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>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
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);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 if (trylock_buffer(bh)) {
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
199 if (buffer_uptodate(bh))
207 * This works like __bread_gfp() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
212 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
216 struct buffer_head *bh;
219 bh = sb_getblk_gfp(sb, block, gfp);
221 return ERR_PTR(-ENOMEM);
222 if (ext4_buffer_uptodate(bh))
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
233 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
239 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
245 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
255 static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
258 if (!ext4_has_feature_metadata_csum(sb))
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
264 static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
273 return cpu_to_le32(csum);
276 static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
279 if (!ext4_has_metadata_csum(sb))
282 return es->s_checksum == ext4_superblock_csum(sb, es);
285 void ext4_superblock_csum_set(struct super_block *sb)
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 if (!ext4_has_metadata_csum(sb))
292 es->s_checksum = ext4_superblock_csum(sb, es);
295 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
296 struct ext4_group_desc *bg)
298 return le32_to_cpu(bg->bg_block_bitmap_lo) |
299 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
300 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
303 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
304 struct ext4_group_desc *bg)
306 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
307 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
308 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
311 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
312 struct ext4_group_desc *bg)
314 return le32_to_cpu(bg->bg_inode_table_lo) |
315 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
316 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
319 __u32 ext4_free_group_clusters(struct super_block *sb,
320 struct ext4_group_desc *bg)
322 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
323 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
324 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
327 __u32 ext4_free_inodes_count(struct super_block *sb,
328 struct ext4_group_desc *bg)
330 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
332 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
335 __u32 ext4_used_dirs_count(struct super_block *sb,
336 struct ext4_group_desc *bg)
338 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
340 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
343 __u32 ext4_itable_unused_count(struct super_block *sb,
344 struct ext4_group_desc *bg)
346 return le16_to_cpu(bg->bg_itable_unused_lo) |
347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
348 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
351 void ext4_block_bitmap_set(struct super_block *sb,
352 struct ext4_group_desc *bg, ext4_fsblk_t blk)
354 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
355 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
356 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
359 void ext4_inode_bitmap_set(struct super_block *sb,
360 struct ext4_group_desc *bg, ext4_fsblk_t blk)
362 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
363 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
364 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
367 void ext4_inode_table_set(struct super_block *sb,
368 struct ext4_group_desc *bg, ext4_fsblk_t blk)
370 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
371 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
372 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
375 void ext4_free_group_clusters_set(struct super_block *sb,
376 struct ext4_group_desc *bg, __u32 count)
378 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
379 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
380 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
383 void ext4_free_inodes_set(struct super_block *sb,
384 struct ext4_group_desc *bg, __u32 count)
386 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
391 void ext4_used_dirs_set(struct super_block *sb,
392 struct ext4_group_desc *bg, __u32 count)
394 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
399 void ext4_itable_unused_set(struct super_block *sb,
400 struct ext4_group_desc *bg, __u32 count)
402 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
404 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
407 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
409 time64_t now = ktime_get_real_seconds();
411 now = clamp_val(now, 0, (1ull << 40) - 1);
413 *lo = cpu_to_le32(lower_32_bits(now));
414 *hi = upper_32_bits(now);
417 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
419 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
421 #define ext4_update_tstamp(es, tstamp) \
422 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
423 #define ext4_get_tstamp(es, tstamp) \
424 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
427 * The del_gendisk() function uninitializes the disk-specific data
428 * structures, including the bdi structure, without telling anyone
429 * else. Once this happens, any attempt to call mark_buffer_dirty()
430 * (for example, by ext4_commit_super), will cause a kernel OOPS.
431 * This is a kludge to prevent these oops until we can put in a proper
432 * hook in del_gendisk() to inform the VFS and file system layers.
434 static int block_device_ejected(struct super_block *sb)
436 struct inode *bd_inode = sb->s_bdev->bd_inode;
437 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
439 return bdi->dev == NULL;
442 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
444 struct super_block *sb = journal->j_private;
445 struct ext4_sb_info *sbi = EXT4_SB(sb);
446 int error = is_journal_aborted(journal);
447 struct ext4_journal_cb_entry *jce;
449 BUG_ON(txn->t_state == T_FINISHED);
451 ext4_process_freed_data(sb, txn->t_tid);
453 spin_lock(&sbi->s_md_lock);
454 while (!list_empty(&txn->t_private_list)) {
455 jce = list_entry(txn->t_private_list.next,
456 struct ext4_journal_cb_entry, jce_list);
457 list_del_init(&jce->jce_list);
458 spin_unlock(&sbi->s_md_lock);
459 jce->jce_func(sb, jce, error);
460 spin_lock(&sbi->s_md_lock);
462 spin_unlock(&sbi->s_md_lock);
466 * This writepage callback for write_cache_pages()
467 * takes care of a few cases after page cleaning.
469 * write_cache_pages() already checks for dirty pages
470 * and calls clear_page_dirty_for_io(), which we want,
471 * to write protect the pages.
473 * However, we may have to redirty a page (see below.)
475 static int ext4_journalled_writepage_callback(struct page *page,
476 struct writeback_control *wbc,
479 transaction_t *transaction = (transaction_t *) data;
480 struct buffer_head *bh, *head;
481 struct journal_head *jh;
483 bh = head = page_buffers(page);
486 * We have to redirty a page in these cases:
487 * 1) If buffer is dirty, it means the page was dirty because it
488 * contains a buffer that needs checkpointing. So the dirty bit
489 * needs to be preserved so that checkpointing writes the buffer
491 * 2) If buffer is not part of the committing transaction
492 * (we may have just accidentally come across this buffer because
493 * inode range tracking is not exact) or if the currently running
494 * transaction already contains this buffer as well, dirty bit
495 * needs to be preserved so that the buffer gets writeprotected
496 * properly on running transaction's commit.
499 if (buffer_dirty(bh) ||
500 (jh && (jh->b_transaction != transaction ||
501 jh->b_next_transaction))) {
502 redirty_page_for_writepage(wbc, page);
505 } while ((bh = bh->b_this_page) != head);
508 return AOP_WRITEPAGE_ACTIVATE;
511 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
513 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
514 struct writeback_control wbc = {
515 .sync_mode = WB_SYNC_ALL,
516 .nr_to_write = LONG_MAX,
517 .range_start = jinode->i_dirty_start,
518 .range_end = jinode->i_dirty_end,
521 return write_cache_pages(mapping, &wbc,
522 ext4_journalled_writepage_callback,
523 jinode->i_transaction);
526 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
530 if (ext4_should_journal_data(jinode->i_vfs_inode))
531 ret = ext4_journalled_submit_inode_data_buffers(jinode);
533 ret = jbd2_journal_submit_inode_data_buffers(jinode);
538 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
542 if (!ext4_should_journal_data(jinode->i_vfs_inode))
543 ret = jbd2_journal_finish_inode_data_buffers(jinode);
548 static bool system_going_down(void)
550 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
551 || system_state == SYSTEM_RESTART;
554 struct ext4_err_translation {
559 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
561 static struct ext4_err_translation err_translation[] = {
562 EXT4_ERR_TRANSLATE(EIO),
563 EXT4_ERR_TRANSLATE(ENOMEM),
564 EXT4_ERR_TRANSLATE(EFSBADCRC),
565 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
566 EXT4_ERR_TRANSLATE(ENOSPC),
567 EXT4_ERR_TRANSLATE(ENOKEY),
568 EXT4_ERR_TRANSLATE(EROFS),
569 EXT4_ERR_TRANSLATE(EFBIG),
570 EXT4_ERR_TRANSLATE(EEXIST),
571 EXT4_ERR_TRANSLATE(ERANGE),
572 EXT4_ERR_TRANSLATE(EOVERFLOW),
573 EXT4_ERR_TRANSLATE(EBUSY),
574 EXT4_ERR_TRANSLATE(ENOTDIR),
575 EXT4_ERR_TRANSLATE(ENOTEMPTY),
576 EXT4_ERR_TRANSLATE(ESHUTDOWN),
577 EXT4_ERR_TRANSLATE(EFAULT),
580 static int ext4_errno_to_code(int errno)
584 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
585 if (err_translation[i].errno == errno)
586 return err_translation[i].code;
587 return EXT4_ERR_UNKNOWN;
590 static void __save_error_info(struct super_block *sb, int error,
591 __u32 ino, __u64 block,
592 const char *func, unsigned int line)
594 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
596 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
597 if (bdev_read_only(sb->s_bdev))
599 /* We default to EFSCORRUPTED error... */
601 error = EFSCORRUPTED;
602 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
603 ext4_update_tstamp(es, s_last_error_time);
604 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
605 es->s_last_error_line = cpu_to_le32(line);
606 es->s_last_error_ino = cpu_to_le32(ino);
607 es->s_last_error_block = cpu_to_le64(block);
608 es->s_last_error_errcode = ext4_errno_to_code(error);
609 if (!es->s_first_error_time) {
610 es->s_first_error_time = es->s_last_error_time;
611 es->s_first_error_time_hi = es->s_last_error_time_hi;
612 strncpy(es->s_first_error_func, func,
613 sizeof(es->s_first_error_func));
614 es->s_first_error_line = cpu_to_le32(line);
615 es->s_first_error_ino = es->s_last_error_ino;
616 es->s_first_error_block = es->s_last_error_block;
617 es->s_first_error_errcode = es->s_last_error_errcode;
620 * Start the daily error reporting function if it hasn't been
623 if (!es->s_error_count)
624 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
625 le32_add_cpu(&es->s_error_count, 1);
628 static void save_error_info(struct super_block *sb, int error,
629 __u32 ino, __u64 block,
630 const char *func, unsigned int line)
632 __save_error_info(sb, error, ino, block, func, line);
633 if (!bdev_read_only(sb->s_bdev))
634 ext4_commit_super(sb, 1);
637 /* Deal with the reporting of failure conditions on a filesystem such as
638 * inconsistencies detected or read IO failures.
640 * On ext2, we can store the error state of the filesystem in the
641 * superblock. That is not possible on ext4, because we may have other
642 * write ordering constraints on the superblock which prevent us from
643 * writing it out straight away; and given that the journal is about to
644 * be aborted, we can't rely on the current, or future, transactions to
645 * write out the superblock safely.
647 * We'll just use the jbd2_journal_abort() error code to record an error in
648 * the journal instead. On recovery, the journal will complain about
649 * that error until we've noted it down and cleared it.
651 * If force_ro is set, we unconditionally force the filesystem into an
652 * ABORT|READONLY state, unless the error response on the fs has been set to
653 * panic in which case we take the easy way out and panic immediately. This is
654 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
655 * at a critical moment in log management.
657 static void ext4_handle_error(struct super_block *sb, bool force_ro)
659 journal_t *journal = EXT4_SB(sb)->s_journal;
661 if (test_opt(sb, WARN_ON_ERROR))
664 if (sb_rdonly(sb) || (!force_ro && test_opt(sb, ERRORS_CONT)))
667 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
669 jbd2_journal_abort(journal, -EIO);
671 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
672 * could panic during 'reboot -f' as the underlying device got already
675 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
676 panic("EXT4-fs (device %s): panic forced after error\n",
679 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
681 * Make sure updated value of ->s_mount_flags will be visible before
685 sb->s_flags |= SB_RDONLY;
688 #define ext4_error_ratelimit(sb) \
689 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
692 void __ext4_error(struct super_block *sb, const char *function,
693 unsigned int line, bool force_ro, int error, __u64 block,
694 const char *fmt, ...)
696 struct va_format vaf;
699 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
702 trace_ext4_error(sb, function, line);
703 if (ext4_error_ratelimit(sb)) {
708 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
709 sb->s_id, function, line, current->comm, &vaf);
712 save_error_info(sb, error, 0, block, function, line);
713 ext4_handle_error(sb, force_ro);
716 void __ext4_error_inode(struct inode *inode, const char *function,
717 unsigned int line, ext4_fsblk_t block, int error,
718 const char *fmt, ...)
721 struct va_format vaf;
723 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
726 trace_ext4_error(inode->i_sb, function, line);
727 if (ext4_error_ratelimit(inode->i_sb)) {
732 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
733 "inode #%lu: block %llu: comm %s: %pV\n",
734 inode->i_sb->s_id, function, line, inode->i_ino,
735 block, current->comm, &vaf);
737 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
738 "inode #%lu: comm %s: %pV\n",
739 inode->i_sb->s_id, function, line, inode->i_ino,
740 current->comm, &vaf);
743 save_error_info(inode->i_sb, error, inode->i_ino, block,
745 ext4_handle_error(inode->i_sb, false);
748 void __ext4_error_file(struct file *file, const char *function,
749 unsigned int line, ext4_fsblk_t block,
750 const char *fmt, ...)
753 struct va_format vaf;
754 struct inode *inode = file_inode(file);
755 char pathname[80], *path;
757 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
760 trace_ext4_error(inode->i_sb, function, line);
761 if (ext4_error_ratelimit(inode->i_sb)) {
762 path = file_path(file, pathname, sizeof(pathname));
770 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
771 "block %llu: comm %s: path %s: %pV\n",
772 inode->i_sb->s_id, function, line, inode->i_ino,
773 block, current->comm, path, &vaf);
776 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
777 "comm %s: path %s: %pV\n",
778 inode->i_sb->s_id, function, line, inode->i_ino,
779 current->comm, path, &vaf);
782 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
784 ext4_handle_error(inode->i_sb, false);
787 const char *ext4_decode_error(struct super_block *sb, int errno,
794 errstr = "Corrupt filesystem";
797 errstr = "Filesystem failed CRC";
800 errstr = "IO failure";
803 errstr = "Out of memory";
806 if (!sb || (EXT4_SB(sb)->s_journal &&
807 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
808 errstr = "Journal has aborted";
810 errstr = "Readonly filesystem";
813 /* If the caller passed in an extra buffer for unknown
814 * errors, textualise them now. Else we just return
817 /* Check for truncated error codes... */
818 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
827 /* __ext4_std_error decodes expected errors from journaling functions
828 * automatically and invokes the appropriate error response. */
830 void __ext4_std_error(struct super_block *sb, const char *function,
831 unsigned int line, int errno)
836 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
839 /* Special case: if the error is EROFS, and we're not already
840 * inside a transaction, then there's really no point in logging
842 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
845 if (ext4_error_ratelimit(sb)) {
846 errstr = ext4_decode_error(sb, errno, nbuf);
847 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
848 sb->s_id, function, line, errstr);
851 save_error_info(sb, -errno, 0, 0, function, line);
852 ext4_handle_error(sb, false);
855 void __ext4_msg(struct super_block *sb,
856 const char *prefix, const char *fmt, ...)
858 struct va_format vaf;
861 atomic_inc(&EXT4_SB(sb)->s_msg_count);
862 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
868 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
872 static int ext4_warning_ratelimit(struct super_block *sb)
874 atomic_inc(&EXT4_SB(sb)->s_warning_count);
875 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
879 void __ext4_warning(struct super_block *sb, const char *function,
880 unsigned int line, const char *fmt, ...)
882 struct va_format vaf;
885 if (!ext4_warning_ratelimit(sb))
891 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
892 sb->s_id, function, line, &vaf);
896 void __ext4_warning_inode(const struct inode *inode, const char *function,
897 unsigned int line, const char *fmt, ...)
899 struct va_format vaf;
902 if (!ext4_warning_ratelimit(inode->i_sb))
908 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
909 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
910 function, line, inode->i_ino, current->comm, &vaf);
914 void __ext4_grp_locked_error(const char *function, unsigned int line,
915 struct super_block *sb, ext4_group_t grp,
916 unsigned long ino, ext4_fsblk_t block,
917 const char *fmt, ...)
921 struct va_format vaf;
924 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
927 trace_ext4_error(sb, function, line);
928 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
930 if (ext4_error_ratelimit(sb)) {
934 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
935 sb->s_id, function, line, grp);
937 printk(KERN_CONT "inode %lu: ", ino);
939 printk(KERN_CONT "block %llu:",
940 (unsigned long long) block);
941 printk(KERN_CONT "%pV\n", &vaf);
945 if (test_opt(sb, WARN_ON_ERROR))
948 if (test_opt(sb, ERRORS_CONT)) {
949 ext4_commit_super(sb, 0);
953 ext4_unlock_group(sb, grp);
954 ext4_commit_super(sb, 1);
955 ext4_handle_error(sb, false);
957 * We only get here in the ERRORS_RO case; relocking the group
958 * may be dangerous, but nothing bad will happen since the
959 * filesystem will have already been marked read/only and the
960 * journal has been aborted. We return 1 as a hint to callers
961 * who might what to use the return value from
962 * ext4_grp_locked_error() to distinguish between the
963 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
964 * aggressively from the ext4 function in question, with a
965 * more appropriate error code.
967 ext4_lock_group(sb, grp);
971 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
975 struct ext4_sb_info *sbi = EXT4_SB(sb);
976 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
977 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
980 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
981 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
984 percpu_counter_sub(&sbi->s_freeclusters_counter,
988 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
989 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
994 count = ext4_free_inodes_count(sb, gdp);
995 percpu_counter_sub(&sbi->s_freeinodes_counter,
1001 void ext4_update_dynamic_rev(struct super_block *sb)
1003 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1005 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1009 "updating to rev %d because of new feature flag, "
1010 "running e2fsck is recommended",
1013 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1014 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1015 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1016 /* leave es->s_feature_*compat flags alone */
1017 /* es->s_uuid will be set by e2fsck if empty */
1020 * The rest of the superblock fields should be zero, and if not it
1021 * means they are likely already in use, so leave them alone. We
1022 * can leave it up to e2fsck to clean up any inconsistencies there.
1027 * Open the external journal device
1029 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1031 struct block_device *bdev;
1033 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1039 ext4_msg(sb, KERN_ERR,
1040 "failed to open journal device unknown-block(%u,%u) %ld",
1041 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1046 * Release the journal device
1048 static void ext4_blkdev_put(struct block_device *bdev)
1050 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1053 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1055 struct block_device *bdev;
1056 bdev = sbi->s_journal_bdev;
1058 ext4_blkdev_put(bdev);
1059 sbi->s_journal_bdev = NULL;
1063 static inline struct inode *orphan_list_entry(struct list_head *l)
1065 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1068 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1070 struct list_head *l;
1072 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1073 le32_to_cpu(sbi->s_es->s_last_orphan));
1075 printk(KERN_ERR "sb_info orphan list:\n");
1076 list_for_each(l, &sbi->s_orphan) {
1077 struct inode *inode = orphan_list_entry(l);
1079 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1080 inode->i_sb->s_id, inode->i_ino, inode,
1081 inode->i_mode, inode->i_nlink,
1082 NEXT_ORPHAN(inode));
1087 static int ext4_quota_off(struct super_block *sb, int type);
1089 static inline void ext4_quota_off_umount(struct super_block *sb)
1093 /* Use our quota_off function to clear inode flags etc. */
1094 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1095 ext4_quota_off(sb, type);
1099 * This is a helper function which is used in the mount/remount
1100 * codepaths (which holds s_umount) to fetch the quota file name.
