.next,
473 struct ext4_journal_cb_entry, jce_list);
474 list_del_init(&jce->jce_list);
475 spin_unlock(&sbi->s_md_lock);
476 jce->jce_func(sb, jce, error);
477 spin_lock(&sbi->s_md_lock);
479 spin_unlock(&sbi->s_md_lock);
483 * This writepage callback for write_cache_pages()
484 * takes care of a few cases after page cleaning.
486 * write_cache_pages() already checks for dirty pages
487 * and calls clear_page_dirty_for_io(), which we want,
488 * to write protect the pages.
490 * However, we may have to redirty a page (see below.)
492 static int ext4_journalled_writepage_callback(struct page *page,
493 struct writeback_control *wbc,
496 transaction_t *transaction = (transaction_t *) data;
497 struct buffer_head *bh, *head;
498 struct journal_head *jh;
500 bh = head = page_buffers(page);
503 * We have to redirty a page in these cases:
504 * 1) If buffer is dirty, it means the page was dirty because it
505 * contains a buffer that needs checkpointing. So the dirty bit
506 * needs to be preserved so that checkpointing writes the buffer
508 * 2) If buffer is not part of the committing transaction
509 * (we may have just accidentally come across this buffer because
510 * inode range tracking is not exact) or if the currently running
511 * transaction already contains this buffer as well, dirty bit
512 * needs to be preserved so that the buffer gets writeprotected
513 * properly on running transaction's commit.
516 if (buffer_dirty(bh) ||
517 (jh && (jh->b_transaction != transaction ||
518 jh->b_next_transaction))) {
519 redirty_page_for_writepage(wbc, page);
522 } while ((bh = bh->b_this_page) != head);
525 return AOP_WRITEPAGE_ACTIVATE;
528 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
530 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
531 struct writeback_control wbc = {
532 .sync_mode = WB_SYNC_ALL,
533 .nr_to_write = LONG_MAX,
534 .range_start = jinode->i_dirty_start,
535 .range_end = jinode->i_dirty_end,
538 return write_cache_pages(mapping, &wbc,
539 ext4_journalled_writepage_callback,
540 jinode->i_transaction);
543 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
547 if (ext4_should_journal_data(jinode->i_vfs_inode))
548 ret = ext4_journalled_submit_inode_data_buffers(jinode);
550 ret = jbd2_journal_submit_inode_data_buffers(jinode);
555 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
559 if (!ext4_should_journal_data(jinode->i_vfs_inode))
560 ret = jbd2_journal_finish_inode_data_buffers(jinode);
565 static bool system_going_down(void)
567 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
568 || system_state == SYSTEM_RESTART;
571 struct ext4_err_translation {
576 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
578 static struct ext4_err_translation err_translation[] = {
579 EXT4_ERR_TRANSLATE(EIO),
580 EXT4_ERR_TRANSLATE(ENOMEM),
581 EXT4_ERR_TRANSLATE(EFSBADCRC),
582 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
583 EXT4_ERR_TRANSLATE(ENOSPC),
584 EXT4_ERR_TRANSLATE(ENOKEY),
585 EXT4_ERR_TRANSLATE(EROFS),
586 EXT4_ERR_TRANSLATE(EFBIG),
587 EXT4_ERR_TRANSLATE(EEXIST),
588 EXT4_ERR_TRANSLATE(ERANGE),
589 EXT4_ERR_TRANSLATE(EOVERFLOW),
590 EXT4_ERR_TRANSLATE(EBUSY),
591 EXT4_ERR_TRANSLATE(ENOTDIR),
592 EXT4_ERR_TRANSLATE(ENOTEMPTY),
593 EXT4_ERR_TRANSLATE(ESHUTDOWN),
594 EXT4_ERR_TRANSLATE(EFAULT),
597 static int ext4_errno_to_code(int errno)
601 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
602 if (err_translation[i].errno == errno)
603 return err_translation[i].code;
604 return EXT4_ERR_UNKNOWN;
607 static void save_error_info(struct super_block *sb, int error,
608 __u32 ino, __u64 block,
609 const char *func, unsigned int line)
611 struct ext4_sb_info *sbi = EXT4_SB(sb);
613 /* We default to EFSCORRUPTED error... */
615 error = EFSCORRUPTED;
617 spin_lock(&sbi->s_error_lock);
618 sbi->s_add_error_count++;
619 sbi->s_last_error_code = error;
620 sbi->s_last_error_line = line;
621 sbi->s_last_error_ino = ino;
622 sbi->s_last_error_block = block;
623 sbi->s_last_error_func = func;
624 sbi->s_last_error_time = ktime_get_real_seconds();
625 if (!sbi->s_first_error_time) {
626 sbi->s_first_error_code = error;
627 sbi->s_first_error_line = line;
628 sbi->s_first_error_ino = ino;
629 sbi->s_first_error_block = block;
630 sbi->s_first_error_func = func;
631 sbi->s_first_error_time = sbi->s_last_error_time;
633 spin_unlock(&sbi->s_error_lock);
636 /* Deal with the reporting of failure conditions on a filesystem such as
637 * inconsistencies detected or read IO failures.
639 * On ext2, we can store the error state of the filesystem in the
640 * superblock. That is not possible on ext4, because we may have other
641 * write ordering constraints on the superblock which prevent us from
642 * writing it out straight away; and given that the journal is about to
643 * be aborted, we can't rely on the current, or future, transactions to
644 * write out the superblock safely.
646 * We'll just use the jbd2_journal_abort() error code to record an error in
647 * the journal instead. On recovery, the journal will complain about
648 * that error until we've noted it down and cleared it.
650 * If force_ro is set, we unconditionally force the filesystem into an
651 * ABORT|READONLY state, unless the error response on the fs has been set to
652 * panic in which case we take the easy way out and panic immediately. This is
653 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
654 * at a critical moment in log management.
656 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
657 __u32 ino, __u64 block,
658 const char *func, unsigned int line)
660 journal_t *journal = EXT4_SB(sb)->s_journal;
661 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
663 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
664 if (test_opt(sb, WARN_ON_ERROR))
667 if (!continue_fs && !sb_rdonly(sb)) {
668 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
670 jbd2_journal_abort(journal, -EIO);
673 if (!bdev_read_only(sb->s_bdev)) {
674 save_error_info(sb, error, ino, block, func, line);
676 * In case the fs should keep running, we need to writeout
677 * superblock through the journal. Due to lock ordering
678 * constraints, it may not be safe to do it right here so we
679 * defer superblock flushing to a workqueue.
681 if (continue_fs && journal)
682 schedule_work(&EXT4_SB(sb)->s_error_work);
684 ext4_commit_super(sb);
688 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
689 * could panic during 'reboot -f' as the underlying device got already
692 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
693 panic("EXT4-fs (device %s): panic forced after error\n",
697 if (sb_rdonly(sb) || continue_fs)
700 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
702 * Make sure updated value of ->s_mount_flags will be visible before
706 sb->s_flags |= SB_RDONLY;
709 static void flush_stashed_error_work(struct work_struct *work)
711 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
713 journal_t *journal = sbi->s_journal;
717 * If the journal is still running, we have to write out superblock
718 * through the journal to avoid collisions of other journalled sb
721 * We use directly jbd2 functions here to avoid recursing back into
722 * ext4 error handling code during handling of previous errors.
724 if (!sb_rdonly(sbi->s_sb) && journal) {
725 struct buffer_head *sbh = sbi->s_sbh;
726 handle = jbd2_journal_start(journal, 1);
729 if (jbd2_journal_get_write_access(handle, sbh)) {
730 jbd2_journal_stop(handle);
733 ext4_update_super(sbi->s_sb);
734 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
735 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
736 "superblock detected");
737 clear_buffer_write_io_error(sbh);
738 set_buffer_uptodate(sbh);
741 if (jbd2_journal_dirty_metadata(handle, sbh)) {
742 jbd2_journal_stop(handle);
745 jbd2_journal_stop(handle);
746 ext4_notify_error_sysfs(sbi);
751 * Write through journal failed. Write sb directly to get error info
752 * out and hope for the best.
754 ext4_commit_super(sbi->s_sb);
755 ext4_notify_error_sysfs(sbi);
758 #define ext4_error_ratelimit(sb) \
759 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
762 void __ext4_error(struct super_block *sb, const char *function,
763 unsigned int line, bool force_ro, int error, __u64 block,
764 const char *fmt, ...)
766 struct va_format vaf;
769 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
772 trace_ext4_error(sb, function, line);
773 if (ext4_error_ratelimit(sb)) {
778 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
779 sb->s_id, function, line, current->comm, &vaf);
782 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
784 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
787 void __ext4_error_inode(struct inode *inode, const char *function,
788 unsigned int line, ext4_fsblk_t block, int error,
789 const char *fmt, ...)
792 struct va_format vaf;
794 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
797 trace_ext4_error(inode->i_sb, function, line);
798 if (ext4_error_ratelimit(inode->i_sb)) {
803 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
804 "inode #%lu: block %llu: comm %s: %pV\n",
805 inode->i_sb->s_id, function, line, inode->i_ino,
806 block, current->comm, &vaf);
808 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
809 "inode #%lu: comm %s: %pV\n",
810 inode->i_sb->s_id, function, line, inode->i_ino,
811 current->comm, &vaf);
814 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
816 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
820 void __ext4_error_file(struct file *file, const char *function,
821 unsigned int line, ext4_fsblk_t block,
822 const char *fmt, ...)
825 struct va_format vaf;
826 struct inode *inode = file_inode(file);
827 char pathname[80], *path;
829 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
832 trace_ext4_error(inode->i_sb, function, line);
833 if (ext4_error_ratelimit(inode->i_sb)) {
834 path = file_path(file, pathname, sizeof(pathname));
842 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
843 "block %llu: comm %s: path %s: %pV\n",
844 inode->i_sb->s_id, function, line, inode->i_ino,
845 block, current->comm, path, &vaf);
848 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
849 "comm %s: path %s: %pV\n",
850 inode->i_sb->s_id, function, line, inode->i_ino,
851 current->comm, path, &vaf);
854 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
856 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
860 const char *ext4_decode_error(struct super_block *sb, int errno,
867 errstr = "Corrupt filesystem";
870 errstr = "Filesystem failed CRC";
873 errstr = "IO failure";
876 errstr = "Out of memory";
879 if (!sb || (EXT4_SB(sb)->s_journal &&
880 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
881 errstr = "Journal has aborted";
883 errstr = "Readonly filesystem";
886 /* If the caller passed in an extra buffer for unknown
887 * errors, textualise them now. Else we just return
890 /* Check for truncated error codes... */
891 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
900 /* __ext4_std_error decodes expected errors from journaling functions
901 * automatically and invokes the appropriate error response. */
903 void __ext4_std_error(struct super_block *sb, const char *function,
904 unsigned int line, int errno)
909 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
912 /* Special case: if the error is EROFS, and we're not already
913 * inside a transaction, then there's really no point in logging
915 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
918 if (ext4_error_ratelimit(sb)) {
919 errstr = ext4_decode_error(sb, errno, nbuf);
920 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
921 sb->s_id, function, line, errstr);
923 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
925 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
928 void __ext4_msg(struct super_block *sb,
929 const char *prefix, const char *fmt, ...)
931 struct va_format vaf;
935 atomic_inc(&EXT4_SB(sb)->s_msg_count);
936 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
945 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
947 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
951 static int ext4_warning_ratelimit(struct super_block *sb)
953 atomic_inc(&EXT4_SB(sb)->s_warning_count);
954 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
958 void __ext4_warning(struct super_block *sb, const char *function,
959 unsigned int line, const char *fmt, ...)
961 struct va_format vaf;
964 if (!ext4_warning_ratelimit(sb))
970 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
971 sb->s_id, function, line, &vaf);
975 void __ext4_warning_inode(const struct inode *inode, const char *function,
976 unsigned int line, const char *fmt, ...)
978 struct va_format vaf;
981 if (!ext4_warning_ratelimit(inode->i_sb))
987 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
988 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
989 function, line, inode->i_ino, current->comm, &vaf);
993 void __ext4_grp_locked_error(const char *function, unsigned int line,
994 struct super_block *sb, ext4_group_t grp,
995 unsigned long ino, ext4_fsblk_t block,
996 const char *fmt, ...)
1000 struct va_format vaf;
1003 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
1006 trace_ext4_error(sb, function, line);
1007 if (ext4_error_ratelimit(sb)) {
1008 va_start(args, fmt);
1011 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1012 sb->s_id, function, line, grp);
1014 printk(KERN_CONT "inode %lu: ", ino);
1016 printk(KERN_CONT "block %llu:",
1017 (unsigned long long) block);
1018 printk(KERN_CONT "%pV\n", &vaf);
1022 if (test_opt(sb, ERRORS_CONT)) {
1023 if (test_opt(sb, WARN_ON_ERROR))
1025 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1026 if (!bdev_read_only(sb->s_bdev)) {
1027 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1029 schedule_work(&EXT4_SB(sb)->s_error_work);
1033 ext4_unlock_group(sb, grp);
1034 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1036 * We only get here in the ERRORS_RO case; relocking the group
1037 * may be dangerous, but nothing bad will happen since the
1038 * filesystem will have already been marked read/only and the
1039 * journal has been aborted. We return 1 as a hint to callers
1040 * who might what to use the return value from
1041 * ext4_grp_locked_error() to distinguish between the
1042 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1043 * aggressively from the ext4 function in question, with a
1044 * more appropriate error code.
1046 ext4_lock_group(sb, grp);
1050 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1054 struct ext4_sb_info *sbi = EXT4_SB(sb);
1055 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1056 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1059 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1060 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1063 percpu_counter_sub(&sbi->s_freeclusters_counter,
1067 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1068 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1073 count = ext4_free_inodes_count(sb, gdp);
1074 percpu_counter_sub(&sbi->s_freeinodes_counter,
1080 void ext4_update_dynamic_rev(struct super_block *sb)
1082 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1084 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1088 "updating to rev %d because of new feature flag, "
1089 "running e2fsck is recommended",
1092 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1093 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1094 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1095 /* leave es->s_feature_*compat flags alone */
1096 /* es->s_uuid will be set by e2fsck if empty */
1099 * The rest of the superblock fields should be zero, and if not it
1100 * means they are likely already in use, so leave them alone. We
1101 * can leave it up to e2fsck to clean up any inconsistencies there.
1106 * Open the external journal device
1108 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1110 struct block_device *bdev;
1112 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1118 ext4_msg(sb, KERN_ERR,
1119 "failed to open journal device unknown-block(%u,%u) %ld",
1120 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1125 * Release the journal device
1127 static void ext4_blkdev_put(struct block_device *bdev)
1129 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1132 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1134 struct block_device *bdev;
1135 bdev = sbi->s_journal_bdev;
1137 ext4_blkdev_put(bdev);
1138 sbi->s_journal_bdev = NULL;
1142 static inline struct inode *orphan_list_entry(struct list_head *l)
1144 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1147 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1149 struct list_head *l;
1151 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1152 le32_to_cpu(sbi->s_es->s_last_orphan));
1154 printk(KERN_ERR "sb_info orphan list:\n");
1155 list_for_each(l, &sbi->s_orphan) {
1156 struct inode *inode = orphan_list_entry(l);
1158 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1159 inode->i_sb->s_id, inode->i_ino, inode,
1160 inode->i_mode, inode->i_nlink,
1161 NEXT_ORPHAN(inode));
1166 static int ext4_quota_off(struct super_block *sb, int type);
1168 static inline void ext4_quota_off_umount(struct super_block *sb)
1172 /* Use our quota_off function to clear inode flags etc. */
1173 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1174 ext4_quota_off(sb, type);
1178 * This is a helper function which is used in the mount/remount
1179 * codepaths (which holds s_umount) to fetch the quota file name.
1181 static inline char *get_qf_name(struct super_block *sb,
1182 struct ext4_sb_info *sbi,
1185 return rcu_dereference_protected(sbi->s_qf_names[type],
1186 lockdep_is_held(&sb->s_umount));
1189 static inline void ext4_quota_off_umount(struct super_block *sb)
1194 static void ext4_put_super(struct super_block *sb)
1196 struct ext4_sb_info *sbi = EXT4_SB(sb);
1197 struct ext4_super_block *es = sbi->s_es;
1198 struct buffer_head **group_desc;
1199 struct flex_groups **flex_groups;
1203 ext4_unregister_li_request(sb);
1204 ext4_quota_off_umount(sb);
1206 flush_work(&sbi->s_error_work);
1207 destroy_workqueue(sbi->rsv_conversion_wq);
1208 ext4_release_orphan_info(sb);
1211 * Unregister sysfs before destroying jbd2 journal.
1212 * Since we could still access attr_journal_task attribute via sysfs
1213 * path which could have sbi->s_journal->j_task as NULL
1215 ext4_unregister_sysfs(sb);
1217 if (sbi->s_journal) {
1218 aborted = is_journal_aborted(sbi->s_journal);
1219 err = jbd2_journal_destroy(sbi->s_journal);
1220 sbi->s_journal = NULL;
1221 if ((err < 0) && !aborted) {
1222 ext4_abort(sb, -err, "Couldn't clean up the journal");
1226 ext4_es_unregister_shrinker(sbi);
1227 del_timer_sync(&sbi->s_err_report);
1228 ext4_release_system_zone(sb);
1229 ext4_mb_release(sb);
1230 ext4_ext_release(sb);
1232 if (!sb_rdonly(sb) && !aborted) {
1233 ext4_clear_feature_journal_needs_recovery(sb);
1234 ext4_clear_feature_orphan_present(sb);
1235 es->s_state = cpu_to_le16(sbi->s_mount_state);
1238 ext4_commit_super(sb);
1241 group_desc = rcu_dereference(sbi->s_group_desc);
1242 for (i = 0; i < sbi->s_gdb_count; i++)
1243 brelse(group_desc[i]);
1245 flex_groups = rcu_dereference(sbi->s_flex_groups);
1247 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1248 kvfree(flex_groups[i]);
1249 kvfree(flex_groups);
1252 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1253 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1254 percpu_counter_destroy(&sbi->s_dirs_counter);
1255 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1256 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1257 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1259 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1260 kfree(get_qf_name(sb, sbi, i));
1263 /* Debugging code just in case the in-memory inode orphan list
1264 * isn't empty. The on-disk one can be non-empty if we've
1265 * detected an error and taken the fs readonly, but the
1266 * in-memory list had better be clean by this point. */
1267 if (!list_empty(&sbi->s_orphan))
1268 dump_orphan_list(sb, sbi);
1269 ASSERT(list_empty(&sbi->s_orphan));
1271 sync_blockdev(sb->s_bdev);
1272 invalidate_bdev(sb->s_bdev);
1273 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1275 * Invalidate the journal device's buffers. We don't want them
1276 * floating about in memory - the physical journal device may
1277 * hotswapped, and it breaks the `ro-after' testing code.
1279 sync_blockdev(sbi->s_journal_bdev);
1280 invalidate_bdev(sbi->s_journal_bdev);
1281 ext4_blkdev_remove(sbi);
1284 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1285 sbi->s_ea_inode_cache = NULL;
1287 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1288 sbi->s_ea_block_cache = NULL;
1290 ext4_stop_mmpd(sbi);
1293 sb->s_fs_info = NULL;
1295 * Now that we are completely done shutting down the
1296 * superblock, we need to actually destroy the kobject.
