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)
426 static void __save_error_info(struct super_block *sb, int error,
427 __u32 ino, __u64 block,
428 const char *func, unsigned int line)
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
433 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
434 if (bdev_read_only(sb->s_bdev))
436 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
437 ext4_update_tstamp(es, s_last_error_time);
438 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
439 es->s_last_error_line = cpu_to_le32(line);
440 es->s_last_error_ino = cpu_to_le32(ino);
441 es->s_last_error_block = cpu_to_le64(block);
447 err = EXT4_ERR_ENOMEM;
450 err = EXT4_ERR_EFSBADCRC;
454 err = EXT4_ERR_EFSCORRUPTED;
457 err = EXT4_ERR_ENOSPC;
460 err = EXT4_ERR_ENOKEY;
463 err = EXT4_ERR_EROFS;
466 err = EXT4_ERR_EFBIG;
469 err = EXT4_ERR_EEXIST;
472 err = EXT4_ERR_ERANGE;
475 err = EXT4_ERR_EOVERFLOW;
478 err = EXT4_ERR_EBUSY;
481 err = EXT4_ERR_ENOTDIR;
484 err = EXT4_ERR_ENOTEMPTY;
487 err = EXT4_ERR_ESHUTDOWN;
490 err = EXT4_ERR_EFAULT;
493 err = EXT4_ERR_UNKNOWN;
495 es->s_last_error_errcode = err;
496 if (!es->s_first_error_time) {
497 es->s_first_error_time = es->s_last_error_time;
498 es->s_first_error_time_hi = es->s_last_error_time_hi;
499 strncpy(es->s_first_error_func, func,
500 sizeof(es->s_first_error_func));
501 es->s_first_error_line = cpu_to_le32(line);
502 es->s_first_error_ino = es->s_last_error_ino;
503 es->s_first_error_block = es->s_last_error_block;
504 es->s_first_error_errcode = es->s_last_error_errcode;
507 * Start the daily error reporting function if it hasn't been
510 if (!es->s_error_count)
511 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
512 le32_add_cpu(&es->s_error_count, 1);
515 static void save_error_info(struct super_block *sb, int error,
516 __u32 ino, __u64 block,
517 const char *func, unsigned int line)
519 __save_error_info(sb, error, ino, block, func, line);
520 if (!bdev_read_only(sb->s_bdev))
521 ext4_commit_super(sb, 1);
525 * The del_gendisk() function uninitializes the disk-specific data
526 * structures, including the bdi structure, without telling anyone
527 * else. Once this happens, any attempt to call mark_buffer_dirty()
528 * (for example, by ext4_commit_super), will cause a kernel OOPS.
529 * This is a kludge to prevent these oops until we can put in a proper
530 * hook in del_gendisk() to inform the VFS and file system layers.
532 static int block_device_ejected(struct super_block *sb)
534 struct inode *bd_inode = sb->s_bdev->bd_inode;
535 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
537 return bdi->dev == NULL;
540 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
542 struct super_block *sb = journal->j_private;
543 struct ext4_sb_info *sbi = EXT4_SB(sb);
544 int error = is_journal_aborted(journal);
545 struct ext4_journal_cb_entry *jce;
547 BUG_ON(txn->t_state == T_FINISHED);
549 ext4_process_freed_data(sb, txn->t_tid);
551 spin_lock(&sbi->s_md_lock);
552 while (!list_empty(&txn->t_private_list)) {
553 jce = list_entry(txn->t_private_list.next,
554 struct ext4_journal_cb_entry, jce_list);
555 list_del_init(&jce->jce_list);
556 spin_unlock(&sbi->s_md_lock);
557 jce->jce_func(sb, jce, error);
558 spin_lock(&sbi->s_md_lock);
560 spin_unlock(&sbi->s_md_lock);
564 * This writepage callback for write_cache_pages()
565 * takes care of a few cases after page cleaning.
567 * write_cache_pages() already checks for dirty pages
568 * and calls clear_page_dirty_for_io(), which we want,
569 * to write protect the pages.
571 * However, we may have to redirty a page (see below.)
573 static int ext4_journalled_writepage_callback(struct page *page,
574 struct writeback_control *wbc,
577 transaction_t *transaction = (transaction_t *) data;
578 struct buffer_head *bh, *head;
579 struct journal_head *jh;
581 bh = head = page_buffers(page);
584 * We have to redirty a page in these cases:
585 * 1) If buffer is dirty, it means the page was dirty because it
586 * contains a buffer that needs checkpointing. So the dirty bit
587 * needs to be preserved so that checkpointing writes the buffer
589 * 2) If buffer is not part of the committing transaction
590 * (we may have just accidentally come across this buffer because
591 * inode range tracking is not exact) or if the currently running
592 * transaction already contains this buffer as well, dirty bit
593 * needs to be preserved so that the buffer gets writeprotected
594 * properly on running transaction's commit.
597 if (buffer_dirty(bh) ||
598 (jh && (jh->b_transaction != transaction ||
599 jh->b_next_transaction))) {
600 redirty_page_for_writepage(wbc, page);
603 } while ((bh = bh->b_this_page) != head);
606 return AOP_WRITEPAGE_ACTIVATE;
609 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
611 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
612 struct writeback_control wbc = {
613 .sync_mode = WB_SYNC_ALL,
614 .nr_to_write = LONG_MAX,
615 .range_start = jinode->i_dirty_start,
616 .range_end = jinode->i_dirty_end,
619 return write_cache_pages(mapping, &wbc,
620 ext4_journalled_writepage_callback,
621 jinode->i_transaction);
624 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
628 if (ext4_should_journal_data(jinode->i_vfs_inode))
629 ret = ext4_journalled_submit_inode_data_buffers(jinode);
631 ret = jbd2_journal_submit_inode_data_buffers(jinode);
636 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
640 if (!ext4_should_journal_data(jinode->i_vfs_inode))
641 ret = jbd2_journal_finish_inode_data_buffers(jinode);
646 static bool system_going_down(void)
648 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
649 || system_state == SYSTEM_RESTART;
652 /* Deal with the reporting of failure conditions on a filesystem such as
653 * inconsistencies detected or read IO failures.
655 * On ext2, we can store the error state of the filesystem in the
656 * superblock. That is not possible on ext4, because we may have other
657 * write ordering constraints on the superblock which prevent us from
658 * writing it out straight away; and given that the journal is about to
659 * be aborted, we can't rely on the current, or future, transactions to
660 * write out the superblock safely.
662 * We'll just use the jbd2_journal_abort() error code to record an error in
663 * the journal instead. On recovery, the journal will complain about
664 * that error until we've noted it down and cleared it.
666 * If force_ro is set, we unconditionally force the filesystem into an
667 * ABORT|READONLY state, unless the error response on the fs has been set to
668 * panic in which case we take the easy way out and panic immediately. This is
669 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
670 * at a critical moment in log management.
672 static void ext4_handle_error(struct super_block *sb, bool force_ro)
674 journal_t *journal = EXT4_SB(sb)->s_journal;
676 if (test_opt(sb, WARN_ON_ERROR))
679 if (sb_rdonly(sb) || (!force_ro && test_opt(sb, ERRORS_CONT)))
682 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
684 jbd2_journal_abort(journal, -EIO);
686 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
687 * could panic during 'reboot -f' as the underlying device got already
690 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
691 panic("EXT4-fs (device %s): panic forced after error\n",
694 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
696 * Make sure updated value of ->s_mount_flags will be visible before
700 sb->s_flags |= SB_RDONLY;
703 #define ext4_error_ratelimit(sb) \
704 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
707 void __ext4_error(struct super_block *sb, const char *function,
708 unsigned int line, bool force_ro, int error, __u64 block,
709 const char *fmt, ...)
711 struct va_format vaf;
714 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
717 trace_ext4_error(sb, function, line);
718 if (ext4_error_ratelimit(sb)) {
723 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
724 sb->s_id, function, line, current->comm, &vaf);
727 save_error_info(sb, error, 0, block, function, line);
728 ext4_handle_error(sb, force_ro);
731 void __ext4_error_inode(struct inode *inode, const char *function,
732 unsigned int line, ext4_fsblk_t block, int error,
733 const char *fmt, ...)
736 struct va_format vaf;
738 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
741 trace_ext4_error(inode->i_sb, function, line);
742 if (ext4_error_ratelimit(inode->i_sb)) {
747 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
748 "inode #%lu: block %llu: comm %s: %pV\n",
749 inode->i_sb->s_id, function, line, inode->i_ino,
750 block, current->comm, &vaf);
752 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
753 "inode #%lu: comm %s: %pV\n",
754 inode->i_sb->s_id, function, line, inode->i_ino,
755 current->comm, &vaf);
758 save_error_info(inode->i_sb, error, inode->i_ino, block,
760 ext4_handle_error(inode->i_sb, false);
763 void __ext4_error_file(struct file *file, const char *function,
764 unsigned int line, ext4_fsblk_t block,
765 const char *fmt, ...)
768 struct va_format vaf;
769 struct inode *inode = file_inode(file);
770 char pathname[80], *path;
772 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
775 trace_ext4_error(inode->i_sb, function, line);
776 if (ext4_error_ratelimit(inode->i_sb)) {
777 path = file_path(file, pathname, sizeof(pathname));
785 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
786 "block %llu: comm %s: path %s: %pV\n",
787 inode->i_sb->s_id, function, line, inode->i_ino,
788 block, current->comm, path, &vaf);
791 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
792 "comm %s: path %s: %pV\n",
793 inode->i_sb->s_id, function, line, inode->i_ino,
794 current->comm, path, &vaf);
797 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
799 ext4_handle_error(inode->i_sb, false);
802 const char *ext4_decode_error(struct super_block *sb, int errno,
809 errstr = "Corrupt filesystem";
812 errstr = "Filesystem failed CRC";
815 errstr = "IO failure";
818 errstr = "Out of memory";
821 if (!sb || (EXT4_SB(sb)->s_journal &&
822 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
823 errstr = "Journal has aborted";
825 errstr = "Readonly filesystem";
828 /* If the caller passed in an extra buffer for unknown
829 * errors, textualise them now. Else we just return
832 /* Check for truncated error codes... */
833 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
842 /* __ext4_std_error decodes expected errors from journaling functions
843 * automatically and invokes the appropriate error response. */
845 void __ext4_std_error(struct super_block *sb, const char *function,
846 unsigned int line, int errno)
851 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
854 /* Special case: if the error is EROFS, and we're not already
855 * inside a transaction, then there's really no point in logging
857 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
860 if (ext4_error_ratelimit(sb)) {
861 errstr = ext4_decode_error(sb, errno, nbuf);
862 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
863 sb->s_id, function, line, errstr);
866 save_error_info(sb, -errno, 0, 0, function, line);
867 ext4_handle_error(sb, false);
870 void __ext4_msg(struct super_block *sb,
871 const char *prefix, const char *fmt, ...)
873 struct va_format vaf;
876 atomic_inc(&EXT4_SB(sb)->s_msg_count);
877 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
883 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
887 static int ext4_warning_ratelimit(struct super_block *sb)
889 atomic_inc(&EXT4_SB(sb)->s_warning_count);
890 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
894 void __ext4_warning(struct super_block *sb, const char *function,
895 unsigned int line, const char *fmt, ...)
897 struct va_format vaf;
900 if (!ext4_warning_ratelimit(sb))
906 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
907 sb->s_id, function, line, &vaf);
911 void __ext4_warning_inode(const struct inode *inode, const char *function,
912 unsigned int line, const char *fmt, ...)
914 struct va_format vaf;
917 if (!ext4_warning_ratelimit(inode->i_sb))
923 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
924 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
925 function, line, inode->i_ino, current->comm, &vaf);
929 void __ext4_grp_locked_error(const char *function, unsigned int line,
930 struct super_block *sb, ext4_group_t grp,
931 unsigned long ino, ext4_fsblk_t block,
932 const char *fmt, ...)