1102 static inline char *get_qf_name(struct super_block *sb,
1103 struct ext4_sb_info *sbi,
1106 return rcu_dereference_protected(sbi->s_qf_names[type],
1107 lockdep_is_held(&sb->s_umount));
1110 static inline void ext4_quota_off_umount(struct super_block *sb)
1115 static void ext4_put_super(struct super_block *sb)
1117 struct ext4_sb_info *sbi = EXT4_SB(sb);
1118 struct ext4_super_block *es = sbi->s_es;
1119 struct buffer_head **group_desc;
1120 struct flex_groups **flex_groups;
1124 ext4_unregister_li_request(sb);
1125 ext4_quota_off_umount(sb);
1127 destroy_workqueue(sbi->rsv_conversion_wq);
1130 * Unregister sysfs before destroying jbd2 journal.
1131 * Since we could still access attr_journal_task attribute via sysfs
1132 * path which could have sbi->s_journal->j_task as NULL
1134 ext4_unregister_sysfs(sb);
1136 if (sbi->s_journal) {
1137 aborted = is_journal_aborted(sbi->s_journal);
1138 err = jbd2_journal_destroy(sbi->s_journal);
1139 sbi->s_journal = NULL;
1140 if ((err < 0) && !aborted) {
1141 ext4_abort(sb, -err, "Couldn't clean up the journal");
1145 ext4_es_unregister_shrinker(sbi);
1146 del_timer_sync(&sbi->s_err_report);
1147 ext4_release_system_zone(sb);
1148 ext4_mb_release(sb);
1149 ext4_ext_release(sb);
1151 if (!sb_rdonly(sb) && !aborted) {
1152 ext4_clear_feature_journal_needs_recovery(sb);
1153 es->s_state = cpu_to_le16(sbi->s_mount_state);
1156 ext4_commit_super(sb, 1);
1159 group_desc = rcu_dereference(sbi->s_group_desc);
1160 for (i = 0; i < sbi->s_gdb_count; i++)
1161 brelse(group_desc[i]);
1163 flex_groups = rcu_dereference(sbi->s_flex_groups);
1165 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1166 kvfree(flex_groups[i]);
1167 kvfree(flex_groups);
1170 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1171 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1172 percpu_counter_destroy(&sbi->s_dirs_counter);
1173 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1174 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1176 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1177 kfree(get_qf_name(sb, sbi, i));
1180 /* Debugging code just in case the in-memory inode orphan list
1181 * isn't empty. The on-disk one can be non-empty if we've
1182 * detected an error and taken the fs readonly, but the
1183 * in-memory list had better be clean by this point. */
1184 if (!list_empty(&sbi->s_orphan))
1185 dump_orphan_list(sb, sbi);
1186 ASSERT(list_empty(&sbi->s_orphan));
1188 sync_blockdev(sb->s_bdev);
1189 invalidate_bdev(sb->s_bdev);
1190 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1192 * Invalidate the journal device's buffers. We don't want them
1193 * floating about in memory - the physical journal device may
1194 * hotswapped, and it breaks the `ro-after' testing code.
1196 sync_blockdev(sbi->s_journal_bdev);
1197 invalidate_bdev(sbi->s_journal_bdev);
1198 ext4_blkdev_remove(sbi);
1201 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1202 sbi->s_ea_inode_cache = NULL;
1204 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1205 sbi->s_ea_block_cache = NULL;
1208 kthread_stop(sbi->s_mmp_tsk);
1210 sb->s_fs_info = NULL;
1212 * Now that we are completely done shutting down the
1213 * superblock, we need to actually destroy the kobject.
1215 kobject_put(&sbi->s_kobj);
1216 wait_for_completion(&sbi->s_kobj_unregister);
1217 if (sbi->s_chksum_driver)
1218 crypto_free_shash(sbi->s_chksum_driver);
1219 kfree(sbi->s_blockgroup_lock);
1220 fs_put_dax(sbi->s_daxdev);
1221 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1222 #ifdef CONFIG_UNICODE
1223 utf8_unload(sb->s_encoding);
1228 static struct kmem_cache *ext4_inode_cachep;
1231 * Called inside transaction, so use GFP_NOFS
1233 static struct inode *ext4_alloc_inode(struct super_block *sb)
1235 struct ext4_inode_info *ei;
1237 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1241 inode_set_iversion(&ei->vfs_inode, 1);
1242 spin_lock_init(&ei->i_raw_lock);
1243 INIT_LIST_HEAD(&ei->i_prealloc_list);
1244 atomic_set(&ei->i_prealloc_active, 0);
1245 spin_lock_init(&ei->i_prealloc_lock);
1246 ext4_es_init_tree(&ei->i_es_tree);
1247 rwlock_init(&ei->i_es_lock);
1248 INIT_LIST_HEAD(&ei->i_es_list);
1249 ei->i_es_all_nr = 0;
1250 ei->i_es_shk_nr = 0;
1251 ei->i_es_shrink_lblk = 0;
1252 ei->i_reserved_data_blocks = 0;
1253 spin_lock_init(&(ei->i_block_reservation_lock));
1254 ext4_init_pending_tree(&ei->i_pending_tree);
1256 ei->i_reserved_quota = 0;
1257 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1260 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1261 spin_lock_init(&ei->i_completed_io_lock);
1263 ei->i_datasync_tid = 0;
1264 atomic_set(&ei->i_unwritten, 0);
1265 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1266 ext4_fc_init_inode(&ei->vfs_inode);
1267 mutex_init(&ei->i_fc_lock);
1268 return &ei->vfs_inode;
1271 static int ext4_drop_inode(struct inode *inode)
1273 int drop = generic_drop_inode(inode);
1276 drop = fscrypt_drop_inode(inode);
1278 trace_ext4_drop_inode(inode, drop);
1282 static void ext4_free_in_core_inode(struct inode *inode)
1284 fscrypt_free_inode(inode);
1285 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1286 pr_warn("%s: inode %ld still in fc list",
1287 __func__, inode->i_ino);
1289 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1292 static void ext4_destroy_inode(struct inode *inode)
1294 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1295 ext4_msg(inode->i_sb, KERN_ERR,
1296 "Inode %lu (%p): orphan list check failed!",
1297 inode->i_ino, EXT4_I(inode));
1298 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1299 EXT4_I(inode), sizeof(struct ext4_inode_info),
1305 static void init_once(void *foo)
1307 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1309 INIT_LIST_HEAD(&ei->i_orphan);
1310 init_rwsem(&ei->xattr_sem);
1311 init_rwsem(&ei->i_data_sem);
1312 init_rwsem(&ei->i_mmap_sem);
1313 inode_init_once(&ei->vfs_inode);
1314 ext4_fc_init_inode(&ei->vfs_inode);
1317 static int __init init_inodecache(void)
1319 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1320 sizeof(struct ext4_inode_info), 0,
1321 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1323 offsetof(struct ext4_inode_info, i_data),
1324 sizeof_field(struct ext4_inode_info, i_data),
1326 if (ext4_inode_cachep == NULL)
1331 static void destroy_inodecache(void)
1334 * Make sure all delayed rcu free inodes are flushed before we
1338 kmem_cache_destroy(ext4_inode_cachep);
1341 void ext4_clear_inode(struct inode *inode)
1344 invalidate_inode_buffers(inode);
1346 ext4_discard_preallocations(inode, 0);
1347 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1349 if (EXT4_I(inode)->jinode) {
1350 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1351 EXT4_I(inode)->jinode);
1352 jbd2_free_inode(EXT4_I(inode)->jinode);
1353 EXT4_I(inode)->jinode = NULL;
1355 fscrypt_put_encryption_info(inode);
1356 fsverity_cleanup_inode(inode);
1359 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1360 u64 ino, u32 generation)
1362 struct inode *inode;
1365 * Currently we don't know the generation for parent directory, so
1366 * a generation of 0 means "accept any"
1368 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1370 return ERR_CAST(inode);
1371 if (generation && inode->i_generation != generation) {
1373 return ERR_PTR(-ESTALE);
1379 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1380 int fh_len, int fh_type)
1382 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1383 ext4_nfs_get_inode);
1386 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1387 int fh_len, int fh_type)
1389 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1390 ext4_nfs_get_inode);
1393 static int ext4_nfs_commit_metadata(struct inode *inode)
1395 struct writeback_control wbc = {
1396 .sync_mode = WB_SYNC_ALL
1399 trace_ext4_nfs_commit_metadata(inode);
1400 return ext4_write_inode(inode, &wbc);
1404 * Try to release metadata pages (indirect blocks, directories) which are
1405 * mapped via the block device. Since these pages could have journal heads
1406 * which would prevent try_to_free_buffers() from freeing them, we must use
1407 * jbd2 layer's try_to_free_buffers() function to release them.
1409 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1412 journal_t *journal = EXT4_SB(sb)->s_journal;
1414 WARN_ON(PageChecked(page));
1415 if (!page_has_buffers(page))
1418 return jbd2_journal_try_to_free_buffers(journal, page);
1420 return try_to_free_buffers(page);
1423 #ifdef CONFIG_FS_ENCRYPTION
1424 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1426 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1427 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1430 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1433 handle_t *handle = fs_data;
1434 int res, res2, credits, retries = 0;
1437 * Encrypting the root directory is not allowed because e2fsck expects
1438 * lost+found to exist and be unencrypted, and encrypting the root
1439 * directory would imply encrypting the lost+found directory as well as
1440 * the filename "lost+found" itself.
1442 if (inode->i_ino == EXT4_ROOT_INO)
1445 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1448 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1451 res = ext4_convert_inline_data(inode);
1456 * If a journal handle was specified, then the encryption context is
1457 * being set on a new inode via inheritance and is part of a larger
1458 * transaction to create the inode. Otherwise the encryption context is
1459 * being set on an existing inode in its own transaction. Only in the
1460 * latter case should the "retry on ENOSPC" logic be used.
1464 res = ext4_xattr_set_handle(handle, inode,
1465 EXT4_XATTR_INDEX_ENCRYPTION,
1466 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1469 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1470 ext4_clear_inode_state(inode,
1471 EXT4_STATE_MAY_INLINE_DATA);
1473 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1474 * S_DAX may be disabled
1476 ext4_set_inode_flags(inode, false);
1481 res = dquot_initialize(inode);
1485 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1490 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1492 return PTR_ERR(handle);
1494 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1498 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1500 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1501 * S_DAX may be disabled
1503 ext4_set_inode_flags(inode, false);
1504 res = ext4_mark_inode_dirty(handle, inode);
1506 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1508 res2 = ext4_journal_stop(handle);
1510 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1517 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1519 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1522 static bool ext4_has_stable_inodes(struct super_block *sb)
1524 return ext4_has_feature_stable_inodes(sb);
1527 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1528 int *ino_bits_ret, int *lblk_bits_ret)
1530 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1531 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1534 static const struct fscrypt_operations ext4_cryptops = {
1535 .key_prefix = "ext4:",
1536 .get_context = ext4_get_context,
1537 .set_context = ext4_set_context,
1538 .get_dummy_policy = ext4_get_dummy_policy,
1539 .empty_dir = ext4_empty_dir,
1540 .max_namelen = EXT4_NAME_LEN,
1541 .has_stable_inodes = ext4_has_stable_inodes,
1542 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1547 static const char * const quotatypes[] = INITQFNAMES;
1548 #define QTYPE2NAME(t) (quotatypes[t])
1550 static int ext4_write_dquot(struct dquot *dquot);
1551 static int ext4_acquire_dquot(struct dquot *dquot);
1552 static int ext4_release_dquot(struct dquot *dquot);
1553 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1554 static int ext4_write_info(struct super_block *sb, int type);
1555 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1556 const struct path *path);
1557 static int ext4_quota_on_mount(struct super_block *sb, int type);
1558 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1559 size_t len, loff_t off);
1560 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1561 const char *data, size_t len, loff_t off);
1562 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1563 unsigned int flags);
1564 static int ext4_enable_quotas(struct super_block *sb);
1566 static struct dquot **ext4_get_dquots(struct inode *inode)
1568 return EXT4_I(inode)->i_dquot;
1571 static const struct dquot_operations ext4_quota_operations = {
1572 .get_reserved_space = ext4_get_reserved_space,
1573 .write_dquot = ext4_write_dquot,
1574 .acquire_dquot = ext4_acquire_dquot,
1575 .release_dquot = ext4_release_dquot,
1576 .mark_dirty = ext4_mark_dquot_dirty,
1577 .write_info = ext4_write_info,
1578 .alloc_dquot = dquot_alloc,
1579 .destroy_dquot = dquot_destroy,
1580 .get_projid = ext4_get_projid,
1581 .get_inode_usage = ext4_get_inode_usage,
1582 .get_next_id = dquot_get_next_id,
1585 static const struct quotactl_ops ext4_qctl_operations = {
1586 .quota_on = ext4_quota_on,
1587 .quota_off = ext4_quota_off,
1588 .quota_sync = dquot_quota_sync,
1589 .get_state = dquot_get_state,
1590 .set_info = dquot_set_dqinfo,
1591 .get_dqblk = dquot_get_dqblk,
1592 .set_dqblk = dquot_set_dqblk,
1593 .get_nextdqblk = dquot_get_next_dqblk,
1597 static const struct super_operations ext4_sops = {
1598 .alloc_inode = ext4_alloc_inode,
1599 .free_inode = ext4_free_in_core_inode,
1600 .destroy_inode = ext4_destroy_inode,
1601 .write_inode = ext4_write_inode,
1602 .dirty_inode = ext4_dirty_inode,
1603 .drop_inode = ext4_drop_inode,
1604 .evict_inode = ext4_evict_inode,
1605 .put_super = ext4_put_super,
1606 .sync_fs = ext4_sync_fs,
1607 .freeze_fs = ext4_freeze,
1608 .unfreeze_fs = ext4_unfreeze,
1609 .statfs = ext4_statfs,
1610 .remount_fs = ext4_remount,
1611 .show_options = ext4_show_options,
1613 .quota_read = ext4_quota_read,
1614 .quota_write = ext4_quota_write,
1615 .get_dquots = ext4_get_dquots,
1617 .bdev_try_to_free_page = bdev_try_to_free_page,
1620 static const struct export_operations ext4_export_ops = {
1621 .fh_to_dentry = ext4_fh_to_dentry,
1622 .fh_to_parent = ext4_fh_to_parent,
1623 .get_parent = ext4_get_parent,
1624 .commit_metadata = ext4_nfs_commit_metadata,
1628 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1629 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1630 Opt_nouid32, Opt_debug, Opt_removed,
1631 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1632 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1633 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1634 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1635 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1636 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1638 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1639 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1640 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1641 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1642 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1643 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1644 Opt_nowarn_on_error, Opt_mblk_io_submit,
1645 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1646 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1647 Opt_inode_readahead_blks, Opt_journal_ioprio,
1648 Opt_dioread_nolock, Opt_dioread_lock,
1649 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1650 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1651 Opt_prefetch_block_bitmaps,
1652 #ifdef CONFIG_EXT4_DEBUG
1653 Opt_fc_debug_max_replay, Opt_fc_debug_force
1657 static const match_table_t tokens = {
1658 {Opt_bsd_df, "bsddf"},
1659 {Opt_minix_df, "minixdf"},
1660 {Opt_grpid, "grpid"},
1661 {Opt_grpid, "bsdgroups"},
1662 {Opt_nogrpid, "nogrpid"},
1663 {Opt_nogrpid, "sysvgroups"},
1664 {Opt_resgid, "resgid=%u"},
1665 {Opt_resuid, "resuid=%u"},
1667 {Opt_err_cont, "errors=continue"},
1668 {Opt_err_panic, "errors=panic"},
1669 {Opt_err_ro, "errors=remount-ro"},
1670 {Opt_nouid32, "nouid32"},
1671 {Opt_debug, "debug"},
1672 {Opt_removed, "oldalloc"},
1673 {Opt_removed, "orlov"},
1674 {Opt_user_xattr, "user_xattr"},
1675 {Opt_nouser_xattr, "nouser_xattr"},
1677 {Opt_noacl, "noacl"},
1678 {Opt_noload, "norecovery"},
1679 {Opt_noload, "noload"},
1680 {Opt_removed, "nobh"},
1681 {Opt_removed, "bh"},
1682 {Opt_commit, "commit=%u"},
1683 {Opt_min_batch_time, "min_batch_time=%u"},
1684 {Opt_max_batch_time, "max_batch_time=%u"},
1685 {Opt_journal_dev, "journal_dev=%u"},
1686 {Opt_journal_path, "journal_path=%s"},
1687 {Opt_journal_checksum, "journal_checksum"},
1688 {Opt_nojournal_checksum, "nojournal_checksum"},
1689 {Opt_journal_async_commit, "journal_async_commit"},
1690 {Opt_abort, "abort"},
1691 {Opt_data_journal, "data=journal"},
1692 {Opt_data_ordered, "data=ordered"},
1693 {Opt_data_writeback, "data=writeback"},
1694 {Opt_data_err_abort, "data_err=abort"},
1695 {Opt_data_err_ignore, "data_err=ignore"},
1696 {Opt_offusrjquota, "usrjquota="},
1697 {Opt_usrjquota, "usrjquota=%s"},
1698 {Opt_offgrpjquota, "grpjquota="},
1699 {Opt_grpjquota, "grpjquota=%s"},
1700 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1701 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1702 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1703 {Opt_grpquota, "grpquota"},
1704 {Opt_noquota, "noquota"},
1705 {Opt_quota, "quota"},
1706 {Opt_usrquota, "usrquota"},
1707 {Opt_prjquota, "prjquota"},
1708 {Opt_barrier, "barrier=%u"},
1709 {Opt_barrier, "barrier"},
1710 {Opt_nobarrier, "nobarrier"},
1711 {Opt_i_version, "i_version"},
1713 {Opt_dax_always, "dax=always"},
1714 {Opt_dax_inode, "dax=inode"},
1715 {Opt_dax_never, "dax=never"},
1716 {Opt_stripe, "stripe=%u"},
1717 {Opt_delalloc, "delalloc"},
1718 {Opt_warn_on_error, "warn_on_error"},
1719 {Opt_nowarn_on_error, "nowarn_on_error"},
1720 {Opt_lazytime, "lazytime"},
1721 {Opt_nolazytime, "nolazytime"},
1722 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1723 {Opt_nodelalloc, "nodelalloc"},
1724 {Opt_removed, "mblk_io_submit"},
1725 {Opt_removed, "nomblk_io_submit"},
1726 {Opt_block_validity, "block_validity"},
1727 {Opt_noblock_validity, "noblock_validity"},
1728 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1729 {Opt_journal_ioprio, "journal_ioprio=%u"},
1730 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1731 {Opt_auto_da_alloc, "auto_da_alloc"},
1732 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1733 {Opt_dioread_nolock, "dioread_nolock"},
1734 {Opt_dioread_lock, "nodioread_nolock"},
1735 {Opt_dioread_lock, "dioread_lock"},
1736 {Opt_discard, "discard"},
1737 {Opt_nodiscard, "nodiscard"},
1738 {Opt_init_itable, "init_itable=%u"},
1739 {Opt_init_itable, "init_itable"},