1298 kobject_put(&sbi->s_kobj);
1299 wait_for_completion(&sbi->s_kobj_unregister);
1300 if (sbi->s_chksum_driver)
1301 crypto_free_shash(sbi->s_chksum_driver);
1302 kfree(sbi->s_blockgroup_lock);
1303 fs_put_dax(sbi->s_daxdev);
1304 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1305 #ifdef CONFIG_UNICODE
1306 utf8_unload(sb->s_encoding);
1311 static struct kmem_cache *ext4_inode_cachep;
1314 * Called inside transaction, so use GFP_NOFS
1316 static struct inode *ext4_alloc_inode(struct super_block *sb)
1318 struct ext4_inode_info *ei;
1320 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1324 inode_set_iversion(&ei->vfs_inode, 1);
1325 spin_lock_init(&ei->i_raw_lock);
1326 INIT_LIST_HEAD(&ei->i_prealloc_list);
1327 atomic_set(&ei->i_prealloc_active, 0);
1328 spin_lock_init(&ei->i_prealloc_lock);
1329 ext4_es_init_tree(&ei->i_es_tree);
1330 rwlock_init(&ei->i_es_lock);
1331 INIT_LIST_HEAD(&ei->i_es_list);
1332 ei->i_es_all_nr = 0;
1333 ei->i_es_shk_nr = 0;
1334 ei->i_es_shrink_lblk = 0;
1335 ei->i_reserved_data_blocks = 0;
1336 spin_lock_init(&(ei->i_block_reservation_lock));
1337 ext4_init_pending_tree(&ei->i_pending_tree);
1339 ei->i_reserved_quota = 0;
1340 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1343 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1344 spin_lock_init(&ei->i_completed_io_lock);
1346 ei->i_datasync_tid = 0;
1347 atomic_set(&ei->i_unwritten, 0);
1348 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1349 ext4_fc_init_inode(&ei->vfs_inode);
1350 mutex_init(&ei->i_fc_lock);
1351 return &ei->vfs_inode;
1354 static int ext4_drop_inode(struct inode *inode)
1356 int drop = generic_drop_inode(inode);
1359 drop = fscrypt_drop_inode(inode);
1361 trace_ext4_drop_inode(inode, drop);
1365 static void ext4_free_in_core_inode(struct inode *inode)
1367 fscrypt_free_inode(inode);
1368 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1369 pr_warn("%s: inode %ld still in fc list",
1370 __func__, inode->i_ino);
1372 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1375 static void ext4_destroy_inode(struct inode *inode)
1377 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1378 ext4_msg(inode->i_sb, KERN_ERR,
1379 "Inode %lu (%p): orphan list check failed!",
1380 inode->i_ino, EXT4_I(inode));
1381 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1382 EXT4_I(inode), sizeof(struct ext4_inode_info),
1387 if (EXT4_I(inode)->i_reserved_data_blocks)
1388 ext4_msg(inode->i_sb, KERN_ERR,
1389 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1390 inode->i_ino, EXT4_I(inode),
1391 EXT4_I(inode)->i_reserved_data_blocks);
1394 static void init_once(void *foo)
1396 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1398 INIT_LIST_HEAD(&ei->i_orphan);
1399 init_rwsem(&ei->xattr_sem);
1400 init_rwsem(&ei->i_data_sem);
1401 inode_init_once(&ei->vfs_inode);
1402 ext4_fc_init_inode(&ei->vfs_inode);
1405 static int __init init_inodecache(void)
1407 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1408 sizeof(struct ext4_inode_info), 0,
1409 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1411 offsetof(struct ext4_inode_info, i_data),
1412 sizeof_field(struct ext4_inode_info, i_data),
1414 if (ext4_inode_cachep == NULL)
1419 static void destroy_inodecache(void)
1422 * Make sure all delayed rcu free inodes are flushed before we
1426 kmem_cache_destroy(ext4_inode_cachep);
1429 void ext4_clear_inode(struct inode *inode)
1432 invalidate_inode_buffers(inode);
1434 ext4_discard_preallocations(inode, 0);
1435 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1437 if (EXT4_I(inode)->jinode) {
1438 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1439 EXT4_I(inode)->jinode);
1440 jbd2_free_inode(EXT4_I(inode)->jinode);
1441 EXT4_I(inode)->jinode = NULL;
1443 fscrypt_put_encryption_info(inode);
1444 fsverity_cleanup_inode(inode);
1447 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1448 u64 ino, u32 generation)
1450 struct inode *inode;
1453 * Currently we don't know the generation for parent directory, so
1454 * a generation of 0 means "accept any"
1456 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1458 return ERR_CAST(inode);
1459 if (generation && inode->i_generation != generation) {
1461 return ERR_PTR(-ESTALE);
1467 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1468 int fh_len, int fh_type)
1470 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1471 ext4_nfs_get_inode);
1474 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1475 int fh_len, int fh_type)
1477 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1478 ext4_nfs_get_inode);
1481 static int ext4_nfs_commit_metadata(struct inode *inode)
1483 struct writeback_control wbc = {
1484 .sync_mode = WB_SYNC_ALL
1487 trace_ext4_nfs_commit_metadata(inode);
1488 return ext4_write_inode(inode, &wbc);
1491 #ifdef CONFIG_FS_ENCRYPTION
1492 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1494 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1498 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1501 handle_t *handle = fs_data;
1502 int res, res2, credits, retries = 0;
1505 * Encrypting the root directory is not allowed because e2fsck expects
1506 * lost+found to exist and be unencrypted, and encrypting the root
1507 * directory would imply encrypting the lost+found directory as well as
1508 * the filename "lost+found" itself.
1510 if (inode->i_ino == EXT4_ROOT_INO)
1513 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1516 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1519 res = ext4_convert_inline_data(inode);
1524 * If a journal handle was specified, then the encryption context is
1525 * being set on a new inode via inheritance and is part of a larger
1526 * transaction to create the inode. Otherwise the encryption context is
1527 * being set on an existing inode in its own transaction. Only in the
1528 * latter case should the "retry on ENOSPC" logic be used.
1532 res = ext4_xattr_set_handle(handle, inode,
1533 EXT4_XATTR_INDEX_ENCRYPTION,
1534 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1537 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1538 ext4_clear_inode_state(inode,
1539 EXT4_STATE_MAY_INLINE_DATA);
1541 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1542 * S_DAX may be disabled
1544 ext4_set_inode_flags(inode, false);
1549 res = dquot_initialize(inode);
1553 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1558 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1560 return PTR_ERR(handle);
1562 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1563 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1566 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1568 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1569 * S_DAX may be disabled
1571 ext4_set_inode_flags(inode, false);
1572 res = ext4_mark_inode_dirty(handle, inode);
1574 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1576 res2 = ext4_journal_stop(handle);
1578 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1585 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1587 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1590 static bool ext4_has_stable_inodes(struct super_block *sb)
1592 return ext4_has_feature_stable_inodes(sb);
1595 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1596 int *ino_bits_ret, int *lblk_bits_ret)
1598 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1599 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1602 static const struct fscrypt_operations ext4_cryptops = {
1603 .key_prefix = "ext4:",
1604 .get_context = ext4_get_context,
1605 .set_context = ext4_set_context,
1606 .get_dummy_policy = ext4_get_dummy_policy,
1607 .empty_dir = ext4_empty_dir,
1608 .has_stable_inodes = ext4_has_stable_inodes,
1609 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1614 static const char * const quotatypes[] = INITQFNAMES;
1615 #define QTYPE2NAME(t) (quotatypes[t])
1617 static int ext4_write_dquot(struct dquot *dquot);
1618 static int ext4_acquire_dquot(struct dquot *dquot);
1619 static int ext4_release_dquot(struct dquot *dquot);
1620 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1621 static int ext4_write_info(struct super_block *sb, int type);
1622 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1623 const struct path *path);
1624 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1625 size_t len, loff_t off);
1626 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1627 const char *data, size_t len, loff_t off);
1628 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1629 unsigned int flags);
1631 static struct dquot **ext4_get_dquots(struct inode *inode)
1633 return EXT4_I(inode)->i_dquot;
1636 static const struct dquot_operations ext4_quota_operations = {
1637 .get_reserved_space = ext4_get_reserved_space,
1638 .write_dquot = ext4_write_dquot,
1639 .acquire_dquot = ext4_acquire_dquot,
1640 .release_dquot = ext4_release_dquot,
1641 .mark_dirty = ext4_mark_dquot_dirty,
1642 .write_info = ext4_write_info,
1643 .alloc_dquot = dquot_alloc,
1644 .destroy_dquot = dquot_destroy,
1645 .get_projid = ext4_get_projid,
1646 .get_inode_usage = ext4_get_inode_usage,
1647 .get_next_id = dquot_get_next_id,
1650 static const struct quotactl_ops ext4_qctl_operations = {
1651 .quota_on = ext4_quota_on,
1652 .quota_off = ext4_quota_off,
1653 .quota_sync = dquot_quota_sync,
1654 .get_state = dquot_get_state,
1655 .set_info = dquot_set_dqinfo,
1656 .get_dqblk = dquot_get_dqblk,
1657 .set_dqblk = dquot_set_dqblk,
1658 .get_nextdqblk = dquot_get_next_dqblk,
1662 static const struct super_operations ext4_sops = {
1663 .alloc_inode = ext4_alloc_inode,
1664 .free_inode = ext4_free_in_core_inode,
1665 .destroy_inode = ext4_destroy_inode,
1666 .write_inode = ext4_write_inode,
1667 .dirty_inode = ext4_dirty_inode,
1668 .drop_inode = ext4_drop_inode,
1669 .evict_inode = ext4_evict_inode,
1670 .put_super = ext4_put_super,
1671 .sync_fs = ext4_sync_fs,
1672 .freeze_fs = ext4_freeze,
1673 .unfreeze_fs = ext4_unfreeze,
1674 .statfs = ext4_statfs,
1675 .show_options = ext4_show_options,
1677 .quota_read = ext4_quota_read,
1678 .quota_write = ext4_quota_write,
1679 .get_dquots = ext4_get_dquots,
1683 static const struct export_operations ext4_export_ops = {
1684 .fh_to_dentry = ext4_fh_to_dentry,
1685 .fh_to_parent = ext4_fh_to_parent,
1686 .get_parent = ext4_get_parent,
1687 .commit_metadata = ext4_nfs_commit_metadata,
1691 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1692 Opt_resgid, Opt_resuid, Opt_sb,
1693 Opt_nouid32, Opt_debug, Opt_removed,
1694 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1695 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1696 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1697 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1698 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1699 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1701 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1702 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1703 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1704 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1705 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1706 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1707 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1708 Opt_inode_readahead_blks, Opt_journal_ioprio,
1709 Opt_dioread_nolock, Opt_dioread_lock,
1710 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1711 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1712 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1713 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1714 #ifdef CONFIG_EXT4_DEBUG
1715 Opt_fc_debug_max_replay, Opt_fc_debug_force
1719 static const struct constant_table ext4_param_errors[] = {
1720 {"continue", EXT4_MOUNT_ERRORS_CONT},
1721 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1722 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1726 static const struct constant_table ext4_param_data[] = {
1727 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1728 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1729 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1733 static const struct constant_table ext4_param_data_err[] = {
1734 {"abort", Opt_data_err_abort},
1735 {"ignore", Opt_data_err_ignore},
1739 static const struct constant_table ext4_param_jqfmt[] = {
1740 {"vfsold", QFMT_VFS_OLD},
1741 {"vfsv0", QFMT_VFS_V0},
1742 {"vfsv1", QFMT_VFS_V1},
1746 static const struct constant_table ext4_param_dax[] = {
1747 {"always", Opt_dax_always},
1748 {"inode", Opt_dax_inode},
1749 {"never", Opt_dax_never},
1753 /* String parameter that allows empty argument */
1754 #define fsparam_string_empty(NAME, OPT) \
1755 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1758 * Mount option specification
1759 * We don't use fsparam_flag_no because of the way we set the
1760 * options and the way we show them in _ext4_show_options(). To
1761 * keep the changes to a minimum, let's keep the negative options
1764 static const struct fs_parameter_spec ext4_param_specs[] = {
1765 fsparam_flag ("bsddf", Opt_bsd_df),
1766 fsparam_flag ("minixdf", Opt_minix_df),
1767 fsparam_flag ("grpid", Opt_grpid),
1768 fsparam_flag ("bsdgroups", Opt_grpid),
1769 fsparam_flag ("nogrpid", Opt_nogrpid),
1770 fsparam_flag ("sysvgroups", Opt_nogrpid),
1771 fsparam_u32 ("resgid", Opt_resgid),
1772 fsparam_u32 ("resuid", Opt_resuid),
1773 fsparam_u32 ("sb", Opt_sb),
1774 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1775 fsparam_flag ("nouid32", Opt_nouid32),
1776 fsparam_flag ("debug", Opt_debug),
1777 fsparam_flag ("oldalloc", Opt_removed),
1778 fsparam_flag ("orlov", Opt_removed),
1779 fsparam_flag ("user_xattr", Opt_user_xattr),
1780 fsparam_flag ("nouser_xattr", Opt_nouser_xattr),
1781 fsparam_flag ("acl", Opt_acl),
1782 fsparam_flag ("noacl", Opt_noacl),
1783 fsparam_flag ("norecovery", Opt_noload),
1784 fsparam_flag ("noload", Opt_noload),
1785 fsparam_flag ("bh", Opt_removed),
1786 fsparam_flag ("nobh", Opt_removed),
1787 fsparam_u32 ("commit", Opt_commit),
1788 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1789 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1790 fsparam_u32 ("journal_dev", Opt_journal_dev),
1791 fsparam_bdev ("journal_path", Opt_journal_path),
1792 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1793 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1794 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1795 fsparam_flag ("abort", Opt_abort),
1796 fsparam_enum ("data", Opt_data, ext4_param_data),
1797 fsparam_enum ("data_err", Opt_data_err,
1798 ext4_param_data_err),
1799 fsparam_string_empty
1800 ("usrjquota", Opt_usrjquota),
1801 fsparam_string_empty
1802 ("grpjquota", Opt_grpjquota),
1803 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1804 fsparam_flag ("grpquota", Opt_grpquota),
1805 fsparam_flag ("quota", Opt_quota),
1806 fsparam_flag ("noquota", Opt_noquota),
1807 fsparam_flag ("usrquota", Opt_usrquota),
1808 fsparam_flag ("prjquota", Opt_prjquota),
1809 fsparam_flag ("barrier", Opt_barrier),
1810 fsparam_u32 ("barrier", Opt_barrier),
1811 fsparam_flag ("nobarrier", Opt_nobarrier),
1812 fsparam_flag ("i_version", Opt_i_version),
1813 fsparam_flag ("dax", Opt_dax),
1814 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1815 fsparam_u32 ("stripe", Opt_stripe),
1816 fsparam_flag ("delalloc", Opt_delalloc),
1817 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1818 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1819 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1820 fsparam_u32 ("debug_want_extra_isize",
1821 Opt_debug_want_extra_isize),
1822 fsparam_flag ("mblk_io_submit", Opt_removed),
1823 fsparam_flag ("nomblk_io_submit", Opt_removed),
1824 fsparam_flag ("block_validity", Opt_block_validity),
1825 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1826 fsparam_u32 ("inode_readahead_blks",
1827 Opt_inode_readahead_blks),
1828 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1829 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1830 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1831 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1832 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1833 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1834 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1835 fsparam_flag ("discard", Opt_discard),
1836 fsparam_flag ("nodiscard", Opt_nodiscard),
1837 fsparam_u32 ("init_itable", Opt_init_itable),
1838 fsparam_flag ("init_itable", Opt_init_itable),
1839 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1840 #ifdef CONFIG_EXT4_DEBUG
1841 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1842 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1844 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1845 fsparam_flag ("test_dummy_encryption",
1846 Opt_test_dummy_encryption),
1847 fsparam_string ("test_dummy_encryption",
1848 Opt_test_dummy_encryption),
1849 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1850 fsparam_flag ("nombcache", Opt_nombcache),
1851 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1852 fsparam_flag ("prefetch_block_bitmaps",
1854 fsparam_flag ("no_prefetch_block_bitmaps",
1855 Opt_no_prefetch_block_bitmaps),
1856 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1857 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1858 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1859 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1860 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1861 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1865 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1866 #define DEFAULT_MB_OPTIMIZE_SCAN (-1)
1868 static const char deprecated_msg[] =
1869 "Mount option \"%s\" will be removed by %s\n"
1870 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1872 #define MOPT_SET 0x0001
1873 #define MOPT_CLEAR 0x0002
1874 #define MOPT_NOSUPPORT 0x0004
1875 #define MOPT_EXPLICIT 0x0008
1878 #define MOPT_QFMT 0x0010
1880 #define MOPT_Q MOPT_NOSUPPORT
1881 #define MOPT_QFMT MOPT_NOSUPPORT
1883 #define MOPT_NO_EXT2 0x0020
1884 #define MOPT_NO_EXT3 0x0040
1885 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1886 #define MOPT_SKIP 0x0080
1887 #define MOPT_2 0x0100
1889 static const struct mount_opts {
1893 } ext4_mount_opts[] = {
1894 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1895 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1896 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1897 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1898 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1899 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1900 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1901 MOPT_EXT4_ONLY | MOPT_SET},
1902 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1903 MOPT_EXT4_ONLY | MOPT_CLEAR},
1904 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1905 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1906 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1907 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1908 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1909 MOPT_EXT4_ONLY | MOPT_CLEAR},
1910 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1911 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1912 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1913 MOPT_EXT4_ONLY | MOPT_CLEAR},
1914 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1915 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1916 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1917 EXT4_MOUNT_JOURNAL_CHECKSUM),
1918 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1919 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1920 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1921 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1922 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1923 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1924 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1925 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1926 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1927 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1928 {Opt_journal_path, 0, MOPT_NO_EXT2},
1929 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1930 {Opt_data, 0, MOPT_NO_EXT2},
1931 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1932 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1933 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1934 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1935 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1937 {Opt_acl, 0, MOPT_NOSUPPORT},
1938 {Opt_noacl, 0, MOPT_NOSUPPORT},
1940 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1941 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1942 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1943 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1945 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1947 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1949 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1950 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1951 MOPT_CLEAR | MOPT_Q},
1952 {Opt_usrjquota, 0, MOPT_Q},
1953 {Opt_grpjquota, 0, MOPT_Q},
1954 {Opt_jqfmt, 0, MOPT_QFMT},
1955 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1956 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1958 #ifdef CONFIG_EXT4_DEBUG
1959 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1960 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1965 #ifdef CONFIG_UNICODE
1966 static const struct ext4_sb_encodings {
1970 } ext4_sb_encoding_map[] = {
1971 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1974 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1975 const struct ext4_sb_encodings **encoding,
1978 __u16 magic = le16_to_cpu(es->s_encoding);
1981 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1982 if (magic == ext4_sb_encoding_map[i].magic)
1985 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1988 *encoding = &ext4_sb_encoding_map[i];
1989 *flags = le16_to_cpu(es->s_encoding_flags);
1995 static int ext4_set_test_dummy_encryption(struct super_block *sb, char *arg)
1997 #ifdef CONFIG_FS_ENCRYPTION
1998 struct ext4_sb_info *sbi = EXT4_SB(sb);
2001 err = fscrypt_set_test_dummy_encryption(sb, arg,
2002 &sbi->s_dummy_enc_policy);
2004 ext4_msg(sb, KERN_WARNING,
2005 "Error while setting test dummy encryption [%d]", err);
2008 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2013 #define EXT4_SPEC_JQUOTA (1 << 0)
2014 #define EXT4_SPEC_JQFMT (1 << 1)
2015 #define EXT4_SPEC_DATAJ (1 << 2)
2016 #define EXT4_SPEC_SB_BLOCK (1 << 3)
2017 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
2018 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
2019 #define EXT4_SPEC_DUMMY_ENCRYPTION (1 << 6)
2020 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
2021 #define EXT4_SPEC_s_max_batch_time (1 << 8)
2022 #define EXT4_SPEC_s_min_batch_time (1 << 9)
2023 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
2024 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
2025 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
2026 #define EXT4_SPEC_s_stripe (1 << 13)
2027 #define EXT4_SPEC_s_resuid (1 << 14)
2028 #define EXT4_SPEC_s_resgid (1 << 15)
2029 #define EXT4_SPEC_s_commit_interval (1 << 16)
2030 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
2031 #define EXT4_SPEC_s_sb_block (1 << 18)
2033 struct ext4_fs_context {
2034 char *s_qf_names[EXT4_MAXQUOTAS];
2035 char *test_dummy_enc_arg;
2036 int s_jquota_fmt; /* Format of quota to use */
2037 int mb_optimize_scan;
2038 #ifdef CONFIG_EXT4_DEBUG
2039 int s_fc_debug_max_replay;
2041 unsigned short qname_spec;
2042 unsigned long vals_s_flags; /* Bits to set in s_flags */
2043 unsigned long mask_s_flags; /* Bits changed in s_flags */
2044 unsigned long journal_devnum;
2045 unsigned long s_commit_interval;
2046 unsigned long s_stripe;
2047 unsigned int s_inode_readahead_blks;
2048 unsigned int s_want_extra_isize;
2049 unsigned int s_li_wait_mult;
2050 unsigned int s_max_dir_size_kb;
2051 unsigned int journal_ioprio;
2052 unsigned int vals_s_mount_opt;
2053 unsigned int mask_s_mount_opt;
2054 unsigned int vals_s_mount_opt2;
2055 unsigned int mask_s_mount_opt2;
2056 unsigned int vals_s_mount_flags;
2057 unsigned int mask_s_mount_flags;
2058 unsigned int opt_flags; /* MOPT flags */
2060 u32 s_max_batch_time;
2061 u32 s_min_batch_time;
2064 ext4_fsblk_t s_sb_block;
2067 static void ext4_fc_free(struct fs_context *fc)
2069 struct ext4_fs_context *ctx = fc->fs_private;
2075 for (i = 0; i < EXT4_MAXQUOTAS; i++)
2076 kfree(ctx->s_qf_names[i]);
2078 kfree(ctx->test_dummy_enc_arg);
2082 int ext4_init_fs_context(struct fs_context *fc)
2084 struct xfs_fs_context *ctx;
2086 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2090 fc->fs_private = ctx;
2091 fc->ops = &ext4_context_ops;
2098 * Note the name of the specified quota file.