936 struct va_format vaf;
939 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
942 trace_ext4_error(sb, function, line);
943 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
945 if (ext4_error_ratelimit(sb)) {
949 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
950 sb->s_id, function, line, grp);
952 printk(KERN_CONT "inode %lu: ", ino);
954 printk(KERN_CONT "block %llu:",
955 (unsigned long long) block);
956 printk(KERN_CONT "%pV\n", &vaf);
960 if (test_opt(sb, WARN_ON_ERROR))
963 if (test_opt(sb, ERRORS_CONT)) {
964 ext4_commit_super(sb, 0);
968 ext4_unlock_group(sb, grp);
969 ext4_commit_super(sb, 1);
970 ext4_handle_error(sb, false);
972 * We only get here in the ERRORS_RO case; relocking the group
973 * may be dangerous, but nothing bad will happen since the
974 * filesystem will have already been marked read/only and the
975 * journal has been aborted. We return 1 as a hint to callers
976 * who might what to use the return value from
977 * ext4_grp_locked_error() to distinguish between the
978 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
979 * aggressively from the ext4 function in question, with a
980 * more appropriate error code.
982 ext4_lock_group(sb, grp);
986 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
990 struct ext4_sb_info *sbi = EXT4_SB(sb);
991 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
992 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
995 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
996 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
999 percpu_counter_sub(&sbi->s_freeclusters_counter,
1003 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1004 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1009 count = ext4_free_inodes_count(sb, gdp);
1010 percpu_counter_sub(&sbi->s_freeinodes_counter,
1016 void ext4_update_dynamic_rev(struct super_block *sb)
1018 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1020 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1024 "updating to rev %d because of new feature flag, "
1025 "running e2fsck is recommended",
1028 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1029 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1030 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1031 /* leave es->s_feature_*compat flags alone */
1032 /* es->s_uuid will be set by e2fsck if empty */
1035 * The rest of the superblock fields should be zero, and if not it
1036 * means they are likely already in use, so leave them alone. We
1037 * can leave it up to e2fsck to clean up any inconsistencies there.
1042 * Open the external journal device
1044 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1046 struct block_device *bdev;
1048 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1054 ext4_msg(sb, KERN_ERR,
1055 "failed to open journal device unknown-block(%u,%u) %ld",
1056 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1061 * Release the journal device
1063 static void ext4_blkdev_put(struct block_device *bdev)
1065 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1068 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1070 struct block_device *bdev;
1071 bdev = sbi->s_journal_bdev;
1073 ext4_blkdev_put(bdev);
1074 sbi->s_journal_bdev = NULL;
1078 static inline struct inode *orphan_list_entry(struct list_head *l)
1080 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1083 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1085 struct list_head *l;
1087 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1088 le32_to_cpu(sbi->s_es->s_last_orphan));
1090 printk(KERN_ERR "sb_info orphan list:\n");
1091 list_for_each(l, &sbi->s_orphan) {
1092 struct inode *inode = orphan_list_entry(l);
1094 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1095 inode->i_sb->s_id, inode->i_ino, inode,
1096 inode->i_mode, inode->i_nlink,
1097 NEXT_ORPHAN(inode));
1102 static int ext4_quota_off(struct super_block *sb, int type);
1104 static inline void ext4_quota_off_umount(struct super_block *sb)
1108 /* Use our quota_off function to clear inode flags etc. */
1109 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1110 ext4_quota_off(sb, type);
1114 * This is a helper function which is used in the mount/remount
1115 * codepaths (which holds s_umount) to fetch the quota file name.
1117 static inline char *get_qf_name(struct super_block *sb,
1118 struct ext4_sb_info *sbi,
1121 return rcu_dereference_protected(sbi->s_qf_names[type],
1122 lockdep_is_held(&sb->s_umount));
1125 static inline void ext4_quota_off_umount(struct super_block *sb)
1130 static void ext4_put_super(struct super_block *sb)
1132 struct ext4_sb_info *sbi = EXT4_SB(sb);
1133 struct ext4_super_block *es = sbi->s_es;
1134 struct buffer_head **group_desc;
1135 struct flex_groups **flex_groups;
1139 ext4_unregister_li_request(sb);
1140 ext4_quota_off_umount(sb);
1142 destroy_workqueue(sbi->rsv_conversion_wq);
1145 * Unregister sysfs before destroying jbd2 journal.
1146 * Since we could still access attr_journal_task attribute via sysfs
1147 * path which could have sbi->s_journal->j_task as NULL
1149 ext4_unregister_sysfs(sb);
1151 if (sbi->s_journal) {
1152 aborted = is_journal_aborted(sbi->s_journal);
1153 err = jbd2_journal_destroy(sbi->s_journal);
1154 sbi->s_journal = NULL;
1155 if ((err < 0) && !aborted) {
1156 ext4_abort(sb, -err, "Couldn't clean up the journal");
1160 ext4_es_unregister_shrinker(sbi);
1161 del_timer_sync(&sbi->s_err_report);
1162 ext4_release_system_zone(sb);
1163 ext4_mb_release(sb);
1164 ext4_ext_release(sb);
1166 if (!sb_rdonly(sb) && !aborted) {
1167 ext4_clear_feature_journal_needs_recovery(sb);
1168 es->s_state = cpu_to_le16(sbi->s_mount_state);
1171 ext4_commit_super(sb, 1);
1174 group_desc = rcu_dereference(sbi->s_group_desc);
1175 for (i = 0; i < sbi->s_gdb_count; i++)
1176 brelse(group_desc[i]);
1178 flex_groups = rcu_dereference(sbi->s_flex_groups);
1180 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1181 kvfree(flex_groups[i]);
1182 kvfree(flex_groups);
1185 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1186 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1187 percpu_counter_destroy(&sbi->s_dirs_counter);
1188 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1189 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1191 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1192 kfree(get_qf_name(sb, sbi, i));
1195 /* Debugging code just in case the in-memory inode orphan list
1196 * isn't empty. The on-disk one can be non-empty if we've
1197 * detected an error and taken the fs readonly, but the
1198 * in-memory list had better be clean by this point. */
1199 if (!list_empty(&sbi->s_orphan))
1200 dump_orphan_list(sb, sbi);
1201 ASSERT(list_empty(&sbi->s_orphan));
1203 sync_blockdev(sb->s_bdev);
1204 invalidate_bdev(sb->s_bdev);
1205 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1207 * Invalidate the journal device's buffers. We don't want them
1208 * floating about in memory - the physical journal device may
1209 * hotswapped, and it breaks the `ro-after' testing code.
1211 sync_blockdev(sbi->s_journal_bdev);
1212 invalidate_bdev(sbi->s_journal_bdev);
1213 ext4_blkdev_remove(sbi);
1216 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1217 sbi->s_ea_inode_cache = NULL;
1219 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1220 sbi->s_ea_block_cache = NULL;
1223 kthread_stop(sbi->s_mmp_tsk);
1225 sb->s_fs_info = NULL;
1227 * Now that we are completely done shutting down the
1228 * superblock, we need to actually destroy the kobject.
1230 kobject_put(&sbi->s_kobj);
1231 wait_for_completion(&sbi->s_kobj_unregister);
1232 if (sbi->s_chksum_driver)
1233 crypto_free_shash(sbi->s_chksum_driver);
1234 kfree(sbi->s_blockgroup_lock);
1235 fs_put_dax(sbi->s_daxdev);
1236 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1237 #ifdef CONFIG_UNICODE
1238 utf8_unload(sb->s_encoding);
1243 static struct kmem_cache *ext4_inode_cachep;
1246 * Called inside transaction, so use GFP_NOFS
1248 static struct inode *ext4_alloc_inode(struct super_block *sb)
1250 struct ext4_inode_info *ei;
1252 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1256 inode_set_iversion(&ei->vfs_inode, 1);
1257 spin_lock_init(&ei->i_raw_lock);
1258 INIT_LIST_HEAD(&ei->i_prealloc_list);
1259 atomic_set(&ei->i_prealloc_active, 0);
1260 spin_lock_init(&ei->i_prealloc_lock);
1261 ext4_es_init_tree(&ei->i_es_tree);
1262 rwlock_init(&ei->i_es_lock);
1263 INIT_LIST_HEAD(&ei->i_es_list);
1264 ei->i_es_all_nr = 0;
1265 ei->i_es_shk_nr = 0;
1266 ei->i_es_shrink_lblk = 0;
1267 ei->i_reserved_data_blocks = 0;
1268 spin_lock_init(&(ei->i_block_reservation_lock));
1269 ext4_init_pending_tree(&ei->i_pending_tree);
1271 ei->i_reserved_quota = 0;
1272 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1275 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1276 spin_lock_init(&ei->i_completed_io_lock);
1278 ei->i_datasync_tid = 0;
1279 atomic_set(&ei->i_unwritten, 0);
1280 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1281 ext4_fc_init_inode(&ei->vfs_inode);
1282 mutex_init(&ei->i_fc_lock);
1283 return &ei->vfs_inode;
1286 static int ext4_drop_inode(struct inode *inode)
1288 int drop = generic_drop_inode(inode);
1291 drop = fscrypt_drop_inode(inode);
1293 trace_ext4_drop_inode(inode, drop);
1297 static void ext4_free_in_core_inode(struct inode *inode)
1299 fscrypt_free_inode(inode);
1300 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1301 pr_warn("%s: inode %ld still in fc list",
1302 __func__, inode->i_ino);
1304 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1307 static void ext4_destroy_inode(struct inode *inode)
1309 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1310 ext4_msg(inode->i_sb, KERN_ERR,
1311 "Inode %lu (%p): orphan list check failed!",
1312 inode->i_ino, EXT4_I(inode));
1313 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1314 EXT4_I(inode), sizeof(struct ext4_inode_info),
1320 static void init_once(void *foo)
1322 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1324 INIT_LIST_HEAD(&ei->i_orphan);
1325 init_rwsem(&ei->xattr_sem);
1326 init_rwsem(&ei->i_data_sem);
1327 init_rwsem(&ei->i_mmap_sem);
1328 inode_init_once(&ei->vfs_inode);
1329 ext4_fc_init_inode(&ei->vfs_inode);
1332 static int __init init_inodecache(void)
1334 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1335 sizeof(struct ext4_inode_info), 0,
1336 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1338 offsetof(struct ext4_inode_info, i_data),
1339 sizeof_field(struct ext4_inode_info, i_data),
1341 if (ext4_inode_cachep == NULL)
1346 static void destroy_inodecache(void)
1349 * Make sure all delayed rcu free inodes are flushed before we
1353 kmem_cache_destroy(ext4_inode_cachep);
1356 void ext4_clear_inode(struct inode *inode)
1359 invalidate_inode_buffers(inode);
1361 ext4_discard_preallocations(inode, 0);
1362 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1364 if (EXT4_I(inode)->jinode) {
1365 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1366 EXT4_I(inode)->jinode);
1367 jbd2_free_inode(EXT4_I(inode)->jinode);
1368 EXT4_I(inode)->jinode = NULL;
1370 fscrypt_put_encryption_info(inode);
1371 fsverity_cleanup_inode(inode);
1374 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1375 u64 ino, u32 generation)
1377 struct inode *inode;
1380 * Currently we don't know the generation for parent directory, so
1381 * a generation of 0 means "accept any"
1383 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1385 return ERR_CAST(inode);
1386 if (generation && inode->i_generation != generation) {
1388 return ERR_PTR(-ESTALE);
1394 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1395 int fh_len, int fh_type)
1397 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1398 ext4_nfs_get_inode);
1401 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1402 int fh_len, int fh_type)
1404 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1405 ext4_nfs_get_inode);
1408 static int ext4_nfs_commit_metadata(struct inode *inode)
1410 struct writeback_control wbc = {
1411 .sync_mode = WB_SYNC_ALL
1414 trace_ext4_nfs_commit_metadata(inode);
1415 return ext4_write_inode(inode, &wbc);
1419 * Try to release metadata pages (indirect blocks, directories) which are
1420 * mapped via the block device. Since these pages could have journal heads
1421 * which would prevent try_to_free_buffers() from freeing them, we must use
1422 * jbd2 layer's try_to_free_buffers() function to release them.