1740 {Opt_noinit_itable, "noinit_itable"},
1741 #ifdef CONFIG_EXT4_DEBUG
1742 {Opt_fc_debug_force, "fc_debug_force"},
1743 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1745 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1746 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1747 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1748 {Opt_inlinecrypt, "inlinecrypt"},
1749 {Opt_nombcache, "nombcache"},
1750 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1751 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1752 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1753 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1754 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1755 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1756 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1760 static ext4_fsblk_t get_sb_block(void **data)
1762 ext4_fsblk_t sb_block;
1763 char *options = (char *) *data;
1765 if (!options || strncmp(options, "sb=", 3) != 0)
1766 return 1; /* Default location */
1769 /* TODO: use simple_strtoll with >32bit ext4 */
1770 sb_block = simple_strtoul(options, &options, 0);
1771 if (*options && *options != ',') {
1772 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1776 if (*options == ',')
1778 *data = (void *) options;
1783 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1784 static const char deprecated_msg[] =
1785 "Mount option \"%s\" will be removed by %s\n"
1786 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1789 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1791 struct ext4_sb_info *sbi = EXT4_SB(sb);
1792 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1795 if (sb_any_quota_loaded(sb) && !old_qname) {
1796 ext4_msg(sb, KERN_ERR,
1797 "Cannot change journaled "
1798 "quota options when quota turned on");
1801 if (ext4_has_feature_quota(sb)) {
1802 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1803 "ignored when QUOTA feature is enabled");
1806 qname = match_strdup(args);
1808 ext4_msg(sb, KERN_ERR,
1809 "Not enough memory for storing quotafile name");
1813 if (strcmp(old_qname, qname) == 0)
1816 ext4_msg(sb, KERN_ERR,
1817 "%s quota file already specified",
1821 if (strchr(qname, '/')) {
1822 ext4_msg(sb, KERN_ERR,
1823 "quotafile must be on filesystem root");
1826 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1834 static int clear_qf_name(struct super_block *sb, int qtype)
1837 struct ext4_sb_info *sbi = EXT4_SB(sb);
1838 char *old_qname = get_qf_name(sb, sbi, qtype);
1840 if (sb_any_quota_loaded(sb) && old_qname) {
1841 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1842 " when quota turned on");
1845 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1852 #define MOPT_SET 0x0001
1853 #define MOPT_CLEAR 0x0002
1854 #define MOPT_NOSUPPORT 0x0004
1855 #define MOPT_EXPLICIT 0x0008
1856 #define MOPT_CLEAR_ERR 0x0010
1857 #define MOPT_GTE0 0x0020
1860 #define MOPT_QFMT 0x0040
1862 #define MOPT_Q MOPT_NOSUPPORT
1863 #define MOPT_QFMT MOPT_NOSUPPORT
1865 #define MOPT_DATAJ 0x0080
1866 #define MOPT_NO_EXT2 0x0100
1867 #define MOPT_NO_EXT3 0x0200
1868 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1869 #define MOPT_STRING 0x0400
1870 #define MOPT_SKIP 0x0800
1871 #define MOPT_2 0x1000
1873 static const struct mount_opts {
1877 } ext4_mount_opts[] = {
1878 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1879 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1880 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1881 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1882 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1883 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1884 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1885 MOPT_EXT4_ONLY | MOPT_SET},
1886 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1887 MOPT_EXT4_ONLY | MOPT_CLEAR},
1888 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1889 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1890 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1891 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1892 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1893 MOPT_EXT4_ONLY | MOPT_CLEAR},
1894 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1895 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1896 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1897 MOPT_EXT4_ONLY | MOPT_CLEAR},
1898 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1899 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1900 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1901 EXT4_MOUNT_JOURNAL_CHECKSUM),
1902 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1903 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1904 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1905 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1906 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1907 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1909 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1911 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1912 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1913 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1914 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1915 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1916 {Opt_commit, 0, MOPT_GTE0},
1917 {Opt_max_batch_time, 0, MOPT_GTE0},
1918 {Opt_min_batch_time, 0, MOPT_GTE0},
1919 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1920 {Opt_init_itable, 0, MOPT_GTE0},
1921 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1922 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1923 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1924 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1925 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1926 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1927 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1928 {Opt_stripe, 0, MOPT_GTE0},
1929 {Opt_resuid, 0, MOPT_GTE0},
1930 {Opt_resgid, 0, MOPT_GTE0},
1931 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1932 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1933 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1934 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1935 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1936 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1937 MOPT_NO_EXT2 | MOPT_DATAJ},
1938 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1939 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1940 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1941 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1942 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1944 {Opt_acl, 0, MOPT_NOSUPPORT},
1945 {Opt_noacl, 0, MOPT_NOSUPPORT},
1947 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1948 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1949 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1950 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1951 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1953 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1955 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1957 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1958 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1959 MOPT_CLEAR | MOPT_Q},
1960 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1961 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1962 {Opt_offusrjquota, 0, MOPT_Q},
1963 {Opt_offgrpjquota, 0, MOPT_Q},
1964 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1965 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1966 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1967 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1968 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1969 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1970 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1972 #ifdef CONFIG_EXT4_DEBUG
1973 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1974 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1975 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
1980 #ifdef CONFIG_UNICODE
1981 static const struct ext4_sb_encodings {
1985 } ext4_sb_encoding_map[] = {
1986 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1989 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1990 const struct ext4_sb_encodings **encoding,
1993 __u16 magic = le16_to_cpu(es->s_encoding);
1996 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1997 if (magic == ext4_sb_encoding_map[i].magic)
2000 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2003 *encoding = &ext4_sb_encoding_map[i];
2004 *flags = le16_to_cpu(es->s_encoding_flags);
2010 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2012 const substring_t *arg,
2015 #ifdef CONFIG_FS_ENCRYPTION
2016 struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 * This mount option is just for testing, and it's not worthwhile to
2021 * implement the extra complexity (e.g. RCU protection) that would be
2022 * needed to allow it to be set or changed during remount. We do allow
2023 * it to be specified during remount, but only if there is no change.
2025 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2026 ext4_msg(sb, KERN_WARNING,
2027 "Can't set test_dummy_encryption on remount");
2030 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2031 &sbi->s_dummy_enc_policy);
2034 ext4_msg(sb, KERN_WARNING,
2035 "Can't change test_dummy_encryption on remount");
2036 else if (err == -EINVAL)
2037 ext4_msg(sb, KERN_WARNING,
2038 "Value of option \"%s\" is unrecognized", opt);
2040 ext4_msg(sb, KERN_WARNING,
2041 "Error processing option \"%s\" [%d]",
2045 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2047 ext4_msg(sb, KERN_WARNING,
2048 "Test dummy encryption mount option ignored");
2053 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2054 substring_t *args, unsigned long *journal_devnum,
2055 unsigned int *journal_ioprio, int is_remount)
2057 struct ext4_sb_info *sbi = EXT4_SB(sb);
2058 const struct mount_opts *m;
2064 if (token == Opt_usrjquota)
2065 return set_qf_name(sb, USRQUOTA, &args[0]);
2066 else if (token == Opt_grpjquota)
2067 return set_qf_name(sb, GRPQUOTA, &args[0]);
2068 else if (token == Opt_offusrjquota)
2069 return clear_qf_name(sb, USRQUOTA);
2070 else if (token == Opt_offgrpjquota)
2071 return clear_qf_name(sb, GRPQUOTA);
2075 case Opt_nouser_xattr:
2076 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2079 return 1; /* handled by get_sb_block() */
2081 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2084 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2087 sb->s_flags |= SB_I_VERSION;
2090 sb->s_flags |= SB_LAZYTIME;
2092 case Opt_nolazytime:
2093 sb->s_flags &= ~SB_LAZYTIME;
2095 case Opt_inlinecrypt:
2096 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2097 sb->s_flags |= SB_INLINECRYPT;
2099 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2104 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2105 if (token == m->token)
2108 if (m->token == Opt_err) {
2109 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2110 "or missing value", opt);
2114 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2115 ext4_msg(sb, KERN_ERR,
2116 "Mount option \"%s\" incompatible with ext2", opt);
2119 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2120 ext4_msg(sb, KERN_ERR,
2121 "Mount option \"%s\" incompatible with ext3", opt);
2125 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2127 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2129 if (m->flags & MOPT_EXPLICIT) {
2130 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2131 set_opt2(sb, EXPLICIT_DELALLOC);
2132 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2133 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2137 if (m->flags & MOPT_CLEAR_ERR)
2138 clear_opt(sb, ERRORS_MASK);
2139 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2140 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2141 "options when quota turned on");
2145 if (m->flags & MOPT_NOSUPPORT) {
2146 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2147 } else if (token == Opt_commit) {
2149 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2150 else if (arg > INT_MAX / HZ) {
2151 ext4_msg(sb, KERN_ERR,
2152 "Invalid commit interval %d, "
2153 "must be smaller than %d",
2157 sbi->s_commit_interval = HZ * arg;
2158 } else if (token == Opt_debug_want_extra_isize) {
2161 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2162 ext4_msg(sb, KERN_ERR,
2163 "Invalid want_extra_isize %d", arg);
2166 sbi->s_want_extra_isize = arg;
2167 } else if (token == Opt_max_batch_time) {
2168 sbi->s_max_batch_time = arg;
2169 } else if (token == Opt_min_batch_time) {
2170 sbi->s_min_batch_time = arg;
2171 } else if (token == Opt_inode_readahead_blks) {
2172 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2173 ext4_msg(sb, KERN_ERR,
2174 "EXT4-fs: inode_readahead_blks must be "
2175 "0 or a power of 2 smaller than 2^31");
2178 sbi->s_inode_readahead_blks = arg;
2179 } else if (token == Opt_init_itable) {
2180 set_opt(sb, INIT_INODE_TABLE);
2182 arg = EXT4_DEF_LI_WAIT_MULT;
2183 sbi->s_li_wait_mult = arg;
2184 } else if (token == Opt_max_dir_size_kb) {
2185 sbi->s_max_dir_size_kb = arg;
2186 #ifdef CONFIG_EXT4_DEBUG
2187 } else if (token == Opt_fc_debug_max_replay) {
2188 sbi->s_fc_debug_max_replay = arg;
2190 } else if (token == Opt_stripe) {
2191 sbi->s_stripe = arg;
2192 } else if (token == Opt_resuid) {
2193 uid = make_kuid(current_user_ns(), arg);
2194 if (!uid_valid(uid)) {
2195 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2198 sbi->s_resuid = uid;
2199 } else if (token == Opt_resgid) {
2200 gid = make_kgid(current_user_ns(), arg);
2201 if (!gid_valid(gid)) {
2202 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2205 sbi->s_resgid = gid;
2206 } else if (token == Opt_journal_dev) {
2208 ext4_msg(sb, KERN_ERR,
2209 "Cannot specify journal on remount");
2212 *journal_devnum = arg;
2213 } else if (token == Opt_journal_path) {
2215 struct inode *journal_inode;
2220 ext4_msg(sb, KERN_ERR,
2221 "Cannot specify journal on remount");
2224 journal_path = match_strdup(&args[0]);
2225 if (!journal_path) {
2226 ext4_msg(sb, KERN_ERR, "error: could not dup "
2227 "journal device string");
2231 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2233 ext4_msg(sb, KERN_ERR, "error: could not find "
2234 "journal device path: error %d", error);
2235 kfree(journal_path);
2239 journal_inode = d_inode(path.dentry);
2240 if (!S_ISBLK(journal_inode->i_mode)) {
2241 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2242 "is not a block device", journal_path);
2244 kfree(journal_path);
2248 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2250 kfree(journal_path);
2251 } else if (token == Opt_journal_ioprio) {
2253 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2258 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2259 } else if (token == Opt_test_dummy_encryption) {
2260 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2262 } else if (m->flags & MOPT_DATAJ) {
2264 if (!sbi->s_journal)
2265 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2266 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2267 ext4_msg(sb, KERN_ERR,
2268 "Cannot change data mode on remount");
2272 clear_opt(sb, DATA_FLAGS);
2273 sbi->s_mount_opt |= m->mount_opt;
2276 } else if (m->flags & MOPT_QFMT) {
2277 if (sb_any_quota_loaded(sb) &&
2278 sbi->s_jquota_fmt != m->mount_opt) {
2279 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2280 "quota options when quota turned on");
2283 if (ext4_has_feature_quota(sb)) {
2284 ext4_msg(sb, KERN_INFO,
2285 "Quota format mount options ignored "
2286 "when QUOTA feature is enabled");
2289 sbi->s_jquota_fmt = m->mount_opt;
2291 } else if (token == Opt_dax || token == Opt_dax_always ||
2292 token == Opt_dax_inode || token == Opt_dax_never) {
2293 #ifdef CONFIG_FS_DAX
2296 case Opt_dax_always:
2298 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2299 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2300 fail_dax_change_remount:
2301 ext4_msg(sb, KERN_ERR, "can't change "
2302 "dax mount option while remounting");
2306 (test_opt(sb, DATA_FLAGS) ==
2307 EXT4_MOUNT_JOURNAL_DATA)) {
2308 ext4_msg(sb, KERN_ERR, "can't mount with "
2309 "both data=journal and dax");
2312 ext4_msg(sb, KERN_WARNING,
2313 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2314 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2315 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2319 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2320 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2321 goto fail_dax_change_remount;
2322 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2323 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2327 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2328 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2329 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2330 goto fail_dax_change_remount;
2331 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2332 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2333 /* Strictly for printing options */
2334 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2338 ext4_msg(sb, KERN_INFO, "dax option not supported");
2339 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2340 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2343 } else if (token == Opt_data_err_abort) {
2344 sbi->s_mount_opt |= m->mount_opt;
2345 } else if (token == Opt_data_err_ignore) {
2346 sbi->s_mount_opt &= ~m->mount_opt;
2350 if (m->flags & MOPT_CLEAR)
2352 else if (unlikely(!(m->flags & MOPT_SET))) {
2353 ext4_msg(sb, KERN_WARNING,
2354 "buggy handling of option %s", opt);
2358 if (m->flags & MOPT_2) {
2360 sbi->s_mount_opt2 |= m->mount_opt;
2362 sbi->s_mount_opt2 &= ~m->mount_opt;
2365 sbi->s_mount_opt |= m->mount_opt;
2367 sbi->s_mount_opt &= ~m->mount_opt;
2373 static int parse_options(char *options, struct super_block *sb,
2374 unsigned long *journal_devnum,
2375 unsigned int *journal_ioprio,
2378 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2379 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2380 substring_t args[MAX_OPT_ARGS];
2386 while ((p = strsep(&options, ",")) != NULL) {
2390 * Initialize args struct so we know whether arg was
2391 * found; some options take optional arguments.
2393 args[0].to = args[0].from = NULL;
2394 token = match_token(p, tokens, args);
2395 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2396 journal_ioprio, is_remount) < 0)
2401 * We do the test below only for project quotas. 'usrquota' and
2402 * 'grpquota' mount options are allowed even without quota feature
2403 * to support legacy quotas in quota files.