2100 static int note_qf_name(struct fs_context *fc, int qtype,
2101 struct fs_parameter *param)
2103 struct ext4_fs_context *ctx = fc->fs_private;
2106 if (param->size < 1) {
2107 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2110 if (strchr(param->string, '/')) {
2111 ext4_msg(NULL, KERN_ERR,
2112 "quotafile must be on filesystem root");
2115 if (ctx->s_qf_names[qtype]) {
2116 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2117 ext4_msg(NULL, KERN_ERR,
2118 "%s quota file already specified",
2125 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2127 ext4_msg(NULL, KERN_ERR,
2128 "Not enough memory for storing quotafile name");
2131 ctx->s_qf_names[qtype] = qname;
2132 ctx->qname_spec |= 1 << qtype;
2133 ctx->spec |= EXT4_SPEC_JQUOTA;
2138 * Clear the name of the specified quota file.
2140 static int unnote_qf_name(struct fs_context *fc, int qtype)
2142 struct ext4_fs_context *ctx = fc->fs_private;
2144 if (ctx->s_qf_names[qtype])
2145 kfree(ctx->s_qf_names[qtype]);
2147 ctx->s_qf_names[qtype] = NULL;
2148 ctx->qname_spec |= 1 << qtype;
2149 ctx->spec |= EXT4_SPEC_JQUOTA;
2154 #define EXT4_SET_CTX(name) \
2155 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2156 unsigned long flag) \
2158 ctx->mask_s_##name |= flag; \
2159 ctx->vals_s_##name |= flag; \
2161 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2162 unsigned long flag) \
2164 ctx->mask_s_##name |= flag; \
2165 ctx->vals_s_##name &= ~flag; \
2167 static inline unsigned long \
2168 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2170 return (ctx->vals_s_##name & flag); \
2173 EXT4_SET_CTX(flags);
2174 EXT4_SET_CTX(mount_opt);
2175 EXT4_SET_CTX(mount_opt2);
2176 EXT4_SET_CTX(mount_flags);
2178 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2180 struct ext4_fs_context *ctx = fc->fs_private;
2181 struct fs_parse_result result;
2182 const struct mount_opts *m;
2188 token = fs_parse(fc, ext4_param_specs, param, &result);
2191 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2193 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2194 if (token == m->token)
2197 ctx->opt_flags |= m->flags;
2199 if (m->flags & MOPT_EXPLICIT) {
2200 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2201 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2202 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2203 ctx_set_mount_opt2(ctx,
2204 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2209 if (m->flags & MOPT_NOSUPPORT) {
2210 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2218 if (!*param->string)
2219 return unnote_qf_name(fc, USRQUOTA);
2221 return note_qf_name(fc, USRQUOTA, param);
2223 if (!*param->string)
2224 return unnote_qf_name(fc, GRPQUOTA);
2226 return note_qf_name(fc, GRPQUOTA, param);
2229 case Opt_nouser_xattr:
2230 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5");
2233 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2234 ext4_msg(NULL, KERN_WARNING,
2235 "Ignoring %s option on remount", param->key);
2237 ctx->s_sb_block = result.uint_32;
2238 ctx->spec |= EXT4_SPEC_s_sb_block;
2242 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2246 ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
2249 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
2250 ext4_msg(NULL, KERN_WARNING, "Use iversion instead\n");
2251 ctx_set_flags(ctx, SB_I_VERSION);
2253 case Opt_inlinecrypt:
2254 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2255 ctx_set_flags(ctx, SB_INLINECRYPT);
2257 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2261 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2262 ctx_set_mount_opt(ctx, result.uint_32);
2266 ctx->s_jquota_fmt = result.uint_32;
2267 ctx->spec |= EXT4_SPEC_JQFMT;
2271 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2272 ctx_set_mount_opt(ctx, result.uint_32);
2273 ctx->spec |= EXT4_SPEC_DATAJ;
2276 if (result.uint_32 == 0)
2277 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
2278 else if (result.uint_32 > INT_MAX / HZ) {
2279 ext4_msg(NULL, KERN_ERR,
2280 "Invalid commit interval %d, "
2281 "must be smaller than %d",
2282 result.uint_32, INT_MAX / HZ);
2285 ctx->s_commit_interval = HZ * result.uint_32;
2286 ctx->spec |= EXT4_SPEC_s_commit_interval;
2288 case Opt_debug_want_extra_isize:
2289 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2290 ext4_msg(NULL, KERN_ERR,
2291 "Invalid want_extra_isize %d", result.uint_32);
2294 ctx->s_want_extra_isize = result.uint_32;
2295 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2297 case Opt_max_batch_time:
2298 ctx->s_max_batch_time = result.uint_32;
2299 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2301 case Opt_min_batch_time:
2302 ctx->s_min_batch_time = result.uint_32;
2303 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2305 case Opt_inode_readahead_blks:
2306 if (result.uint_32 &&
2307 (result.uint_32 > (1 << 30) ||
2308 !is_power_of_2(result.uint_32))) {
2309 ext4_msg(NULL, KERN_ERR,
2310 "EXT4-fs: inode_readahead_blks must be "
2311 "0 or a power of 2 smaller than 2^31");
2314 ctx->s_inode_readahead_blks = result.uint_32;
2315 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2317 case Opt_init_itable:
2318 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2319 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2320 if (param->type == fs_value_is_string)
2321 ctx->s_li_wait_mult = result.uint_32;
2322 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2324 case Opt_max_dir_size_kb:
2325 ctx->s_max_dir_size_kb = result.uint_32;
2326 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2328 #ifdef CONFIG_EXT4_DEBUG
2329 case Opt_fc_debug_max_replay:
2330 ctx->s_fc_debug_max_replay = result.uint_32;
2331 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2335 ctx->s_stripe = result.uint_32;
2336 ctx->spec |= EXT4_SPEC_s_stripe;
2339 uid = make_kuid(current_user_ns(), result.uint_32);
2340 if (!uid_valid(uid)) {
2341 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2345 ctx->s_resuid = uid;
2346 ctx->spec |= EXT4_SPEC_s_resuid;
2349 gid = make_kgid(current_user_ns(), result.uint_32);
2350 if (!gid_valid(gid)) {
2351 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2355 ctx->s_resgid = gid;
2356 ctx->spec |= EXT4_SPEC_s_resgid;
2358 case Opt_journal_dev:
2360 ext4_msg(NULL, KERN_ERR,
2361 "Cannot specify journal on remount");
2364 ctx->journal_devnum = result.uint_32;
2365 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2367 case Opt_journal_path:
2369 struct inode *journal_inode;
2374 ext4_msg(NULL, KERN_ERR,
2375 "Cannot specify journal on remount");
2379 error = fs_lookup_param(fc, param, 1, &path);
2381 ext4_msg(NULL, KERN_ERR, "error: could not find "
2382 "journal device path");
2386 journal_inode = d_inode(path.dentry);
2387 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2388 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2392 case Opt_journal_ioprio:
2393 if (result.uint_32 > 7) {
2394 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2398 ctx->journal_ioprio =
2399 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2400 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2402 case Opt_test_dummy_encryption:
2403 #ifdef CONFIG_FS_ENCRYPTION
2404 if (param->type == fs_value_is_flag) {
2405 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION;
2406 ctx->test_dummy_enc_arg = NULL;
2409 if (*param->string &&
2410 !(!strcmp(param->string, "v1") ||
2411 !strcmp(param->string, "v2"))) {
2412 ext4_msg(NULL, KERN_WARNING,
2413 "Value of option \"%s\" is unrecognized",
2417 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION;
2418 ctx->test_dummy_enc_arg = kmemdup_nul(param->string, param->size,
2421 ext4_msg(NULL, KERN_WARNING,
2422 "Test dummy encryption mount option ignored");
2427 #ifdef CONFIG_FS_DAX
2429 int type = (token == Opt_dax) ?
2430 Opt_dax : result.uint_32;
2434 case Opt_dax_always:
2435 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2436 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2439 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2440 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2443 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2444 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2445 /* Strictly for printing options */
2446 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2452 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2456 if (result.uint_32 == Opt_data_err_abort)
2457 ctx_set_mount_opt(ctx, m->mount_opt);
2458 else if (result.uint_32 == Opt_data_err_ignore)
2459 ctx_clear_mount_opt(ctx, m->mount_opt);
2461 case Opt_mb_optimize_scan:
2462 if (result.int_32 != 0 && result.int_32 != 1) {
2463 ext4_msg(NULL, KERN_WARNING,
2464 "mb_optimize_scan should be set to 0 or 1.");
2467 ctx->mb_optimize_scan = result.int_32;
2472 * At this point we should only be getting options requiring MOPT_SET,
2473 * or MOPT_CLEAR. Anything else is a bug
2475 if (m->token == Opt_err) {
2476 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2483 unsigned int set = 0;
2485 if ((param->type == fs_value_is_flag) ||
2489 if (m->flags & MOPT_CLEAR)
2491 else if (unlikely(!(m->flags & MOPT_SET))) {
2492 ext4_msg(NULL, KERN_WARNING,
2493 "buggy handling of option %s",
2498 if (m->flags & MOPT_2) {
2500 ctx_set_mount_opt2(ctx, m->mount_opt);
2502 ctx_clear_mount_opt2(ctx, m->mount_opt);
2505 ctx_set_mount_opt(ctx, m->mount_opt);
2507 ctx_clear_mount_opt(ctx, m->mount_opt);
2514 static int parse_options(struct fs_context *fc, char *options)
2516 struct fs_parameter param;
2523 while ((key = strsep(&options, ",")) != NULL) {
2526 char *value = strchr(key, '=');
2528 param.type = fs_value_is_flag;
2529 param.string = NULL;
2536 v_len = strlen(value);
2537 param.string = kmemdup_nul(value, v_len,
2541 param.type = fs_value_is_string;
2547 ret = ext4_parse_param(fc, ¶m);
2549 kfree(param.string);
2555 ret = ext4_validate_options(fc);
2562 static int parse_apply_sb_mount_options(struct super_block *sb,
2563 struct ext4_fs_context *m_ctx)
2565 struct ext4_sb_info *sbi = EXT4_SB(sb);
2566 char *s_mount_opts = NULL;
2567 struct ext4_fs_context *s_ctx = NULL;
2568 struct fs_context *fc = NULL;
2571 if (!sbi->s_es->s_mount_opts[0])
2574 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2575 sizeof(sbi->s_es->s_mount_opts),
2580 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2584 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2588 fc->fs_private = s_ctx;
2589 fc->s_fs_info = sbi;
2591 ret = parse_options(fc, s_mount_opts);
2595 ret = ext4_check_opt_consistency(fc, sb);
2598 ext4_msg(sb, KERN_WARNING,
2599 "failed to parse options in superblock: %s",
2605 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2606 m_ctx->journal_devnum = s_ctx->journal_devnum;
2607 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2608 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2610 ret = ext4_apply_options(fc, sb);
2615 kfree(s_mount_opts);
2619 static void ext4_apply_quota_options(struct fs_context *fc,
2620 struct super_block *sb)
2623 bool quota_feature = ext4_has_feature_quota(sb);
2624 struct ext4_fs_context *ctx = fc->fs_private;
2625 struct ext4_sb_info *sbi = EXT4_SB(sb);
2632 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2633 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2634 if (!(ctx->qname_spec & (1 << i)))
2637 qname = ctx->s_qf_names[i]; /* May be NULL */
2638 ctx->s_qf_names[i] = NULL;
2639 kfree(sbi->s_qf_names[i]);
2640 rcu_assign_pointer(sbi->s_qf_names[i], qname);
2645 if (ctx->spec & EXT4_SPEC_JQFMT)
2646 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2651 * Check quota settings consistency.
2653 static int ext4_check_quota_consistency(struct fs_context *fc,
2654 struct super_block *sb)
2657 struct ext4_fs_context *ctx = fc->fs_private;
2658 struct ext4_sb_info *sbi = EXT4_SB(sb);
2659 bool quota_feature = ext4_has_feature_quota(sb);
2660 bool quota_loaded = sb_any_quota_loaded(sb);
2661 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2665 * We do the test below only for project quotas. 'usrquota' and
2666 * 'grpquota' mount options are allowed even without quota feature
2667 * to support legacy quotas in quota files.
2669 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2670 !ext4_has_feature_project(sb)) {
2671 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2672 "Cannot enable project quota enforcement.");
2676 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2677 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2679 ctx->mask_s_mount_opt & quota_flags &&
2680 !ctx_test_mount_opt(ctx, quota_flags))
2681 goto err_quota_change;
2683 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2685 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2686 if (!(ctx->qname_spec & (1 << i)))
2690 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2691 goto err_jquota_change;
2693 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2694 strcmp(sbi->s_qf_names[i],
2695 ctx->s_qf_names[i]) != 0)
2696 goto err_jquota_specified;
2699 if (quota_feature) {
2700 ext4_msg(NULL, KERN_INFO,
2701 "Journaled quota options ignored when "
2702 "QUOTA feature is enabled");
2707 if (ctx->spec & EXT4_SPEC_JQFMT) {
2708 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2709 goto err_jquota_change;
2710 if (quota_feature) {
2711 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2712 "ignored when QUOTA feature is enabled");
2717 /* Make sure we don't mix old and new quota format */
2718 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2719 ctx->s_qf_names[USRQUOTA]);
2720 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2721 ctx->s_qf_names[GRPQUOTA]);
2723 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2724 test_opt(sb, USRQUOTA));
2726 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2727 test_opt(sb, GRPQUOTA));
2730 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2734 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2738 if (usr_qf_name || grp_qf_name) {
2739 if (usrquota || grpquota) {
2740 ext4_msg(NULL, KERN_ERR, "old and new quota "
2745 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2746 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2755 ext4_msg(NULL, KERN_ERR,
2756 "Cannot change quota options when quota turned on");
2759 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2760 "options when quota turned on");
2762 err_jquota_specified:
2763 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2771 static int ext4_check_opt_consistency(struct fs_context *fc,
2772 struct super_block *sb)
2774 struct ext4_fs_context *ctx = fc->fs_private;
2775 struct ext4_sb_info *sbi = fc->s_fs_info;
2776 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2778 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2779 ext4_msg(NULL, KERN_ERR,
2780 "Mount option(s) incompatible with ext2");
2783 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2784 ext4_msg(NULL, KERN_ERR,
2785 "Mount option(s) incompatible with ext3");
2789 if (ctx->s_want_extra_isize >
2790 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2791 ext4_msg(NULL, KERN_ERR,
2792 "Invalid want_extra_isize %d",
2793 ctx->s_want_extra_isize);
2797 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2799 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2800 if (blocksize < PAGE_SIZE)
2801 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2802 "experimental mount option 'dioread_nolock' "
2803 "for blocksize < PAGE_SIZE");
2806 #ifdef CONFIG_FS_ENCRYPTION
2808 * This mount option is just for testing, and it's not worthwhile to
2809 * implement the extra complexity (e.g. RCU protection) that would be
2810 * needed to allow it to be set or changed during remount. We do allow
2811 * it to be specified during remount, but only if there is no change.