1424 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1427 journal_t *journal = EXT4_SB(sb)->s_journal;
1429 WARN_ON(PageChecked(page));
1430 if (!page_has_buffers(page))
1433 return jbd2_journal_try_to_free_buffers(journal, page);
1435 return try_to_free_buffers(page);
1438 #ifdef CONFIG_FS_ENCRYPTION
1439 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1441 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1442 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1445 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1448 handle_t *handle = fs_data;
1449 int res, res2, credits, retries = 0;
1452 * Encrypting the root directory is not allowed because e2fsck expects
1453 * lost+found to exist and be unencrypted, and encrypting the root
1454 * directory would imply encrypting the lost+found directory as well as
1455 * the filename "lost+found" itself.
1457 if (inode->i_ino == EXT4_ROOT_INO)
1460 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1463 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1466 res = ext4_convert_inline_data(inode);
1471 * If a journal handle was specified, then the encryption context is
1472 * being set on a new inode via inheritance and is part of a larger
1473 * transaction to create the inode. Otherwise the encryption context is
1474 * being set on an existing inode in its own transaction. Only in the
1475 * latter case should the "retry on ENOSPC" logic be used.
1479 res = ext4_xattr_set_handle(handle, inode,
1480 EXT4_XATTR_INDEX_ENCRYPTION,
1481 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1484 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1485 ext4_clear_inode_state(inode,
1486 EXT4_STATE_MAY_INLINE_DATA);
1488 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1489 * S_DAX may be disabled
1491 ext4_set_inode_flags(inode, false);
1496 res = dquot_initialize(inode);
1500 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1505 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1507 return PTR_ERR(handle);
1509 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1510 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1513 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1515 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1516 * S_DAX may be disabled
1518 ext4_set_inode_flags(inode, false);
1519 res = ext4_mark_inode_dirty(handle, inode);
1521 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1523 res2 = ext4_journal_stop(handle);
1525 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1532 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1534 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1537 static bool ext4_has_stable_inodes(struct super_block *sb)
1539 return ext4_has_feature_stable_inodes(sb);
1542 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1543 int *ino_bits_ret, int *lblk_bits_ret)
1545 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1546 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1549 static const struct fscrypt_operations ext4_cryptops = {
1550 .key_prefix = "ext4:",
1551 .get_context = ext4_get_context,
1552 .set_context = ext4_set_context,
1553 .get_dummy_policy = ext4_get_dummy_policy,
1554 .empty_dir = ext4_empty_dir,
1555 .max_namelen = EXT4_NAME_LEN,
1556 .has_stable_inodes = ext4_has_stable_inodes,
1557 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1562 static const char * const quotatypes[] = INITQFNAMES;
1563 #define QTYPE2NAME(t) (quotatypes[t])
1565 static int ext4_write_dquot(struct dquot *dquot);
1566 static int ext4_acquire_dquot(struct dquot *dquot);
1567 static int ext4_release_dquot(struct dquot *dquot);
1568 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1569 static int ext4_write_info(struct super_block *sb, int type);
1570 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1571 const struct path *path);
1572 static int ext4_quota_on_mount(struct super_block *sb, int type);
1573 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1574 size_t len, loff_t off);
1575 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1576 const char *data, size_t len, loff_t off);
1577 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1578 unsigned int flags);
1579 static int ext4_enable_quotas(struct super_block *sb);
1581 static struct dquot **ext4_get_dquots(struct inode *inode)
1583 return EXT4_I(inode)->i_dquot;
1586 static const struct dquot_operations ext4_quota_operations = {
1587 .get_reserved_space = ext4_get_reserved_space,
1588 .write_dquot = ext4_write_dquot,
1589 .acquire_dquot = ext4_acquire_dquot,
1590 .release_dquot = ext4_release_dquot,
1591 .mark_dirty = ext4_mark_dquot_dirty,
1592 .write_info = ext4_write_info,
1593 .alloc_dquot = dquot_alloc,
1594 .destroy_dquot = dquot_destroy,
1595 .get_projid = ext4_get_projid,
1596 .get_inode_usage = ext4_get_inode_usage,
1597 .get_next_id = dquot_get_next_id,
1600 static const struct quotactl_ops ext4_qctl_operations = {
1601 .quota_on = ext4_quota_on,
1602 .quota_off = ext4_quota_off,
1603 .quota_sync = dquot_quota_sync,
1604 .get_state = dquot_get_state,
1605 .set_info = dquot_set_dqinfo,
1606 .get_dqblk = dquot_get_dqblk,
1607 .set_dqblk = dquot_set_dqblk,
1608 .get_nextdqblk = dquot_get_next_dqblk,
1612 static const struct super_operations ext4_sops = {
1613 .alloc_inode = ext4_alloc_inode,
1614 .free_inode = ext4_free_in_core_inode,
1615 .destroy_inode = ext4_destroy_inode,
1616 .write_inode = ext4_write_inode,
1617 .dirty_inode = ext4_dirty_inode,
1618 .drop_inode = ext4_drop_inode,
1619 .evict_inode = ext4_evict_inode,
1620 .put_super = ext4_put_super,
1621 .sync_fs = ext4_sync_fs,
1622 .freeze_fs = ext4_freeze,
1623 .unfreeze_fs = ext4_unfreeze,
1624 .statfs = ext4_statfs,
1625 .remount_fs = ext4_remount,
1626 .show_options = ext4_show_options,
1628 .quota_read = ext4_quota_read,
1629 .quota_write = ext4_quota_write,
1630 .get_dquots = ext4_get_dquots,
1632 .bdev_try_to_free_page = bdev_try_to_free_page,
1635 static const struct export_operations ext4_export_ops = {
1636 .fh_to_dentry = ext4_fh_to_dentry,
1637 .fh_to_parent = ext4_fh_to_parent,
1638 .get_parent = ext4_get_parent,
1639 .commit_metadata = ext4_nfs_commit_metadata,
1643 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1644 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1645 Opt_nouid32, Opt_debug, Opt_removed,
1646 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1647 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1648 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1649 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1650 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1651 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1653 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1654 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1655 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1656 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1657 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1658 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1659 Opt_nowarn_on_error, Opt_mblk_io_submit,
1660 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1661 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1662 Opt_inode_readahead_blks, Opt_journal_ioprio,
1663 Opt_dioread_nolock, Opt_dioread_lock,
1664 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1665 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1666 Opt_prefetch_block_bitmaps,
1667 #ifdef CONFIG_EXT4_DEBUG
1668 Opt_fc_debug_max_replay, Opt_fc_debug_force
1672 static const match_table_t tokens = {
1673 {Opt_bsd_df, "bsddf"},
1674 {Opt_minix_df, "minixdf"},
1675 {Opt_grpid, "grpid"},
1676 {Opt_grpid, "bsdgroups"},
1677 {Opt_nogrpid, "nogrpid"},
1678 {Opt_nogrpid, "sysvgroups"},
1679 {Opt_resgid, "resgid=%u"},
1680 {Opt_resuid, "resuid=%u"},
1682 {Opt_err_cont, "errors=continue"},
1683 {Opt_err_panic, "errors=panic"},
1684 {Opt_err_ro, "errors=remount-ro"},
1685 {Opt_nouid32, "nouid32"},
1686 {Opt_debug, "debug"},
1687 {Opt_removed, "oldalloc"},
1688 {Opt_removed, "orlov"},
1689 {Opt_user_xattr, "user_xattr"},
1690 {Opt_nouser_xattr, "nouser_xattr"},
1692 {Opt_noacl, "noacl"},
1693 {Opt_noload, "norecovery"},
1694 {Opt_noload, "noload"},
1695 {Opt_removed, "nobh"},
1696 {Opt_removed, "bh"},
1697 {Opt_commit, "commit=%u"},
1698 {Opt_min_batch_time, "min_batch_time=%u"},
1699 {Opt_max_batch_time, "max_batch_time=%u"},
1700 {Opt_journal_dev, "journal_dev=%u"},
1701 {Opt_journal_path, "journal_path=%s"},
1702 {Opt_journal_checksum, "journal_checksum"},
1703 {Opt_nojournal_checksum, "nojournal_checksum"},
1704 {Opt_journal_async_commit, "journal_async_commit"},
1705 {Opt_abort, "abort"},
1706 {Opt_data_journal, "data=journal"},
1707 {Opt_data_ordered, "data=ordered"},
1708 {Opt_data_writeback, "data=writeback"},
1709 {Opt_data_err_abort, "data_err=abort"},
1710 {Opt_data_err_ignore, "data_err=ignore"},
1711 {Opt_offusrjquota, "usrjquota="},
1712 {Opt_usrjquota, "usrjquota=%s"},
1713 {Opt_offgrpjquota, "grpjquota="},
1714 {Opt_grpjquota, "grpjquota=%s"},
1715 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1716 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1717 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1718 {Opt_grpquota, "grpquota"},
1719 {Opt_noquota, "noquota"},
1720 {Opt_quota, "quota"},
1721 {Opt_usrquota, "usrquota"},
1722 {Opt_prjquota, "prjquota"},
1723 {Opt_barrier, "barrier=%u"},
1724 {Opt_barrier, "barrier"},
1725 {Opt_nobarrier, "nobarrier"},
1726 {Opt_i_version, "i_version"},
1728 {Opt_dax_always, "dax=always"},
1729 {Opt_dax_inode, "dax=inode"},
1730 {Opt_dax_never, "dax=never"},
1731 {Opt_stripe, "stripe=%u"},
1732 {Opt_delalloc, "delalloc"},
1733 {Opt_warn_on_error, "warn_on_error"},
1734 {Opt_nowarn_on_error, "nowarn_on_error"},
1735 {Opt_lazytime, "lazytime"},
1736 {Opt_nolazytime, "nolazytime"},
1737 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1738 {Opt_nodelalloc, "nodelalloc"},
1739 {Opt_removed, "mblk_io_submit"},
1740 {Opt_removed, "nomblk_io_submit"},
1741 {Opt_block_validity, "block_validity"},
1742 {Opt_noblock_validity, "noblock_validity"},
1743 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1744 {Opt_journal_ioprio, "journal_ioprio=%u"},
1745 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1746 {Opt_auto_da_alloc, "auto_da_alloc"},
1747 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1748 {Opt_dioread_nolock, "dioread_nolock"},
1749 {Opt_dioread_lock, "nodioread_nolock"},
1750 {Opt_dioread_lock, "dioread_lock"},
1751 {Opt_discard, "discard"},
1752 {Opt_nodiscard, "nodiscard"},
1753 {Opt_init_itable, "init_itable=%u"},
1754 {Opt_init_itable, "init_itable"},
1755 {Opt_noinit_itable, "noinit_itable"},
1756 #ifdef CONFIG_EXT4_DEBUG
1757 {Opt_fc_debug_force, "fc_debug_force"},
1758 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1760 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1761 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1762 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1763 {Opt_inlinecrypt, "inlinecrypt"},
1764 {Opt_nombcache, "nombcache"},
1765 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1766 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1767 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1768 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1769 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1770 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1771 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1775 static ext4_fsblk_t get_sb_block(void **data)
1777 ext4_fsblk_t sb_block;
1778 char *options = (char *) *data;
1780 if (!options || strncmp(options, "sb=", 3) != 0)
1781 return 1; /* Default location */
1784 /* TODO: use simple_strtoll with >32bit ext4 */
1785 sb_block = simple_strtoul(options, &options, 0);
1786 if (*options && *options != ',') {
1787 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1791 if (*options == ',')
1793 *data = (void *) options;
1798 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1799 static const char deprecated_msg[] =
1800 "Mount option \"%s\" will be removed by %s\n"
1801 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1804 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1806 struct ext4_sb_info *sbi = EXT4_SB(sb);
1807 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1810 if (sb_any_quota_loaded(sb) && !old_qname) {
1811 ext4_msg(sb, KERN_ERR,
1812 "Cannot change journaled "
1813 "quota options when quota turned on");
1816 if (ext4_has_feature_quota(sb)) {
1817 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1818 "ignored when QUOTA feature is enabled");
1821 qname = match_strdup(args);
1823 ext4_msg(sb, KERN_ERR,
1824 "Not enough memory for storing quotafile name");
1828 if (strcmp(old_qname, qname) == 0)
1831 ext4_msg(sb, KERN_ERR,
1832 "%s quota file already specified",
1836 if (strchr(qname, '/')) {
1837 ext4_msg(sb, KERN_ERR,
1838 "quotafile must be on filesystem root");
1841 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1849 static int clear_qf_name(struct super_block *sb, int qtype)
1852 struct ext4_sb_info *sbi = EXT4_SB(sb);
1853 char *old_qname = get_qf_name(sb, sbi, qtype);
1855 if (sb_any_quota_loaded(sb) && old_qname) {
1856 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1857 " when quota turned on");
1860 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1867 #define MOPT_SET 0x0001
1868 #define MOPT_CLEAR 0x0002
1869 #define MOPT_NOSUPPORT 0x0004
1870 #define MOPT_EXPLICIT 0x0008
1871 #define MOPT_CLEAR_ERR 0x0010
1872 #define MOPT_GTE0 0x0020
1875 #define MOPT_QFMT 0x0040
1877 #define MOPT_Q MOPT_NOSUPPORT
1878 #define MOPT_QFMT MOPT_NOSUPPORT
1880 #define MOPT_DATAJ 0x0080
1881 #define MOPT_NO_EXT2 0x0100
1882 #define MOPT_NO_EXT3 0x0200
1883 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1884 #define MOPT_STRING 0x0400
1885 #define MOPT_SKIP 0x0800
1886 #define MOPT_2 0x1000
1888 static const struct mount_opts {
1892 } ext4_mount_opts[] = {
1893 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1894 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1895 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1896 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1897 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1898 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1899 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1900 MOPT_EXT4_ONLY | MOPT_SET},
1901 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1902 MOPT_EXT4_ONLY | MOPT_CLEAR},