2405 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2406 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2407 "Cannot enable project quota enforcement.");
2410 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2411 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2412 if (usr_qf_name || grp_qf_name) {
2413 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2414 clear_opt(sb, USRQUOTA);
2416 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2417 clear_opt(sb, GRPQUOTA);
2419 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2420 ext4_msg(sb, KERN_ERR, "old and new quota "
2425 if (!sbi->s_jquota_fmt) {
2426 ext4_msg(sb, KERN_ERR, "journaled quota format "
2432 if (test_opt(sb, DIOREAD_NOLOCK)) {
2434 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2435 if (blocksize < PAGE_SIZE)
2436 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2437 "experimental mount option 'dioread_nolock' "
2438 "for blocksize < PAGE_SIZE");
2443 static inline void ext4_show_quota_options(struct seq_file *seq,
2444 struct super_block *sb)
2446 #if defined(CONFIG_QUOTA)
2447 struct ext4_sb_info *sbi = EXT4_SB(sb);
2448 char *usr_qf_name, *grp_qf_name;
2450 if (sbi->s_jquota_fmt) {
2453 switch (sbi->s_jquota_fmt) {
2464 seq_printf(seq, ",jqfmt=%s", fmtname);
2468 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2469 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2471 seq_show_option(seq, "usrjquota", usr_qf_name);
2473 seq_show_option(seq, "grpjquota", grp_qf_name);
2478 static const char *token2str(int token)
2480 const struct match_token *t;
2482 for (t = tokens; t->token != Opt_err; t++)
2483 if (t->token == token && !strchr(t->pattern, '='))
2490 * - it's set to a non-default value OR
2491 * - if the per-sb default is different from the global default
2493 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2496 struct ext4_sb_info *sbi = EXT4_SB(sb);
2497 struct ext4_super_block *es = sbi->s_es;
2498 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2499 const struct mount_opts *m;
2500 char sep = nodefs ? '\n' : ',';
2502 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2503 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2505 if (sbi->s_sb_block != 1)
2506 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2508 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2509 int want_set = m->flags & MOPT_SET;
2510 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2511 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2513 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2514 continue; /* skip if same as the default */
2516 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2517 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2518 continue; /* select Opt_noFoo vs Opt_Foo */
2519 SEQ_OPTS_PRINT("%s", token2str(m->token));
2522 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2523 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2524 SEQ_OPTS_PRINT("resuid=%u",
2525 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2526 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2527 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2528 SEQ_OPTS_PRINT("resgid=%u",
2529 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2530 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2531 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2532 SEQ_OPTS_PUTS("errors=remount-ro");
2533 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2534 SEQ_OPTS_PUTS("errors=continue");
2535 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2536 SEQ_OPTS_PUTS("errors=panic");
2537 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2538 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2539 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2540 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2541 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2542 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2543 if (sb->s_flags & SB_I_VERSION)
2544 SEQ_OPTS_PUTS("i_version");
2545 if (nodefs || sbi->s_stripe)
2546 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2547 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2548 (sbi->s_mount_opt ^ def_mount_opt)) {
2549 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2550 SEQ_OPTS_PUTS("data=journal");
2551 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2552 SEQ_OPTS_PUTS("data=ordered");
2553 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2554 SEQ_OPTS_PUTS("data=writeback");
2557 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2558 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2559 sbi->s_inode_readahead_blks);
2561 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2562 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2563 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2564 if (nodefs || sbi->s_max_dir_size_kb)
2565 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2566 if (test_opt(sb, DATA_ERR_ABORT))
2567 SEQ_OPTS_PUTS("data_err=abort");
2569 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2571 if (sb->s_flags & SB_INLINECRYPT)
2572 SEQ_OPTS_PUTS("inlinecrypt");
2574 if (test_opt(sb, DAX_ALWAYS)) {
2576 SEQ_OPTS_PUTS("dax");
2578 SEQ_OPTS_PUTS("dax=always");
2579 } else if (test_opt2(sb, DAX_NEVER)) {
2580 SEQ_OPTS_PUTS("dax=never");
2581 } else if (test_opt2(sb, DAX_INODE)) {
2582 SEQ_OPTS_PUTS("dax=inode");
2584 ext4_show_quota_options(seq, sb);
2588 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2590 return _ext4_show_options(seq, root->d_sb, 0);
2593 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2595 struct super_block *sb = seq->private;
2598 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2599 rc = _ext4_show_options(seq, sb, 1);
2600 seq_puts(seq, "\n");
2604 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2607 struct ext4_sb_info *sbi = EXT4_SB(sb);
2610 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2611 ext4_msg(sb, KERN_ERR, "revision level too high, "
2612 "forcing read-only mode");
2618 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2619 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2620 "running e2fsck is recommended");
2621 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2622 ext4_msg(sb, KERN_WARNING,
2623 "warning: mounting fs with errors, "
2624 "running e2fsck is recommended");
2625 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2626 le16_to_cpu(es->s_mnt_count) >=
2627 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2628 ext4_msg(sb, KERN_WARNING,
2629 "warning: maximal mount count reached, "
2630 "running e2fsck is recommended");
2631 else if (le32_to_cpu(es->s_checkinterval) &&
2632 (ext4_get_tstamp(es, s_lastcheck) +
2633 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2634 ext4_msg(sb, KERN_WARNING,
2635 "warning: checktime reached, "
2636 "running e2fsck is recommended");
2637 if (!sbi->s_journal)
2638 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2639 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2640 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2641 le16_add_cpu(&es->s_mnt_count, 1);
2642 ext4_update_tstamp(es, s_mtime);
2644 ext4_set_feature_journal_needs_recovery(sb);
2646 err = ext4_commit_super(sb, 1);
2648 if (test_opt(sb, DEBUG))
2649 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2650 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2652 sbi->s_groups_count,
2653 EXT4_BLOCKS_PER_GROUP(sb),
2654 EXT4_INODES_PER_GROUP(sb),
2655 sbi->s_mount_opt, sbi->s_mount_opt2);
2657 cleancache_init_fs(sb);
2661 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2663 struct ext4_sb_info *sbi = EXT4_SB(sb);
2664 struct flex_groups **old_groups, **new_groups;
2667 if (!sbi->s_log_groups_per_flex)
2670 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2671 if (size <= sbi->s_flex_groups_allocated)
2674 new_groups = kvzalloc(roundup_pow_of_two(size *
2675 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2677 ext4_msg(sb, KERN_ERR,
2678 "not enough memory for %d flex group pointers", size);
2681 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2682 new_groups[i] = kvzalloc(roundup_pow_of_two(
2683 sizeof(struct flex_groups)),
2685 if (!new_groups[i]) {
2686 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2687 kvfree(new_groups[j]);
2689 ext4_msg(sb, KERN_ERR,
2690 "not enough memory for %d flex groups", size);
2695 old_groups = rcu_dereference(sbi->s_flex_groups);
2697 memcpy(new_groups, old_groups,
2698 (sbi->s_flex_groups_allocated *
2699 sizeof(struct flex_groups *)));
2701 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2702 sbi->s_flex_groups_allocated = size;
2704 ext4_kvfree_array_rcu(old_groups);
2708 static int ext4_fill_flex_info(struct super_block *sb)
2710 struct ext4_sb_info *sbi = EXT4_SB(sb);
2711 struct ext4_group_desc *gdp = NULL;
2712 struct flex_groups *fg;
2713 ext4_group_t flex_group;
2716 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2717 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2718 sbi->s_log_groups_per_flex = 0;
2722 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2726 for (i = 0; i < sbi->s_groups_count; i++) {
2727 gdp = ext4_get_group_desc(sb, i, NULL);
2729 flex_group = ext4_flex_group(sbi, i);
2730 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2731 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2732 atomic64_add(ext4_free_group_clusters(sb, gdp),
2733 &fg->free_clusters);
2734 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2742 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2743 struct ext4_group_desc *gdp)
2745 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2747 __le32 le_group = cpu_to_le32(block_group);
2748 struct ext4_sb_info *sbi = EXT4_SB(sb);
2750 if (ext4_has_metadata_csum(sbi->s_sb)) {
2751 /* Use new metadata_csum algorithm */
2753 __u16 dummy_csum = 0;
2755 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2757 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2758 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2759 sizeof(dummy_csum));
2760 offset += sizeof(dummy_csum);
2761 if (offset < sbi->s_desc_size)
2762 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2763 sbi->s_desc_size - offset);
2765 crc = csum32 & 0xFFFF;
2769 /* old crc16 code */
2770 if (!ext4_has_feature_gdt_csum(sb))
2773 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2774 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2775 crc = crc16(crc, (__u8 *)gdp, offset);
2776 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2777 /* for checksum of struct ext4_group_desc do the rest...*/
2778 if (ext4_has_feature_64bit(sb) &&
2779 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2780 crc = crc16(crc, (__u8 *)gdp + offset,
2781 le16_to_cpu(sbi->s_es->s_desc_size) -
2785 return cpu_to_le16(crc);
2788 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2789 struct ext4_group_desc *gdp)
2791 if (ext4_has_group_desc_csum(sb) &&
2792 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2798 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2799 struct ext4_group_desc *gdp)
2801 if (!ext4_has_group_desc_csum(sb))
2803 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2806 /* Called at mount-time, super-block is locked */
2807 static int ext4_check_descriptors(struct super_block *sb,
2808 ext4_fsblk_t sb_block,
2809 ext4_group_t *first_not_zeroed)
2811 struct ext4_sb_info *sbi = EXT4_SB(sb);
2812 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2813 ext4_fsblk_t last_block;
2814 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2815 ext4_fsblk_t block_bitmap;
2816 ext4_fsblk_t inode_bitmap;
2817 ext4_fsblk_t inode_table;
2818 int flexbg_flag = 0;
2819 ext4_group_t i, grp = sbi->s_groups_count;
2821 if (ext4_has_feature_flex_bg(sb))
2824 ext4_debug("Checking group descriptors");
2826 for (i = 0; i < sbi->s_groups_count; i++) {
2827 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2829 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2830 last_block = ext4_blocks_count(sbi->s_es) - 1;
2832 last_block = first_block +
2833 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2835 if ((grp == sbi->s_groups_count) &&
2836 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2839 block_bitmap = ext4_block_bitmap(sb, gdp);
2840 if (block_bitmap == sb_block) {
2841 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2842 "Block bitmap for group %u overlaps "
2847 if (block_bitmap >= sb_block + 1 &&
2848 block_bitmap <= last_bg_block) {
2849 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2850 "Block bitmap for group %u overlaps "
2851 "block group descriptors", i);
2855 if (block_bitmap < first_block || block_bitmap > last_block) {
2856 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2857 "Block bitmap for group %u not in group "
2858 "(block %llu)!", i, block_bitmap);
2861 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2862 if (inode_bitmap == sb_block) {
2863 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2864 "Inode bitmap for group %u overlaps "
2869 if (inode_bitmap >= sb_block + 1 &&
2870 inode_bitmap <= last_bg_block) {
2871 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2872 "Inode bitmap for group %u overlaps "
2873 "block group descriptors", i);
2877 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2878 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2879 "Inode bitmap for group %u not in group "
2880 "(block %llu)!", i, inode_bitmap);
2883 inode_table = ext4_inode_table(sb, gdp);
2884 if (inode_table == sb_block) {
2885 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2886 "Inode table for group %u overlaps "
2891 if (inode_table >= sb_block + 1 &&
2892 inode_table <= last_bg_block) {
2893 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2894 "Inode table for group %u overlaps "
2895 "block group descriptors", i);
2899 if (inode_table < first_block ||
2900 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2901 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2902 "Inode table for group %u not in group "
2903 "(block %llu)!", i, inode_table);
2906 ext4_lock_group(sb, i);
2907 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2908 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2909 "Checksum for group %u failed (%u!=%u)",
2910 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2911 gdp)), le16_to_cpu(gdp->bg_checksum));
2912 if (!sb_rdonly(sb)) {
2913 ext4_unlock_group(sb, i);
2917 ext4_unlock_group(sb, i);
2919 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2921 if (NULL != first_not_zeroed)
2922 *first_not_zeroed = grp;
2926 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2927 * the superblock) which were deleted from all directories, but held open by
2928 * a process at the time of a crash. We walk the list and try to delete these
2929 * inodes at recovery time (only with a read-write filesystem).
2931 * In order to keep the orphan inode chain consistent during traversal (in
2932 * case of crash during recovery), we link each inode into the superblock
2933 * orphan list_head and handle it the same way as an inode deletion during
2934 * normal operation (which journals the operations for us).
2936 * We only do an iget() and an iput() on each inode, which is very safe if we
2937 * accidentally point at an in-use or already deleted inode. The worst that
2938 * can happen in this case is that we get a "bit already cleared" message from
2939 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2940 * e2fsck was run on this filesystem, and it must have already done the orphan
2941 * inode cleanup for us, so we can safely abort without any further action.
2943 static void ext4_orphan_cleanup(struct super_block *sb,
2944 struct ext4_super_block *es)
2946 unsigned int s_flags = sb->s_flags;
2947 int ret, nr_orphans = 0, nr_truncates = 0;
2949 int quota_update = 0;
2952 if (!es->s_last_orphan) {
2953 jbd_debug(4, "no orphan inodes to clean up\n");
2957 if (bdev_read_only(sb->s_bdev)) {
2958 ext4_msg(sb, KERN_ERR, "write access "
2959 "unavailable, skipping orphan cleanup");
2963 /* Check if feature set would not allow a r/w mount */
2964 if (!ext4_feature_set_ok(sb, 0)) {
2965 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2966 "unknown ROCOMPAT features");
2970 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2971 /* don't clear list on RO mount w/ errors */
2972 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2973 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2974 "clearing orphan list.\n");
2975 es->s_last_orphan = 0;
2977 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2981 if (s_flags & SB_RDONLY) {
2982 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2983 sb->s_flags &= ~SB_RDONLY;
2986 /* Needed for iput() to work correctly and not trash data */
2987 sb->s_flags |= SB_ACTIVE;
2990 * Turn on quotas which were not enabled for read-only mounts if
2991 * filesystem has quota feature, so that they are updated correctly.
2993 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2994 int ret = ext4_enable_quotas(sb);
2999 ext4_msg(sb, KERN_ERR,
3000 "Cannot turn on quotas: error %d", ret);
3003 /* Turn on journaled quotas used for old sytle */
3004 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3005 if (EXT4_SB(sb)->s_qf_names[i]) {
3006 int ret = ext4_quota_on_mount(sb, i);
3011 ext4_msg(sb, KERN_ERR,
3012 "Cannot turn on journaled "
3013 "quota: type %d: error %d", i, ret);
3018 while (es->s_last_orphan) {
3019 struct inode *inode;
3022 * We may have encountered an error during cleanup; if
3023 * so, skip the rest.
3025 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3026 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3027 es->s_last_orphan = 0;
3031 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3032 if (IS_ERR(inode)) {
3033 es->s_last_orphan = 0;
3037 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3038 dquot_initialize(inode);
3039 if (inode->i_nlink) {
3040 if (test_opt(sb, DEBUG))
3041 ext4_msg(sb, KERN_DEBUG,
3042 "%s: truncating inode %lu to %lld bytes",
3043 __func__, inode->i_ino, inode->i_size);
3044 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3045 inode->i_ino, inode->i_size);
3047 truncate_inode_pages(inode->i_mapping, inode->i_size);
3048 ret = ext4_truncate(inode);
3050 ext4_std_error(inode->i_sb, ret);
3051 inode_unlock(inode);
3054 if (test_opt(sb, DEBUG))
3055 ext4_msg(sb, KERN_DEBUG,
3056 "%s: deleting unreferenced inode %lu",
3057 __func__, inode->i_ino);
3058 jbd_debug(2, "deleting unreferenced inode %lu\n",
3062 iput(inode); /* The delete magic happens here! */
3065 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3068 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3069 PLURAL(nr_orphans));
3071 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3072 PLURAL(nr_truncates));
3074 /* Turn off quotas if they were enabled for orphan cleanup */
3076 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3077 if (sb_dqopt(sb)->files[i])
3078 dquot_quota_off(sb, i);
3082 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3086 * Maximal extent format file size.
3087 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3088 * extent format containers, within a sector_t, and within i_blocks
3089 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3090 * so that won't be a limiting factor.
3092 * However there is other limiting factor. We do store extents in the form
3093 * of starting block and length, hence the resulting length of the extent
3094 * covering maximum file size must fit into on-disk format containers as
3095 * well. Given that length is always by 1 unit bigger than max unit (because
3096 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3098 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3100 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3103 loff_t upper_limit = MAX_LFS_FILESIZE;
3105 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3107 if (!has_huge_files) {
3108 upper_limit = (1LL << 32) - 1;
3110 /* total blocks in file system block size */
3111 upper_limit >>= (blkbits - 9);
3112 upper_limit <<= blkbits;
3116 * 32-bit extent-start container, ee_block. We lower the maxbytes
3117 * by one fs block, so ee_len can cover the extent of maximum file
3120 res = (1LL << 32) - 1;
3123 /* Sanity check against vm- & vfs- imposed limits */
3124 if (res > upper_limit)
3131 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3132 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3133 * We need to be 1 filesystem block less than the 2^48 sector limit.
3135 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3137 loff_t res = EXT4_NDIR_BLOCKS;
3140 /* This is calculated to be the largest file size for a dense, block
3141 * mapped file such that the file's total number of 512-byte sectors,
3142 * including data and all indirect blocks, does not exceed (2^48 - 1).
3144 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3145 * number of 512-byte sectors of the file.
3148 if (!has_huge_files) {
3150 * !has_huge_files or implies that the inode i_block field
3151 * represents total file blocks in 2^32 512-byte sectors ==
3152 * size of vfs inode i_blocks * 8
3154 upper_limit = (1LL << 32) - 1;
3156 /* total blocks in file system block size */
3157 upper_limit >>= (bits - 9);
3161 * We use 48 bit ext4_inode i_blocks
3162 * With EXT4_HUGE_FILE_FL set the i_blocks
3163 * represent total number of blocks in
3164 * file system block size
3166 upper_limit = (1LL << 48) - 1;
3170 /* indirect blocks */
3172 /* double indirect blocks */
3173 meta_blocks += 1 + (1LL << (bits-2));
3174 /* tripple indirect blocks */
3175 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3177 upper_limit -= meta_blocks;
3178 upper_limit <<= bits;
3180 res += 1LL << (bits-2);
3181 res += 1LL << (2*(bits-2));
3182 res += 1LL << (3*(bits-2));
3184 if (res > upper_limit)
3187 if (res > MAX_LFS_FILESIZE)
3188 res = MAX_LFS_FILESIZE;
3193 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3194 ext4_fsblk_t logical_sb_block, int nr)
3196 struct ext4_sb_info *sbi = EXT4_SB(sb);
3197 ext4_group_t bg, first_meta_bg;
3200 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3202 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3203 return logical_sb_block + nr + 1;
3204 bg = sbi->s_desc_per_block * nr;
3205 if (ext4_bg_has_super(sb, bg))
3209 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3210 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3211 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3214 if (sb->s_blocksize == 1024 && nr == 0 &&
3215 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3218 return (has_super + ext4_group_first_block_no(sb, bg));
3222 * ext4_get_stripe_size: Get the stripe size.
3223 * @sbi: In memory super block info
3225 * If we have specified it via mount option, then
3226 * use the mount option value. If the value specified at mount time is
3227 * greater than the blocks per group use the super block value.
3228 * If the super block value is greater than blocks per group return 0.
3229 * Allocator needs it be less than blocks per group.
3232 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3234 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3235 unsigned long stripe_width =
3236 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3239 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3240 ret = sbi->s_stripe;
3241 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3243 else if (stride && stride <= sbi->s_blocks_per_group)
3249 * If the stripe width is 1, this makes no sense and
3250 * we set it to 0 to turn off stripe handling code.
3259 * Check whether this filesystem can be mounted based on
3260 * the features present and the RDONLY/RDWR mount requested.
3261 * Returns 1 if this filesystem can be mounted as requested,
3262 * 0 if it cannot be.