2813 if ((ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) &&
2814 is_remount && !sbi->s_dummy_enc_policy.policy) {
2815 ext4_msg(NULL, KERN_WARNING,
2816 "Can't set test_dummy_encryption on remount");
2821 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2822 if (!sbi->s_journal) {
2823 ext4_msg(NULL, KERN_WARNING,
2824 "Remounting file system with no journal "
2825 "so ignoring journalled data option");
2826 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2827 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2828 test_opt(sb, DATA_FLAGS)) {
2829 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2836 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2837 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2838 ext4_msg(NULL, KERN_ERR, "can't mount with "
2839 "both data=journal and dax");
2843 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2844 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2845 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2846 fail_dax_change_remount:
2847 ext4_msg(NULL, KERN_ERR, "can't change "
2848 "dax mount option while remounting");
2850 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2851 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2852 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2853 goto fail_dax_change_remount;
2854 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2855 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2856 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2857 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2858 goto fail_dax_change_remount;
2862 return ext4_check_quota_consistency(fc, sb);
2865 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2867 struct ext4_fs_context *ctx = fc->fs_private;
2868 struct ext4_sb_info *sbi = fc->s_fs_info;
2871 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2872 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2873 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2874 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2875 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2876 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2877 sb->s_flags &= ~ctx->mask_s_flags;
2878 sb->s_flags |= ctx->vals_s_flags;
2881 * i_version differs from common mount option iversion so we have
2882 * to let vfs know that it was set, otherwise it would get cleared
2885 if (ctx->mask_s_flags & SB_I_VERSION)
2886 fc->sb_flags |= SB_I_VERSION;
2888 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2889 APPLY(s_commit_interval);
2891 APPLY(s_max_batch_time);
2892 APPLY(s_min_batch_time);
2893 APPLY(s_want_extra_isize);
2894 APPLY(s_inode_readahead_blks);
2895 APPLY(s_max_dir_size_kb);
2896 APPLY(s_li_wait_mult);
2900 #ifdef CONFIG_EXT4_DEBUG
2901 APPLY(s_fc_debug_max_replay);
2904 ext4_apply_quota_options(fc, sb);
2906 if (ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION)
2907 ret = ext4_set_test_dummy_encryption(sb, ctx->test_dummy_enc_arg);
2913 static int ext4_validate_options(struct fs_context *fc)
2916 struct ext4_fs_context *ctx = fc->fs_private;
2917 char *usr_qf_name, *grp_qf_name;
2919 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2920 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2922 if (usr_qf_name || grp_qf_name) {
2923 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2924 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2926 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2927 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2929 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2930 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2931 ext4_msg(NULL, KERN_ERR, "old and new quota "
2940 static inline void ext4_show_quota_options(struct seq_file *seq,
2941 struct super_block *sb)
2943 #if defined(CONFIG_QUOTA)
2944 struct ext4_sb_info *sbi = EXT4_SB(sb);
2945 char *usr_qf_name, *grp_qf_name;
2947 if (sbi->s_jquota_fmt) {
2950 switch (sbi->s_jquota_fmt) {
2961 seq_printf(seq, ",jqfmt=%s", fmtname);
2965 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2966 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2968 seq_show_option(seq, "usrjquota", usr_qf_name);
2970 seq_show_option(seq, "grpjquota", grp_qf_name);
2975 static const char *token2str(int token)
2977 const struct fs_parameter_spec *spec;
2979 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2980 if (spec->opt == token && !spec->type)
2987 * - it's set to a non-default value OR
2988 * - if the per-sb default is different from the global default
2990 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2993 struct ext4_sb_info *sbi = EXT4_SB(sb);
2994 struct ext4_super_block *es = sbi->s_es;
2995 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2996 const struct mount_opts *m;
2997 char sep = nodefs ? '\n' : ',';
2999 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
3000 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
3002 if (sbi->s_sb_block != 1)
3003 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
3005 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
3006 int want_set = m->flags & MOPT_SET;
3007 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
3008 m->flags & MOPT_SKIP)
3010 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
3011 continue; /* skip if same as the default */
3013 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
3014 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
3015 continue; /* select Opt_noFoo vs Opt_Foo */
3016 SEQ_OPTS_PRINT("%s", token2str(m->token));
3019 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
3020 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
3021 SEQ_OPTS_PRINT("resuid=%u",
3022 from_kuid_munged(&init_user_ns, sbi->s_resuid));
3023 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
3024 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
3025 SEQ_OPTS_PRINT("resgid=%u",
3026 from_kgid_munged(&init_user_ns, sbi->s_resgid));
3027 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
3028 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3029 SEQ_OPTS_PUTS("errors=remount-ro");
3030 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
3031 SEQ_OPTS_PUTS("errors=continue");
3032 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
3033 SEQ_OPTS_PUTS("errors=panic");
3034 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
3035 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
3036 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
3037 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
3038 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
3039 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3040 if (sb->s_flags & SB_I_VERSION)
3041 SEQ_OPTS_PUTS("i_version");
3042 if (nodefs || sbi->s_stripe)
3043 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3044 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3045 (sbi->s_mount_opt ^ def_mount_opt)) {
3046 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3047 SEQ_OPTS_PUTS("data=journal");
3048 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3049 SEQ_OPTS_PUTS("data=ordered");
3050 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3051 SEQ_OPTS_PUTS("data=writeback");
3054 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3055 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3056 sbi->s_inode_readahead_blks);
3058 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3059 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3060 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3061 if (nodefs || sbi->s_max_dir_size_kb)
3062 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3063 if (test_opt(sb, DATA_ERR_ABORT))
3064 SEQ_OPTS_PUTS("data_err=abort");
3066 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3068 if (sb->s_flags & SB_INLINECRYPT)
3069 SEQ_OPTS_PUTS("inlinecrypt");
3071 if (test_opt(sb, DAX_ALWAYS)) {
3073 SEQ_OPTS_PUTS("dax");
3075 SEQ_OPTS_PUTS("dax=always");
3076 } else if (test_opt2(sb, DAX_NEVER)) {
3077 SEQ_OPTS_PUTS("dax=never");
3078 } else if (test_opt2(sb, DAX_INODE)) {
3079 SEQ_OPTS_PUTS("dax=inode");
3081 ext4_show_quota_options(seq, sb);
3085 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3087 return _ext4_show_options(seq, root->d_sb, 0);
3090 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3092 struct super_block *sb = seq->private;
3095 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3096 rc = _ext4_show_options(seq, sb, 1);
3097 seq_puts(seq, "\n");
3101 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3107 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3108 ext4_msg(sb, KERN_ERR, "revision level too high, "
3109 "forcing read-only mode");
3115 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3116 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3117 "running e2fsck is recommended");
3118 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3119 ext4_msg(sb, KERN_WARNING,
3120 "warning: mounting fs with errors, "
3121 "running e2fsck is recommended");
3122 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3123 le16_to_cpu(es->s_mnt_count) >=
3124 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3125 ext4_msg(sb, KERN_WARNING,
3126 "warning: maximal mount count reached, "
3127 "running e2fsck is recommended");
3128 else if (le32_to_cpu(es->s_checkinterval) &&
3129 (ext4_get_tstamp(es, s_lastcheck) +
3130 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3131 ext4_msg(sb, KERN_WARNING,
3132 "warning: checktime reached, "
3133 "running e2fsck is recommended");
3134 if (!sbi->s_journal)
3135 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3136 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3137 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3138 le16_add_cpu(&es->s_mnt_count, 1);
3139 ext4_update_tstamp(es, s_mtime);
3140 if (sbi->s_journal) {
3141 ext4_set_feature_journal_needs_recovery(sb);
3142 if (ext4_has_feature_orphan_file(sb))
3143 ext4_set_feature_orphan_present(sb);
3146 err = ext4_commit_super(sb);
3148 if (test_opt(sb, DEBUG))
3149 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3150 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3152 sbi->s_groups_count,
3153 EXT4_BLOCKS_PER_GROUP(sb),
3154 EXT4_INODES_PER_GROUP(sb),
3155 sbi->s_mount_opt, sbi->s_mount_opt2);
3157 cleancache_init_fs(sb);
3161 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3163 struct ext4_sb_info *sbi = EXT4_SB(sb);
3164 struct flex_groups **old_groups, **new_groups;
3167 if (!sbi->s_log_groups_per_flex)
3170 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3171 if (size <= sbi->s_flex_groups_allocated)
3174 new_groups = kvzalloc(roundup_pow_of_two(size *
3175 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3177 ext4_msg(sb, KERN_ERR,
3178 "not enough memory for %d flex group pointers", size);
3181 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3182 new_groups[i] = kvzalloc(roundup_pow_of_two(
3183 sizeof(struct flex_groups)),
3185 if (!new_groups[i]) {
3186 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3187 kvfree(new_groups[j]);
3189 ext4_msg(sb, KERN_ERR,
3190 "not enough memory for %d flex groups", size);
3195 old_groups = rcu_dereference(sbi->s_flex_groups);
3197 memcpy(new_groups, old_groups,
3198 (sbi->s_flex_groups_allocated *
3199 sizeof(struct flex_groups *)));
3201 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3202 sbi->s_flex_groups_allocated = size;
3204 ext4_kvfree_array_rcu(old_groups);
3208 static int ext4_fill_flex_info(struct super_block *sb)
3210 struct ext4_sb_info *sbi = EXT4_SB(sb);
3211 struct ext4_group_desc *gdp = NULL;
3212 struct flex_groups *fg;
3213 ext4_group_t flex_group;
3216 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3217 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3218 sbi->s_log_groups_per_flex = 0;
3222 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3226 for (i = 0; i < sbi->s_groups_count; i++) {
3227 gdp = ext4_get_group_desc(sb, i, NULL);
3229 flex_group = ext4_flex_group(sbi, i);
3230 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3231 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3232 atomic64_add(ext4_free_group_clusters(sb, gdp),
3233 &fg->free_clusters);
3234 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3242 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3243 struct ext4_group_desc *gdp)
3245 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3247 __le32 le_group = cpu_to_le32(block_group);
3248 struct ext4_sb_info *sbi = EXT4_SB(sb);
3250 if (ext4_has_metadata_csum(sbi->s_sb)) {
3251 /* Use new metadata_csum algorithm */
3253 __u16 dummy_csum = 0;
3255 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3257 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3258 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3259 sizeof(dummy_csum));
3260 offset += sizeof(dummy_csum);
3261 if (offset < sbi->s_desc_size)
3262 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3263 sbi->s_desc_size - offset);
3265 crc = csum32 & 0xFFFF;
3269 /* old crc16 code */
3270 if (!ext4_has_feature_gdt_csum(sb))
3273 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3274 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3275 crc = crc16(crc, (__u8 *)gdp, offset);
3276 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3277 /* for checksum of struct ext4_group_desc do the rest...*/
3278 if (ext4_has_feature_64bit(sb) &&
3279 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3280 crc = crc16(crc, (__u8 *)gdp + offset,
3281 le16_to_cpu(sbi->s_es->s_desc_size) -
3285 return cpu_to_le16(crc);
3288 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3289 struct ext4_group_desc *gdp)
3291 if (ext4_has_group_desc_csum(sb) &&
3292 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3298 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3299 struct ext4_group_desc *gdp)
3301 if (!ext4_has_group_desc_csum(sb))
3303 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3306 /* Called at mount-time, super-block is locked */
3307 static int ext4_check_descriptors(struct super_block *sb,
3308 ext4_fsblk_t sb_block,
3309 ext4_group_t *first_not_zeroed)
3311 struct ext4_sb_info *sbi = EXT4_SB(sb);
3312 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3313 ext4_fsblk_t last_block;
3314 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3315 ext4_fsblk_t block_bitmap;
3316 ext4_fsblk_t inode_bitmap;
3317 ext4_fsblk_t inode_table;
3318 int flexbg_flag = 0;
3319 ext4_group_t i, grp = sbi->s_groups_count;
3321 if (ext4_has_feature_flex_bg(sb))
3324 ext4_debug("Checking group descriptors");
3326 for (i = 0; i < sbi->s_groups_count; i++) {
3327 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3329 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3330 last_block = ext4_blocks_count(sbi->s_es) - 1;
3332 last_block = first_block +
3333 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3335 if ((grp == sbi->s_groups_count) &&
3336 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3339 block_bitmap = ext4_block_bitmap(sb, gdp);
3340 if (block_bitmap == sb_block) {
3341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3342 "Block bitmap for group %u overlaps "
3347 if (block_bitmap >= sb_block + 1 &&
3348 block_bitmap <= last_bg_block) {
3349 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3350 "Block bitmap for group %u overlaps "
3351 "block group descriptors", i);
3355 if (block_bitmap < first_block || block_bitmap > last_block) {
3356 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3357 "Block bitmap for group %u not in group "
3358 "(block %llu)!", i, block_bitmap);
3361 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3362 if (inode_bitmap == sb_block) {
3363 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3364 "Inode bitmap for group %u overlaps "
3369 if (inode_bitmap >= sb_block + 1 &&
3370 inode_bitmap <= last_bg_block) {
3371 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3372 "Inode bitmap for group %u overlaps "
3373 "block group descriptors", i);
3377 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3378 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3379 "Inode bitmap for group %u not in group "
3380 "(block %llu)!", i, inode_bitmap);
3383 inode_table = ext4_inode_table(sb, gdp);
3384 if (inode_table == sb_block) {
3385 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3386 "Inode table for group %u overlaps "
3391 if (inode_table >= sb_block + 1 &&
3392 inode_table <= last_bg_block) {
3393 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3394 "Inode table for group %u overlaps "
3395 "block group descriptors", i);
3399 if (inode_table < first_block ||
3400 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3401 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3402 "Inode table for group %u not in group "
3403 "(block %llu)!", i, inode_table);
3406 ext4_lock_group(sb, i);
3407 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3408 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3409 "Checksum for group %u failed (%u!=%u)",
3410 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3411 gdp)), le16_to_cpu(gdp->bg_checksum));
3412 if (!sb_rdonly(sb)) {
3413 ext4_unlock_group(sb, i);
3417 ext4_unlock_group(sb, i);
3419 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3421 if (NULL != first_not_zeroed)
3422 *first_not_zeroed = grp;
3427 * Maximal extent format file size.
3428 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3429 * extent format containers, within a sector_t, and within i_blocks
3430 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3431 * so that won't be a limiting factor.
3433 * However there is other limiting factor. We do store extents in the form
3434 * of starting block and length, hence the resulting length of the extent
3435 * covering maximum file size must fit into on-disk format containers as
3436 * well. Given that length is always by 1 unit bigger than max unit (because
3437 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3439 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3441 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3444 loff_t upper_limit = MAX_LFS_FILESIZE;
3446 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3448 if (!has_huge_files) {
3449 upper_limit = (1LL << 32) - 1;
3451 /* total blocks in file system block size */
3452 upper_limit >>= (blkbits - 9);
3453 upper_limit <<= blkbits;
3457 * 32-bit extent-start container, ee_block. We lower the maxbytes
3458 * by one fs block, so ee_len can cover the extent of maximum file
3461 res = (1LL << 32) - 1;
3464 /* Sanity check against vm- & vfs- imposed limits */
3465 if (res > upper_limit)
3472 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3473 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3474 * We need to be 1 filesystem block less than the 2^48 sector limit.
3476 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3478 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS;
3482 * This is calculated to be the largest file size for a dense, block
3483 * mapped file such that the file's total number of 512-byte sectors,
3484 * including data and all indirect blocks, does not exceed (2^48 - 1).
3486 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3487 * number of 512-byte sectors of the file.
3489 if (!has_huge_files) {
3491 * !has_huge_files or implies that the inode i_block field
3492 * represents total file blocks in 2^32 512-byte sectors ==
3493 * size of vfs inode i_blocks * 8
3495 upper_limit = (1LL << 32) - 1;
3497 /* total blocks in file system block size */
3498 upper_limit >>= (bits - 9);
3502 * We use 48 bit ext4_inode i_blocks
3503 * With EXT4_HUGE_FILE_FL set the i_blocks
3504 * represent total number of blocks in
3505 * file system block size
3507 upper_limit = (1LL << 48) - 1;
3511 /* indirect blocks */
3513 /* double indirect blocks */
3514 meta_blocks += 1 + (1LL << (bits-2));
3515 /* tripple indirect blocks */
3516 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3518 upper_limit -= meta_blocks;
3519 upper_limit <<= bits;
3521 res += 1LL << (bits-2);
3522 res += 1LL << (2*(bits-2));
3523 res += 1LL << (3*(bits-2));
3525 if (res > upper_limit)
3528 if (res > MAX_LFS_FILESIZE)
3529 res = MAX_LFS_FILESIZE;
3534 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3535 ext4_fsblk_t logical_sb_block, int nr)
3537 struct ext4_sb_info *sbi = EXT4_SB(sb);
3538 ext4_group_t bg, first_meta_bg;
3541 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3543 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3544 return logical_sb_block + nr + 1;
3545 bg = sbi->s_desc_per_block * nr;
3546 if (ext4_bg_has_super(sb, bg))
3550 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3551 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3552 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3555 if (sb->s_blocksize == 1024 && nr == 0 &&
3556 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3559 return (has_super + ext4_group_first_block_no(sb, bg));
3563 * ext4_get_stripe_size: Get the stripe size.
3564 * @sbi: In memory super block info
3566 * If we have specified it via mount option, then
3567 * use the mount option value. If the value specified at mount time is
3568 * greater than the blocks per group use the super block value.
3569 * If the super block value is greater than blocks per group return 0.
3570 * Allocator needs it be less than blocks per group.
3573 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3575 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3576 unsigned long stripe_width =
3577 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3580 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3581 ret = sbi->s_stripe;
3582 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3584 else if (stride && stride <= sbi->s_blocks_per_group)
3590 * If the stripe width is 1, this makes no sense and
3591 * we set it to 0 to turn off stripe handling code.
3600 * Check whether this filesystem can be mounted based on
3601 * the features present and the RDONLY/RDWR mount requested.
3602 * Returns 1 if this filesystem can be mounted as requested,
3603 * 0 if it cannot be.
3605 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3607 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3608 ext4_msg(sb, KERN_ERR,
3609 "Couldn't mount because of "
3610 "unsupported optional features (%x)",
3611 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3612 ~EXT4_FEATURE_INCOMPAT_SUPP));
3616 #ifndef CONFIG_UNICODE
3617 if (ext4_has_feature_casefold(sb)) {
3618 ext4_msg(sb, KERN_ERR,
3619 "Filesystem with casefold feature cannot be "
3620 "mounted without CONFIG_UNICODE");
3628 if (ext4_has_feature_readonly(sb)) {
3629 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3630 sb->s_flags |= SB_RDONLY;
3634 /* Check that feature set is OK for a read-write mount */
3635 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3636 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3637 "unsupported optional features (%x)",
3638 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3639 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3642 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3643 ext4_msg(sb, KERN_ERR,
3644 "Can't support bigalloc feature without "
3645 "extents feature\n");
3649 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3650 if (!readonly && (ext4_has_feature_quota(sb) ||
3651 ext4_has_feature_project(sb))) {
3652 ext4_msg(sb, KERN_ERR,
3653 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3656 #endif /* CONFIG_QUOTA */
3661 * This function is called once a day if we have errors logged
3662 * on the file system
3664 static void print_daily_error_info(struct timer_list *t)
3666 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3667 struct super_block *sb = sbi->s_sb;
3668 struct ext4_super_block *es = sbi->s_es;
3670 if (es->s_error_count)
3671 /* fsck newer than v1.41.13 is needed to clean this condition. */
3672 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3673 le32_to_cpu(es->s_error_count));
3674 if (es->s_first_error_time) {
3675 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3677 ext4_get_tstamp(es, s_first_error_time),
3678 (int) sizeof(es->s_first_error_func),
3679 es->s_first_error_func,
3680 le32_to_cpu(es->s_first_error_line));
3681 if (es->s_first_error_ino)
3682 printk(KERN_CONT ": inode %u",
3683 le32_to_cpu(es->s_first_error_ino));
3684 if (es->s_first_error_block)
3685 printk(KERN_CONT ": block %llu", (unsigned long long)
3686 le64_to_cpu(es->s_first_error_block));
3687 printk(KERN_CONT "\n");
3689 if (es->s_last_error_time) {
3690 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3692 ext4_get_tstamp(es, s_last_error_time),
3693 (int) sizeof(es->s_last_error_func),
3694 es->s_last_error_func,
3695 le32_to_cpu(es->s_last_error_line));
3696 if (es->s_last_error_ino)
3697 printk(KERN_CONT ": inode %u",
3698 le32_to_cpu(es->s_last_error_ino));
3699 if (es->s_last_error_block)
3700 printk(KERN_CONT ": block %llu", (unsigned long long)
3701 le64_to_cpu(es->s_last_error_block));
3702 printk(KERN_CONT "\n");
3704 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3707 /* Find next suitable group and run ext4_init_inode_table */
3708 static int ext4_run_li_request(struct ext4_li_request *elr)
3710 struct ext4_group_desc *gdp = NULL;
3711 struct super_block *sb = elr->lr_super;
3712 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3713 ext4_group_t group = elr->lr_next_group;
3714 unsigned int prefetch_ios = 0;
3718 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3719 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3720 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3722 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3724 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3726 if (group >= elr->lr_next_group) {
3728 if (elr->lr_first_not_zeroed != ngroups &&
3729 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3730 elr->lr_next_group = elr->lr_first_not_zeroed;
3731 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3738 for (; group < ngroups; group++) {
3739 gdp = ext4_get_group_desc(sb, group, NULL);
3745 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3749 if (group >= ngroups)
3753 start_time = ktime_get_real_ns();
3754 ret = ext4_init_inode_table(sb, group,
3755 elr->lr_timeout ? 0 : 1);
3756 trace_ext4_lazy_itable_init(sb, group);
3757 if (elr->lr_timeout == 0) {
3758 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3759 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3761 elr->lr_next_sched = jiffies + elr->lr_timeout;
3762 elr->lr_next_group = group + 1;
3768 * Remove lr_request from the list_request and free the
3769 * request structure. Should be called with li_list_mtx held
3771 static void ext4_remove_li_request(struct ext4_li_request *elr)
3776 list_del(&elr->lr_request);
3777 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3781 static void ext4_unregister_li_request(struct super_block *sb)
3783 mutex_lock(&ext4_li_mtx);
3784 if (!ext4_li_info) {
3785 mutex_unlock(&ext4_li_mtx);
3789 mutex_lock(&ext4_li_info->li_list_mtx);
3790 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3791 mutex_unlock(&ext4_li_info->li_list_mtx);
3792 mutex_unlock(&ext4_li_mtx);
3795 static struct task_struct *ext4_lazyinit_task;
3798 * This is the function where ext4lazyinit thread lives. It walks
3799 * through the request list searching for next scheduled filesystem.