1903 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1904 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1905 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1906 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1907 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1908 MOPT_EXT4_ONLY | MOPT_CLEAR},
1909 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1910 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1911 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1912 MOPT_EXT4_ONLY | MOPT_CLEAR},
1913 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1914 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1915 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1916 EXT4_MOUNT_JOURNAL_CHECKSUM),
1917 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1918 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1919 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1920 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1921 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1922 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1924 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1926 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1927 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1928 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1929 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1930 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1931 {Opt_commit, 0, MOPT_GTE0},
1932 {Opt_max_batch_time, 0, MOPT_GTE0},
1933 {Opt_min_batch_time, 0, MOPT_GTE0},
1934 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1935 {Opt_init_itable, 0, MOPT_GTE0},
1936 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1937 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1938 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1939 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1940 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1941 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1942 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1943 {Opt_stripe, 0, MOPT_GTE0},
1944 {Opt_resuid, 0, MOPT_GTE0},
1945 {Opt_resgid, 0, MOPT_GTE0},
1946 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1947 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1948 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1949 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1950 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1951 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1952 MOPT_NO_EXT2 | MOPT_DATAJ},
1953 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1954 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1955 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1956 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1957 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1959 {Opt_acl, 0, MOPT_NOSUPPORT},
1960 {Opt_noacl, 0, MOPT_NOSUPPORT},
1962 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1963 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1964 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1965 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1966 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1968 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1970 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1972 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1973 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1974 MOPT_CLEAR | MOPT_Q},
1975 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1976 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1977 {Opt_offusrjquota, 0, MOPT_Q},
1978 {Opt_offgrpjquota, 0, MOPT_Q},
1979 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1980 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1981 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1982 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1983 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1984 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1985 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1987 #ifdef CONFIG_EXT4_DEBUG
1988 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1989 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1990 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
1995 #ifdef CONFIG_UNICODE
1996 static const struct ext4_sb_encodings {
2000 } ext4_sb_encoding_map[] = {
2001 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2004 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2005 const struct ext4_sb_encodings **encoding,
2008 __u16 magic = le16_to_cpu(es->s_encoding);
2011 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2012 if (magic == ext4_sb_encoding_map[i].magic)
2015 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2018 *encoding = &ext4_sb_encoding_map[i];
2019 *flags = le16_to_cpu(es->s_encoding_flags);
2025 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2027 const substring_t *arg,
2030 #ifdef CONFIG_FS_ENCRYPTION
2031 struct ext4_sb_info *sbi = EXT4_SB(sb);
2035 * This mount option is just for testing, and it's not worthwhile to
2036 * implement the extra complexity (e.g. RCU protection) that would be
2037 * needed to allow it to be set or changed during remount. We do allow
2038 * it to be specified during remount, but only if there is no change.
2040 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2041 ext4_msg(sb, KERN_WARNING,
2042 "Can't set test_dummy_encryption on remount");
2045 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2046 &sbi->s_dummy_enc_policy);
2049 ext4_msg(sb, KERN_WARNING,
2050 "Can't change test_dummy_encryption on remount");
2051 else if (err == -EINVAL)
2052 ext4_msg(sb, KERN_WARNING,
2053 "Value of option \"%s\" is unrecognized", opt);
2055 ext4_msg(sb, KERN_WARNING,
2056 "Error processing option \"%s\" [%d]",
2060 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2062 ext4_msg(sb, KERN_WARNING,
2063 "Test dummy encryption mount option ignored");
2068 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2069 substring_t *args, unsigned long *journal_devnum,
2070 unsigned int *journal_ioprio, int is_remount)
2072 struct ext4_sb_info *sbi = EXT4_SB(sb);
2073 const struct mount_opts *m;
2079 if (token == Opt_usrjquota)
2080 return set_qf_name(sb, USRQUOTA, &args[0]);
2081 else if (token == Opt_grpjquota)
2082 return set_qf_name(sb, GRPQUOTA, &args[0]);
2083 else if (token == Opt_offusrjquota)
2084 return clear_qf_name(sb, USRQUOTA);
2085 else if (token == Opt_offgrpjquota)
2086 return clear_qf_name(sb, GRPQUOTA);
2090 case Opt_nouser_xattr:
2091 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2094 return 1; /* handled by get_sb_block() */
2096 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2099 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2102 sb->s_flags |= SB_I_VERSION;
2105 sb->s_flags |= SB_LAZYTIME;
2107 case Opt_nolazytime:
2108 sb->s_flags &= ~SB_LAZYTIME;
2110 case Opt_inlinecrypt:
2111 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2112 sb->s_flags |= SB_INLINECRYPT;
2114 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2119 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2120 if (token == m->token)
2123 if (m->token == Opt_err) {
2124 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2125 "or missing value", opt);
2129 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2130 ext4_msg(sb, KERN_ERR,
2131 "Mount option \"%s\" incompatible with ext2", opt);
2134 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2135 ext4_msg(sb, KERN_ERR,
2136 "Mount option \"%s\" incompatible with ext3", opt);
2140 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2142 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2144 if (m->flags & MOPT_EXPLICIT) {
2145 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2146 set_opt2(sb, EXPLICIT_DELALLOC);
2147 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2148 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2152 if (m->flags & MOPT_CLEAR_ERR)
2153 clear_opt(sb, ERRORS_MASK);
2154 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2155 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2156 "options when quota turned on");
2160 if (m->flags & MOPT_NOSUPPORT) {
2161 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2162 } else if (token == Opt_commit) {
2164 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2165 else if (arg > INT_MAX / HZ) {
2166 ext4_msg(sb, KERN_ERR,
2167 "Invalid commit interval %d, "
2168 "must be smaller than %d",
2172 sbi->s_commit_interval = HZ * arg;
2173 } else if (token == Opt_debug_want_extra_isize) {
2176 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2177 ext4_msg(sb, KERN_ERR,
2178 "Invalid want_extra_isize %d", arg);
2181 sbi->s_want_extra_isize = arg;
2182 } else if (token == Opt_max_batch_time) {
2183 sbi->s_max_batch_time = arg;
2184 } else if (token == Opt_min_batch_time) {
2185 sbi->s_min_batch_time = arg;
2186 } else if (token == Opt_inode_readahead_blks) {
2187 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2188 ext4_msg(sb, KERN_ERR,
2189 "EXT4-fs: inode_readahead_blks must be "
2190 "0 or a power of 2 smaller than 2^31");
2193 sbi->s_inode_readahead_blks = arg;
2194 } else if (token == Opt_init_itable) {
2195 set_opt(sb, INIT_INODE_TABLE);
2197 arg = EXT4_DEF_LI_WAIT_MULT;
2198 sbi->s_li_wait_mult = arg;
2199 } else if (token == Opt_max_dir_size_kb) {
2200 sbi->s_max_dir_size_kb = arg;
2201 #ifdef CONFIG_EXT4_DEBUG
2202 } else if (token == Opt_fc_debug_max_replay) {
2203 sbi->s_fc_debug_max_replay = arg;
2205 } else if (token == Opt_stripe) {
2206 sbi->s_stripe = arg;
2207 } else if (token == Opt_resuid) {
2208 uid = make_kuid(current_user_ns(), arg);
2209 if (!uid_valid(uid)) {
2210 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2213 sbi->s_resuid = uid;
2214 } else if (token == Opt_resgid) {
2215 gid = make_kgid(current_user_ns(), arg);
2216 if (!gid_valid(gid)) {
2217 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2220 sbi->s_resgid = gid;
2221 } else if (token == Opt_journal_dev) {
2223 ext4_msg(sb, KERN_ERR,
2224 "Cannot specify journal on remount");
2227 *journal_devnum = arg;
2228 } else if (token == Opt_journal_path) {
2230 struct inode *journal_inode;
2235 ext4_msg(sb, KERN_ERR,
2236 "Cannot specify journal on remount");
2239 journal_path = match_strdup(&args[0]);
2240 if (!journal_path) {
2241 ext4_msg(sb, KERN_ERR, "error: could not dup "
2242 "journal device string");
2246 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2248 ext4_msg(sb, KERN_ERR, "error: could not find "
2249 "journal device path: error %d", error);
2250 kfree(journal_path);
2254 journal_inode = d_inode(path.dentry);
2255 if (!S_ISBLK(journal_inode->i_mode)) {
2256 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2257 "is not a block device", journal_path);
2259 kfree(journal_path);
2263 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2265 kfree(journal_path);
2266 } else if (token == Opt_journal_ioprio) {
2268 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2273 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2274 } else if (token == Opt_test_dummy_encryption) {
2275 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2277 } else if (m->flags & MOPT_DATAJ) {
2279 if (!sbi->s_journal)
2280 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2281 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2282 ext4_msg(sb, KERN_ERR,
2283 "Cannot change data mode on remount");
2287 clear_opt(sb, DATA_FLAGS);
2288 sbi->s_mount_opt |= m->mount_opt;
2291 } else if (m->flags & MOPT_QFMT) {
2292 if (sb_any_quota_loaded(sb) &&
2293 sbi->s_jquota_fmt != m->mount_opt) {
2294 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2295 "quota options when quota turned on");
2298 if (ext4_has_feature_quota(sb)) {
2299 ext4_msg(sb, KERN_INFO,
2300 "Quota format mount options ignored "
2301 "when QUOTA feature is enabled");
2304 sbi->s_jquota_fmt = m->mount_opt;
2306 } else if (token == Opt_dax || token == Opt_dax_always ||
2307 token == Opt_dax_inode || token == Opt_dax_never) {
2308 #ifdef CONFIG_FS_DAX
2311 case Opt_dax_always:
2313 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2314 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2315 fail_dax_change_remount:
2316 ext4_msg(sb, KERN_ERR, "can't change "
2317 "dax mount option while remounting");
2321 (test_opt(sb, DATA_FLAGS) ==
2322 EXT4_MOUNT_JOURNAL_DATA)) {
2323 ext4_msg(sb, KERN_ERR, "can't mount with "
2324 "both data=journal and dax");
2327 ext4_msg(sb, KERN_WARNING,
2328 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2329 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2330 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2334 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2335 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2336 goto fail_dax_change_remount;
2337 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2338 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2342 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2343 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2344 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2345 goto fail_dax_change_remount;
2346 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2347 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2348 /* Strictly for printing options */
2349 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2353 ext4_msg(sb, KERN_INFO, "dax option not supported");
2354 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2355 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2358 } else if (token == Opt_data_err_abort) {
2359 sbi->s_mount_opt |= m->mount_opt;
2360 } else if (token == Opt_data_err_ignore) {
2361 sbi->s_mount_opt &= ~m->mount_opt;
2365 if (m->flags & MOPT_CLEAR)
2367 else if (unlikely(!(m->flags & MOPT_SET))) {
2368 ext4_msg(sb, KERN_WARNING,
2369 "buggy handling of option %s", opt);
2373 if (m->flags & MOPT_2) {
2375 sbi->s_mount_opt2 |= m->mount_opt;
2377 sbi->s_mount_opt2 &= ~m->mount_opt;
2380 sbi->s_mount_opt |= m->mount_opt;
2382 sbi->s_mount_opt &= ~m->mount_opt;
2388 static int parse_options(char *options, struct super_block *sb,
2389 unsigned long *journal_devnum,
2390 unsigned int *journal_ioprio,
2393 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2394 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2395 substring_t args[MAX_OPT_ARGS];
2401 while ((p = strsep(&options, ",")) != NULL) {
2405 * Initialize args struct so we know whether arg was
2406 * found; some options take optional arguments.