3264 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3266 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3267 ext4_msg(sb, KERN_ERR,
3268 "Couldn't mount because of "
3269 "unsupported optional features (%x)",
3270 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3271 ~EXT4_FEATURE_INCOMPAT_SUPP));
3275 #ifndef CONFIG_UNICODE
3276 if (ext4_has_feature_casefold(sb)) {
3277 ext4_msg(sb, KERN_ERR,
3278 "Filesystem with casefold feature cannot be "
3279 "mounted without CONFIG_UNICODE");
3287 if (ext4_has_feature_readonly(sb)) {
3288 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3289 sb->s_flags |= SB_RDONLY;
3293 /* Check that feature set is OK for a read-write mount */
3294 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3295 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3296 "unsupported optional features (%x)",
3297 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3298 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3301 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3302 ext4_msg(sb, KERN_ERR,
3303 "Can't support bigalloc feature without "
3304 "extents feature\n");
3308 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3309 if (!readonly && (ext4_has_feature_quota(sb) ||
3310 ext4_has_feature_project(sb))) {
3311 ext4_msg(sb, KERN_ERR,
3312 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3315 #endif /* CONFIG_QUOTA */
3320 * This function is called once a day if we have errors logged
3321 * on the file system
3323 static void print_daily_error_info(struct timer_list *t)
3325 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3326 struct super_block *sb = sbi->s_sb;
3327 struct ext4_super_block *es = sbi->s_es;
3329 if (es->s_error_count)
3330 /* fsck newer than v1.41.13 is needed to clean this condition. */
3331 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3332 le32_to_cpu(es->s_error_count));
3333 if (es->s_first_error_time) {
3334 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3336 ext4_get_tstamp(es, s_first_error_time),
3337 (int) sizeof(es->s_first_error_func),
3338 es->s_first_error_func,
3339 le32_to_cpu(es->s_first_error_line));
3340 if (es->s_first_error_ino)
3341 printk(KERN_CONT ": inode %u",
3342 le32_to_cpu(es->s_first_error_ino));
3343 if (es->s_first_error_block)
3344 printk(KERN_CONT ": block %llu", (unsigned long long)
3345 le64_to_cpu(es->s_first_error_block));
3346 printk(KERN_CONT "\n");
3348 if (es->s_last_error_time) {
3349 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3351 ext4_get_tstamp(es, s_last_error_time),
3352 (int) sizeof(es->s_last_error_func),
3353 es->s_last_error_func,
3354 le32_to_cpu(es->s_last_error_line));
3355 if (es->s_last_error_ino)
3356 printk(KERN_CONT ": inode %u",
3357 le32_to_cpu(es->s_last_error_ino));
3358 if (es->s_last_error_block)
3359 printk(KERN_CONT ": block %llu", (unsigned long long)
3360 le64_to_cpu(es->s_last_error_block));
3361 printk(KERN_CONT "\n");
3363 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3366 /* Find next suitable group and run ext4_init_inode_table */
3367 static int ext4_run_li_request(struct ext4_li_request *elr)
3369 struct ext4_group_desc *gdp = NULL;
3370 struct super_block *sb = elr->lr_super;
3371 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3372 ext4_group_t group = elr->lr_next_group;
3373 unsigned long timeout = 0;
3374 unsigned int prefetch_ios = 0;
3377 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3378 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3379 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3381 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3383 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3385 if (group >= elr->lr_next_group) {
3387 if (elr->lr_first_not_zeroed != ngroups &&
3388 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3389 elr->lr_next_group = elr->lr_first_not_zeroed;
3390 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3397 for (; group < ngroups; group++) {
3398 gdp = ext4_get_group_desc(sb, group, NULL);
3404 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3408 if (group >= ngroups)
3413 ret = ext4_init_inode_table(sb, group,
3414 elr->lr_timeout ? 0 : 1);
3415 trace_ext4_lazy_itable_init(sb, group);
3416 if (elr->lr_timeout == 0) {
3417 timeout = (jiffies - timeout) *
3418 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3419 elr->lr_timeout = timeout;
3421 elr->lr_next_sched = jiffies + elr->lr_timeout;
3422 elr->lr_next_group = group + 1;
3428 * Remove lr_request from the list_request and free the
3429 * request structure. Should be called with li_list_mtx held
3431 static void ext4_remove_li_request(struct ext4_li_request *elr)
3436 list_del(&elr->lr_request);
3437 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3441 static void ext4_unregister_li_request(struct super_block *sb)
3443 mutex_lock(&ext4_li_mtx);
3444 if (!ext4_li_info) {
3445 mutex_unlock(&ext4_li_mtx);
3449 mutex_lock(&ext4_li_info->li_list_mtx);
3450 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3451 mutex_unlock(&ext4_li_info->li_list_mtx);
3452 mutex_unlock(&ext4_li_mtx);
3455 static struct task_struct *ext4_lazyinit_task;
3458 * This is the function where ext4lazyinit thread lives. It walks
3459 * through the request list searching for next scheduled filesystem.
3460 * When such a fs is found, run the lazy initialization request
3461 * (ext4_rn_li_request) and keep track of the time spend in this
3462 * function. Based on that time we compute next schedule time of
3463 * the request. When walking through the list is complete, compute
3464 * next waking time and put itself into sleep.
3466 static int ext4_lazyinit_thread(void *arg)
3468 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3469 struct list_head *pos, *n;
3470 struct ext4_li_request *elr;
3471 unsigned long next_wakeup, cur;
3473 BUG_ON(NULL == eli);
3477 next_wakeup = MAX_JIFFY_OFFSET;
3479 mutex_lock(&eli->li_list_mtx);
3480 if (list_empty(&eli->li_request_list)) {
3481 mutex_unlock(&eli->li_list_mtx);
3484 list_for_each_safe(pos, n, &eli->li_request_list) {
3487 elr = list_entry(pos, struct ext4_li_request,
3490 if (time_before(jiffies, elr->lr_next_sched)) {
3491 if (time_before(elr->lr_next_sched, next_wakeup))
3492 next_wakeup = elr->lr_next_sched;
3495 if (down_read_trylock(&elr->lr_super->s_umount)) {
3496 if (sb_start_write_trylock(elr->lr_super)) {
3499 * We hold sb->s_umount, sb can not
3500 * be removed from the list, it is
3501 * now safe to drop li_list_mtx
3503 mutex_unlock(&eli->li_list_mtx);
3504 err = ext4_run_li_request(elr);
3505 sb_end_write(elr->lr_super);
3506 mutex_lock(&eli->li_list_mtx);
3509 up_read((&elr->lr_super->s_umount));
3511 /* error, remove the lazy_init job */
3513 ext4_remove_li_request(elr);
3517 elr->lr_next_sched = jiffies +
3519 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3521 if (time_before(elr->lr_next_sched, next_wakeup))
3522 next_wakeup = elr->lr_next_sched;
3524 mutex_unlock(&eli->li_list_mtx);
3529 if ((time_after_eq(cur, next_wakeup)) ||
3530 (MAX_JIFFY_OFFSET == next_wakeup)) {
3535 schedule_timeout_interruptible(next_wakeup - cur);
3537 if (kthread_should_stop()) {
3538 ext4_clear_request_list();
3545 * It looks like the request list is empty, but we need
3546 * to check it under the li_list_mtx lock, to prevent any
3547 * additions into it, and of course we should lock ext4_li_mtx
3548 * to atomically free the list and ext4_li_info, because at
3549 * this point another ext4 filesystem could be registering
3552 mutex_lock(&ext4_li_mtx);
3553 mutex_lock(&eli->li_list_mtx);
3554 if (!list_empty(&eli->li_request_list)) {
3555 mutex_unlock(&eli->li_list_mtx);
3556 mutex_unlock(&ext4_li_mtx);
3559 mutex_unlock(&eli->li_list_mtx);
3560 kfree(ext4_li_info);
3561 ext4_li_info = NULL;
3562 mutex_unlock(&ext4_li_mtx);
3567 static void ext4_clear_request_list(void)
3569 struct list_head *pos, *n;
3570 struct ext4_li_request *elr;
3572 mutex_lock(&ext4_li_info->li_list_mtx);
3573 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3574 elr = list_entry(pos, struct ext4_li_request,
3576 ext4_remove_li_request(elr);
3578 mutex_unlock(&ext4_li_info->li_list_mtx);
3581 static int ext4_run_lazyinit_thread(void)
3583 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3584 ext4_li_info, "ext4lazyinit");
3585 if (IS_ERR(ext4_lazyinit_task)) {
3586 int err = PTR_ERR(ext4_lazyinit_task);
3587 ext4_clear_request_list();
3588 kfree(ext4_li_info);
3589 ext4_li_info = NULL;
3590 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3591 "initialization thread\n",
3595 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3600 * Check whether it make sense to run itable init. thread or not.
3601 * If there is at least one uninitialized inode table, return
3602 * corresponding group number, else the loop goes through all
3603 * groups and return total number of groups.
3605 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3607 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3608 struct ext4_group_desc *gdp = NULL;
3610 if (!ext4_has_group_desc_csum(sb))
3613 for (group = 0; group < ngroups; group++) {
3614 gdp = ext4_get_group_desc(sb, group, NULL);
3618 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3625 static int ext4_li_info_new(void)
3627 struct ext4_lazy_init *eli = NULL;
3629 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3633 INIT_LIST_HEAD(&eli->li_request_list);
3634 mutex_init(&eli->li_list_mtx);
3636 eli->li_state |= EXT4_LAZYINIT_QUIT;
3643 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3646 struct ext4_li_request *elr;
3648 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3653 elr->lr_first_not_zeroed = start;
3654 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3655 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3657 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3658 elr->lr_next_group = start;
3662 * Randomize first schedule time of the request to
3663 * spread the inode table initialization requests
3666 elr->lr_next_sched = jiffies + (prandom_u32() %
3667 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3671 int ext4_register_li_request(struct super_block *sb,
3672 ext4_group_t first_not_zeroed)
3674 struct ext4_sb_info *sbi = EXT4_SB(sb);
3675 struct ext4_li_request *elr = NULL;
3676 ext4_group_t ngroups = sbi->s_groups_count;
3679 mutex_lock(&ext4_li_mtx);
3680 if (sbi->s_li_request != NULL) {
3682 * Reset timeout so it can be computed again, because
3683 * s_li_wait_mult might have changed.
3685 sbi->s_li_request->lr_timeout = 0;
3689 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3690 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3691 !test_opt(sb, INIT_INODE_TABLE)))
3694 elr = ext4_li_request_new(sb, first_not_zeroed);
3700 if (NULL == ext4_li_info) {
3701 ret = ext4_li_info_new();
3706 mutex_lock(&ext4_li_info->li_list_mtx);
3707 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3708 mutex_unlock(&ext4_li_info->li_list_mtx);
3710 sbi->s_li_request = elr;
3712 * set elr to NULL here since it has been inserted to
3713 * the request_list and the removal and free of it is
3714 * handled by ext4_clear_request_list from now on.
3718 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3719 ret = ext4_run_lazyinit_thread();
3724 mutex_unlock(&ext4_li_mtx);
3731 * We do not need to lock anything since this is called on
3734 static void ext4_destroy_lazyinit_thread(void)
3737 * If thread exited earlier
3738 * there's nothing to be done.
3740 if (!ext4_li_info || !ext4_lazyinit_task)
3743 kthread_stop(ext4_lazyinit_task);
3746 static int set_journal_csum_feature_set(struct super_block *sb)
3749 int compat, incompat;
3750 struct ext4_sb_info *sbi = EXT4_SB(sb);
3752 if (ext4_has_metadata_csum(sb)) {
3753 /* journal checksum v3 */
3755 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3757 /* journal checksum v1 */
3758 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3762 jbd2_journal_clear_features(sbi->s_journal,
3763 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3764 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3765 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3766 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3767 ret = jbd2_journal_set_features(sbi->s_journal,
3769 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3771 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3772 ret = jbd2_journal_set_features(sbi->s_journal,
3775 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3776 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3778 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3779 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3786 * Note: calculating the overhead so we can be compatible with
3787 * historical BSD practice is quite difficult in the face of
3788 * clusters/bigalloc. This is because multiple metadata blocks from
3789 * different block group can end up in the same allocation cluster.
3790 * Calculating the exact overhead in the face of clustered allocation
3791 * requires either O(all block bitmaps) in memory or O(number of block
3792 * groups**2) in time. We will still calculate the superblock for
3793 * older file systems --- and if we come across with a bigalloc file
3794 * system with zero in s_overhead_clusters the estimate will be close to
3795 * correct especially for very large cluster sizes --- but for newer
3796 * file systems, it's better to calculate this figure once at mkfs
3797 * time, and store it in the superblock. If the superblock value is
3798 * present (even for non-bigalloc file systems), we will use it.
3800 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3803 struct ext4_sb_info *sbi = EXT4_SB(sb);
3804 struct ext4_group_desc *gdp;
3805 ext4_fsblk_t first_block, last_block, b;
3806 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3807 int s, j, count = 0;
3809 if (!ext4_has_feature_bigalloc(sb))
3810 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3811 sbi->s_itb_per_group + 2);
3813 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3814 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3815 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3816 for (i = 0; i < ngroups; i++) {
3817 gdp = ext4_get_group_desc(sb, i, NULL);
3818 b = ext4_block_bitmap(sb, gdp);
3819 if (b >= first_block && b <= last_block) {
3820 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3823 b = ext4_inode_bitmap(sb, gdp);
3824 if (b >= first_block && b <= last_block) {
3825 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3828 b = ext4_inode_table(sb, gdp);
3829 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3830 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3831 int c = EXT4_B2C(sbi, b - first_block);
3832 ext4_set_bit(c, buf);
3838 if (ext4_bg_has_super(sb, grp)) {
3839 ext4_set_bit(s++, buf);
3842 j = ext4_bg_num_gdb(sb, grp);
3843 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3844 ext4_error(sb, "Invalid number of block group "
3845 "descriptor blocks: %d", j);
3846 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3850 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3854 return EXT4_CLUSTERS_PER_GROUP(sb) -
3855 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3859 * Compute the overhead and stash it in sbi->s_overhead
3861 int ext4_calculate_overhead(struct super_block *sb)
3863 struct ext4_sb_info *sbi = EXT4_SB(sb);
3864 struct ext4_super_block *es = sbi->s_es;
3865 struct inode *j_inode;
3866 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3867 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3868 ext4_fsblk_t overhead = 0;
3869 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3875 * Compute the overhead (FS structures). This is constant
3876 * for a given filesystem unless the number of block groups
3877 * changes so we cache the previous value until it does.
3881 * All of the blocks before first_data_block are overhead
3883 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3886 * Add the overhead found in each block group
3888 for (i = 0; i < ngroups; i++) {
3891 blks = count_overhead(sb, i, buf);
3894 memset(buf, 0, PAGE_SIZE);
3899 * Add the internal journal blocks whether the journal has been
3902 if (sbi->s_journal && !sbi->s_journal_bdev)
3903 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3904 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3905 /* j_inum for internal journal is non-zero */
3906 j_inode = ext4_get_journal_inode(sb, j_inum);
3908 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3909 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3912 ext4_msg(sb, KERN_ERR, "can't get journal size");
3915 sbi->s_overhead = overhead;
3917 free_page((unsigned long) buf);
3921 static void ext4_set_resv_clusters(struct super_block *sb)
3923 ext4_fsblk_t resv_clusters;
3924 struct ext4_sb_info *sbi = EXT4_SB(sb);
3927 * There's no need to reserve anything when we aren't using extents.
3928 * The space estimates are exact, there are no unwritten extents,
3929 * hole punching doesn't need new metadata... This is needed especially
3930 * to keep ext2/3 backward compatibility.
3932 if (!ext4_has_feature_extents(sb))
3935 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3936 * This should cover the situations where we can not afford to run
3937 * out of space like for example punch hole, or converting
3938 * unwritten extents in delalloc path. In most cases such
3939 * allocation would require 1, or 2 blocks, higher numbers are
3942 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3943 sbi->s_cluster_bits);
3945 do_div(resv_clusters, 50);
3946 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3948 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3951 static const char *ext4_quota_mode(struct super_block *sb)
3954 if (!ext4_quota_capable(sb))
3957 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3958 return "journalled";
3966 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3968 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3969 char *orig_data = kstrdup(data, GFP_KERNEL);
3970 struct buffer_head *bh, **group_desc;
3971 struct ext4_super_block *es = NULL;
3972 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3973 struct flex_groups **flex_groups;
3975 ext4_fsblk_t sb_block = get_sb_block(&data);
3976 ext4_fsblk_t logical_sb_block;
3977 unsigned long offset = 0;
3978 unsigned long journal_devnum = 0;
3979 unsigned long def_mount_opts;
3983 int blocksize, clustersize;
3984 unsigned int db_count;
3986 int needs_recovery, has_huge_files;
3989 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3990 ext4_group_t first_not_zeroed;
3992 if ((data && !orig_data) || !sbi)
3995 sbi->s_daxdev = dax_dev;
3996 sbi->s_blockgroup_lock =
3997 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3998 if (!sbi->s_blockgroup_lock)
4001 sb->s_fs_info = sbi;
4003 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4004 sbi->s_sb_block = sb_block;
4005 if (sb->s_bdev->bd_part)
4006 sbi->s_sectors_written_start =
4007 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
4009 /* Cleanup superblock name */
4010 strreplace(sb->s_id, '/', '!');
4012 /* -EINVAL is default */
4014 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4016 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4021 * The ext4 superblock will not be buffer aligned for other than 1kB
4022 * block sizes. We need to calculate the offset from buffer start.
4024 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4025 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4026 offset = do_div(logical_sb_block, blocksize);
4028 logical_sb_block = sb_block;
4031 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4033 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4038 * Note: s_es must be initialized as soon as possible because
4039 * some ext4 macro-instructions depend on its value
4041 es = (struct ext4_super_block *) (bh->b_data + offset);
4043 sb->s_magic = le16_to_cpu(es->s_magic);
4044 if (sb->s_magic != EXT4_SUPER_MAGIC)
4046 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4048 /* Warn if metadata_csum and gdt_csum are both set. */
4049 if (ext4_has_feature_metadata_csum(sb) &&
4050 ext4_has_feature_gdt_csum(sb))
4051 ext4_warning(sb, "metadata_csum and uninit_bg are "
4052 "redundant flags; please run fsck.");
4054 /* Check for a known checksum algorithm */
4055 if (!ext4_verify_csum_type(sb, es)) {
4056 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4057 "unknown checksum algorithm.");
4062 /* Load the checksum driver */
4063 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4064 if (IS_ERR(sbi->s_chksum_driver)) {
4065 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4066 ret = PTR_ERR(sbi->s_chksum_driver);
4067 sbi->s_chksum_driver = NULL;
4071 /* Check superblock checksum */
4072 if (!ext4_superblock_csum_verify(sb, es)) {
4073 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4074 "invalid superblock checksum. Run e2fsck?");
4080 /* Precompute checksum seed for all metadata */
4081 if (ext4_has_feature_csum_seed(sb))
4082 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4083 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4084 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4085 sizeof(es->s_uuid));
4087 /* Set defaults before we parse the mount options */
4088 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4089 set_opt(sb, INIT_INODE_TABLE);
4090 if (def_mount_opts & EXT4_DEFM_DEBUG)
4092 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4094 if (def_mount_opts & EXT4_DEFM_UID16)
4095 set_opt(sb, NO_UID32);
4096 /* xattr user namespace & acls are now defaulted on */
4097 set_opt(sb, XATTR_USER);
4098 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4099 set_opt(sb, POSIX_ACL);
4101 if (ext4_has_feature_fast_commit(sb))
4102 set_opt2(sb, JOURNAL_FAST_COMMIT);
4103 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4104 if (ext4_has_metadata_csum(sb))
4105 set_opt(sb, JOURNAL_CHECKSUM);
4107 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4108 set_opt(sb, JOURNAL_DATA);
4109 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4110 set_opt(sb, ORDERED_DATA);
4111 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4112 set_opt(sb, WRITEBACK_DATA);
4114 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4115 set_opt(sb, ERRORS_PANIC);
4116 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4117 set_opt(sb, ERRORS_CONT);
4119 set_opt(sb, ERRORS_RO);
4120 /* block_validity enabled by default; disable with noblock_validity */
4121 set_opt(sb, BLOCK_VALIDITY);
4122 if (def_mount_opts & EXT4_DEFM_DISCARD)
4123 set_opt(sb, DISCARD);
4125 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4126 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4127 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4128 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4129 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4131 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4132 set_opt(sb, BARRIER);
4135 * enable delayed allocation by default
4136 * Use -o nodelalloc to turn it off
4138 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4139 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4140 set_opt(sb, DELALLOC);
4143 * set default s_li_wait_mult for lazyinit, for the case there is
4144 * no mount option specified.