3800 * When such a fs is found, run the lazy initialization request
3801 * (ext4_rn_li_request) and keep track of the time spend in this
3802 * function. Based on that time we compute next schedule time of
3803 * the request. When walking through the list is complete, compute
3804 * next waking time and put itself into sleep.
3806 static int ext4_lazyinit_thread(void *arg)
3808 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3809 struct list_head *pos, *n;
3810 struct ext4_li_request *elr;
3811 unsigned long next_wakeup, cur;
3813 BUG_ON(NULL == eli);
3817 next_wakeup = MAX_JIFFY_OFFSET;
3819 mutex_lock(&eli->li_list_mtx);
3820 if (list_empty(&eli->li_request_list)) {
3821 mutex_unlock(&eli->li_list_mtx);
3824 list_for_each_safe(pos, n, &eli->li_request_list) {
3827 elr = list_entry(pos, struct ext4_li_request,
3830 if (time_before(jiffies, elr->lr_next_sched)) {
3831 if (time_before(elr->lr_next_sched, next_wakeup))
3832 next_wakeup = elr->lr_next_sched;
3835 if (down_read_trylock(&elr->lr_super->s_umount)) {
3836 if (sb_start_write_trylock(elr->lr_super)) {
3839 * We hold sb->s_umount, sb can not
3840 * be removed from the list, it is
3841 * now safe to drop li_list_mtx
3843 mutex_unlock(&eli->li_list_mtx);
3844 err = ext4_run_li_request(elr);
3845 sb_end_write(elr->lr_super);
3846 mutex_lock(&eli->li_list_mtx);
3849 up_read((&elr->lr_super->s_umount));
3851 /* error, remove the lazy_init job */
3853 ext4_remove_li_request(elr);
3857 elr->lr_next_sched = jiffies +
3859 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3861 if (time_before(elr->lr_next_sched, next_wakeup))
3862 next_wakeup = elr->lr_next_sched;
3864 mutex_unlock(&eli->li_list_mtx);
3869 if ((time_after_eq(cur, next_wakeup)) ||
3870 (MAX_JIFFY_OFFSET == next_wakeup)) {
3875 schedule_timeout_interruptible(next_wakeup - cur);
3877 if (kthread_should_stop()) {
3878 ext4_clear_request_list();
3885 * It looks like the request list is empty, but we need
3886 * to check it under the li_list_mtx lock, to prevent any
3887 * additions into it, and of course we should lock ext4_li_mtx
3888 * to atomically free the list and ext4_li_info, because at
3889 * this point another ext4 filesystem could be registering
3892 mutex_lock(&ext4_li_mtx);
3893 mutex_lock(&eli->li_list_mtx);
3894 if (!list_empty(&eli->li_request_list)) {
3895 mutex_unlock(&eli->li_list_mtx);
3896 mutex_unlock(&ext4_li_mtx);
3899 mutex_unlock(&eli->li_list_mtx);
3900 kfree(ext4_li_info);
3901 ext4_li_info = NULL;
3902 mutex_unlock(&ext4_li_mtx);
3907 static void ext4_clear_request_list(void)
3909 struct list_head *pos, *n;
3910 struct ext4_li_request *elr;
3912 mutex_lock(&ext4_li_info->li_list_mtx);
3913 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3914 elr = list_entry(pos, struct ext4_li_request,
3916 ext4_remove_li_request(elr);
3918 mutex_unlock(&ext4_li_info->li_list_mtx);
3921 static int ext4_run_lazyinit_thread(void)
3923 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3924 ext4_li_info, "ext4lazyinit");
3925 if (IS_ERR(ext4_lazyinit_task)) {
3926 int err = PTR_ERR(ext4_lazyinit_task);
3927 ext4_clear_request_list();
3928 kfree(ext4_li_info);
3929 ext4_li_info = NULL;
3930 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3931 "initialization thread\n",
3935 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3940 * Check whether it make sense to run itable init. thread or not.
3941 * If there is at least one uninitialized inode table, return
3942 * corresponding group number, else the loop goes through all
3943 * groups and return total number of groups.
3945 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3947 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3948 struct ext4_group_desc *gdp = NULL;
3950 if (!ext4_has_group_desc_csum(sb))
3953 for (group = 0; group < ngroups; group++) {
3954 gdp = ext4_get_group_desc(sb, group, NULL);
3958 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3965 static int ext4_li_info_new(void)
3967 struct ext4_lazy_init *eli = NULL;
3969 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3973 INIT_LIST_HEAD(&eli->li_request_list);
3974 mutex_init(&eli->li_list_mtx);
3976 eli->li_state |= EXT4_LAZYINIT_QUIT;
3983 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3986 struct ext4_li_request *elr;
3988 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3993 elr->lr_first_not_zeroed = start;
3994 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3995 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3996 elr->lr_next_group = start;
3998 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
4002 * Randomize first schedule time of the request to
4003 * spread the inode table initialization requests
4006 elr->lr_next_sched = jiffies + (prandom_u32() %
4007 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
4011 int ext4_register_li_request(struct super_block *sb,
4012 ext4_group_t first_not_zeroed)
4014 struct ext4_sb_info *sbi = EXT4_SB(sb);
4015 struct ext4_li_request *elr = NULL;
4016 ext4_group_t ngroups = sbi->s_groups_count;
4019 mutex_lock(&ext4_li_mtx);
4020 if (sbi->s_li_request != NULL) {
4022 * Reset timeout so it can be computed again, because
4023 * s_li_wait_mult might have changed.
4025 sbi->s_li_request->lr_timeout = 0;
4029 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4030 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
4031 !test_opt(sb, INIT_INODE_TABLE)))
4034 elr = ext4_li_request_new(sb, first_not_zeroed);
4040 if (NULL == ext4_li_info) {
4041 ret = ext4_li_info_new();
4046 mutex_lock(&ext4_li_info->li_list_mtx);
4047 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4048 mutex_unlock(&ext4_li_info->li_list_mtx);
4050 sbi->s_li_request = elr;
4052 * set elr to NULL here since it has been inserted to
4053 * the request_list and the removal and free of it is
4054 * handled by ext4_clear_request_list from now on.
4058 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4059 ret = ext4_run_lazyinit_thread();
4064 mutex_unlock(&ext4_li_mtx);
4071 * We do not need to lock anything since this is called on
4074 static void ext4_destroy_lazyinit_thread(void)
4077 * If thread exited earlier
4078 * there's nothing to be done.
4080 if (!ext4_li_info || !ext4_lazyinit_task)
4083 kthread_stop(ext4_lazyinit_task);
4086 static int set_journal_csum_feature_set(struct super_block *sb)
4089 int compat, incompat;
4090 struct ext4_sb_info *sbi = EXT4_SB(sb);
4092 if (ext4_has_metadata_csum(sb)) {
4093 /* journal checksum v3 */
4095 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4097 /* journal checksum v1 */
4098 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4102 jbd2_journal_clear_features(sbi->s_journal,
4103 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4104 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4105 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4106 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4107 ret = jbd2_journal_set_features(sbi->s_journal,
4109 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4111 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4112 ret = jbd2_journal_set_features(sbi->s_journal,
4115 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4116 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4118 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4119 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4126 * Note: calculating the overhead so we can be compatible with
4127 * historical BSD practice is quite difficult in the face of
4128 * clusters/bigalloc. This is because multiple metadata blocks from
4129 * different block group can end up in the same allocation cluster.
4130 * Calculating the exact overhead in the face of clustered allocation
4131 * requires either O(all block bitmaps) in memory or O(number of block
4132 * groups**2) in time. We will still calculate the superblock for
4133 * older file systems --- and if we come across with a bigalloc file
4134 * system with zero in s_overhead_clusters the estimate will be close to
4135 * correct especially for very large cluster sizes --- but for newer
4136 * file systems, it's better to calculate this figure once at mkfs
4137 * time, and store it in the superblock. If the superblock value is
4138 * present (even for non-bigalloc file systems), we will use it.
4140 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4143 struct ext4_sb_info *sbi = EXT4_SB(sb);
4144 struct ext4_group_desc *gdp;
4145 ext4_fsblk_t first_block, last_block, b;
4146 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4147 int s, j, count = 0;
4149 if (!ext4_has_feature_bigalloc(sb))
4150 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
4151 sbi->s_itb_per_group + 2);
4153 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4154 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4155 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4156 for (i = 0; i < ngroups; i++) {
4157 gdp = ext4_get_group_desc(sb, i, NULL);
4158 b = ext4_block_bitmap(sb, gdp);
4159 if (b >= first_block && b <= last_block) {
4160 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4163 b = ext4_inode_bitmap(sb, gdp);
4164 if (b >= first_block && b <= last_block) {
4165 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4168 b = ext4_inode_table(sb, gdp);
4169 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4170 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4171 int c = EXT4_B2C(sbi, b - first_block);
4172 ext4_set_bit(c, buf);
4178 if (ext4_bg_has_super(sb, grp)) {
4179 ext4_set_bit(s++, buf);
4182 j = ext4_bg_num_gdb(sb, grp);
4183 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4184 ext4_error(sb, "Invalid number of block group "
4185 "descriptor blocks: %d", j);
4186 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4190 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4194 return EXT4_CLUSTERS_PER_GROUP(sb) -
4195 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4199 * Compute the overhead and stash it in sbi->s_overhead
4201 int ext4_calculate_overhead(struct super_block *sb)
4203 struct ext4_sb_info *sbi = EXT4_SB(sb);
4204 struct ext4_super_block *es = sbi->s_es;
4205 struct inode *j_inode;
4206 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4207 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4208 ext4_fsblk_t overhead = 0;
4209 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4215 * Compute the overhead (FS structures). This is constant
4216 * for a given filesystem unless the number of block groups
4217 * changes so we cache the previous value until it does.
4221 * All of the blocks before first_data_block are overhead
4223 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4226 * Add the overhead found in each block group
4228 for (i = 0; i < ngroups; i++) {
4231 blks = count_overhead(sb, i, buf);
4234 memset(buf, 0, PAGE_SIZE);
4239 * Add the internal journal blocks whether the journal has been
4242 if (sbi->s_journal && !sbi->s_journal_bdev)
4243 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4244 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4245 /* j_inum for internal journal is non-zero */
4246 j_inode = ext4_get_journal_inode(sb, j_inum);
4248 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4249 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4252 ext4_msg(sb, KERN_ERR, "can't get journal size");
4255 sbi->s_overhead = overhead;
4257 free_page((unsigned long) buf);
4261 static void ext4_set_resv_clusters(struct super_block *sb)
4263 ext4_fsblk_t resv_clusters;
4264 struct ext4_sb_info *sbi = EXT4_SB(sb);
4267 * There's no need to reserve anything when we aren't using extents.
4268 * The space estimates are exact, there are no unwritten extents,
4269 * hole punching doesn't need new metadata... This is needed especially
4270 * to keep ext2/3 backward compatibility.
4272 if (!ext4_has_feature_extents(sb))
4275 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4276 * This should cover the situations where we can not afford to run
4277 * out of space like for example punch hole, or converting
4278 * unwritten extents in delalloc path. In most cases such
4279 * allocation would require 1, or 2 blocks, higher numbers are
4282 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4283 sbi->s_cluster_bits);
4285 do_div(resv_clusters, 50);
4286 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4288 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4291 static const char *ext4_quota_mode(struct super_block *sb)
4294 if (!ext4_quota_capable(sb))
4297 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4298 return "journalled";
4306 static void ext4_setup_csum_trigger(struct super_block *sb,
4307 enum ext4_journal_trigger_type type,
4309 struct jbd2_buffer_trigger_type *type,
4310 struct buffer_head *bh,
4314 struct ext4_sb_info *sbi = EXT4_SB(sb);
4316 sbi->s_journal_triggers[type].sb = sb;
4317 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4320 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4325 kfree(sbi->s_blockgroup_lock);
4326 fs_put_dax(sbi->s_daxdev);
4330 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4332 struct ext4_sb_info *sbi;
4334 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4338 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev);
4340 sbi->s_blockgroup_lock =
4341 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4343 if (!sbi->s_blockgroup_lock)
4346 sb->s_fs_info = sbi;
4350 fs_put_dax(sbi->s_daxdev);
4355 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
4357 struct buffer_head *bh, **group_desc;
4358 struct ext4_super_block *es = NULL;
4359 struct ext4_sb_info *sbi = EXT4_SB(sb);
4360 struct flex_groups **flex_groups;
4362 ext4_fsblk_t logical_sb_block;
4363 unsigned long offset = 0;
4364 unsigned long def_mount_opts;
4367 int blocksize, clustersize;
4368 unsigned int db_count;
4370 int needs_recovery, has_huge_files;
4373 ext4_group_t first_not_zeroed;
4374 struct ext4_fs_context *ctx = fc->fs_private;
4375 int silent = fc->sb_flags & SB_SILENT;
4377 /* Set defaults for the variables that will be set during parsing */
4378 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4379 ctx->mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
4381 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4382 sbi->s_sectors_written_start =
4383 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4385 /* -EINVAL is default */
4387 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4389 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4394 * The ext4 superblock will not be buffer aligned for other than 1kB
4395 * block sizes. We need to calculate the offset from buffer start.
4397 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4398 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4399 offset = do_div(logical_sb_block, blocksize);
4401 logical_sb_block = sbi->s_sb_block;
4404 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4406 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4411 * Note: s_es must be initialized as soon as possible because
4412 * some ext4 macro-instructions depend on its value
4414 es = (struct ext4_super_block *) (bh->b_data + offset);
4416 sb->s_magic = le16_to_cpu(es->s_magic);
4417 if (sb->s_magic != EXT4_SUPER_MAGIC)
4419 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4421 /* Warn if metadata_csum and gdt_csum are both set. */
4422 if (ext4_has_feature_metadata_csum(sb) &&
4423 ext4_has_feature_gdt_csum(sb))
4424 ext4_warning(sb, "metadata_csum and uninit_bg are "
4425 "redundant flags; please run fsck.");
4427 /* Check for a known checksum algorithm */
4428 if (!ext4_verify_csum_type(sb, es)) {
4429 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4430 "unknown checksum algorithm.");
4434 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4435 ext4_orphan_file_block_trigger);
4437 /* Load the checksum driver */
4438 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4439 if (IS_ERR(sbi->s_chksum_driver)) {
4440 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4441 ret = PTR_ERR(sbi->s_chksum_driver);
4442 sbi->s_chksum_driver = NULL;
4446 /* Check superblock checksum */
4447 if (!ext4_superblock_csum_verify(sb, es)) {
4448 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4449 "invalid superblock checksum. Run e2fsck?");
4455 /* Precompute checksum seed for all metadata */
4456 if (ext4_has_feature_csum_seed(sb))
4457 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4458 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4459 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4460 sizeof(es->s_uuid));
4462 /* Set defaults before we parse the mount options */
4463 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4464 set_opt(sb, INIT_INODE_TABLE);
4465 if (def_mount_opts & EXT4_DEFM_DEBUG)
4467 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4469 if (def_mount_opts & EXT4_DEFM_UID16)
4470 set_opt(sb, NO_UID32);
4471 /* xattr user namespace & acls are now defaulted on */
4472 set_opt(sb, XATTR_USER);
4473 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4474 set_opt(sb, POSIX_ACL);
4476 if (ext4_has_feature_fast_commit(sb))
4477 set_opt2(sb, JOURNAL_FAST_COMMIT);
4478 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4479 if (ext4_has_metadata_csum(sb))
4480 set_opt(sb, JOURNAL_CHECKSUM);
4482 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4483 set_opt(sb, JOURNAL_DATA);
4484 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4485 set_opt(sb, ORDERED_DATA);
4486 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4487 set_opt(sb, WRITEBACK_DATA);
4489 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4490 set_opt(sb, ERRORS_PANIC);
4491 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4492 set_opt(sb, ERRORS_CONT);
4494 set_opt(sb, ERRORS_RO);
4495 /* block_validity enabled by default; disable with noblock_validity */
4496 set_opt(sb, BLOCK_VALIDITY);
4497 if (def_mount_opts & EXT4_DEFM_DISCARD)
4498 set_opt(sb, DISCARD);
4500 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4501 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4502 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4503 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4504 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4506 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4507 set_opt(sb, BARRIER);
4510 * enable delayed allocation by default
4511 * Use -o nodelalloc to turn it off
4513 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4514 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4515 set_opt(sb, DELALLOC);
4518 * set default s_li_wait_mult for lazyinit, for the case there is
4519 * no mount option specified.