2408 args[0].to = args[0].from = NULL;
2409 token = match_token(p, tokens, args);
2410 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2411 journal_ioprio, is_remount) < 0)
2416 * We do the test below only for project quotas. 'usrquota' and
2417 * 'grpquota' mount options are allowed even without quota feature
2418 * to support legacy quotas in quota files.
2420 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2421 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2422 "Cannot enable project quota enforcement.");
2425 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2426 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2427 if (usr_qf_name || grp_qf_name) {
2428 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2429 clear_opt(sb, USRQUOTA);
2431 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2432 clear_opt(sb, GRPQUOTA);
2434 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2435 ext4_msg(sb, KERN_ERR, "old and new quota "
2440 if (!sbi->s_jquota_fmt) {
2441 ext4_msg(sb, KERN_ERR, "journaled quota format "
2447 if (test_opt(sb, DIOREAD_NOLOCK)) {
2449 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2450 if (blocksize < PAGE_SIZE)
2451 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2452 "experimental mount option 'dioread_nolock' "
2453 "for blocksize < PAGE_SIZE");
2458 static inline void ext4_show_quota_options(struct seq_file *seq,
2459 struct super_block *sb)
2461 #if defined(CONFIG_QUOTA)
2462 struct ext4_sb_info *sbi = EXT4_SB(sb);
2463 char *usr_qf_name, *grp_qf_name;
2465 if (sbi->s_jquota_fmt) {
2468 switch (sbi->s_jquota_fmt) {
2479 seq_printf(seq, ",jqfmt=%s", fmtname);
2483 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2484 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2486 seq_show_option(seq, "usrjquota", usr_qf_name);
2488 seq_show_option(seq, "grpjquota", grp_qf_name);
2493 static const char *token2str(int token)
2495 const struct match_token *t;
2497 for (t = tokens; t->token != Opt_err; t++)
2498 if (t->token == token && !strchr(t->pattern, '='))
2505 * - it's set to a non-default value OR
2506 * - if the per-sb default is different from the global default
2508 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2511 struct ext4_sb_info *sbi = EXT4_SB(sb);
2512 struct ext4_super_block *es = sbi->s_es;
2513 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2514 const struct mount_opts *m;
2515 char sep = nodefs ? '\n' : ',';
2517 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2518 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2520 if (sbi->s_sb_block != 1)
2521 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2523 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2524 int want_set = m->flags & MOPT_SET;
2525 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2526 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2528 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2529 continue; /* skip if same as the default */
2531 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2532 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2533 continue; /* select Opt_noFoo vs Opt_Foo */
2534 SEQ_OPTS_PRINT("%s", token2str(m->token));
2537 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2538 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2539 SEQ_OPTS_PRINT("resuid=%u",
2540 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2541 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2542 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2543 SEQ_OPTS_PRINT("resgid=%u",
2544 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2545 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2546 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2547 SEQ_OPTS_PUTS("errors=remount-ro");
2548 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2549 SEQ_OPTS_PUTS("errors=continue");
2550 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2551 SEQ_OPTS_PUTS("errors=panic");
2552 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2553 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2554 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2555 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2556 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2557 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2558 if (sb->s_flags & SB_I_VERSION)
2559 SEQ_OPTS_PUTS("i_version");
2560 if (nodefs || sbi->s_stripe)
2561 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2562 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2563 (sbi->s_mount_opt ^ def_mount_opt)) {
2564 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2565 SEQ_OPTS_PUTS("data=journal");
2566 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2567 SEQ_OPTS_PUTS("data=ordered");
2568 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2569 SEQ_OPTS_PUTS("data=writeback");
2572 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2573 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2574 sbi->s_inode_readahead_blks);
2576 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2577 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2578 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2579 if (nodefs || sbi->s_max_dir_size_kb)
2580 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2581 if (test_opt(sb, DATA_ERR_ABORT))
2582 SEQ_OPTS_PUTS("data_err=abort");
2584 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2586 if (sb->s_flags & SB_INLINECRYPT)
2587 SEQ_OPTS_PUTS("inlinecrypt");
2589 if (test_opt(sb, DAX_ALWAYS)) {
2591 SEQ_OPTS_PUTS("dax");
2593 SEQ_OPTS_PUTS("dax=always");
2594 } else if (test_opt2(sb, DAX_NEVER)) {
2595 SEQ_OPTS_PUTS("dax=never");
2596 } else if (test_opt2(sb, DAX_INODE)) {
2597 SEQ_OPTS_PUTS("dax=inode");
2599 ext4_show_quota_options(seq, sb);
2603 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2605 return _ext4_show_options(seq, root->d_sb, 0);
2608 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2610 struct super_block *sb = seq->private;
2613 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2614 rc = _ext4_show_options(seq, sb, 1);
2615 seq_puts(seq, "\n");
2619 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2622 struct ext4_sb_info *sbi = EXT4_SB(sb);
2625 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2626 ext4_msg(sb, KERN_ERR, "revision level too high, "
2627 "forcing read-only mode");
2633 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2634 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2635 "running e2fsck is recommended");
2636 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2637 ext4_msg(sb, KERN_WARNING,
2638 "warning: mounting fs with errors, "
2639 "running e2fsck is recommended");
2640 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2641 le16_to_cpu(es->s_mnt_count) >=
2642 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2643 ext4_msg(sb, KERN_WARNING,
2644 "warning: maximal mount count reached, "
2645 "running e2fsck is recommended");
2646 else if (le32_to_cpu(es->s_checkinterval) &&
2647 (ext4_get_tstamp(es, s_lastcheck) +
2648 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2649 ext4_msg(sb, KERN_WARNING,
2650 "warning: checktime reached, "
2651 "running e2fsck is recommended");
2652 if (!sbi->s_journal)
2653 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2654 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2655 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2656 le16_add_cpu(&es->s_mnt_count, 1);
2657 ext4_update_tstamp(es, s_mtime);
2659 ext4_set_feature_journal_needs_recovery(sb);
2661 err = ext4_commit_super(sb, 1);
2663 if (test_opt(sb, DEBUG))
2664 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2665 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2667 sbi->s_groups_count,
2668 EXT4_BLOCKS_PER_GROUP(sb),
2669 EXT4_INODES_PER_GROUP(sb),
2670 sbi->s_mount_opt, sbi->s_mount_opt2);
2672 cleancache_init_fs(sb);
2676 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2678 struct ext4_sb_info *sbi = EXT4_SB(sb);
2679 struct flex_groups **old_groups, **new_groups;
2682 if (!sbi->s_log_groups_per_flex)
2685 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2686 if (size <= sbi->s_flex_groups_allocated)
2689 new_groups = kvzalloc(roundup_pow_of_two(size *
2690 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2692 ext4_msg(sb, KERN_ERR,
2693 "not enough memory for %d flex group pointers", size);
2696 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2697 new_groups[i] = kvzalloc(roundup_pow_of_two(
2698 sizeof(struct flex_groups)),
2700 if (!new_groups[i]) {
2701 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2702 kvfree(new_groups[j]);
2704 ext4_msg(sb, KERN_ERR,
2705 "not enough memory for %d flex groups", size);
2710 old_groups = rcu_dereference(sbi->s_flex_groups);
2712 memcpy(new_groups, old_groups,
2713 (sbi->s_flex_groups_allocated *
2714 sizeof(struct flex_groups *)));
2716 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2717 sbi->s_flex_groups_allocated = size;
2719 ext4_kvfree_array_rcu(old_groups);
2723 static int ext4_fill_flex_info(struct super_block *sb)
2725 struct ext4_sb_info *sbi = EXT4_SB(sb);
2726 struct ext4_group_desc *gdp = NULL;
2727 struct flex_groups *fg;
2728 ext4_group_t flex_group;
2731 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2732 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2733 sbi->s_log_groups_per_flex = 0;
2737 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2741 for (i = 0; i < sbi->s_groups_count; i++) {
2742 gdp = ext4_get_group_desc(sb, i, NULL);
2744 flex_group = ext4_flex_group(sbi, i);
2745 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2746 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2747 atomic64_add(ext4_free_group_clusters(sb, gdp),
2748 &fg->free_clusters);
2749 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2757 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2758 struct ext4_group_desc *gdp)
2760 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2762 __le32 le_group = cpu_to_le32(block_group);
2763 struct ext4_sb_info *sbi = EXT4_SB(sb);
2765 if (ext4_has_metadata_csum(sbi->s_sb)) {
2766 /* Use new metadata_csum algorithm */
2768 __u16 dummy_csum = 0;
2770 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2772 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2773 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2774 sizeof(dummy_csum));
2775 offset += sizeof(dummy_csum);
2776 if (offset < sbi->s_desc_size)
2777 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2778 sbi->s_desc_size - offset);
2780 crc = csum32 & 0xFFFF;
2784 /* old crc16 code */
2785 if (!ext4_has_feature_gdt_csum(sb))
2788 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2789 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2790 crc = crc16(crc, (__u8 *)gdp, offset);
2791 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2792 /* for checksum of struct ext4_group_desc do the rest...*/
2793 if (ext4_has_feature_64bit(sb) &&
2794 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2795 crc = crc16(crc, (__u8 *)gdp + offset,
2796 le16_to_cpu(sbi->s_es->s_desc_size) -
2800 return cpu_to_le16(crc);
2803 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2804 struct ext4_group_desc *gdp)
2806 if (ext4_has_group_desc_csum(sb) &&
2807 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2813 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2814 struct ext4_group_desc *gdp)
2816 if (!ext4_has_group_desc_csum(sb))
2818 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2821 /* Called at mount-time, super-block is locked */
2822 static int ext4_check_descriptors(struct super_block *sb,
2823 ext4_fsblk_t sb_block,
2824 ext4_group_t *first_not_zeroed)
2826 struct ext4_sb_info *sbi = EXT4_SB(sb);
2827 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2828 ext4_fsblk_t last_block;
2829 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2830 ext4_fsblk_t block_bitmap;
2831 ext4_fsblk_t inode_bitmap;
2832 ext4_fsblk_t inode_table;
2833 int flexbg_flag = 0;
2834 ext4_group_t i, grp = sbi->s_groups_count;
2836 if (ext4_has_feature_flex_bg(sb))
2839 ext4_debug("Checking group descriptors");
2841 for (i = 0; i < sbi->s_groups_count; i++) {
2842 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2844 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2845 last_block = ext4_blocks_count(sbi->s_es) - 1;
2847 last_block = first_block +
2848 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2850 if ((grp == sbi->s_groups_count) &&
2851 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2854 block_bitmap = ext4_block_bitmap(sb, gdp);
2855 if (block_bitmap == sb_block) {
2856 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2857 "Block bitmap for group %u overlaps "
2862 if (block_bitmap >= sb_block + 1 &&
2863 block_bitmap <= last_bg_block) {
2864 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2865 "Block bitmap for group %u overlaps "
2866 "block group descriptors", i);
2870 if (block_bitmap < first_block || block_bitmap > last_block) {
2871 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2872 "Block bitmap for group %u not in group "
2873 "(block %llu)!", i, block_bitmap);
2876 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2877 if (inode_bitmap == sb_block) {
2878 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2879 "Inode bitmap for group %u overlaps "
2884 if (inode_bitmap >= sb_block + 1 &&
2885 inode_bitmap <= last_bg_block) {
2886 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2887 "Inode bitmap for group %u overlaps "
2888 "block group descriptors", i);
2892 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2893 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2894 "Inode bitmap for group %u not in group "
2895 "(block %llu)!", i, inode_bitmap);
2898 inode_table = ext4_inode_table(sb, gdp);
2899 if (inode_table == sb_block) {
2900 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901 "Inode table for group %u overlaps "
2906 if (inode_table >= sb_block + 1 &&
2907 inode_table <= last_bg_block) {
2908 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2909 "Inode table for group %u overlaps "
2910 "block group descriptors", i);
2914 if (inode_table < first_block ||
2915 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2916 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2917 "Inode table for group %u not in group "
2918 "(block %llu)!", i, inode_table);
2921 ext4_lock_group(sb, i);
2922 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2923 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2924 "Checksum for group %u failed (%u!=%u)",
2925 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2926 gdp)), le16_to_cpu(gdp->bg_checksum));
2927 if (!sb_rdonly(sb)) {
2928 ext4_unlock_group(sb, i);
2932 ext4_unlock_group(sb, i);
2934 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2936 if (NULL != first_not_zeroed)
2937 *first_not_zeroed = grp;
2941 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2942 * the superblock) which were deleted from all directories, but held open by
2943 * a process at the time of a crash. We walk the list and try to delete these
2944 * inodes at recovery time (only with a read-write filesystem).