4146 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4148 if (le32_to_cpu(es->s_log_block_size) >
4149 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4150 ext4_msg(sb, KERN_ERR,
4151 "Invalid log block size: %u",
4152 le32_to_cpu(es->s_log_block_size));
4155 if (le32_to_cpu(es->s_log_cluster_size) >
4156 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4157 ext4_msg(sb, KERN_ERR,
4158 "Invalid log cluster size: %u",
4159 le32_to_cpu(es->s_log_cluster_size));
4163 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4165 if (blocksize == PAGE_SIZE)
4166 set_opt(sb, DIOREAD_NOLOCK);
4168 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4169 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4170 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4172 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4173 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4174 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4175 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4179 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4180 (!is_power_of_2(sbi->s_inode_size)) ||
4181 (sbi->s_inode_size > blocksize)) {
4182 ext4_msg(sb, KERN_ERR,
4183 "unsupported inode size: %d",
4185 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4189 * i_atime_extra is the last extra field available for
4190 * [acm]times in struct ext4_inode. Checking for that
4191 * field should suffice to ensure we have extra space
4194 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4195 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4196 sb->s_time_gran = 1;
4197 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4199 sb->s_time_gran = NSEC_PER_SEC;
4200 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4202 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4204 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4205 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4206 EXT4_GOOD_OLD_INODE_SIZE;
4207 if (ext4_has_feature_extra_isize(sb)) {
4208 unsigned v, max = (sbi->s_inode_size -
4209 EXT4_GOOD_OLD_INODE_SIZE);
4211 v = le16_to_cpu(es->s_want_extra_isize);
4213 ext4_msg(sb, KERN_ERR,
4214 "bad s_want_extra_isize: %d", v);
4217 if (sbi->s_want_extra_isize < v)
4218 sbi->s_want_extra_isize = v;
4220 v = le16_to_cpu(es->s_min_extra_isize);
4222 ext4_msg(sb, KERN_ERR,
4223 "bad s_min_extra_isize: %d", v);
4226 if (sbi->s_want_extra_isize < v)
4227 sbi->s_want_extra_isize = v;
4231 if (sbi->s_es->s_mount_opts[0]) {
4232 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4233 sizeof(sbi->s_es->s_mount_opts),
4237 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4238 &journal_ioprio, 0)) {
4239 ext4_msg(sb, KERN_WARNING,
4240 "failed to parse options in superblock: %s",
4243 kfree(s_mount_opts);
4245 sbi->s_def_mount_opt = sbi->s_mount_opt;
4246 if (!parse_options((char *) data, sb, &journal_devnum,
4247 &journal_ioprio, 0))
4250 #ifdef CONFIG_UNICODE
4251 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4252 const struct ext4_sb_encodings *encoding_info;
4253 struct unicode_map *encoding;
4254 __u16 encoding_flags;
4256 if (ext4_has_feature_encrypt(sb)) {
4257 ext4_msg(sb, KERN_ERR,
4258 "Can't mount with encoding and encryption");
4262 if (ext4_sb_read_encoding(es, &encoding_info,
4264 ext4_msg(sb, KERN_ERR,
4265 "Encoding requested by superblock is unknown");
4269 encoding = utf8_load(encoding_info->version);
4270 if (IS_ERR(encoding)) {
4271 ext4_msg(sb, KERN_ERR,
4272 "can't mount with superblock charset: %s-%s "
4273 "not supported by the kernel. flags: 0x%x.",
4274 encoding_info->name, encoding_info->version,
4278 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4279 "%s-%s with flags 0x%hx", encoding_info->name,
4280 encoding_info->version?:"\b", encoding_flags);
4282 sb->s_encoding = encoding;
4283 sb->s_encoding_flags = encoding_flags;
4287 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4288 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4289 /* can't mount with both data=journal and dioread_nolock. */
4290 clear_opt(sb, DIOREAD_NOLOCK);
4291 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4292 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4293 ext4_msg(sb, KERN_ERR, "can't mount with "
4294 "both data=journal and delalloc");
4297 if (test_opt(sb, DAX_ALWAYS)) {
4298 ext4_msg(sb, KERN_ERR, "can't mount with "
4299 "both data=journal and dax");
4302 if (ext4_has_feature_encrypt(sb)) {
4303 ext4_msg(sb, KERN_WARNING,
4304 "encrypted files will use data=ordered "
4305 "instead of data journaling mode");
4307 if (test_opt(sb, DELALLOC))
4308 clear_opt(sb, DELALLOC);
4310 sb->s_iflags |= SB_I_CGROUPWB;
4313 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4314 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4316 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4317 (ext4_has_compat_features(sb) ||
4318 ext4_has_ro_compat_features(sb) ||
4319 ext4_has_incompat_features(sb)))
4320 ext4_msg(sb, KERN_WARNING,
4321 "feature flags set on rev 0 fs, "
4322 "running e2fsck is recommended");
4324 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4325 set_opt2(sb, HURD_COMPAT);
4326 if (ext4_has_feature_64bit(sb)) {
4327 ext4_msg(sb, KERN_ERR,
4328 "The Hurd can't support 64-bit file systems");
4333 * ea_inode feature uses l_i_version field which is not
4334 * available in HURD_COMPAT mode.
4336 if (ext4_has_feature_ea_inode(sb)) {
4337 ext4_msg(sb, KERN_ERR,
4338 "ea_inode feature is not supported for Hurd");
4343 if (IS_EXT2_SB(sb)) {
4344 if (ext2_feature_set_ok(sb))
4345 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4346 "using the ext4 subsystem");
4349 * If we're probing be silent, if this looks like
4350 * it's actually an ext[34] filesystem.
4352 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4354 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4355 "to feature incompatibilities");
4360 if (IS_EXT3_SB(sb)) {
4361 if (ext3_feature_set_ok(sb))
4362 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4363 "using the ext4 subsystem");
4366 * If we're probing be silent, if this looks like
4367 * it's actually an ext4 filesystem.
4369 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4371 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4372 "to feature incompatibilities");
4378 * Check feature flags regardless of the revision level, since we
4379 * previously didn't change the revision level when setting the flags,
4380 * so there is a chance incompat flags are set on a rev 0 filesystem.
4382 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4385 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4386 ext4_msg(sb, KERN_ERR,
4387 "Number of reserved GDT blocks insanely large: %d",
4388 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4392 if (bdev_dax_supported(sb->s_bdev, blocksize))
4393 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4395 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4396 if (ext4_has_feature_inline_data(sb)) {
4397 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4398 " that may contain inline data");
4401 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4402 ext4_msg(sb, KERN_ERR,
4403 "DAX unsupported by block device.");
4408 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4409 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4410 es->s_encryption_level);
4414 if (sb->s_blocksize != blocksize) {
4415 /* Validate the filesystem blocksize */
4416 if (!sb_set_blocksize(sb, blocksize)) {
4417 ext4_msg(sb, KERN_ERR, "bad block size %d",
4423 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4424 offset = do_div(logical_sb_block, blocksize);
4425 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4427 ext4_msg(sb, KERN_ERR,
4428 "Can't read superblock on 2nd try");
4433 es = (struct ext4_super_block *)(bh->b_data + offset);
4435 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4436 ext4_msg(sb, KERN_ERR,
4437 "Magic mismatch, very weird!");
4442 has_huge_files = ext4_has_feature_huge_file(sb);
4443 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4445 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4447 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4448 if (ext4_has_feature_64bit(sb)) {
4449 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4450 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4451 !is_power_of_2(sbi->s_desc_size)) {
4452 ext4_msg(sb, KERN_ERR,
4453 "unsupported descriptor size %lu",
4458 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4460 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4461 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4463 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4464 if (sbi->s_inodes_per_block == 0)
4466 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4467 sbi->s_inodes_per_group > blocksize * 8) {
4468 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4469 sbi->s_inodes_per_group);
4472 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4473 sbi->s_inodes_per_block;
4474 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4476 sbi->s_mount_state = le16_to_cpu(es->s_state);
4477 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4478 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4480 for (i = 0; i < 4; i++)
4481 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4482 sbi->s_def_hash_version = es->s_def_hash_version;
4483 if (ext4_has_feature_dir_index(sb)) {
4484 i = le32_to_cpu(es->s_flags);
4485 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4486 sbi->s_hash_unsigned = 3;
4487 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4488 #ifdef __CHAR_UNSIGNED__
4491 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4492 sbi->s_hash_unsigned = 3;
4496 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4501 /* Handle clustersize */
4502 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4503 if (ext4_has_feature_bigalloc(sb)) {
4504 if (clustersize < blocksize) {
4505 ext4_msg(sb, KERN_ERR,
4506 "cluster size (%d) smaller than "
4507 "block size (%d)", clustersize, blocksize);
4510 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4511 le32_to_cpu(es->s_log_block_size);
4512 sbi->s_clusters_per_group =
4513 le32_to_cpu(es->s_clusters_per_group);
4514 if (sbi->s_clusters_per_group > blocksize * 8) {
4515 ext4_msg(sb, KERN_ERR,
4516 "#clusters per group too big: %lu",
4517 sbi->s_clusters_per_group);
4520 if (sbi->s_blocks_per_group !=
4521 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4522 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4523 "clusters per group (%lu) inconsistent",
4524 sbi->s_blocks_per_group,
4525 sbi->s_clusters_per_group);
4529 if (clustersize != blocksize) {
4530 ext4_msg(sb, KERN_ERR,
4531 "fragment/cluster size (%d) != "
4532 "block size (%d)", clustersize, blocksize);
4535 if (sbi->s_blocks_per_group > blocksize * 8) {
4536 ext4_msg(sb, KERN_ERR,
4537 "#blocks per group too big: %lu",
4538 sbi->s_blocks_per_group);
4541 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4542 sbi->s_cluster_bits = 0;
4544 sbi->s_cluster_ratio = clustersize / blocksize;
4546 /* Do we have standard group size of clustersize * 8 blocks ? */
4547 if (sbi->s_blocks_per_group == clustersize << 3)
4548 set_opt2(sb, STD_GROUP_SIZE);
4551 * Test whether we have more sectors than will fit in sector_t,
4552 * and whether the max offset is addressable by the page cache.
4554 err = generic_check_addressable(sb->s_blocksize_bits,
4555 ext4_blocks_count(es));
4557 ext4_msg(sb, KERN_ERR, "filesystem"
4558 " too large to mount safely on this system");
4562 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4565 /* check blocks count against device size */
4566 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4567 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4568 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4569 "exceeds size of device (%llu blocks)",
4570 ext4_blocks_count(es), blocks_count);
4575 * It makes no sense for the first data block to be beyond the end
4576 * of the filesystem.
4578 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4579 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4580 "block %u is beyond end of filesystem (%llu)",
4581 le32_to_cpu(es->s_first_data_block),
4582 ext4_blocks_count(es));
4585 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4586 (sbi->s_cluster_ratio == 1)) {
4587 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4588 "block is 0 with a 1k block and cluster size");
4592 blocks_count = (ext4_blocks_count(es) -
4593 le32_to_cpu(es->s_first_data_block) +
4594 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4595 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4596 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4597 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4598 "(block count %llu, first data block %u, "
4599 "blocks per group %lu)", blocks_count,
4600 ext4_blocks_count(es),
4601 le32_to_cpu(es->s_first_data_block),
4602 EXT4_BLOCKS_PER_GROUP(sb));
4605 sbi->s_groups_count = blocks_count;
4606 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4607 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4608 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4609 le32_to_cpu(es->s_inodes_count)) {
4610 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4611 le32_to_cpu(es->s_inodes_count),
4612 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4616 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4617 EXT4_DESC_PER_BLOCK(sb);
4618 if (ext4_has_feature_meta_bg(sb)) {
4619 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4620 ext4_msg(sb, KERN_WARNING,
4621 "first meta block group too large: %u "
4622 "(group descriptor block count %u)",
4623 le32_to_cpu(es->s_first_meta_bg), db_count);
4627 rcu_assign_pointer(sbi->s_group_desc,
4628 kvmalloc_array(db_count,
4629 sizeof(struct buffer_head *),
4631 if (sbi->s_group_desc == NULL) {
4632 ext4_msg(sb, KERN_ERR, "not enough memory");
4637 bgl_lock_init(sbi->s_blockgroup_lock);
4639 /* Pre-read the descriptors into the buffer cache */
4640 for (i = 0; i < db_count; i++) {
4641 block = descriptor_loc(sb, logical_sb_block, i);
4642 ext4_sb_breadahead_unmovable(sb, block);
4645 for (i = 0; i < db_count; i++) {
4646 struct buffer_head *bh;
4648 block = descriptor_loc(sb, logical_sb_block, i);
4649 bh = ext4_sb_bread_unmovable(sb, block);
4651 ext4_msg(sb, KERN_ERR,
4652 "can't read group descriptor %d", i);
4658 rcu_dereference(sbi->s_group_desc)[i] = bh;
4661 sbi->s_gdb_count = db_count;
4662 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4663 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4664 ret = -EFSCORRUPTED;
4668 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4670 /* Register extent status tree shrinker */
4671 if (ext4_es_register_shrinker(sbi))
4674 sbi->s_stripe = ext4_get_stripe_size(sbi);
4675 sbi->s_extent_max_zeroout_kb = 32;
4678 * set up enough so that it can read an inode
4680 sb->s_op = &ext4_sops;
4681 sb->s_export_op = &ext4_export_ops;
4682 sb->s_xattr = ext4_xattr_handlers;
4683 #ifdef CONFIG_FS_ENCRYPTION
4684 sb->s_cop = &ext4_cryptops;
4686 #ifdef CONFIG_FS_VERITY
4687 sb->s_vop = &ext4_verityops;
4690 sb->dq_op = &ext4_quota_operations;
4691 if (ext4_has_feature_quota(sb))
4692 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4694 sb->s_qcop = &ext4_qctl_operations;
4695 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4697 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4699 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4700 mutex_init(&sbi->s_orphan_lock);
4702 /* Initialize fast commit stuff */
4703 atomic_set(&sbi->s_fc_subtid, 0);
4704 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4705 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4706 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4707 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4708 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4709 sbi->s_fc_bytes = 0;
4710 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4711 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4712 spin_lock_init(&sbi->s_fc_lock);
4713 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4714 sbi->s_fc_replay_state.fc_regions = NULL;
4715 sbi->s_fc_replay_state.fc_regions_size = 0;
4716 sbi->s_fc_replay_state.fc_regions_used = 0;
4717 sbi->s_fc_replay_state.fc_regions_valid = 0;
4718 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4719 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4720 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4724 needs_recovery = (es->s_last_orphan != 0 ||
4725 ext4_has_feature_journal_needs_recovery(sb));
4727 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4728 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4729 goto failed_mount3a;
4732 * The first inode we look at is the journal inode. Don't try
4733 * root first: it may be modified in the journal!
4735 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4736 err = ext4_load_journal(sb, es, journal_devnum);
4738 goto failed_mount3a;
4739 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4740 ext4_has_feature_journal_needs_recovery(sb)) {
4741 ext4_msg(sb, KERN_ERR, "required journal recovery "
4742 "suppressed and not mounted read-only");
4743 goto failed_mount_wq;
4745 /* Nojournal mode, all journal mount options are illegal */
4746 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4747 ext4_msg(sb, KERN_ERR, "can't mount with "
4748 "journal_checksum, fs mounted w/o journal");
4749 goto failed_mount_wq;
4751 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4752 ext4_msg(sb, KERN_ERR, "can't mount with "
4753 "journal_async_commit, fs mounted w/o journal");
4754 goto failed_mount_wq;
4756 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4757 ext4_msg(sb, KERN_ERR, "can't mount with "
4758 "commit=%lu, fs mounted w/o journal",
4759 sbi->s_commit_interval / HZ);
4760 goto failed_mount_wq;
4762 if (EXT4_MOUNT_DATA_FLAGS &
4763 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4764 ext4_msg(sb, KERN_ERR, "can't mount with "
4765 "data=, fs mounted w/o journal");
4766 goto failed_mount_wq;
4768 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4769 clear_opt(sb, JOURNAL_CHECKSUM);
4770 clear_opt(sb, DATA_FLAGS);
4771 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4772 sbi->s_journal = NULL;
4777 if (ext4_has_feature_64bit(sb) &&
4778 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4779 JBD2_FEATURE_INCOMPAT_64BIT)) {
4780 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4781 goto failed_mount_wq;
4784 if (!set_journal_csum_feature_set(sb)) {
4785 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4787 goto failed_mount_wq;
4790 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4791 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4792 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4793 ext4_msg(sb, KERN_ERR,
4794 "Failed to set fast commit journal feature");
4795 goto failed_mount_wq;
4798 /* We have now updated the journal if required, so we can
4799 * validate the data journaling mode. */
4800 switch (test_opt(sb, DATA_FLAGS)) {
4802 /* No mode set, assume a default based on the journal
4803 * capabilities: ORDERED_DATA if the journal can
4804 * cope, else JOURNAL_DATA
4806 if (jbd2_journal_check_available_features
4807 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4808 set_opt(sb, ORDERED_DATA);
4809 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4811 set_opt(sb, JOURNAL_DATA);
4812 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4816 case EXT4_MOUNT_ORDERED_DATA:
4817 case EXT4_MOUNT_WRITEBACK_DATA:
4818 if (!jbd2_journal_check_available_features
4819 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4820 ext4_msg(sb, KERN_ERR, "Journal does not support "
4821 "requested data journaling mode");
4822 goto failed_mount_wq;
4829 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4830 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4831 ext4_msg(sb, KERN_ERR, "can't mount with "
4832 "journal_async_commit in data=ordered mode");
4833 goto failed_mount_wq;
4836 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4838 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4839 sbi->s_journal->j_submit_inode_data_buffers =
4840 ext4_journal_submit_inode_data_buffers;
4841 sbi->s_journal->j_finish_inode_data_buffers =
4842 ext4_journal_finish_inode_data_buffers;
4845 if (!test_opt(sb, NO_MBCACHE)) {
4846 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4847 if (!sbi->s_ea_block_cache) {
4848 ext4_msg(sb, KERN_ERR,
4849 "Failed to create ea_block_cache");
4850 goto failed_mount_wq;
4853 if (ext4_has_feature_ea_inode(sb)) {
4854 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4855 if (!sbi->s_ea_inode_cache) {
4856 ext4_msg(sb, KERN_ERR,
4857 "Failed to create ea_inode_cache");
4858 goto failed_mount_wq;
4863 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4864 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4865 goto failed_mount_wq;
4868 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4869 !ext4_has_feature_encrypt(sb)) {
4870 ext4_set_feature_encrypt(sb);
4871 ext4_commit_super(sb, 1);
4875 * Get the # of file system overhead blocks from the
4876 * superblock if present.
4878 if (es->s_overhead_clusters)
4879 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4881 err = ext4_calculate_overhead(sb);
4883 goto failed_mount_wq;
4887 * The maximum number of concurrent works can be high and
4888 * concurrency isn't really necessary. Limit it to 1.
4890 EXT4_SB(sb)->rsv_conversion_wq =
4891 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4892 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4893 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4899 * The jbd2_journal_load will have done any necessary log recovery,
4900 * so we can safely mount the rest of the filesystem now.