4521 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4523 if (le32_to_cpu(es->s_log_block_size) >
4524 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4525 ext4_msg(sb, KERN_ERR,
4526 "Invalid log block size: %u",
4527 le32_to_cpu(es->s_log_block_size));
4530 if (le32_to_cpu(es->s_log_cluster_size) >
4531 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4532 ext4_msg(sb, KERN_ERR,
4533 "Invalid log cluster size: %u",
4534 le32_to_cpu(es->s_log_cluster_size));
4538 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4540 if (blocksize == PAGE_SIZE)
4541 set_opt(sb, DIOREAD_NOLOCK);
4543 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4544 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4545 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4547 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4548 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4549 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4550 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4554 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4555 (!is_power_of_2(sbi->s_inode_size)) ||
4556 (sbi->s_inode_size > blocksize)) {
4557 ext4_msg(sb, KERN_ERR,
4558 "unsupported inode size: %d",
4560 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4564 * i_atime_extra is the last extra field available for
4565 * [acm]times in struct ext4_inode. Checking for that
4566 * field should suffice to ensure we have extra space
4569 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4570 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4571 sb->s_time_gran = 1;
4572 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4574 sb->s_time_gran = NSEC_PER_SEC;
4575 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4577 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4579 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4580 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4581 EXT4_GOOD_OLD_INODE_SIZE;
4582 if (ext4_has_feature_extra_isize(sb)) {
4583 unsigned v, max = (sbi->s_inode_size -
4584 EXT4_GOOD_OLD_INODE_SIZE);
4586 v = le16_to_cpu(es->s_want_extra_isize);
4588 ext4_msg(sb, KERN_ERR,
4589 "bad s_want_extra_isize: %d", v);
4592 if (sbi->s_want_extra_isize < v)
4593 sbi->s_want_extra_isize = v;
4595 v = le16_to_cpu(es->s_min_extra_isize);
4597 ext4_msg(sb, KERN_ERR,
4598 "bad s_min_extra_isize: %d", v);
4601 if (sbi->s_want_extra_isize < v)
4602 sbi->s_want_extra_isize = v;
4606 err = parse_apply_sb_mount_options(sb, ctx);
4610 sbi->s_def_mount_opt = sbi->s_mount_opt;
4612 err = ext4_check_opt_consistency(fc, sb);
4616 err = ext4_apply_options(fc, sb);
4620 #ifdef CONFIG_UNICODE
4621 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4622 const struct ext4_sb_encodings *encoding_info;
4623 struct unicode_map *encoding;
4624 __u16 encoding_flags;
4626 if (ext4_sb_read_encoding(es, &encoding_info,
4628 ext4_msg(sb, KERN_ERR,
4629 "Encoding requested by superblock is unknown");
4633 encoding = utf8_load(encoding_info->version);
4634 if (IS_ERR(encoding)) {
4635 ext4_msg(sb, KERN_ERR,
4636 "can't mount with superblock charset: %s-%s "
4637 "not supported by the kernel. flags: 0x%x.",
4638 encoding_info->name, encoding_info->version,
4642 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4643 "%s-%s with flags 0x%hx", encoding_info->name,
4644 encoding_info->version?:"\b", encoding_flags);
4646 sb->s_encoding = encoding;
4647 sb->s_encoding_flags = encoding_flags;
4651 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4652 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4653 /* can't mount with both data=journal and dioread_nolock. */
4654 clear_opt(sb, DIOREAD_NOLOCK);
4655 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4656 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4657 ext4_msg(sb, KERN_ERR, "can't mount with "
4658 "both data=journal and delalloc");
4661 if (test_opt(sb, DAX_ALWAYS)) {
4662 ext4_msg(sb, KERN_ERR, "can't mount with "
4663 "both data=journal and dax");
4666 if (ext4_has_feature_encrypt(sb)) {
4667 ext4_msg(sb, KERN_WARNING,
4668 "encrypted files will use data=ordered "
4669 "instead of data journaling mode");
4671 if (test_opt(sb, DELALLOC))
4672 clear_opt(sb, DELALLOC);
4674 sb->s_iflags |= SB_I_CGROUPWB;
4677 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4678 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4680 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4681 (ext4_has_compat_features(sb) ||
4682 ext4_has_ro_compat_features(sb) ||
4683 ext4_has_incompat_features(sb)))
4684 ext4_msg(sb, KERN_WARNING,
4685 "feature flags set on rev 0 fs, "
4686 "running e2fsck is recommended");
4688 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4689 set_opt2(sb, HURD_COMPAT);
4690 if (ext4_has_feature_64bit(sb)) {
4691 ext4_msg(sb, KERN_ERR,
4692 "The Hurd can't support 64-bit file systems");
4697 * ea_inode feature uses l_i_version field which is not
4698 * available in HURD_COMPAT mode.
4700 if (ext4_has_feature_ea_inode(sb)) {
4701 ext4_msg(sb, KERN_ERR,
4702 "ea_inode feature is not supported for Hurd");
4707 if (IS_EXT2_SB(sb)) {
4708 if (ext2_feature_set_ok(sb))
4709 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4710 "using the ext4 subsystem");
4713 * If we're probing be silent, if this looks like
4714 * it's actually an ext[34] filesystem.
4716 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4718 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4719 "to feature incompatibilities");
4724 if (IS_EXT3_SB(sb)) {
4725 if (ext3_feature_set_ok(sb))
4726 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4727 "using the ext4 subsystem");
4730 * If we're probing be silent, if this looks like
4731 * it's actually an ext4 filesystem.
4733 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4735 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4736 "to feature incompatibilities");
4742 * Check feature flags regardless of the revision level, since we
4743 * previously didn't change the revision level when setting the flags,
4744 * so there is a chance incompat flags are set on a rev 0 filesystem.
4746 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4749 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4750 ext4_msg(sb, KERN_ERR,
4751 "Number of reserved GDT blocks insanely large: %d",
4752 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4756 if (dax_supported(sbi->s_daxdev, sb->s_bdev, blocksize, 0,
4757 bdev_nr_sectors(sb->s_bdev)))
4758 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4760 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4761 if (ext4_has_feature_inline_data(sb)) {
4762 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4763 " that may contain inline data");
4766 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4767 ext4_msg(sb, KERN_ERR,
4768 "DAX unsupported by block device.");
4773 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4774 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4775 es->s_encryption_level);
4779 if (sb->s_blocksize != blocksize) {
4781 * bh must be released before kill_bdev(), otherwise
4782 * it won't be freed and its page also. kill_bdev()
4783 * is called by sb_set_blocksize().
4786 /* Validate the filesystem blocksize */
4787 if (!sb_set_blocksize(sb, blocksize)) {
4788 ext4_msg(sb, KERN_ERR, "bad block size %d",
4794 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4795 offset = do_div(logical_sb_block, blocksize);
4796 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4798 ext4_msg(sb, KERN_ERR,
4799 "Can't read superblock on 2nd try");
4804 es = (struct ext4_super_block *)(bh->b_data + offset);
4806 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4807 ext4_msg(sb, KERN_ERR,
4808 "Magic mismatch, very weird!");
4813 has_huge_files = ext4_has_feature_huge_file(sb);
4814 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4816 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4818 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4819 if (ext4_has_feature_64bit(sb)) {
4820 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4821 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4822 !is_power_of_2(sbi->s_desc_size)) {
4823 ext4_msg(sb, KERN_ERR,
4824 "unsupported descriptor size %lu",
4829 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4831 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4832 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4834 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4835 if (sbi->s_inodes_per_block == 0)
4837 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4838 sbi->s_inodes_per_group > blocksize * 8) {
4839 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4840 sbi->s_inodes_per_group);
4843 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4844 sbi->s_inodes_per_block;
4845 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4847 sbi->s_mount_state = le16_to_cpu(es->s_state);
4848 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4849 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4851 for (i = 0; i < 4; i++)
4852 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4853 sbi->s_def_hash_version = es->s_def_hash_version;
4854 if (ext4_has_feature_dir_index(sb)) {
4855 i = le32_to_cpu(es->s_flags);
4856 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4857 sbi->s_hash_unsigned = 3;
4858 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4859 #ifdef __CHAR_UNSIGNED__
4862 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4863 sbi->s_hash_unsigned = 3;
4867 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4872 /* Handle clustersize */
4873 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4874 if (ext4_has_feature_bigalloc(sb)) {
4875 if (clustersize < blocksize) {
4876 ext4_msg(sb, KERN_ERR,
4877 "cluster size (%d) smaller than "
4878 "block size (%d)", clustersize, blocksize);
4881 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4882 le32_to_cpu(es->s_log_block_size);
4883 sbi->s_clusters_per_group =
4884 le32_to_cpu(es->s_clusters_per_group);
4885 if (sbi->s_clusters_per_group > blocksize * 8) {
4886 ext4_msg(sb, KERN_ERR,
4887 "#clusters per group too big: %lu",
4888 sbi->s_clusters_per_group);
4891 if (sbi->s_blocks_per_group !=
4892 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4893 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4894 "clusters per group (%lu) inconsistent",
4895 sbi->s_blocks_per_group,
4896 sbi->s_clusters_per_group);
4900 if (clustersize != blocksize) {
4901 ext4_msg(sb, KERN_ERR,
4902 "fragment/cluster size (%d) != "
4903 "block size (%d)", clustersize, blocksize);
4906 if (sbi->s_blocks_per_group > blocksize * 8) {
4907 ext4_msg(sb, KERN_ERR,
4908 "#blocks per group too big: %lu",
4909 sbi->s_blocks_per_group);
4912 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4913 sbi->s_cluster_bits = 0;
4915 sbi->s_cluster_ratio = clustersize / blocksize;
4917 /* Do we have standard group size of clustersize * 8 blocks ? */
4918 if (sbi->s_blocks_per_group == clustersize << 3)
4919 set_opt2(sb, STD_GROUP_SIZE);
4922 * Test whether we have more sectors than will fit in sector_t,
4923 * and whether the max offset is addressable by the page cache.
4925 err = generic_check_addressable(sb->s_blocksize_bits,
4926 ext4_blocks_count(es));
4928 ext4_msg(sb, KERN_ERR, "filesystem"
4929 " too large to mount safely on this system");
4933 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4936 /* check blocks count against device size */
4937 blocks_count = sb_bdev_nr_blocks(sb);
4938 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4939 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4940 "exceeds size of device (%llu blocks)",
4941 ext4_blocks_count(es), blocks_count);
4946 * It makes no sense for the first data block to be beyond the end
4947 * of the filesystem.
4949 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4950 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4951 "block %u is beyond end of filesystem (%llu)",
4952 le32_to_cpu(es->s_first_data_block),
4953 ext4_blocks_count(es));
4956 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4957 (sbi->s_cluster_ratio == 1)) {
4958 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4959 "block is 0 with a 1k block and cluster size");
4963 blocks_count = (ext4_blocks_count(es) -
4964 le32_to_cpu(es->s_first_data_block) +
4965 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4966 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4967 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4968 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4969 "(block count %llu, first data block %u, "
4970 "blocks per group %lu)", blocks_count,
4971 ext4_blocks_count(es),
4972 le32_to_cpu(es->s_first_data_block),
4973 EXT4_BLOCKS_PER_GROUP(sb));
4976 sbi->s_groups_count = blocks_count;
4977 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4978 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4979 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4980 le32_to_cpu(es->s_inodes_count)) {
4981 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4982 le32_to_cpu(es->s_inodes_count),
4983 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4987 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4988 EXT4_DESC_PER_BLOCK(sb);
4989 if (ext4_has_feature_meta_bg(sb)) {
4990 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4991 ext4_msg(sb, KERN_WARNING,
4992 "first meta block group too large: %u "
4993 "(group descriptor block count %u)",
4994 le32_to_cpu(es->s_first_meta_bg), db_count);
4998 rcu_assign_pointer(sbi->s_group_desc,
4999 kvmalloc_array(db_count,
5000 sizeof(struct buffer_head *),
5002 if (sbi->s_group_desc == NULL) {
5003 ext4_msg(sb, KERN_ERR, "not enough memory");
5008 bgl_lock_init(sbi->s_blockgroup_lock);
5010 /* Pre-read the descriptors into the buffer cache */
5011 for (i = 0; i < db_count; i++) {
5012 block = descriptor_loc(sb, logical_sb_block, i);
5013 ext4_sb_breadahead_unmovable(sb, block);
5016 for (i = 0; i < db_count; i++) {
5017 struct buffer_head *bh;
5019 block = descriptor_loc(sb, logical_sb_block, i);
5020 bh = ext4_sb_bread_unmovable(sb, block);
5022 ext4_msg(sb, KERN_ERR,
5023 "can't read group descriptor %d", i);
5029 rcu_dereference(sbi->s_group_desc)[i] = bh;
5032 sbi->s_gdb_count = db_count;
5033 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
5034 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
5035 ret = -EFSCORRUPTED;
5039 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5040 spin_lock_init(&sbi->s_error_lock);
5041 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5043 /* Register extent status tree shrinker */
5044 if (ext4_es_register_shrinker(sbi))
5047 sbi->s_stripe = ext4_get_stripe_size(sbi);
5048 sbi->s_extent_max_zeroout_kb = 32;
5051 * set up enough so that it can read an inode
5053 sb->s_op = &ext4_sops;
5054 sb->s_export_op = &ext4_export_ops;
5055 sb->s_xattr = ext4_xattr_handlers;
5056 #ifdef CONFIG_FS_ENCRYPTION
5057 sb->s_cop = &ext4_cryptops;
5059 #ifdef CONFIG_FS_VERITY
5060 sb->s_vop = &ext4_verityops;
5063 sb->dq_op = &ext4_quota_operations;
5064 if (ext4_has_feature_quota(sb))
5065 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5067 sb->s_qcop = &ext4_qctl_operations;
5068 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5070 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5072 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5073 mutex_init(&sbi->s_orphan_lock);
5075 /* Initialize fast commit stuff */
5076 atomic_set(&sbi->s_fc_subtid, 0);
5077 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
5078 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
5079 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
5080 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
5081 sbi->s_fc_bytes = 0;
5082 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
5083 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
5084 spin_lock_init(&sbi->s_fc_lock);
5085 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
5086 sbi->s_fc_replay_state.fc_regions = NULL;
5087 sbi->s_fc_replay_state.fc_regions_size = 0;
5088 sbi->s_fc_replay_state.fc_regions_used = 0;
5089 sbi->s_fc_replay_state.fc_regions_valid = 0;
5090 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
5091 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
5092 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
5096 needs_recovery = (es->s_last_orphan != 0 ||
5097 ext4_has_feature_orphan_present(sb) ||
5098 ext4_has_feature_journal_needs_recovery(sb));
5100 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5101 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5102 goto failed_mount3a;
5105 * The first inode we look at is the journal inode. Don't try
5106 * root first: it may be modified in the journal!
5108 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5109 err = ext4_load_journal(sb, es, ctx->journal_devnum);
5111 goto failed_mount3a;
5112 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5113 ext4_has_feature_journal_needs_recovery(sb)) {
5114 ext4_msg(sb, KERN_ERR, "required journal recovery "
5115 "suppressed and not mounted read-only");
5116 goto failed_mount_wq;
5118 /* Nojournal mode, all journal mount options are illegal */
5119 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5120 ext4_msg(sb, KERN_ERR, "can't mount with "
5121 "journal_checksum, fs mounted w/o journal");
5122 goto failed_mount_wq;
5124 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5125 ext4_msg(sb, KERN_ERR, "can't mount with "
5126 "journal_async_commit, fs mounted w/o journal");
5127 goto failed_mount_wq;
5129 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5130 ext4_msg(sb, KERN_ERR, "can't mount with "
5131 "commit=%lu, fs mounted w/o journal",
5132 sbi->s_commit_interval / HZ);
5133 goto failed_mount_wq;
5135 if (EXT4_MOUNT_DATA_FLAGS &
5136 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5137 ext4_msg(sb, KERN_ERR, "can't mount with "
5138 "data=, fs mounted w/o journal");
5139 goto failed_mount_wq;
5141 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5142 clear_opt(sb, JOURNAL_CHECKSUM);
5143 clear_opt(sb, DATA_FLAGS);
5144 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5145 sbi->s_journal = NULL;
5150 if (ext4_has_feature_64bit(sb) &&
5151 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5152 JBD2_FEATURE_INCOMPAT_64BIT)) {
5153 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
5154 goto failed_mount_wq;
5157 if (!set_journal_csum_feature_set(sb)) {
5158 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
5160 goto failed_mount_wq;
5163 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
5164 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5165 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
5166 ext4_msg(sb, KERN_ERR,
5167 "Failed to set fast commit journal feature");
5168 goto failed_mount_wq;
5171 /* We have now updated the journal if required, so we can
5172 * validate the data journaling mode. */
5173 switch (test_opt(sb, DATA_FLAGS)) {
5175 /* No mode set, assume a default based on the journal
5176 * capabilities: ORDERED_DATA if the journal can
5177 * cope, else JOURNAL_DATA
5179 if (jbd2_journal_check_available_features
5180 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5181 set_opt(sb, ORDERED_DATA);
5182 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
5184 set_opt(sb, JOURNAL_DATA);
5185 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
5189 case EXT4_MOUNT_ORDERED_DATA:
5190 case EXT4_MOUNT_WRITEBACK_DATA:
5191 if (!jbd2_journal_check_available_features
5192 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
5193 ext4_msg(sb, KERN_ERR, "Journal does not support "
5194 "requested data journaling mode");
5195 goto failed_mount_wq;
5202 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
5203 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5204 ext4_msg(sb, KERN_ERR, "can't mount with "
5205 "journal_async_commit in data=ordered mode");
5206 goto failed_mount_wq;
5209 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
5211 sbi->s_journal->j_submit_inode_data_buffers =
5212 ext4_journal_submit_inode_data_buffers;
5213 sbi->s_journal->j_finish_inode_data_buffers =
5214 ext4_journal_finish_inode_data_buffers;
5217 if (!test_opt(sb, NO_MBCACHE)) {
5218 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5219 if (!sbi->s_ea_block_cache) {
5220 ext4_msg(sb, KERN_ERR,
5221 "Failed to create ea_block_cache");
5222 goto failed_mount_wq;
5225 if (ext4_has_feature_ea_inode(sb)) {
5226 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5227 if (!sbi->s_ea_inode_cache) {
5228 ext4_msg(sb, KERN_ERR,
5229 "Failed to create ea_inode_cache");
5230 goto failed_mount_wq;
5235 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
5236 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5237 goto failed_mount_wq;
5240 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
5241 !ext4_has_feature_encrypt(sb)) {
5242 ext4_set_feature_encrypt(sb);
5243 ext4_commit_super(sb);
5247 * Get the # of file system overhead blocks from the
5248 * superblock if present.
5250 if (es->s_overhead_clusters)
5251 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5253 err = ext4_calculate_overhead(sb);
5255 goto failed_mount_wq;
5259 * The maximum number of concurrent works can be high and
5260 * concurrency isn't really necessary. Limit it to 1.
5262 EXT4_SB(sb)->rsv_conversion_wq =
5263 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5264 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5265 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5271 * The jbd2_journal_load will have done any necessary log recovery,
5272 * so we can safely mount the rest of the filesystem now.
5275 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5277 ext4_msg(sb, KERN_ERR, "get root inode failed");
5278 ret = PTR_ERR(root);
5282 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5283 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5288 sb->s_root = d_make_root(root);
5290 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5295 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5296 if (ret == -EROFS) {
5297 sb->s_flags |= SB_RDONLY;
5300 goto failed_mount4a;
5302 ext4_set_resv_clusters(sb);
5304 if (test_opt(sb, BLOCK_VALIDITY)) {
5305 err = ext4_setup_system_zone(sb);
5307 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5309 goto failed_mount4a;
5312 ext4_fc_replay_cleanup(sb);
5317 * Enable optimize_scan if number of groups is > threshold. This can be
5318 * turned off by passing "mb_optimize_scan=0". This can also be
5319 * turned on forcefully by passing "mb_optimize_scan=1".
5321 if (ctx->mb_optimize_scan == 1)
5322 set_opt2(sb, MB_OPTIMIZE_SCAN);
5323 else if (ctx->mb_optimize_scan == 0)
5324 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5325 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5326 set_opt2(sb, MB_OPTIMIZE_SCAN);
5328 err = ext4_mb_init(sb);
5330 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5336 * We can only set up the journal commit callback once
5337 * mballoc is initialized
5340 sbi->s_journal->j_commit_callback =
5341 ext4_journal_commit_callback;
5343 block = ext4_count_free_clusters(sb);
5344 ext4_free_blocks_count_set(sbi->s_es,
5345 EXT4_C2B(sbi, block));
5346 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5349 unsigned long freei = ext4_count_free_inodes(sb);
5350 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5351 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5355 * Update the checksum after updating free space/inode
5356 * counters. Otherwise the superblock can have an incorrect
5357 * checksum in the buffer cache until it is written out and
5358 * e2fsprogs programs trying to open a file system immediately
5359 * after it is mounted can fail.