2946 * In order to keep the orphan inode chain consistent during traversal (in
2947 * case of crash during recovery), we link each inode into the superblock
2948 * orphan list_head and handle it the same way as an inode deletion during
2949 * normal operation (which journals the operations for us).
2951 * We only do an iget() and an iput() on each inode, which is very safe if we
2952 * accidentally point at an in-use or already deleted inode. The worst that
2953 * can happen in this case is that we get a "bit already cleared" message from
2954 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2955 * e2fsck was run on this filesystem, and it must have already done the orphan
2956 * inode cleanup for us, so we can safely abort without any further action.
2958 static void ext4_orphan_cleanup(struct super_block *sb,
2959 struct ext4_super_block *es)
2961 unsigned int s_flags = sb->s_flags;
2962 int ret, nr_orphans = 0, nr_truncates = 0;
2964 int quota_update = 0;
2967 if (!es->s_last_orphan) {
2968 jbd_debug(4, "no orphan inodes to clean up\n");
2972 if (bdev_read_only(sb->s_bdev)) {
2973 ext4_msg(sb, KERN_ERR, "write access "
2974 "unavailable, skipping orphan cleanup");
2978 /* Check if feature set would not allow a r/w mount */
2979 if (!ext4_feature_set_ok(sb, 0)) {
2980 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2981 "unknown ROCOMPAT features");
2985 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2986 /* don't clear list on RO mount w/ errors */
2987 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2988 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2989 "clearing orphan list.\n");
2990 es->s_last_orphan = 0;
2992 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2996 if (s_flags & SB_RDONLY) {
2997 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2998 sb->s_flags &= ~SB_RDONLY;
3001 /* Needed for iput() to work correctly and not trash data */
3002 sb->s_flags |= SB_ACTIVE;
3005 * Turn on quotas which were not enabled for read-only mounts if
3006 * filesystem has quota feature, so that they are updated correctly.
3008 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
3009 int ret = ext4_enable_quotas(sb);
3014 ext4_msg(sb, KERN_ERR,
3015 "Cannot turn on quotas: error %d", ret);
3018 /* Turn on journaled quotas used for old sytle */
3019 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3020 if (EXT4_SB(sb)->s_qf_names[i]) {
3021 int ret = ext4_quota_on_mount(sb, i);
3026 ext4_msg(sb, KERN_ERR,
3027 "Cannot turn on journaled "
3028 "quota: type %d: error %d", i, ret);
3033 while (es->s_last_orphan) {
3034 struct inode *inode;
3037 * We may have encountered an error during cleanup; if
3038 * so, skip the rest.
3040 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3041 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3042 es->s_last_orphan = 0;
3046 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3047 if (IS_ERR(inode)) {
3048 es->s_last_orphan = 0;
3052 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3053 dquot_initialize(inode);
3054 if (inode->i_nlink) {
3055 if (test_opt(sb, DEBUG))
3056 ext4_msg(sb, KERN_DEBUG,
3057 "%s: truncating inode %lu to %lld bytes",
3058 __func__, inode->i_ino, inode->i_size);
3059 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3060 inode->i_ino, inode->i_size);
3062 truncate_inode_pages(inode->i_mapping, inode->i_size);
3063 ret = ext4_truncate(inode);
3065 ext4_std_error(inode->i_sb, ret);
3066 inode_unlock(inode);
3069 if (test_opt(sb, DEBUG))
3070 ext4_msg(sb, KERN_DEBUG,
3071 "%s: deleting unreferenced inode %lu",
3072 __func__, inode->i_ino);
3073 jbd_debug(2, "deleting unreferenced inode %lu\n",
3077 iput(inode); /* The delete magic happens here! */
3080 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3083 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3084 PLURAL(nr_orphans));
3086 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3087 PLURAL(nr_truncates));
3089 /* Turn off quotas if they were enabled for orphan cleanup */
3091 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3092 if (sb_dqopt(sb)->files[i])
3093 dquot_quota_off(sb, i);
3097 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3101 * Maximal extent format file size.
3102 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3103 * extent format containers, within a sector_t, and within i_blocks
3104 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3105 * so that won't be a limiting factor.
3107 * However there is other limiting factor. We do store extents in the form
3108 * of starting block and length, hence the resulting length of the extent
3109 * covering maximum file size must fit into on-disk format containers as
3110 * well. Given that length is always by 1 unit bigger than max unit (because
3111 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3113 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3115 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3118 loff_t upper_limit = MAX_LFS_FILESIZE;
3120 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3122 if (!has_huge_files) {
3123 upper_limit = (1LL << 32) - 1;
3125 /* total blocks in file system block size */
3126 upper_limit >>= (blkbits - 9);
3127 upper_limit <<= blkbits;
3131 * 32-bit extent-start container, ee_block. We lower the maxbytes
3132 * by one fs block, so ee_len can cover the extent of maximum file
3135 res = (1LL << 32) - 1;
3138 /* Sanity check against vm- & vfs- imposed limits */
3139 if (res > upper_limit)
3146 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3147 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3148 * We need to be 1 filesystem block less than the 2^48 sector limit.
3150 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3152 loff_t res = EXT4_NDIR_BLOCKS;
3155 /* This is calculated to be the largest file size for a dense, block
3156 * mapped file such that the file's total number of 512-byte sectors,
3157 * including data and all indirect blocks, does not exceed (2^48 - 1).
3159 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3160 * number of 512-byte sectors of the file.
3163 if (!has_huge_files) {
3165 * !has_huge_files or implies that the inode i_block field
3166 * represents total file blocks in 2^32 512-byte sectors ==
3167 * size of vfs inode i_blocks * 8
3169 upper_limit = (1LL << 32) - 1;
3171 /* total blocks in file system block size */
3172 upper_limit >>= (bits - 9);
3176 * We use 48 bit ext4_inode i_blocks
3177 * With EXT4_HUGE_FILE_FL set the i_blocks
3178 * represent total number of blocks in
3179 * file system block size
3181 upper_limit = (1LL << 48) - 1;
3185 /* indirect blocks */
3187 /* double indirect blocks */
3188 meta_blocks += 1 + (1LL << (bits-2));
3189 /* tripple indirect blocks */
3190 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3192 upper_limit -= meta_blocks;
3193 upper_limit <<= bits;
3195 res += 1LL << (bits-2);
3196 res += 1LL << (2*(bits-2));
3197 res += 1LL << (3*(bits-2));
3199 if (res > upper_limit)
3202 if (res > MAX_LFS_FILESIZE)
3203 res = MAX_LFS_FILESIZE;
3208 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3209 ext4_fsblk_t logical_sb_block, int nr)
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3212 ext4_group_t bg, first_meta_bg;
3215 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3217 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3218 return logical_sb_block + nr + 1;
3219 bg = sbi->s_desc_per_block * nr;
3220 if (ext4_bg_has_super(sb, bg))
3224 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3225 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3226 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3229 if (sb->s_blocksize == 1024 && nr == 0 &&
3230 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3233 return (has_super + ext4_group_first_block_no(sb, bg));
3237 * ext4_get_stripe_size: Get the stripe size.
3238 * @sbi: In memory super block info
3240 * If we have specified it via mount option, then
3241 * use the mount option value. If the value specified at mount time is
3242 * greater than the blocks per group use the super block value.
3243 * If the super block value is greater than blocks per group return 0.
3244 * Allocator needs it be less than blocks per group.
3247 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3249 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3250 unsigned long stripe_width =
3251 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3254 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3255 ret = sbi->s_stripe;
3256 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3258 else if (stride && stride <= sbi->s_blocks_per_group)
3264 * If the stripe width is 1, this makes no sense and
3265 * we set it to 0 to turn off stripe handling code.
3274 * Check whether this filesystem can be mounted based on
3275 * the features present and the RDONLY/RDWR mount requested.
3276 * Returns 1 if this filesystem can be mounted as requested,
3277 * 0 if it cannot be.