4903 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4905 ext4_msg(sb, KERN_ERR, "get root inode failed");
4906 ret = PTR_ERR(root);
4910 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4911 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4916 #ifdef CONFIG_UNICODE
4918 sb->s_d_op = &ext4_dentry_ops;
4921 sb->s_root = d_make_root(root);
4923 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4928 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4929 if (ret == -EROFS) {
4930 sb->s_flags |= SB_RDONLY;
4933 goto failed_mount4a;
4935 ext4_set_resv_clusters(sb);
4937 if (test_opt(sb, BLOCK_VALIDITY)) {
4938 err = ext4_setup_system_zone(sb);
4940 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4942 goto failed_mount4a;
4945 ext4_fc_replay_cleanup(sb);
4948 err = ext4_mb_init(sb);
4950 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4955 block = ext4_count_free_clusters(sb);
4956 ext4_free_blocks_count_set(sbi->s_es,
4957 EXT4_C2B(sbi, block));
4958 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4961 unsigned long freei = ext4_count_free_inodes(sb);
4962 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4963 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4967 err = percpu_counter_init(&sbi->s_dirs_counter,
4968 ext4_count_dirs(sb), GFP_KERNEL);
4970 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4973 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4976 ext4_msg(sb, KERN_ERR, "insufficient memory");
4980 if (ext4_has_feature_flex_bg(sb))
4981 if (!ext4_fill_flex_info(sb)) {
4982 ext4_msg(sb, KERN_ERR,
4983 "unable to initialize "
4984 "flex_bg meta info!");
4988 err = ext4_register_li_request(sb, first_not_zeroed);
4992 err = ext4_register_sysfs(sb);
4997 /* Enable quota usage during mount. */
4998 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4999 err = ext4_enable_quotas(sb);
5003 #endif /* CONFIG_QUOTA */
5006 * Save the original bdev mapping's wb_err value which could be
5007 * used to detect the metadata async write error.
5009 spin_lock_init(&sbi->s_bdev_wb_lock);
5010 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5011 &sbi->s_bdev_wb_err);
5012 sb->s_bdev->bd_super = sb;
5013 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5014 ext4_orphan_cleanup(sb, es);
5015 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5016 if (needs_recovery) {
5017 ext4_msg(sb, KERN_INFO, "recovery complete");
5018 err = ext4_mark_recovery_complete(sb, es);
5022 if (EXT4_SB(sb)->s_journal) {
5023 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5024 descr = " journalled data mode";
5025 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5026 descr = " ordered data mode";
5028 descr = " writeback data mode";
5030 descr = "out journal";
5032 if (test_opt(sb, DISCARD)) {
5033 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5034 if (!blk_queue_discard(q))
5035 ext4_msg(sb, KERN_WARNING,
5036 "mounting with \"discard\" option, but "
5037 "the device does not support discard");
5040 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5041 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5042 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5043 (int) sizeof(sbi->s_es->s_mount_opts),
5044 sbi->s_es->s_mount_opts,
5045 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5046 ext4_quota_mode(sb));
5048 if (es->s_error_count)
5049 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5051 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5052 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5053 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5054 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5055 atomic_set(&sbi->s_warning_count, 0);
5056 atomic_set(&sbi->s_msg_count, 0);
5063 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5067 ext4_unregister_sysfs(sb);
5068 kobject_put(&sbi->s_kobj);
5070 ext4_unregister_li_request(sb);
5072 ext4_mb_release(sb);
5074 flex_groups = rcu_dereference(sbi->s_flex_groups);
5076 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5077 kvfree(flex_groups[i]);
5078 kvfree(flex_groups);
5081 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5082 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5083 percpu_counter_destroy(&sbi->s_dirs_counter);
5084 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5085 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5087 ext4_ext_release(sb);
5088 ext4_release_system_zone(sb);
5093 ext4_msg(sb, KERN_ERR, "mount failed");
5094 if (EXT4_SB(sb)->rsv_conversion_wq)
5095 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5097 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5098 sbi->s_ea_inode_cache = NULL;
5100 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5101 sbi->s_ea_block_cache = NULL;
5103 if (sbi->s_journal) {
5104 jbd2_journal_destroy(sbi->s_journal);
5105 sbi->s_journal = NULL;
5108 ext4_es_unregister_shrinker(sbi);
5110 del_timer_sync(&sbi->s_err_report);
5112 kthread_stop(sbi->s_mmp_tsk);
5115 group_desc = rcu_dereference(sbi->s_group_desc);
5116 for (i = 0; i < db_count; i++)
5117 brelse(group_desc[i]);
5121 if (sbi->s_chksum_driver)
5122 crypto_free_shash(sbi->s_chksum_driver);
5124 #ifdef CONFIG_UNICODE
5125 utf8_unload(sb->s_encoding);
5129 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5130 kfree(get_qf_name(sb, sbi, i));
5132 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5133 ext4_blkdev_remove(sbi);
5136 sb->s_fs_info = NULL;
5137 kfree(sbi->s_blockgroup_lock);
5141 fs_put_dax(dax_dev);
5142 return err ? err : ret;
5146 * Setup any per-fs journal parameters now. We'll do this both on
5147 * initial mount, once the journal has been initialised but before we've
5148 * done any recovery; and again on any subsequent remount.
5150 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5152 struct ext4_sb_info *sbi = EXT4_SB(sb);
5154 journal->j_commit_interval = sbi->s_commit_interval;
5155 journal->j_min_batch_time = sbi->s_min_batch_time;
5156 journal->j_max_batch_time = sbi->s_max_batch_time;
5157 ext4_fc_init(sb, journal);
5159 write_lock(&journal->j_state_lock);
5160 if (test_opt(sb, BARRIER))
5161 journal->j_flags |= JBD2_BARRIER;
5163 journal->j_flags &= ~JBD2_BARRIER;
5164 if (test_opt(sb, DATA_ERR_ABORT))
5165 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5167 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5168 write_unlock(&journal->j_state_lock);
5171 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5172 unsigned int journal_inum)
5174 struct inode *journal_inode;
5177 * Test for the existence of a valid inode on disk. Bad things
5178 * happen if we iget() an unused inode, as the subsequent iput()
5179 * will try to delete it.
5181 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5182 if (IS_ERR(journal_inode)) {
5183 ext4_msg(sb, KERN_ERR, "no journal found");
5186 if (!journal_inode->i_nlink) {
5187 make_bad_inode(journal_inode);
5188 iput(journal_inode);
5189 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5193 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5194 journal_inode, journal_inode->i_size);
5195 if (!S_ISREG(journal_inode->i_mode)) {
5196 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5197 iput(journal_inode);
5200 return journal_inode;
5203 static journal_t *ext4_get_journal(struct super_block *sb,
5204 unsigned int journal_inum)
5206 struct inode *journal_inode;
5209 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5212 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5216 journal = jbd2_journal_init_inode(journal_inode);
5218 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5219 iput(journal_inode);
5222 journal->j_private = sb;
5223 ext4_init_journal_params(sb, journal);
5227 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5230 struct buffer_head *bh;
5234 int hblock, blocksize;
5235 ext4_fsblk_t sb_block;
5236 unsigned long offset;
5237 struct ext4_super_block *es;
5238 struct block_device *bdev;
5240 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5243 bdev = ext4_blkdev_get(j_dev, sb);
5247 blocksize = sb->s_blocksize;
5248 hblock = bdev_logical_block_size(bdev);
5249 if (blocksize < hblock) {
5250 ext4_msg(sb, KERN_ERR,
5251 "blocksize too small for journal device");
5255 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5256 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5257 set_blocksize(bdev, blocksize);
5258 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5259 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5260 "external journal");
5264 es = (struct ext4_super_block *) (bh->b_data + offset);
5265 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5266 !(le32_to_cpu(es->s_feature_incompat) &
5267 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5268 ext4_msg(sb, KERN_ERR, "external journal has "
5274 if ((le32_to_cpu(es->s_feature_ro_compat) &
5275 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5276 es->s_checksum != ext4_superblock_csum(sb, es)) {
5277 ext4_msg(sb, KERN_ERR, "external journal has "
5278 "corrupt superblock");
5283 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5284 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5289 len = ext4_blocks_count(es);
5290 start = sb_block + 1;
5291 brelse(bh); /* we're done with the superblock */
5293 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5294 start, len, blocksize);
5296 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5299 journal->j_private = sb;
5300 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5301 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5304 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5305 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5306 "user (unsupported) - %d",
5307 be32_to_cpu(journal->j_superblock->s_nr_users));
5310 EXT4_SB(sb)->s_journal_bdev = bdev;
5311 ext4_init_journal_params(sb, journal);
5315 jbd2_journal_destroy(journal);
5317 ext4_blkdev_put(bdev);
5321 static int ext4_load_journal(struct super_block *sb,
5322 struct ext4_super_block *es,
5323 unsigned long journal_devnum)
5326 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5329 int really_read_only;
5332 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5333 return -EFSCORRUPTED;
5335 if (journal_devnum &&
5336 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5337 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5338 "numbers have changed");
5339 journal_dev = new_decode_dev(journal_devnum);
5341 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5343 if (journal_inum && journal_dev) {
5344 ext4_msg(sb, KERN_ERR,
5345 "filesystem has both journal inode and journal device!");
5350 journal = ext4_get_journal(sb, journal_inum);
5354 journal = ext4_get_dev_journal(sb, journal_dev);
5359 journal_dev_ro = bdev_read_only(journal->j_dev);
5360 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5362 if (journal_dev_ro && !sb_rdonly(sb)) {
5363 ext4_msg(sb, KERN_ERR,
5364 "journal device read-only, try mounting with '-o ro'");
5370 * Are we loading a blank journal or performing recovery after a
5371 * crash? For recovery, we need to check in advance whether we
5372 * can get read-write access to the device.
5374 if (ext4_has_feature_journal_needs_recovery(sb)) {
5375 if (sb_rdonly(sb)) {
5376 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5377 "required on readonly filesystem");
5378 if (really_read_only) {
5379 ext4_msg(sb, KERN_ERR, "write access "
5380 "unavailable, cannot proceed "
5381 "(try mounting with noload)");
5385 ext4_msg(sb, KERN_INFO, "write access will "
5386 "be enabled during recovery");
5390 if (!(journal->j_flags & JBD2_BARRIER))
5391 ext4_msg(sb, KERN_INFO, "barriers disabled");
5393 if (!ext4_has_feature_journal_needs_recovery(sb))
5394 err = jbd2_journal_wipe(journal, !really_read_only);
5396 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5398 memcpy(save, ((char *) es) +
5399 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5400 err = jbd2_journal_load(journal);
5402 memcpy(((char *) es) + EXT4_S_ERR_START,
5403 save, EXT4_S_ERR_LEN);
5408 ext4_msg(sb, KERN_ERR, "error loading journal");
5412 EXT4_SB(sb)->s_journal = journal;
5413 err = ext4_clear_journal_err(sb, es);
5415 EXT4_SB(sb)->s_journal = NULL;
5416 jbd2_journal_destroy(journal);
5420 if (!really_read_only && journal_devnum &&
5421 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5422 es->s_journal_dev = cpu_to_le32(journal_devnum);
5424 /* Make sure we flush the recovery flag to disk. */
5425 ext4_commit_super(sb, 1);
5431 jbd2_journal_destroy(journal);
5435 static int ext4_commit_super(struct super_block *sb, int sync)
5437 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5438 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5441 if (!sbh || block_device_ejected(sb))
5445 * If the file system is mounted read-only, don't update the
5446 * superblock write time. This avoids updating the superblock
5447 * write time when we are mounting the root file system
5448 * read/only but we need to replay the journal; at that point,
5449 * for people who are east of GMT and who make their clock
5450 * tick in localtime for Windows bug-for-bug compatibility,
5451 * the clock is set in the future, and this will cause e2fsck
5452 * to complain and force a full file system check.
5454 if (!(sb->s_flags & SB_RDONLY))
5455 ext4_update_tstamp(es, s_wtime);
5456 if (sb->s_bdev->bd_part)
5457 es->s_kbytes_written =
5458 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5459 ((part_stat_read(sb->s_bdev->bd_part,
5460 sectors[STAT_WRITE]) -
5461 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5463 es->s_kbytes_written =
5464 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5465 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5466 ext4_free_blocks_count_set(es,
5467 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5468 &EXT4_SB(sb)->s_freeclusters_counter)));
5469 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5470 es->s_free_inodes_count =
5471 cpu_to_le32(percpu_counter_sum_positive(
5472 &EXT4_SB(sb)->s_freeinodes_counter));
5473 BUFFER_TRACE(sbh, "marking dirty");
5474 ext4_superblock_csum_set(sb);
5477 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5479 * Oh, dear. A previous attempt to write the
5480 * superblock failed. This could happen because the
5481 * USB device was yanked out. Or it could happen to
5482 * be a transient write error and maybe the block will
5483 * be remapped. Nothing we can do but to retry the
5484 * write and hope for the best.
5486 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5487 "superblock detected");
5488 clear_buffer_write_io_error(sbh);
5489 set_buffer_uptodate(sbh);
5491 mark_buffer_dirty(sbh);
5494 error = __sync_dirty_buffer(sbh,
5495 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5496 if (buffer_write_io_error(sbh)) {
5497 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5499 clear_buffer_write_io_error(sbh);
5500 set_buffer_uptodate(sbh);
5507 * Have we just finished recovery? If so, and if we are mounting (or
5508 * remounting) the filesystem readonly, then we will end up with a
5509 * consistent fs on disk. Record that fact.
5511 static int ext4_mark_recovery_complete(struct super_block *sb,
5512 struct ext4_super_block *es)
5515 journal_t *journal = EXT4_SB(sb)->s_journal;
5517 if (!ext4_has_feature_journal(sb)) {
5518 if (journal != NULL) {
5519 ext4_error(sb, "Journal got removed while the fs was "
5521 return -EFSCORRUPTED;
5525 jbd2_journal_lock_updates(journal);
5526 err = jbd2_journal_flush(journal);
5530 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5531 ext4_clear_feature_journal_needs_recovery(sb);
5532 ext4_commit_super(sb, 1);
5535 jbd2_journal_unlock_updates(journal);
5540 * If we are mounting (or read-write remounting) a filesystem whose journal
5541 * has recorded an error from a previous lifetime, move that error to the
5542 * main filesystem now.
5544 static int ext4_clear_journal_err(struct super_block *sb,
5545 struct ext4_super_block *es)
5551 if (!ext4_has_feature_journal(sb)) {
5552 ext4_error(sb, "Journal got removed while the fs was mounted!");
5553 return -EFSCORRUPTED;
5556 journal = EXT4_SB(sb)->s_journal;
5559 * Now check for any error status which may have been recorded in the
5560 * journal by a prior ext4_error() or ext4_abort()
5563 j_errno = jbd2_journal_errno(journal);
5567 errstr = ext4_decode_error(sb, j_errno, nbuf);
5568 ext4_warning(sb, "Filesystem error recorded "
5569 "from previous mount: %s", errstr);
5570 ext4_warning(sb, "Marking fs in need of filesystem check.");
5572 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5573 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5574 ext4_commit_super(sb, 1);
5576 jbd2_journal_clear_err(journal);
5577 jbd2_journal_update_sb_errno(journal);
5583 * Force the running and committing transactions to commit,
5584 * and wait on the commit.
5586 int ext4_force_commit(struct super_block *sb)
5593 journal = EXT4_SB(sb)->s_journal;
5594 return ext4_journal_force_commit(journal);
5597 static int ext4_sync_fs(struct super_block *sb, int wait)
5601 bool needs_barrier = false;
5602 struct ext4_sb_info *sbi = EXT4_SB(sb);
5604 if (unlikely(ext4_forced_shutdown(sbi)))
5607 trace_ext4_sync_fs(sb, wait);
5608 flush_workqueue(sbi->rsv_conversion_wq);
5610 * Writeback quota in non-journalled quota case - journalled quota has
5613 dquot_writeback_dquots(sb, -1);
5615 * Data writeback is possible w/o journal transaction, so barrier must
5616 * being sent at the end of the function. But we can skip it if
5617 * transaction_commit will do it for us.
5619 if (sbi->s_journal) {
5620 target = jbd2_get_latest_transaction(sbi->s_journal);
5621 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5622 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5623 needs_barrier = true;
5625 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5627 ret = jbd2_log_wait_commit(sbi->s_journal,
5630 } else if (wait && test_opt(sb, BARRIER))
5631 needs_barrier = true;
5632 if (needs_barrier) {
5634 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5643 * LVM calls this function before a (read-only) snapshot is created. This
5644 * gives us a chance to flush the journal completely and mark the fs clean.
5646 * Note that only this function cannot bring a filesystem to be in a clean
5647 * state independently. It relies on upper layer to stop all data & metadata
5650 static int ext4_freeze(struct super_block *sb)
5658 journal = EXT4_SB(sb)->s_journal;
5661 /* Now we set up the journal barrier. */
5662 jbd2_journal_lock_updates(journal);
5665 * Don't clear the needs_recovery flag if we failed to
5666 * flush the journal.
5668 error = jbd2_journal_flush(journal);
5672 /* Journal blocked and flushed, clear needs_recovery flag. */
5673 ext4_clear_feature_journal_needs_recovery(sb);
5676 error = ext4_commit_super(sb, 1);
5679 /* we rely on upper layer to stop further updates */
5680 jbd2_journal_unlock_updates(journal);
5685 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5686 * flag here, even though the filesystem is not technically dirty yet.