5361 ext4_superblock_csum_set(sb);
5363 err = percpu_counter_init(&sbi->s_dirs_counter,
5364 ext4_count_dirs(sb), GFP_KERNEL);
5366 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5369 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5372 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5375 ext4_msg(sb, KERN_ERR, "insufficient memory");
5379 if (ext4_has_feature_flex_bg(sb))
5380 if (!ext4_fill_flex_info(sb)) {
5381 ext4_msg(sb, KERN_ERR,
5382 "unable to initialize "
5383 "flex_bg meta info!");
5388 err = ext4_register_li_request(sb, first_not_zeroed);
5392 err = ext4_register_sysfs(sb);
5396 err = ext4_init_orphan_info(sb);
5400 /* Enable quota usage during mount. */
5401 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5402 err = ext4_enable_quotas(sb);
5406 #endif /* CONFIG_QUOTA */
5409 * Save the original bdev mapping's wb_err value which could be
5410 * used to detect the metadata async write error.
5412 spin_lock_init(&sbi->s_bdev_wb_lock);
5413 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5414 &sbi->s_bdev_wb_err);
5415 sb->s_bdev->bd_super = sb;
5416 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5417 ext4_orphan_cleanup(sb, es);
5418 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5419 if (needs_recovery) {
5420 ext4_msg(sb, KERN_INFO, "recovery complete");
5421 err = ext4_mark_recovery_complete(sb, es);
5426 if (test_opt(sb, DISCARD)) {
5427 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5428 if (!blk_queue_discard(q))
5429 ext4_msg(sb, KERN_WARNING,
5430 "mounting with \"discard\" option, but "
5431 "the device does not support discard");
5434 if (es->s_error_count)
5435 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5437 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5438 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5439 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5440 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5441 atomic_set(&sbi->s_warning_count, 0);
5442 atomic_set(&sbi->s_msg_count, 0);
5448 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5452 ext4_release_orphan_info(sb);
5454 ext4_unregister_sysfs(sb);
5455 kobject_put(&sbi->s_kobj);
5457 ext4_unregister_li_request(sb);
5459 ext4_mb_release(sb);
5461 flex_groups = rcu_dereference(sbi->s_flex_groups);
5463 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5464 kvfree(flex_groups[i]);
5465 kvfree(flex_groups);
5468 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5469 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5470 percpu_counter_destroy(&sbi->s_dirs_counter);
5471 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5472 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5473 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5475 ext4_ext_release(sb);
5476 ext4_release_system_zone(sb);
5481 ext4_msg(sb, KERN_ERR, "mount failed");
5482 if (EXT4_SB(sb)->rsv_conversion_wq)
5483 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5485 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5486 sbi->s_ea_inode_cache = NULL;
5488 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5489 sbi->s_ea_block_cache = NULL;
5491 if (sbi->s_journal) {
5492 /* flush s_error_work before journal destroy. */
5493 flush_work(&sbi->s_error_work);
5494 jbd2_journal_destroy(sbi->s_journal);
5495 sbi->s_journal = NULL;
5498 ext4_es_unregister_shrinker(sbi);
5500 /* flush s_error_work before sbi destroy */
5501 flush_work(&sbi->s_error_work);
5502 del_timer_sync(&sbi->s_err_report);
5503 ext4_stop_mmpd(sbi);
5506 group_desc = rcu_dereference(sbi->s_group_desc);
5507 for (i = 0; i < db_count; i++)
5508 brelse(group_desc[i]);
5512 if (sbi->s_chksum_driver)
5513 crypto_free_shash(sbi->s_chksum_driver);
5515 #ifdef CONFIG_UNICODE
5516 utf8_unload(sb->s_encoding);
5520 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5521 kfree(get_qf_name(sb, sbi, i));
5523 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5524 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5526 ext4_blkdev_remove(sbi);
5528 sb->s_fs_info = NULL;
5529 return err ? err : ret;
5532 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5534 struct ext4_fs_context *ctx = fc->fs_private;
5535 struct ext4_sb_info *sbi;
5539 sbi = ext4_alloc_sbi(sb);
5543 fc->s_fs_info = sbi;
5545 /* Cleanup superblock name */
5546 strreplace(sb->s_id, '/', '!');
5548 sbi->s_sb_block = 1; /* Default super block location */
5549 if (ctx->spec & EXT4_SPEC_s_sb_block)
5550 sbi->s_sb_block = ctx->s_sb_block;
5552 ret = __ext4_fill_super(fc, sb);
5556 if (sbi->s_journal) {
5557 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5558 descr = " journalled data mode";
5559 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5560 descr = " ordered data mode";
5562 descr = " writeback data mode";
5564 descr = "out journal";
5566 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5567 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5568 "Quota mode: %s.", descr, ext4_quota_mode(sb));
5574 fc->s_fs_info = NULL;
5578 static int ext4_get_tree(struct fs_context *fc)
5580 return get_tree_bdev(fc, ext4_fill_super);
5584 * Setup any per-fs journal parameters now. We'll do this both on
5585 * initial mount, once the journal has been initialised but before we've
5586 * done any recovery; and again on any subsequent remount.
5588 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5590 struct ext4_sb_info *sbi = EXT4_SB(sb);
5592 journal->j_commit_interval = sbi->s_commit_interval;
5593 journal->j_min_batch_time = sbi->s_min_batch_time;
5594 journal->j_max_batch_time = sbi->s_max_batch_time;
5595 ext4_fc_init(sb, journal);
5597 write_lock(&journal->j_state_lock);
5598 if (test_opt(sb, BARRIER))
5599 journal->j_flags |= JBD2_BARRIER;
5601 journal->j_flags &= ~JBD2_BARRIER;
5602 if (test_opt(sb, DATA_ERR_ABORT))
5603 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5605 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5606 write_unlock(&journal->j_state_lock);
5609 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5610 unsigned int journal_inum)
5612 struct inode *journal_inode;
5615 * Test for the existence of a valid inode on disk. Bad things
5616 * happen if we iget() an unused inode, as the subsequent iput()
5617 * will try to delete it.
5619 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5620 if (IS_ERR(journal_inode)) {
5621 ext4_msg(sb, KERN_ERR, "no journal found");
5624 if (!journal_inode->i_nlink) {
5625 make_bad_inode(journal_inode);
5626 iput(journal_inode);
5627 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5631 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5632 journal_inode, journal_inode->i_size);
5633 if (!S_ISREG(journal_inode->i_mode)) {
5634 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5635 iput(journal_inode);
5638 return journal_inode;
5641 static journal_t *ext4_get_journal(struct super_block *sb,
5642 unsigned int journal_inum)
5644 struct inode *journal_inode;
5647 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5650 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5654 journal = jbd2_journal_init_inode(journal_inode);
5656 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5657 iput(journal_inode);
5660 journal->j_private = sb;
5661 ext4_init_journal_params(sb, journal);
5665 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5668 struct buffer_head *bh;
5672 int hblock, blocksize;
5673 ext4_fsblk_t sb_block;
5674 unsigned long offset;
5675 struct ext4_super_block *es;
5676 struct block_device *bdev;
5678 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5681 bdev = ext4_blkdev_get(j_dev, sb);
5685 blocksize = sb->s_blocksize;
5686 hblock = bdev_logical_block_size(bdev);
5687 if (blocksize < hblock) {
5688 ext4_msg(sb, KERN_ERR,
5689 "blocksize too small for journal device");
5693 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5694 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5695 set_blocksize(bdev, blocksize);
5696 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5697 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5698 "external journal");
5702 es = (struct ext4_super_block *) (bh->b_data + offset);
5703 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5704 !(le32_to_cpu(es->s_feature_incompat) &
5705 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5706 ext4_msg(sb, KERN_ERR, "external journal has "
5712 if ((le32_to_cpu(es->s_feature_ro_compat) &
5713 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5714 es->s_checksum != ext4_superblock_csum(sb, es)) {
5715 ext4_msg(sb, KERN_ERR, "external journal has "
5716 "corrupt superblock");
5721 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5722 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5727 len = ext4_blocks_count(es);
5728 start = sb_block + 1;
5729 brelse(bh); /* we're done with the superblock */
5731 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5732 start, len, blocksize);
5734 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5737 journal->j_private = sb;
5738 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5739 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5742 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5743 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5744 "user (unsupported) - %d",
5745 be32_to_cpu(journal->j_superblock->s_nr_users));
5748 EXT4_SB(sb)->s_journal_bdev = bdev;
5749 ext4_init_journal_params(sb, journal);
5753 jbd2_journal_destroy(journal);
5755 ext4_blkdev_put(bdev);
5759 static int ext4_load_journal(struct super_block *sb,
5760 struct ext4_super_block *es,
5761 unsigned long journal_devnum)
5764 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5767 int really_read_only;
5770 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5771 return -EFSCORRUPTED;
5773 if (journal_devnum &&
5774 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5775 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5776 "numbers have changed");
5777 journal_dev = new_decode_dev(journal_devnum);
5779 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5781 if (journal_inum && journal_dev) {
5782 ext4_msg(sb, KERN_ERR,
5783 "filesystem has both journal inode and journal device!");
5788 journal = ext4_get_journal(sb, journal_inum);
5792 journal = ext4_get_dev_journal(sb, journal_dev);
5797 journal_dev_ro = bdev_read_only(journal->j_dev);
5798 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5800 if (journal_dev_ro && !sb_rdonly(sb)) {
5801 ext4_msg(sb, KERN_ERR,
5802 "journal device read-only, try mounting with '-o ro'");
5808 * Are we loading a blank journal or performing recovery after a
5809 * crash? For recovery, we need to check in advance whether we
5810 * can get read-write access to the device.
5812 if (ext4_has_feature_journal_needs_recovery(sb)) {
5813 if (sb_rdonly(sb)) {
5814 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5815 "required on readonly filesystem");
5816 if (really_read_only) {
5817 ext4_msg(sb, KERN_ERR, "write access "
5818 "unavailable, cannot proceed "
5819 "(try mounting with noload)");
5823 ext4_msg(sb, KERN_INFO, "write access will "
5824 "be enabled during recovery");
5828 if (!(journal->j_flags & JBD2_BARRIER))
5829 ext4_msg(sb, KERN_INFO, "barriers disabled");
5831 if (!ext4_has_feature_journal_needs_recovery(sb))
5832 err = jbd2_journal_wipe(journal, !really_read_only);
5834 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5836 memcpy(save, ((char *) es) +
5837 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5838 err = jbd2_journal_load(journal);
5840 memcpy(((char *) es) + EXT4_S_ERR_START,
5841 save, EXT4_S_ERR_LEN);
5846 ext4_msg(sb, KERN_ERR, "error loading journal");
5850 EXT4_SB(sb)->s_journal = journal;
5851 err = ext4_clear_journal_err(sb, es);
5853 EXT4_SB(sb)->s_journal = NULL;
5854 jbd2_journal_destroy(journal);
5858 if (!really_read_only && journal_devnum &&
5859 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5860 es->s_journal_dev = cpu_to_le32(journal_devnum);
5862 /* Make sure we flush the recovery flag to disk. */
5863 ext4_commit_super(sb);
5869 jbd2_journal_destroy(journal);
5873 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5874 static void ext4_update_super(struct super_block *sb)
5876 struct ext4_sb_info *sbi = EXT4_SB(sb);
5877 struct ext4_super_block *es = sbi->s_es;
5878 struct buffer_head *sbh = sbi->s_sbh;
5882 * If the file system is mounted read-only, don't update the
5883 * superblock write time. This avoids updating the superblock
5884 * write time when we are mounting the root file system
5885 * read/only but we need to replay the journal; at that point,
5886 * for people who are east of GMT and who make their clock
5887 * tick in localtime for Windows bug-for-bug compatibility,
5888 * the clock is set in the future, and this will cause e2fsck
5889 * to complain and force a full file system check.
5891 if (!(sb->s_flags & SB_RDONLY))
5892 ext4_update_tstamp(es, s_wtime);
5893 es->s_kbytes_written =
5894 cpu_to_le64(sbi->s_kbytes_written +
5895 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5896 sbi->s_sectors_written_start) >> 1));
5897 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5898 ext4_free_blocks_count_set(es,
5899 EXT4_C2B(sbi, percpu_counter_sum_positive(
5900 &sbi->s_freeclusters_counter)));
5901 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5902 es->s_free_inodes_count =
5903 cpu_to_le32(percpu_counter_sum_positive(
5904 &sbi->s_freeinodes_counter));
5905 /* Copy error information to the on-disk superblock */
5906 spin_lock(&sbi->s_error_lock);
5907 if (sbi->s_add_error_count > 0) {
5908 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5909 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5910 __ext4_update_tstamp(&es->s_first_error_time,
5911 &es->s_first_error_time_hi,
5912 sbi->s_first_error_time);
5913 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5914 sizeof(es->s_first_error_func));
5915 es->s_first_error_line =
5916 cpu_to_le32(sbi->s_first_error_line);
5917 es->s_first_error_ino =
5918 cpu_to_le32(sbi->s_first_error_ino);
5919 es->s_first_error_block =
5920 cpu_to_le64(sbi->s_first_error_block);
5921 es->s_first_error_errcode =
5922 ext4_errno_to_code(sbi->s_first_error_code);
5924 __ext4_update_tstamp(&es->s_last_error_time,
5925 &es->s_last_error_time_hi,
5926 sbi->s_last_error_time);
5927 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5928 sizeof(es->s_last_error_func));
5929 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5930 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5931 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5932 es->s_last_error_errcode =
5933 ext4_errno_to_code(sbi->s_last_error_code);
5935 * Start the daily error reporting function if it hasn't been
5938 if (!es->s_error_count)
5939 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5940 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5941 sbi->s_add_error_count = 0;
5943 spin_unlock(&sbi->s_error_lock);
5945 ext4_superblock_csum_set(sb);
5949 static int ext4_commit_super(struct super_block *sb)
5951 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5956 if (block_device_ejected(sb))
5959 ext4_update_super(sb);
5961 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5963 * Oh, dear. A previous attempt to write the
5964 * superblock failed. This could happen because the
5965 * USB device was yanked out. Or it could happen to
5966 * be a transient write error and maybe the block will
5967 * be remapped. Nothing we can do but to retry the
5968 * write and hope for the best.
5970 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5971 "superblock detected");
5972 clear_buffer_write_io_error(sbh);
5973 set_buffer_uptodate(sbh);
5975 BUFFER_TRACE(sbh, "marking dirty");
5976 mark_buffer_dirty(sbh);
5977 error = __sync_dirty_buffer(sbh,
5978 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5979 if (buffer_write_io_error(sbh)) {
5980 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5982 clear_buffer_write_io_error(sbh);
5983 set_buffer_uptodate(sbh);
5989 * Have we just finished recovery? If so, and if we are mounting (or
5990 * remounting) the filesystem readonly, then we will end up with a
5991 * consistent fs on disk. Record that fact.
5993 static int ext4_mark_recovery_complete(struct super_block *sb,
5994 struct ext4_super_block *es)
5997 journal_t *journal = EXT4_SB(sb)->s_journal;
5999 if (!ext4_has_feature_journal(sb)) {
6000 if (journal != NULL) {
6001 ext4_error(sb, "Journal got removed while the fs was "
6003 return -EFSCORRUPTED;
6007 jbd2_journal_lock_updates(journal);
6008 err = jbd2_journal_flush(journal, 0);
6012 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6013 ext4_has_feature_orphan_present(sb))) {
6014 if (!ext4_orphan_file_empty(sb)) {
6015 ext4_error(sb, "Orphan file not empty on read-only fs.");
6016 err = -EFSCORRUPTED;
6019 ext4_clear_feature_journal_needs_recovery(sb);
6020 ext4_clear_feature_orphan_present(sb);
6021 ext4_commit_super(sb);
6024 jbd2_journal_unlock_updates(journal);
6029 * If we are mounting (or read-write remounting) a filesystem whose journal
6030 * has recorded an error from a previous lifetime, move that error to the
6031 * main filesystem now.
6033 static int ext4_clear_journal_err(struct super_block *sb,
6034 struct ext4_super_block *es)
6040 if (!ext4_has_feature_journal(sb)) {
6041 ext4_error(sb, "Journal got removed while the fs was mounted!");
6042 return -EFSCORRUPTED;
6045 journal = EXT4_SB(sb)->s_journal;
6048 * Now check for any error status which may have been recorded in the
6049 * journal by a prior ext4_error() or ext4_abort()
6052 j_errno = jbd2_journal_errno(journal);
6056 errstr = ext4_decode_error(sb, j_errno, nbuf);
6057 ext4_warning(sb, "Filesystem error recorded "
6058 "from previous mount: %s", errstr);
6059 ext4_warning(sb, "Marking fs in need of filesystem check.");
6061 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6062 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6063 ext4_commit_super(sb);
6065 jbd2_journal_clear_err(journal);
6066 jbd2_journal_update_sb_errno(journal);
6072 * Force the running and committing transactions to commit,
6073 * and wait on the commit.
6075 int ext4_force_commit(struct super_block *sb)
6082 journal = EXT4_SB(sb)->s_journal;
6083 return ext4_journal_force_commit(journal);
6086 static int ext4_sync_fs(struct super_block *sb, int wait)
6090 bool needs_barrier = false;
6091 struct ext4_sb_info *sbi = EXT4_SB(sb);
6093 if (unlikely(ext4_forced_shutdown(sbi)))
6096 trace_ext4_sync_fs(sb, wait);
6097 flush_workqueue(sbi->rsv_conversion_wq);
6099 * Writeback quota in non-journalled quota case - journalled quota has
6102 dquot_writeback_dquots(sb, -1);
6104 * Data writeback is possible w/o journal transaction, so barrier must
6105 * being sent at the end of the function. But we can skip it if
6106 * transaction_commit will do it for us.
6108 if (sbi->s_journal) {
6109 target = jbd2_get_latest_transaction(sbi->s_journal);
6110 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6111 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6112 needs_barrier = true;
6114 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6116 ret = jbd2_log_wait_commit(sbi->s_journal,
6119 } else if (wait && test_opt(sb, BARRIER))
6120 needs_barrier = true;
6121 if (needs_barrier) {
6123 err = blkdev_issue_flush(sb->s_bdev);
6132 * LVM calls this function before a (read-only) snapshot is created. This
6133 * gives us a chance to flush the journal completely and mark the fs clean.
6135 * Note that only this function cannot bring a filesystem to be in a clean
6136 * state independently. It relies on upper layer to stop all data & metadata
6139 static int ext4_freeze(struct super_block *sb)
6147 journal = EXT4_SB(sb)->s_journal;
6150 /* Now we set up the journal barrier. */
6151 jbd2_journal_lock_updates(journal);
6154 * Don't clear the needs_recovery flag if we failed to
6155 * flush the journal.
6157 error = jbd2_journal_flush(journal, 0);
6161 /* Journal blocked and flushed, clear needs_recovery flag. */
6162 ext4_clear_feature_journal_needs_recovery(sb);
6163 if (ext4_orphan_file_empty(sb))
6164 ext4_clear_feature_orphan_present(sb);
6167 error = ext4_commit_super(sb);
6170 /* we rely on upper layer to stop further updates */
6171 jbd2_journal_unlock_updates(journal);
6176 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6177 * flag here, even though the filesystem is not technically dirty yet.