3279 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3281 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3282 ext4_msg(sb, KERN_ERR,
3283 "Couldn't mount because of "
3284 "unsupported optional features (%x)",
3285 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3286 ~EXT4_FEATURE_INCOMPAT_SUPP));
3290 #ifndef CONFIG_UNICODE
3291 if (ext4_has_feature_casefold(sb)) {
3292 ext4_msg(sb, KERN_ERR,
3293 "Filesystem with casefold feature cannot be "
3294 "mounted without CONFIG_UNICODE");
3302 if (ext4_has_feature_readonly(sb)) {
3303 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3304 sb->s_flags |= SB_RDONLY;
3308 /* Check that feature set is OK for a read-write mount */
3309 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3310 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3311 "unsupported optional features (%x)",
3312 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3313 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3316 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3317 ext4_msg(sb, KERN_ERR,
3318 "Can't support bigalloc feature without "
3319 "extents feature\n");
3323 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3324 if (!readonly && (ext4_has_feature_quota(sb) ||
3325 ext4_has_feature_project(sb))) {
3326 ext4_msg(sb, KERN_ERR,
3327 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3330 #endif /* CONFIG_QUOTA */
3335 * This function is called once a day if we have errors logged
3336 * on the file system
3338 static void print_daily_error_info(struct timer_list *t)
3340 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3341 struct super_block *sb = sbi->s_sb;
3342 struct ext4_super_block *es = sbi->s_es;
3344 if (es->s_error_count)
3345 /* fsck newer than v1.41.13 is needed to clean this condition. */
3346 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3347 le32_to_cpu(es->s_error_count));
3348 if (es->s_first_error_time) {
3349 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3351 ext4_get_tstamp(es, s_first_error_time),
3352 (int) sizeof(es->s_first_error_func),
3353 es->s_first_error_func,
3354 le32_to_cpu(es->s_first_error_line));
3355 if (es->s_first_error_ino)
3356 printk(KERN_CONT ": inode %u",
3357 le32_to_cpu(es->s_first_error_ino));
3358 if (es->s_first_error_block)
3359 printk(KERN_CONT ": block %llu", (unsigned long long)
3360 le64_to_cpu(es->s_first_error_block));
3361 printk(KERN_CONT "\n");
3363 if (es->s_last_error_time) {
3364 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3366 ext4_get_tstamp(es, s_last_error_time),
3367 (int) sizeof(es->s_last_error_func),
3368 es->s_last_error_func,
3369 le32_to_cpu(es->s_last_error_line));
3370 if (es->s_last_error_ino)
3371 printk(KERN_CONT ": inode %u",
3372 le32_to_cpu(es->s_last_error_ino));
3373 if (es->s_last_error_block)
3374 printk(KERN_CONT ": block %llu", (unsigned long long)
3375 le64_to_cpu(es->s_last_error_block));
3376 printk(KERN_CONT "\n");
3378 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3381 /* Find next suitable group and run ext4_init_inode_table */
3382 static int ext4_run_li_request(struct ext4_li_request *elr)
3384 struct ext4_group_desc *gdp = NULL;
3385 struct super_block *sb = elr->lr_super;
3386 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3387 ext4_group_t group = elr->lr_next_group;
3388 unsigned long timeout = 0;
3389 unsigned int prefetch_ios = 0;
3392 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3393 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3394 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3396 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3398 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3400 if (group >= elr->lr_next_group) {
3402 if (elr->lr_first_not_zeroed != ngroups &&
3403 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3404 elr->lr_next_group = elr->lr_first_not_zeroed;
3405 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3412 for (; group < ngroups; group++) {
3413 gdp = ext4_get_group_desc(sb, group, NULL);
3419 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3423 if (group >= ngroups)
3428 ret = ext4_init_inode_table(sb, group,
3429 elr->lr_timeout ? 0 : 1);
3430 trace_ext4_lazy_itable_init(sb, group);
3431 if (elr->lr_timeout == 0) {
3432 timeout = (jiffies - timeout) *
3433 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3434 elr->lr_timeout = timeout;
3436 elr->lr_next_sched = jiffies + elr->lr_timeout;
3437 elr->lr_next_group = group + 1;
3443 * Remove lr_request from the list_request and free the
3444 * request structure. Should be called with li_list_mtx held
3446 static void ext4_remove_li_request(struct ext4_li_request *elr)
3451 list_del(&elr->lr_request);
3452 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3456 static void ext4_unregister_li_request(struct super_block *sb)
3458 mutex_lock(&ext4_li_mtx);
3459 if (!ext4_li_info) {
3460 mutex_unlock(&ext4_li_mtx);
3464 mutex_lock(&ext4_li_info->li_list_mtx);
3465 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3466 mutex_unlock(&ext4_li_info->li_list_mtx);
3467 mutex_unlock(&ext4_li_mtx);
3470 static struct task_struct *ext4_lazyinit_task;
3473 * This is the function where ext4lazyinit thread lives. It walks
3474 * through the request list searching for next scheduled filesystem.
3475 * When such a fs is found, run the lazy initialization request
3476 * (ext4_rn_li_request) and keep track of the time spend in this
3477 * function. Based on that time we compute next schedule time of
3478 * the request. When walking through the list is complete, compute
3479 * next waking time and put itself into sleep.
3481 static int ext4_lazyinit_thread(void *arg)
3483 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3484 struct list_head *pos, *n;
3485 struct ext4_li_request *elr;
3486 unsigned long next_wakeup, cur;
3488 BUG_ON(NULL == eli);
3492 next_wakeup = MAX_JIFFY_OFFSET;
3494 mutex_lock(&eli->li_list_mtx);
3495 if (list_empty(&eli->li_request_list)) {
3496 mutex_unlock(&eli->li_list_mtx);
3499 list_for_each_safe(pos, n, &eli->li_request_list) {
3502 elr = list_entry(pos, struct ext4_li_request,
3505 if (time_before(jiffies, elr->lr_next_sched)) {
3506 if (time_before(elr->lr_next_sched, next_wakeup))
3507 next_wakeup = elr->lr_next_sched;
3510 if (down_read_trylock(&elr->lr_super->s_umount)) {
3511 if (sb_start_write_trylock(elr->lr_super)) {
3514 * We hold sb->s_umount, sb can not
3515 * be removed from the list, it is
3516 * now safe to drop li_list_mtx
3518 mutex_unlock(&eli->li_list_mtx);
3519 err = ext4_run_li_request(elr);
3520 sb_end_write(elr->lr_super);
3521 mutex_lock(&eli->li_list_mtx);
3524 up_read((&elr->lr_super->s_umount));
3526 /* error, remove the lazy_init job */
3528 ext4_remove_li_request(elr);
3532 elr->lr_next_sched = jiffies +
3534 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3536 if (time_before(elr->lr_next_sched, next_wakeup))
3537 next_wakeup = elr->lr_next_sched;
3539 mutex_unlock(&eli->li_list_mtx);
3544 if ((time_after_eq(cur, next_wakeup)) ||
3545 (MAX_JIFFY_OFFSET == next_wakeup)) {
3550 schedule_timeout_interruptible(next_wakeup - cur);
3552 if (kthread_should_stop()) {
3553 ext4_clear_request_list();
3560 * It looks like the request list is empty, but we need
3561 * to check it under the li_list_mtx lock, to prevent any
3562 * additions into it, and of course we should lock ext4_li_mtx
3563 * to atomically free the list and ext4_li_info, because at
3564 * this point another ext4 filesystem could be registering
3567 mutex_lock(&ext4_li_mtx);
3568 mutex_lock(&eli->li_list_mtx);
3569 if (!list_empty(&eli->li_request_list)) {
3570 mutex_unlock(&eli->li_list_mtx);
3571 mutex_unlock(&ext4_li_mtx);
3574 mutex_unlock(&eli->li_list_mtx);
3575 kfree(ext4_li_info);
3576 ext4_li_info = NULL;
3577 mutex_unlock(&ext4_li_mtx);
3582 static void ext4_clear_request_list(void)
3584 struct list_head *pos, *n;
3585 struct ext4_li_request *elr;
3587 mutex_lock(&ext4_li_info->li_list_mtx);
3588 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3589 elr = list_entry(pos, struct ext4_li_request,
3591 ext4_remove_li_request(elr);
3593 mutex_unlock(&ext4_li_info->li_list_mtx);
3596 static int ext4_run_lazyinit_thread(void)
3598 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3599 ext4_li_info, "ext4lazyinit");
3600 if (IS_ERR(ext4_lazyinit_task)) {
3601 int err = PTR_ERR(ext4_lazyinit_task);
3602 ext4_clear_request_list();
3603 kfree(ext4_li_info);
3604 ext4_li_info = NULL;
3605 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3606 "initialization thread\n",
3610 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3615 * Check whether it make sense to run itable init. thread or not.
3616 * If there is at least one uninitialized inode table, return
3617 * corresponding group number, else the loop goes through all
3618 * groups and return total number of groups.
3620 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3622 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3623 struct ext4_group_desc *gdp = NULL;
3625 if (!ext4_has_group_desc_csum(sb))
3628 for (group = 0; group < ngroups; group++) {
3629 gdp = ext4_get_group_desc(sb, group, NULL);
3633 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3640 static int ext4_li_info_new(void)
3642 struct ext4_lazy_init *eli = NULL;
3644 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3648 INIT_LIST_HEAD(&eli->li_request_list);
3649 mutex_init(&eli->li_list_mtx);
3651 eli->li_state |= EXT4_LAZYINIT_QUIT;
3658 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3661 struct ext4_li_request *elr;
3663 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3668 elr->lr_first_not_zeroed = start;
3669 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3670 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3672 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3673 elr->lr_next_group = start;
3677 * Randomize first schedule time of the request to
3678 * spread the inode table initialization requests
3681 elr->lr_next_sched = jiffies + (prandom_u32() %
3682 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3686 int ext4_register_li_request(struct super_block *sb,
3687 ext4_group_t first_not_zeroed)
3689 struct ext4_sb_info *sbi = EXT4_SB(sb);
3690 struct ext4_li_request *elr = NULL;
3691 ext4_group_t ngroups = sbi->s_groups_count;
3694 mutex_lock(&ext4_li_mtx);
3695 if (sbi->s_li_request != NULL) {
3697 * Reset timeout so it can be computed again, because
3698 * s_li_wait_mult might have changed.
3700 sbi->s_li_request->lr_timeout = 0;
3704 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3705 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3706 !test_opt(sb, INIT_INODE_TABLE)))
3709 elr = ext4_li_request_new(sb, first_not_zeroed);
3715 if (NULL == ext4_li_info) {
3716 ret = ext4_li_info_new();
3721 mutex_lock(&ext4_li_info->li_list_mtx);
3722 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3723 mutex_unlock(&ext4_li_info->li_list_mtx);
3725 sbi->s_li_request = elr;
3727 * set elr to NULL here since it has been inserted to
3728 * the request_list and the removal and free of it is
3729 * handled by ext4_clear_request_list from now on.
3733 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3734 ret = ext4_run_lazyinit_thread();
3739 mutex_unlock(&ext4_li_mtx);
3746 * We do not need to lock anything since this is called on
3749 static void ext4_destroy_lazyinit_thread(void)
3752 * If thread exited earlier
3753 * there's nothing to be done.
3755 if (!ext4_li_info || !ext4_lazyinit_task)
3758 kthread_stop(ext4_lazyinit_task);
3761 static int set_journal_csum_feature_set(struct super_block *sb)
3764 int compat, incompat;
3765 struct ext4_sb_info *sbi = EXT4_SB(sb);
3767 if (ext4_has_metadata_csum(sb)) {
3768 /* journal checksum v3 */
3770 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3772 /* journal checksum v1 */
3773 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3777 jbd2_journal_clear_features(sbi->s_journal,
3778 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3779 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3780 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3781 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3782 ret = jbd2_journal_set_features(sbi->s_journal,
3784 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3786 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3787 ret = jbd2_journal_set_features(sbi->s_journal,
3790 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3791 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3793 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3794 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3801 * Note: calculating the overhead so we can be compatible with
3802 * historical BSD practice is quite difficult in the face of
3803 * clusters/bigalloc. This is because multiple metadata blocks from
3804 * different block group can end up in the same allocation cluster.
3805 * Calculating the exact overhead in the face of clustered allocation
3806 * requires either O(all block bitmaps) in memory or O(number of block
3807 * groups**2) in time. We will still calculate the superblock for
3808 * older file systems --- and if we come across with a bigalloc file
3809 * system with zero in s_overhead_clusters the estimate will be close to
3810 * correct especially for very large cluster sizes --- but for newer
3811 * file systems, it's better to calculate this figure once at mkfs
3812 * time, and store it in the superblock. If the superblock value is
3813 * present (even for non-bigalloc file systems), we will use it.