5688 static int ext4_unfreeze(struct super_block *sb)
5690 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5693 if (EXT4_SB(sb)->s_journal) {
5694 /* Reset the needs_recovery flag before the fs is unlocked. */
5695 ext4_set_feature_journal_needs_recovery(sb);
5698 ext4_commit_super(sb, 1);
5703 * Structure to save mount options for ext4_remount's benefit
5705 struct ext4_mount_options {
5706 unsigned long s_mount_opt;
5707 unsigned long s_mount_opt2;
5710 unsigned long s_commit_interval;
5711 u32 s_min_batch_time, s_max_batch_time;
5714 char *s_qf_names[EXT4_MAXQUOTAS];
5718 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5720 struct ext4_super_block *es;
5721 struct ext4_sb_info *sbi = EXT4_SB(sb);
5722 unsigned long old_sb_flags, vfs_flags;
5723 struct ext4_mount_options old_opts;
5724 int enable_quota = 0;
5726 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5730 char *to_free[EXT4_MAXQUOTAS];
5732 char *orig_data = kstrdup(data, GFP_KERNEL);
5734 if (data && !orig_data)
5737 /* Store the original options */
5738 old_sb_flags = sb->s_flags;
5739 old_opts.s_mount_opt = sbi->s_mount_opt;
5740 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5741 old_opts.s_resuid = sbi->s_resuid;
5742 old_opts.s_resgid = sbi->s_resgid;
5743 old_opts.s_commit_interval = sbi->s_commit_interval;
5744 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5745 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5747 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5748 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5749 if (sbi->s_qf_names[i]) {
5750 char *qf_name = get_qf_name(sb, sbi, i);
5752 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5753 if (!old_opts.s_qf_names[i]) {
5754 for (j = 0; j < i; j++)
5755 kfree(old_opts.s_qf_names[j]);
5760 old_opts.s_qf_names[i] = NULL;
5762 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5763 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5766 * Some options can be enabled by ext4 and/or by VFS mount flag
5767 * either way we need to make sure it matches in both *flags and
5768 * s_flags. Copy those selected flags from *flags to s_flags
5770 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5771 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5773 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5778 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5779 test_opt(sb, JOURNAL_CHECKSUM)) {
5780 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5781 "during remount not supported; ignoring");
5782 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5785 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5786 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5787 ext4_msg(sb, KERN_ERR, "can't mount with "
5788 "both data=journal and delalloc");
5792 if (test_opt(sb, DIOREAD_NOLOCK)) {
5793 ext4_msg(sb, KERN_ERR, "can't mount with "
5794 "both data=journal and dioread_nolock");
5798 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5799 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5800 ext4_msg(sb, KERN_ERR, "can't mount with "
5801 "journal_async_commit in data=ordered mode");
5807 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5808 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5813 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5814 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5816 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5817 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5821 if (sbi->s_journal) {
5822 ext4_init_journal_params(sb, sbi->s_journal);
5823 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5826 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5827 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5832 if (*flags & SB_RDONLY) {
5833 err = sync_filesystem(sb);
5836 err = dquot_suspend(sb, -1);
5841 * First of all, the unconditional stuff we have to do
5842 * to disable replay of the journal when we next remount
5844 sb->s_flags |= SB_RDONLY;
5847 * OK, test if we are remounting a valid rw partition
5848 * readonly, and if so set the rdonly flag and then
5849 * mark the partition as valid again.
5851 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5852 (sbi->s_mount_state & EXT4_VALID_FS))
5853 es->s_state = cpu_to_le16(sbi->s_mount_state);
5855 if (sbi->s_journal) {
5857 * We let remount-ro finish even if marking fs
5858 * as clean failed...
5860 ext4_mark_recovery_complete(sb, es);
5863 kthread_stop(sbi->s_mmp_tsk);
5865 /* Make sure we can mount this feature set readwrite */
5866 if (ext4_has_feature_readonly(sb) ||
5867 !ext4_feature_set_ok(sb, 0)) {
5872 * Make sure the group descriptor checksums
5873 * are sane. If they aren't, refuse to remount r/w.
5875 for (g = 0; g < sbi->s_groups_count; g++) {
5876 struct ext4_group_desc *gdp =
5877 ext4_get_group_desc(sb, g, NULL);
5879 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5880 ext4_msg(sb, KERN_ERR,
5881 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5882 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5883 le16_to_cpu(gdp->bg_checksum));
5890 * If we have an unprocessed orphan list hanging
5891 * around from a previously readonly bdev mount,
5892 * require a full umount/remount for now.
5894 if (es->s_last_orphan) {
5895 ext4_msg(sb, KERN_WARNING, "Couldn't "
5896 "remount RDWR because of unprocessed "
5897 "orphan inode list. Please "
5898 "umount/remount instead");
5904 * Mounting a RDONLY partition read-write, so reread
5905 * and store the current valid flag. (It may have
5906 * been changed by e2fsck since we originally mounted
5909 if (sbi->s_journal) {
5910 err = ext4_clear_journal_err(sb, es);
5914 sbi->s_mount_state = le16_to_cpu(es->s_state);
5916 err = ext4_setup_super(sb, es, 0);
5920 sb->s_flags &= ~SB_RDONLY;
5921 if (ext4_has_feature_mmp(sb))
5922 if (ext4_multi_mount_protect(sb,
5923 le64_to_cpu(es->s_mmp_block))) {
5932 * Reinitialize lazy itable initialization thread based on
5935 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5936 ext4_unregister_li_request(sb);
5938 ext4_group_t first_not_zeroed;
5939 first_not_zeroed = ext4_has_uninit_itable(sb);
5940 ext4_register_li_request(sb, first_not_zeroed);
5944 * Handle creation of system zone data early because it can fail.
5945 * Releasing of existing data is done when we are sure remount will
5948 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5949 err = ext4_setup_system_zone(sb);
5954 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5955 err = ext4_commit_super(sb, 1);
5961 /* Release old quota file names */
5962 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5963 kfree(old_opts.s_qf_names[i]);
5965 if (sb_any_quota_suspended(sb))
5966 dquot_resume(sb, -1);
5967 else if (ext4_has_feature_quota(sb)) {
5968 err = ext4_enable_quotas(sb);
5974 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5975 ext4_release_system_zone(sb);
5978 * Some options can be enabled by ext4 and/or by VFS mount flag
5979 * either way we need to make sure it matches in both *flags and
5980 * s_flags. Copy those selected flags from s_flags to *flags
5982 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5984 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
5985 orig_data, ext4_quota_mode(sb));
5990 sb->s_flags = old_sb_flags;
5991 sbi->s_mount_opt = old_opts.s_mount_opt;
5992 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5993 sbi->s_resuid = old_opts.s_resuid;
5994 sbi->s_resgid = old_opts.s_resgid;
5995 sbi->s_commit_interval = old_opts.s_commit_interval;
5996 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5997 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5998 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5999 ext4_release_system_zone(sb);
6001 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6002 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6003 to_free[i] = get_qf_name(sb, sbi, i);
6004 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6007 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6015 static int ext4_statfs_project(struct super_block *sb,
6016 kprojid_t projid, struct kstatfs *buf)
6019 struct dquot *dquot;
6023 qid = make_kqid_projid(projid);
6024 dquot = dqget(sb, qid);
6026 return PTR_ERR(dquot);
6027 spin_lock(&dquot->dq_dqb_lock);
6029 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6030 dquot->dq_dqb.dqb_bhardlimit);
6031 limit >>= sb->s_blocksize_bits;
6033 if (limit && buf->f_blocks > limit) {
6034 curblock = (dquot->dq_dqb.dqb_curspace +
6035 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6036 buf->f_blocks = limit;
6037 buf->f_bfree = buf->f_bavail =
6038 (buf->f_blocks > curblock) ?
6039 (buf->f_blocks - curblock) : 0;
6042 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6043 dquot->dq_dqb.dqb_ihardlimit);
6044 if (limit && buf->f_files > limit) {
6045 buf->f_files = limit;
6047 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6048 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6051 spin_unlock(&dquot->dq_dqb_lock);
6057 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6059 struct super_block *sb = dentry->d_sb;
6060 struct ext4_sb_info *sbi = EXT4_SB(sb);
6061 struct ext4_super_block *es = sbi->s_es;
6062 ext4_fsblk_t overhead = 0, resv_blocks;
6065 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6067 if (!test_opt(sb, MINIX_DF))
6068 overhead = sbi->s_overhead;
6070 buf->f_type = EXT4_SUPER_MAGIC;
6071 buf->f_bsize = sb->s_blocksize;
6072 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6073 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6074 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6075 /* prevent underflow in case that few free space is available */
6076 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6077 buf->f_bavail = buf->f_bfree -
6078 (ext4_r_blocks_count(es) + resv_blocks);
6079 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6081 buf->f_files = le32_to_cpu(es->s_inodes_count);
6082 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6083 buf->f_namelen = EXT4_NAME_LEN;
6084 fsid = le64_to_cpup((void *)es->s_uuid) ^
6085 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6086 buf->f_fsid = u64_to_fsid(fsid);
6089 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6090 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6091 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6100 * Helper functions so that transaction is started before we acquire dqio_sem
6101 * to keep correct lock ordering of transaction > dqio_sem
6103 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6105 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6108 static int ext4_write_dquot(struct dquot *dquot)
6112 struct inode *inode;
6114 inode = dquot_to_inode(dquot);
6115 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6116 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6118 return PTR_ERR(handle);
6119 ret = dquot_commit(dquot);
6120 err = ext4_journal_stop(handle);
6126 static int ext4_acquire_dquot(struct dquot *dquot)
6131 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6132 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6134 return PTR_ERR(handle);
6135 ret = dquot_acquire(dquot);
6136 err = ext4_journal_stop(handle);
6142 static int ext4_release_dquot(struct dquot *dquot)
6147 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6148 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6149 if (IS_ERR(handle)) {
6150 /* Release dquot anyway to avoid endless cycle in dqput() */
6151 dquot_release(dquot);
6152 return PTR_ERR(handle);
6154 ret = dquot_release(dquot);
6155 err = ext4_journal_stop(handle);
6161 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6163 struct super_block *sb = dquot->dq_sb;
6165 if (ext4_is_quota_journalled(sb)) {
6166 dquot_mark_dquot_dirty(dquot);
6167 return ext4_write_dquot(dquot);
6169 return dquot_mark_dquot_dirty(dquot);
6173 static int ext4_write_info(struct super_block *sb, int type)
6178 /* Data block + inode block */
6179 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6181 return PTR_ERR(handle);
6182 ret = dquot_commit_info(sb, type);
6183 err = ext4_journal_stop(handle);
6190 * Turn on quotas during mount time - we need to find
6191 * the quota file and such...
6193 static int ext4_quota_on_mount(struct super_block *sb, int type)
6195 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6196 EXT4_SB(sb)->s_jquota_fmt, type);
6199 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6201 struct ext4_inode_info *ei = EXT4_I(inode);
6203 /* The first argument of lockdep_set_subclass has to be
6204 * *exactly* the same as the argument to init_rwsem() --- in
6205 * this case, in init_once() --- or lockdep gets unhappy
6206 * because the name of the lock is set using the
6207 * stringification of the argument to init_rwsem().
6209 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6210 lockdep_set_subclass(&ei->i_data_sem, subclass);
6214 * Standard function to be called on quota_on
6216 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6217 const struct path *path)
6221 if (!test_opt(sb, QUOTA))
6224 /* Quotafile not on the same filesystem? */
6225 if (path->dentry->d_sb != sb)
6228 /* Quota already enabled for this file? */
6229 if (IS_NOQUOTA(d_inode(path->dentry)))
6232 /* Journaling quota? */
6233 if (EXT4_SB(sb)->s_qf_names[type]) {
6234 /* Quotafile not in fs root? */
6235 if (path->dentry->d_parent != sb->s_root)
6236 ext4_msg(sb, KERN_WARNING,
6237 "Quota file not on filesystem root. "
6238 "Journaled quota will not work");
6239 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6242 * Clear the flag just in case mount options changed since
6245 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6249 * When we journal data on quota file, we have to flush journal to see
6250 * all updates to the file when we bypass pagecache...
6252 if (EXT4_SB(sb)->s_journal &&
6253 ext4_should_journal_data(d_inode(path->dentry))) {
6255 * We don't need to lock updates but journal_flush() could
6256 * otherwise be livelocked...
6258 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6259 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6260 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6265 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6266 err = dquot_quota_on(sb, type, format_id, path);
6268 lockdep_set_quota_inode(path->dentry->d_inode,
6271 struct inode *inode = d_inode(path->dentry);
6275 * Set inode flags to prevent userspace from messing with quota
6276 * files. If this fails, we return success anyway since quotas
6277 * are already enabled and this is not a hard failure.
6280 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6283 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6284 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6285 S_NOATIME | S_IMMUTABLE);
6286 err = ext4_mark_inode_dirty(handle, inode);
6287 ext4_journal_stop(handle);
6289 inode_unlock(inode);
6294 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6298 struct inode *qf_inode;
6299 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6300 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6301 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6302 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6305 BUG_ON(!ext4_has_feature_quota(sb));
6307 if (!qf_inums[type])
6310 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6311 if (IS_ERR(qf_inode)) {
6312 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6313 return PTR_ERR(qf_inode);
6316 /* Don't account quota for quota files to avoid recursion */
6317 qf_inode->i_flags |= S_NOQUOTA;
6318 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6319 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6321 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6327 /* Enable usage tracking for all quota types. */
6328 static int ext4_enable_quotas(struct super_block *sb)
6331 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6332 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6333 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6334 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6336 bool quota_mopt[EXT4_MAXQUOTAS] = {
6337 test_opt(sb, USRQUOTA),
6338 test_opt(sb, GRPQUOTA),
6339 test_opt(sb, PRJQUOTA),
6342 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6343 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6344 if (qf_inums[type]) {
6345 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6346 DQUOT_USAGE_ENABLED |
6347 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6350 "Failed to enable quota tracking "
6351 "(type=%d, err=%d). Please run "
6352 "e2fsck to fix.", type, err);
6353 for (type--; type >= 0; type--)
6354 dquot_quota_off(sb, type);
6363 static int ext4_quota_off(struct super_block *sb, int type)
6365 struct inode *inode = sb_dqopt(sb)->files[type];
6369 /* Force all delayed allocation blocks to be allocated.
6370 * Caller already holds s_umount sem */
6371 if (test_opt(sb, DELALLOC))
6372 sync_filesystem(sb);
6374 if (!inode || !igrab(inode))
6377 err = dquot_quota_off(sb, type);
6378 if (err || ext4_has_feature_quota(sb))
6383 * Update modification times of quota files when userspace can
6384 * start looking at them. If we fail, we return success anyway since
6385 * this is not a hard failure and quotas are already disabled.
6387 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6388 if (IS_ERR(handle)) {
6389 err = PTR_ERR(handle);
6392 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6393 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6394 inode->i_mtime = inode->i_ctime = current_time(inode);
6395 err = ext4_mark_inode_dirty(handle, inode);
6396 ext4_journal_stop(handle);
6398 inode_unlock(inode);
6400 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6404 return dquot_quota_off(sb, type);
6407 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6408 * acquiring the locks... As quota files are never truncated and quota code
6409 * itself serializes the operations (and no one else should touch the files)
6410 * we don't have to be afraid of races */
6411 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6412 size_t len, loff_t off)
6414 struct inode *inode = sb_dqopt(sb)->files[type];
6415 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6416 int offset = off & (sb->s_blocksize - 1);
6419 struct buffer_head *bh;
6420 loff_t i_size = i_size_read(inode);
6424 if (off+len > i_size)
6427 while (toread > 0) {
6428 tocopy = sb->s_blocksize - offset < toread ?
6429 sb->s_blocksize - offset : toread;
6430 bh = ext4_bread(NULL, inode, blk, 0);
6433 if (!bh) /* A hole? */
6434 memset(data, 0, tocopy);
6436 memcpy(data, bh->b_data+offset, tocopy);
6446 /* Write to quotafile (we know the transaction is already started and has
6447 * enough credits) */
6448 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6449 const char *data, size_t len, loff_t off)
6451 struct inode *inode = sb_dqopt(sb)->files[type];
6452 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6453 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6455 struct buffer_head *bh;
6456 handle_t *handle = journal_current_handle();
6458 if (EXT4_SB(sb)->s_journal && !handle) {
6459 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6460 " cancelled because transaction is not started",
6461 (unsigned long long)off, (unsigned long long)len);
6465 * Since we account only one data block in transaction credits,
6466 * then it is impossible to cross a block boundary.
6468 if (sb->s_blocksize - offset < len) {
6469 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6470 " cancelled because not block aligned",
6471 (unsigned long long)off, (unsigned long long)len);
6476 bh = ext4_bread(handle, inode, blk,
6477 EXT4_GET_BLOCKS_CREATE |
6478 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6479 } while (PTR_ERR(bh) == -ENOSPC &&
6480 ext4_should_retry_alloc(inode->i_sb, &retries));
6485 BUFFER_TRACE(bh, "get write access");
6486 err = ext4_journal_get_write_access(handle, bh);
6492 memcpy(bh->b_data+offset, data, len);
6493 flush_dcache_page(bh->b_page);
6495 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6498 if (inode->i_size < off + len) {
6499 i_size_write(inode, off + len);
6500 EXT4_I(inode)->i_disksize = inode->i_size;
6501 err2 = ext4_mark_inode_dirty(handle, inode);
6502 if (unlikely(err2 && !err))
6505 return err ? err : len;
6509 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6510 const char *dev_name, void *data)
6512 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6515 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6516 static inline void register_as_ext2(void)
6518 int err = register_filesystem(&ext2_fs_type);
6521 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6524 static inline void unregister_as_ext2(void)
6526 unregister_filesystem(&ext2_fs_type);
6529 static inline int ext2_feature_set_ok(struct super_block *sb)
6531 if (ext4_has_unknown_ext2_incompat_features(sb))
6535 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6540 static inline void register_as_ext2(void) { }
6541 static inline void unregister_as_ext2(void) { }
6542 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6545 static inline void register_as_ext3(void)
6547 int err = register_filesystem(&ext3_fs_type);
6550 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6553 static inline void unregister_as_ext3(void)
6555 unregister_filesystem(&ext3_fs_type);
6558 static inline int ext3_feature_set_ok(struct super_block *sb)
6560 if (ext4_has_unknown_ext3_incompat_features(sb))
6562 if (!ext4_has_feature_journal(sb))
6566 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6571 static struct file_system_type ext4_fs_type = {
6572 .owner = THIS_MODULE,
6574 .mount = ext4_mount,
6575 .kill_sb = kill_block_super,
6576 .fs_flags = FS_REQUIRES_DEV,
6578 MODULE_ALIAS_FS("ext4");
6580 /* Shared across all ext4 file systems */
6581 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6583 static int __init ext4_init_fs(void)
6587 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6588 ext4_li_info = NULL;
6589 mutex_init(&ext4_li_mtx);
6591 /* Build-time check for flags consistency */
6592 ext4_check_flag_values();
6594 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6595 init_waitqueue_head(&ext4__ioend_wq[i]);
6597 err = ext4_init_es();
6601 err = ext4_init_pending();
6605 err = ext4_init_post_read_processing();
6609 err = ext4_init_pageio();
6613 err = ext4_init_system_zone();
6617 err = ext4_init_sysfs();
6621 err = ext4_init_mballoc();
6624 err = init_inodecache();
6628 err = ext4_fc_init_dentry_cache();
6634 err = register_filesystem(&ext4_fs_type);
6640 unregister_as_ext2();
6641 unregister_as_ext3();
6643 destroy_inodecache();
6645 ext4_exit_mballoc();
6649 ext4_exit_system_zone();
6653 ext4_exit_post_read_processing();
6655 ext4_exit_pending();
6662 static void __exit ext4_exit_fs(void)
6664 ext4_destroy_lazyinit_thread();
6665 unregister_as_ext2();
6666 unregister_as_ext3();
6667 unregister_filesystem(&ext4_fs_type);
6668 destroy_inodecache();
6669 ext4_exit_mballoc();
6671 ext4_exit_system_zone();
6673 ext4_exit_post_read_processing();
6675 ext4_exit_pending();
6678 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6679 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6680 MODULE_LICENSE("GPL");
6681 MODULE_SOFTDEP("pre: crc32c");
6682 module_init(ext4_init_fs)
6683 module_exit(ext4_exit_fs)