6179 static int ext4_unfreeze(struct super_block *sb)
6181 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6184 if (EXT4_SB(sb)->s_journal) {
6185 /* Reset the needs_recovery flag before the fs is unlocked. */
6186 ext4_set_feature_journal_needs_recovery(sb);
6187 if (ext4_has_feature_orphan_file(sb))
6188 ext4_set_feature_orphan_present(sb);
6191 ext4_commit_super(sb);
6196 * Structure to save mount options for ext4_remount's benefit
6198 struct ext4_mount_options {
6199 unsigned long s_mount_opt;
6200 unsigned long s_mount_opt2;
6203 unsigned long s_commit_interval;
6204 u32 s_min_batch_time, s_max_batch_time;
6207 char *s_qf_names[EXT4_MAXQUOTAS];
6211 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6213 struct ext4_fs_context *ctx = fc->fs_private;
6214 struct ext4_super_block *es;
6215 struct ext4_sb_info *sbi = EXT4_SB(sb);
6216 unsigned long old_sb_flags;
6217 struct ext4_mount_options old_opts;
6221 int enable_quota = 0;
6223 char *to_free[EXT4_MAXQUOTAS];
6226 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6228 /* Store the original options */
6229 old_sb_flags = sb->s_flags;
6230 old_opts.s_mount_opt = sbi->s_mount_opt;
6231 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6232 old_opts.s_resuid = sbi->s_resuid;
6233 old_opts.s_resgid = sbi->s_resgid;
6234 old_opts.s_commit_interval = sbi->s_commit_interval;
6235 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6236 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6238 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6239 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6240 if (sbi->s_qf_names[i]) {
6241 char *qf_name = get_qf_name(sb, sbi, i);
6243 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6244 if (!old_opts.s_qf_names[i]) {
6245 for (j = 0; j < i; j++)
6246 kfree(old_opts.s_qf_names[j]);
6250 old_opts.s_qf_names[i] = NULL;
6252 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6253 ctx->journal_ioprio =
6254 sbi->s_journal->j_task->io_context->ioprio;
6256 ext4_apply_options(fc, sb);
6258 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6259 test_opt(sb, JOURNAL_CHECKSUM)) {
6260 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6261 "during remount not supported; ignoring");
6262 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6265 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6266 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6267 ext4_msg(sb, KERN_ERR, "can't mount with "
6268 "both data=journal and delalloc");
6272 if (test_opt(sb, DIOREAD_NOLOCK)) {
6273 ext4_msg(sb, KERN_ERR, "can't mount with "
6274 "both data=journal and dioread_nolock");
6278 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6279 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6280 ext4_msg(sb, KERN_ERR, "can't mount with "
6281 "journal_async_commit in data=ordered mode");
6287 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6288 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6293 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6294 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6296 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6297 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6301 if (sbi->s_journal) {
6302 ext4_init_journal_params(sb, sbi->s_journal);
6303 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6306 /* Flush outstanding errors before changing fs state */
6307 flush_work(&sbi->s_error_work);
6309 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6310 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6315 if (fc->sb_flags & SB_RDONLY) {
6316 err = sync_filesystem(sb);
6319 err = dquot_suspend(sb, -1);
6324 * First of all, the unconditional stuff we have to do
6325 * to disable replay of the journal when we next remount
6327 sb->s_flags |= SB_RDONLY;
6330 * OK, test if we are remounting a valid rw partition
6331 * readonly, and if so set the rdonly flag and then
6332 * mark the partition as valid again.
6334 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6335 (sbi->s_mount_state & EXT4_VALID_FS))
6336 es->s_state = cpu_to_le16(sbi->s_mount_state);
6338 if (sbi->s_journal) {
6340 * We let remount-ro finish even if marking fs
6341 * as clean failed...
6343 ext4_mark_recovery_complete(sb, es);
6346 /* Make sure we can mount this feature set readwrite */
6347 if (ext4_has_feature_readonly(sb) ||
6348 !ext4_feature_set_ok(sb, 0)) {
6353 * Make sure the group descriptor checksums
6354 * are sane. If they aren't, refuse to remount r/w.
6356 for (g = 0; g < sbi->s_groups_count; g++) {
6357 struct ext4_group_desc *gdp =
6358 ext4_get_group_desc(sb, g, NULL);
6360 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6361 ext4_msg(sb, KERN_ERR,
6362 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6363 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6364 le16_to_cpu(gdp->bg_checksum));
6371 * If we have an unprocessed orphan list hanging
6372 * around from a previously readonly bdev mount,
6373 * require a full umount/remount for now.
6375 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6376 ext4_msg(sb, KERN_WARNING, "Couldn't "
6377 "remount RDWR because of unprocessed "
6378 "orphan inode list. Please "
6379 "umount/remount instead");
6385 * Mounting a RDONLY partition read-write, so reread
6386 * and store the current valid flag. (It may have
6387 * been changed by e2fsck since we originally mounted
6390 if (sbi->s_journal) {
6391 err = ext4_clear_journal_err(sb, es);
6395 sbi->s_mount_state = le16_to_cpu(es->s_state);
6397 err = ext4_setup_super(sb, es, 0);
6401 sb->s_flags &= ~SB_RDONLY;
6402 if (ext4_has_feature_mmp(sb))
6403 if (ext4_multi_mount_protect(sb,
6404 le64_to_cpu(es->s_mmp_block))) {
6415 * Reinitialize lazy itable initialization thread based on
6418 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6419 ext4_unregister_li_request(sb);
6421 ext4_group_t first_not_zeroed;
6422 first_not_zeroed = ext4_has_uninit_itable(sb);
6423 ext4_register_li_request(sb, first_not_zeroed);
6427 * Handle creation of system zone data early because it can fail.
6428 * Releasing of existing data is done when we are sure remount will
6431 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6432 err = ext4_setup_system_zone(sb);
6437 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6438 err = ext4_commit_super(sb);
6444 /* Release old quota file names */
6445 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6446 kfree(old_opts.s_qf_names[i]);
6448 if (sb_any_quota_suspended(sb))
6449 dquot_resume(sb, -1);
6450 else if (ext4_has_feature_quota(sb)) {
6451 err = ext4_enable_quotas(sb);
6457 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6458 ext4_release_system_zone(sb);
6460 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6461 ext4_stop_mmpd(sbi);
6466 sb->s_flags = old_sb_flags;
6467 sbi->s_mount_opt = old_opts.s_mount_opt;
6468 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6469 sbi->s_resuid = old_opts.s_resuid;
6470 sbi->s_resgid = old_opts.s_resgid;
6471 sbi->s_commit_interval = old_opts.s_commit_interval;
6472 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6473 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6474 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6475 ext4_release_system_zone(sb);
6477 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6478 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6479 to_free[i] = get_qf_name(sb, sbi, i);
6480 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6483 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6486 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6487 ext4_stop_mmpd(sbi);
6491 static int ext4_reconfigure(struct fs_context *fc)
6493 struct super_block *sb = fc->root->d_sb;
6496 fc->s_fs_info = EXT4_SB(sb);
6498 ret = ext4_check_opt_consistency(fc, sb);
6502 ret = __ext4_remount(fc, sb);
6506 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.",
6507 ext4_quota_mode(sb));
6513 static int ext4_statfs_project(struct super_block *sb,
6514 kprojid_t projid, struct kstatfs *buf)
6517 struct dquot *dquot;
6521 qid = make_kqid_projid(projid);
6522 dquot = dqget(sb, qid);
6524 return PTR_ERR(dquot);
6525 spin_lock(&dquot->dq_dqb_lock);
6527 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6528 dquot->dq_dqb.dqb_bhardlimit);
6529 limit >>= sb->s_blocksize_bits;
6531 if (limit && buf->f_blocks > limit) {
6532 curblock = (dquot->dq_dqb.dqb_curspace +
6533 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6534 buf->f_blocks = limit;
6535 buf->f_bfree = buf->f_bavail =
6536 (buf->f_blocks > curblock) ?
6537 (buf->f_blocks - curblock) : 0;
6540 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6541 dquot->dq_dqb.dqb_ihardlimit);
6542 if (limit && buf->f_files > limit) {
6543 buf->f_files = limit;
6545 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6546 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6549 spin_unlock(&dquot->dq_dqb_lock);
6555 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6557 struct super_block *sb = dentry->d_sb;
6558 struct ext4_sb_info *sbi = EXT4_SB(sb);
6559 struct ext4_super_block *es = sbi->s_es;
6560 ext4_fsblk_t overhead = 0, resv_blocks;
6562 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6564 if (!test_opt(sb, MINIX_DF))
6565 overhead = sbi->s_overhead;
6567 buf->f_type = EXT4_SUPER_MAGIC;
6568 buf->f_bsize = sb->s_blocksize;
6569 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6570 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6571 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6572 /* prevent underflow in case that few free space is available */
6573 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6574 buf->f_bavail = buf->f_bfree -
6575 (ext4_r_blocks_count(es) + resv_blocks);
6576 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6578 buf->f_files = le32_to_cpu(es->s_inodes_count);
6579 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6580 buf->f_namelen = EXT4_NAME_LEN;
6581 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6584 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6585 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6586 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6595 * Helper functions so that transaction is started before we acquire dqio_sem
6596 * to keep correct lock ordering of transaction > dqio_sem
6598 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6600 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6603 static int ext4_write_dquot(struct dquot *dquot)
6607 struct inode *inode;
6609 inode = dquot_to_inode(dquot);
6610 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6611 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6613 return PTR_ERR(handle);
6614 ret = dquot_commit(dquot);
6615 err = ext4_journal_stop(handle);
6621 static int ext4_acquire_dquot(struct dquot *dquot)
6626 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6627 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6629 return PTR_ERR(handle);
6630 ret = dquot_acquire(dquot);
6631 err = ext4_journal_stop(handle);
6637 static int ext4_release_dquot(struct dquot *dquot)
6642 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6643 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6644 if (IS_ERR(handle)) {
6645 /* Release dquot anyway to avoid endless cycle in dqput() */
6646 dquot_release(dquot);
6647 return PTR_ERR(handle);
6649 ret = dquot_release(dquot);
6650 err = ext4_journal_stop(handle);
6656 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6658 struct super_block *sb = dquot->dq_sb;
6660 if (ext4_is_quota_journalled(sb)) {
6661 dquot_mark_dquot_dirty(dquot);
6662 return ext4_write_dquot(dquot);
6664 return dquot_mark_dquot_dirty(dquot);
6668 static int ext4_write_info(struct super_block *sb, int type)
6673 /* Data block + inode block */
6674 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6676 return PTR_ERR(handle);
6677 ret = dquot_commit_info(sb, type);
6678 err = ext4_journal_stop(handle);
6684 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6686 struct ext4_inode_info *ei = EXT4_I(inode);
6688 /* The first argument of lockdep_set_subclass has to be
6689 * *exactly* the same as the argument to init_rwsem() --- in
6690 * this case, in init_once() --- or lockdep gets unhappy
6691 * because the name of the lock is set using the
6692 * stringification of the argument to init_rwsem().
6694 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6695 lockdep_set_subclass(&ei->i_data_sem, subclass);
6699 * Standard function to be called on quota_on
6701 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6702 const struct path *path)
6706 if (!test_opt(sb, QUOTA))
6709 /* Quotafile not on the same filesystem? */
6710 if (path->dentry->d_sb != sb)
6713 /* Quota already enabled for this file? */
6714 if (IS_NOQUOTA(d_inode(path->dentry)))
6717 /* Journaling quota? */
6718 if (EXT4_SB(sb)->s_qf_names[type]) {
6719 /* Quotafile not in fs root? */
6720 if (path->dentry->d_parent != sb->s_root)
6721 ext4_msg(sb, KERN_WARNING,
6722 "Quota file not on filesystem root. "
6723 "Journaled quota will not work");
6724 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6727 * Clear the flag just in case mount options changed since
6730 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6734 * When we journal data on quota file, we have to flush journal to see
6735 * all updates to the file when we bypass pagecache...
6737 if (EXT4_SB(sb)->s_journal &&
6738 ext4_should_journal_data(d_inode(path->dentry))) {
6740 * We don't need to lock updates but journal_flush() could
6741 * otherwise be livelocked...
6743 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6744 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6745 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6750 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6751 err = dquot_quota_on(sb, type, format_id, path);
6753 struct inode *inode = d_inode(path->dentry);
6757 * Set inode flags to prevent userspace from messing with quota
6758 * files. If this fails, we return success anyway since quotas
6759 * are already enabled and this is not a hard failure.
6762 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6765 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6766 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6767 S_NOATIME | S_IMMUTABLE);
6768 err = ext4_mark_inode_dirty(handle, inode);
6769 ext4_journal_stop(handle);
6771 inode_unlock(inode);
6773 dquot_quota_off(sb, type);
6776 lockdep_set_quota_inode(path->dentry->d_inode,
6781 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6785 struct inode *qf_inode;
6786 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6787 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6788 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6789 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6792 BUG_ON(!ext4_has_feature_quota(sb));
6794 if (!qf_inums[type])
6797 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6798 if (IS_ERR(qf_inode)) {
6799 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6800 return PTR_ERR(qf_inode);
6803 /* Don't account quota for quota files to avoid recursion */
6804 qf_inode->i_flags |= S_NOQUOTA;
6805 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6806 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6808 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6814 /* Enable usage tracking for all quota types. */
6815 int ext4_enable_quotas(struct super_block *sb)
6818 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6819 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6820 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6821 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6823 bool quota_mopt[EXT4_MAXQUOTAS] = {
6824 test_opt(sb, USRQUOTA),
6825 test_opt(sb, GRPQUOTA),
6826 test_opt(sb, PRJQUOTA),
6829 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6830 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6831 if (qf_inums[type]) {
6832 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6833 DQUOT_USAGE_ENABLED |
6834 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6837 "Failed to enable quota tracking "
6838 "(type=%d, err=%d). Please run "
6839 "e2fsck to fix.", type, err);
6840 for (type--; type >= 0; type--) {
6841 struct inode *inode;
6843 inode = sb_dqopt(sb)->files[type];
6845 inode = igrab(inode);
6846 dquot_quota_off(sb, type);
6848 lockdep_set_quota_inode(inode,
6861 static int ext4_quota_off(struct super_block *sb, int type)
6863 struct inode *inode = sb_dqopt(sb)->files[type];
6867 /* Force all delayed allocation blocks to be allocated.
6868 * Caller already holds s_umount sem */
6869 if (test_opt(sb, DELALLOC))
6870 sync_filesystem(sb);
6872 if (!inode || !igrab(inode))
6875 err = dquot_quota_off(sb, type);
6876 if (err || ext4_has_feature_quota(sb))
6881 * Update modification times of quota files when userspace can
6882 * start looking at them. If we fail, we return success anyway since
6883 * this is not a hard failure and quotas are already disabled.
6885 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6886 if (IS_ERR(handle)) {
6887 err = PTR_ERR(handle);
6890 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6891 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6892 inode->i_mtime = inode->i_ctime = current_time(inode);
6893 err = ext4_mark_inode_dirty(handle, inode);
6894 ext4_journal_stop(handle);
6896 inode_unlock(inode);
6898 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6902 return dquot_quota_off(sb, type);
6905 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6906 * acquiring the locks... As quota files are never truncated and quota code
6907 * itself serializes the operations (and no one else should touch the files)
6908 * we don't have to be afraid of races */
6909 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6910 size_t len, loff_t off)
6912 struct inode *inode = sb_dqopt(sb)->files[type];
6913 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6914 int offset = off & (sb->s_blocksize - 1);
6917 struct buffer_head *bh;
6918 loff_t i_size = i_size_read(inode);
6922 if (off+len > i_size)
6925 while (toread > 0) {
6926 tocopy = sb->s_blocksize - offset < toread ?
6927 sb->s_blocksize - offset : toread;
6928 bh = ext4_bread(NULL, inode, blk, 0);
6931 if (!bh) /* A hole? */
6932 memset(data, 0, tocopy);
6934 memcpy(data, bh->b_data+offset, tocopy);
6944 /* Write to quotafile (we know the transaction is already started and has
6945 * enough credits) */
6946 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6947 const char *data, size_t len, loff_t off)
6949 struct inode *inode = sb_dqopt(sb)->files[type];
6950 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6951 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6953 struct buffer_head *bh;
6954 handle_t *handle = journal_current_handle();
6957 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6958 " cancelled because transaction is not started",
6959 (unsigned long long)off, (unsigned long long)len);
6963 * Since we account only one data block in transaction credits,
6964 * then it is impossible to cross a block boundary.
6966 if (sb->s_blocksize - offset < len) {
6967 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6968 " cancelled because not block aligned",
6969 (unsigned long long)off, (unsigned long long)len);
6974 bh = ext4_bread(handle, inode, blk,
6975 EXT4_GET_BLOCKS_CREATE |
6976 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6977 } while (PTR_ERR(bh) == -ENOSPC &&
6978 ext4_should_retry_alloc(inode->i_sb, &retries));
6983 BUFFER_TRACE(bh, "get write access");
6984 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6990 memcpy(bh->b_data+offset, data, len);
6991 flush_dcache_page(bh->b_page);
6993 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6996 if (inode->i_size < off + len) {
6997 i_size_write(inode, off + len);
6998 EXT4_I(inode)->i_disksize = inode->i_size;
6999 err2 = ext4_mark_inode_dirty(handle, inode);
7000 if (unlikely(err2 && !err))
7003 return err ? err : len;
7007 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7008 static inline void register_as_ext2(void)
7010 int err = register_filesystem(&ext2_fs_type);
7013 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7016 static inline void unregister_as_ext2(void)
7018 unregister_filesystem(&ext2_fs_type);
7021 static inline int ext2_feature_set_ok(struct super_block *sb)
7023 if (ext4_has_unknown_ext2_incompat_features(sb))
7027 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7032 static inline void register_as_ext2(void) { }
7033 static inline void unregister_as_ext2(void) { }
7034 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7037 static inline void register_as_ext3(void)
7039 int err = register_filesystem(&ext3_fs_type);
7042 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7045 static inline void unregister_as_ext3(void)
7047 unregister_filesystem(&ext3_fs_type);
7050 static inline int ext3_feature_set_ok(struct super_block *sb)
7052 if (ext4_has_unknown_ext3_incompat_features(sb))
7054 if (!ext4_has_feature_journal(sb))
7058 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7063 static struct file_system_type ext4_fs_type = {
7064 .owner = THIS_MODULE,
7066 .init_fs_context = ext4_init_fs_context,
7067 .parameters = ext4_param_specs,
7068 .kill_sb = kill_block_super,
7069 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7071 MODULE_ALIAS_FS("ext4");
7073 /* Shared across all ext4 file systems */
7074 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7076 static int __init ext4_init_fs(void)
7080 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7081 ext4_li_info = NULL;
7083 /* Build-time check for flags consistency */
7084 ext4_check_flag_values();
7086 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7087 init_waitqueue_head(&ext4__ioend_wq[i]);
7089 err = ext4_init_es();
7093 err = ext4_init_pending();
7097 err = ext4_init_post_read_processing();
7101 err = ext4_init_pageio();
7105 err = ext4_init_system_zone();
7109 err = ext4_init_sysfs();
7113 err = ext4_init_mballoc();
7116 err = init_inodecache();
7120 err = ext4_fc_init_dentry_cache();
7126 err = register_filesystem(&ext4_fs_type);
7132 unregister_as_ext2();
7133 unregister_as_ext3();
7134 ext4_fc_destroy_dentry_cache();
7136 destroy_inodecache();
7138 ext4_exit_mballoc();
7142 ext4_exit_system_zone();
7146 ext4_exit_post_read_processing();
7148 ext4_exit_pending();
7155 static void __exit ext4_exit_fs(void)
7157 ext4_destroy_lazyinit_thread();
7158 unregister_as_ext2();
7159 unregister_as_ext3();
7160 unregister_filesystem(&ext4_fs_type);
7161 ext4_fc_destroy_dentry_cache();
7162 destroy_inodecache();
7163 ext4_exit_mballoc();
7165 ext4_exit_system_zone();
7167 ext4_exit_post_read_processing();
7169 ext4_exit_pending();
7172 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7173 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7174 MODULE_LICENSE("GPL");
7175 MODULE_SOFTDEP("pre: crc32c");
7176 module_init(ext4_init_fs)
7177 module_exit(ext4_exit_fs)