3815 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3818 struct ext4_sb_info *sbi = EXT4_SB(sb);
3819 struct ext4_group_desc *gdp;
3820 ext4_fsblk_t first_block, last_block, b;
3821 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3822 int s, j, count = 0;
3824 if (!ext4_has_feature_bigalloc(sb))
3825 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3826 sbi->s_itb_per_group + 2);
3828 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3829 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3830 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3831 for (i = 0; i < ngroups; i++) {
3832 gdp = ext4_get_group_desc(sb, i, NULL);
3833 b = ext4_block_bitmap(sb, gdp);
3834 if (b >= first_block && b <= last_block) {
3835 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3838 b = ext4_inode_bitmap(sb, gdp);
3839 if (b >= first_block && b <= last_block) {
3840 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3843 b = ext4_inode_table(sb, gdp);
3844 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3845 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3846 int c = EXT4_B2C(sbi, b - first_block);
3847 ext4_set_bit(c, buf);
3853 if (ext4_bg_has_super(sb, grp)) {
3854 ext4_set_bit(s++, buf);
3857 j = ext4_bg_num_gdb(sb, grp);
3858 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3859 ext4_error(sb, "Invalid number of block group "
3860 "descriptor blocks: %d", j);
3861 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3865 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3869 return EXT4_CLUSTERS_PER_GROUP(sb) -
3870 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3874 * Compute the overhead and stash it in sbi->s_overhead
3876 int ext4_calculate_overhead(struct super_block *sb)
3878 struct ext4_sb_info *sbi = EXT4_SB(sb);
3879 struct ext4_super_block *es = sbi->s_es;
3880 struct inode *j_inode;
3881 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3882 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3883 ext4_fsblk_t overhead = 0;
3884 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3890 * Compute the overhead (FS structures). This is constant
3891 * for a given filesystem unless the number of block groups
3892 * changes so we cache the previous value until it does.
3896 * All of the blocks before first_data_block are overhead
3898 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3901 * Add the overhead found in each block group
3903 for (i = 0; i < ngroups; i++) {
3906 blks = count_overhead(sb, i, buf);
3909 memset(buf, 0, PAGE_SIZE);
3914 * Add the internal journal blocks whether the journal has been
3917 if (sbi->s_journal && !sbi->s_journal_bdev)
3918 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3919 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3920 /* j_inum for internal journal is non-zero */
3921 j_inode = ext4_get_journal_inode(sb, j_inum);
3923 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3924 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3927 ext4_msg(sb, KERN_ERR, "can't get journal size");
3930 sbi->s_overhead = overhead;
3932 free_page((unsigned long) buf);
3936 static void ext4_set_resv_clusters(struct super_block *sb)
3938 ext4_fsblk_t resv_clusters;
3939 struct ext4_sb_info *sbi = EXT4_SB(sb);
3942 * There's no need to reserve anything when we aren't using extents.
3943 * The space estimates are exact, there are no unwritten extents,
3944 * hole punching doesn't need new metadata... This is needed especially
3945 * to keep ext2/3 backward compatibility.
3947 if (!ext4_has_feature_extents(sb))
3950 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3951 * This should cover the situations where we can not afford to run
3952 * out of space like for example punch hole, or converting
3953 * unwritten extents in delalloc path. In most cases such
3954 * allocation would require 1, or 2 blocks, higher numbers are
3957 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3958 sbi->s_cluster_bits);
3960 do_div(resv_clusters, 50);
3961 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3963 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3966 static const char *ext4_quota_mode(struct super_block *sb)
3969 if (!ext4_quota_capable(sb))
3972 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3973 return "journalled";
3981 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3983 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3984 char *orig_data = kstrdup(data, GFP_KERNEL);
3985 struct buffer_head *bh, **group_desc;
3986 struct ext4_super_block *es = NULL;
3987 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3988 struct flex_groups **flex_groups;
3990 ext4_fsblk_t sb_block = get_sb_block(&data);
3991 ext4_fsblk_t logical_sb_block;
3992 unsigned long offset = 0;
3993 unsigned long journal_devnum = 0;
3994 unsigned long def_mount_opts;
3998 int blocksize, clustersize;
3999 unsigned int db_count;
4001 int needs_recovery, has_huge_files;
4004 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4005 ext4_group_t first_not_zeroed;
4007 if ((data && !orig_data) || !sbi)
4010 sbi->s_daxdev = dax_dev;
4011 sbi->s_blockgroup_lock =
4012 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4013 if (!sbi->s_blockgroup_lock)
4016 sb->s_fs_info = sbi;
4018 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4019 sbi->s_sb_block = sb_block;
4020 if (sb->s_bdev->bd_part)
4021 sbi->s_sectors_written_start =
4022 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
4024 /* Cleanup superblock name */
4025 strreplace(sb->s_id, '/', '!');
4027 /* -EINVAL is default */
4029 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4031 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4036 * The ext4 superblock will not be buffer aligned for other than 1kB
4037 * block sizes. We need to calculate the offset from buffer start.
4039 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4040 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4041 offset = do_div(logical_sb_block, blocksize);
4043 logical_sb_block = sb_block;
4046 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4048 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4053 * Note: s_es must be initialized as soon as possible because
4054 * some ext4 macro-instructions depend on its value
4056 es = (struct ext4_super_block *) (bh->b_data + offset);
4058 sb->s_magic = le16_to_cpu(es->s_magic);
4059 if (sb->s_magic != EXT4_SUPER_MAGIC)
4061 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4063 /* Warn if metadata_csum and gdt_csum are both set. */
4064 if (ext4_has_feature_metadata_csum(sb) &&
4065 ext4_has_feature_gdt_csum(sb))
4066 ext4_warning(sb, "metadata_csum and uninit_bg are "
4067 "redundant flags; please run fsck.");
4069 /* Check for a known checksum algorithm */
4070 if (!ext4_verify_csum_type(sb, es)) {
4071 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4072 "unknown checksum algorithm.");
4077 /* Load the checksum driver */
4078 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4079 if (IS_ERR(sbi->s_chksum_driver)) {
4080 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4081 ret = PTR_ERR(sbi->s_chksum_driver);
4082 sbi->s_chksum_driver = NULL;
4086 /* Check superblock checksum */
4087 if (!ext4_superblock_csum_verify(sb, es)) {
4088 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4089 "invalid superblock checksum. Run e2fsck?");
4095 /* Precompute checksum seed for all metadata */
4096 if (ext4_has_feature_csum_seed(sb))
4097 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4098 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4099 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4100 sizeof(es->s_uuid));
4102 /* Set defaults before we parse the mount options */
4103 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4104 set_opt(sb, INIT_INODE_TABLE);
4105 if (def_mount_opts & EXT4_DEFM_DEBUG)
4107 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4109 if (def_mount_opts & EXT4_DEFM_UID16)
4110 set_opt(sb, NO_UID32);
4111 /* xattr user namespace & acls are now defaulted on */
4112 set_opt(sb, XATTR_USER);
4113 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4114 set_opt(sb, POSIX_ACL);
4116 if (ext4_has_feature_fast_commit(sb))
4117 set_opt2(sb, JOURNAL_FAST_COMMIT);
4118 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4119 if (ext4_has_metadata_csum(sb))
4120 set_opt(sb, JOURNAL_CHECKSUM);
4122 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4123 set_opt(sb, JOURNAL_DATA);
4124 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4125 set_opt(sb, ORDERED_DATA);
4126 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4127 set_opt(sb, WRITEBACK_DATA);
4129 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4130 set_opt(sb, ERRORS_PANIC);
4131 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4132 set_opt(sb, ERRORS_CONT);
4134 set_opt(sb, ERRORS_RO);
4135 /* block_validity enabled by default; disable with noblock_validity */
4136 set_opt(sb, BLOCK_VALIDITY);
4137 if (def_mount_opts & EXT4_DEFM_DISCARD)
4138 set_opt(sb, DISCARD);
4140 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4141 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4142 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4143 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4144 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4146 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4147 set_opt(sb, BARRIER);
4150 * enable delayed allocation by default
4151 * Use -o nodelalloc to turn it off
4153 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4154 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4155 set_opt(sb, DELALLOC);
4158 * set default s_li_wait_mult for lazyinit, for the case there is
4159 * no mount option specified.
4161 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4163 if (le32_to_cpu(es->s_log_block_size) >
4164 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4165 ext4_msg(sb, KERN_ERR,
4166 "Invalid log block size: %u",
4167 le32_to_cpu(es->s_log_block_size));
4170 if (le32_to_cpu(es->s_log_cluster_size) >
4171 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4172 ext4_msg(sb, KERN_ERR,
4173 "Invalid log cluster size: %u",
4174 le32_to_cpu(es->s_log_cluster_size));
4178 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4180 if (blocksize == PAGE_SIZE)
4181 set_opt(sb, DIOREAD_NOLOCK);
4183 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4184 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4185 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4187 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4188 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4189 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4190 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4194 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4195 (!is_power_of_2(sbi->s_inode_size)) ||
4196 (sbi->s_inode_size > blocksize)) {
4197 ext4_msg(sb, KERN_ERR,
4198 "unsupported inode size: %d",
4200 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4204 * i_atime_extra is the last extra field available for
4205 * [acm]times in struct ext4_inode. Checking for that
4206 * field should suffice to ensure we have extra space
4209 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4210 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4211 sb->s_time_gran = 1;
4212 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4214 sb->s_time_gran = NSEC_PER_SEC;
4215 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4217 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4219 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4220 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4221 EXT4_GOOD_OLD_INODE_SIZE;
4222 if (ext4_has_feature_extra_isize(sb)) {
4223 unsigned v, max = (sbi->s_inode_size -
4224 EXT4_GOOD_OLD_INODE_SIZE);
4226 v = le16_to_cpu(es->s_want_extra_isize);
4228 ext4_msg(sb, KERN_ERR,
4229 "bad s_want_extra_isize: %d", v);
4232 if (sbi->s_want_extra_isize < v)
4233 sbi->s_want_extra_isize = v;
4235 v = le16_to_cpu(es->s_min_extra_isize);
4237 ext4_msg(sb, KERN_ERR,
4238 "bad s_min_extra_isize: %d", v);
4241 if (sbi->s_want_extra_isize < v)
4242 sbi->s_want_extra_isize = v;
4246 if (sbi->s_es->s_mount_opts[0]) {
4247 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4248 sizeof(sbi->s_es->s_mount_opts),
4252 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4253 &journal_ioprio, 0)) {
4254 ext4_msg(sb, KERN_WARNING,
4255 "failed to parse options in superblock: %s",
4258 kfree(s_mount_opts);
4260 sbi->s_def_mount_opt = sbi->s_mount_opt;
4261 if (!parse_options((char *) data, sb, &journal_devnum,
4262 &journal_ioprio, 0))
4265 #ifdef CONFIG_UNICODE
4266 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4267 const struct ext4_sb_encodings *encoding_info;
4268 struct unicode_map *encoding;
4269 __u16 encoding_flags;
4271 if (ext4_has_feature_encrypt(sb)) {
4272 ext4_msg(sb, KERN_ERR,
4273 "Can't mount with encoding and encryption");
4277 if (ext4_sb_read_encoding(es, &encoding_info,
4279 ext4_msg(sb, KERN_ERR,
4280 "Encoding requested by superblock is unknown");
4284 encoding = utf8_load(encoding_info->version);
4285 if (IS_ERR(encoding)) {
4286 ext4_msg(sb, KERN_ERR,
4287 "can't mount with superblock charset: %s-%s "
4288 "not supported by the kernel. flags: 0x%x.",
4289 encoding_info->name, encoding_info->version,
4293 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4294 "%s-%s with flags 0x%hx", encoding_info->name,
4295 encoding_info->version?:"\b", encoding_flags);
4297 sb->s_encoding = encoding;
4298 sb->s_encoding_flags = encoding_flags;
4302 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4303 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4304 /* can't mount with both data=journal and dioread_nolock. */
4305 clear_opt(sb, DIOREAD_NOLOCK);
4306 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4307 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4308 ext4_msg(sb, KERN_ERR, "can't mount with "
4309 "both data=journal and delalloc");
4312 if (test_opt(sb, DAX_ALWAYS)) {
4313 ext4_msg(sb, KERN_ERR, "can't mount with "
4314 "both data=journal and dax");
4317 if (ext4_has_feature_encrypt(sb)) {
4318 ext4_msg(sb, KERN_WARNING,
4319 "encrypted files will use data=ordered "
4320 "instead of data journaling mode");
4322 if (test_opt(sb, DELALLOC))
4323 clear_opt(sb, DELALLOC);
4325 sb->s_iflags |= SB_I_CGROUPWB;
4328 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4329 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4331 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4332 (ext4_has_compat_features(sb) ||
4333 ext4_has_ro_compat_features(sb) ||
4334 ext4_has_incompat_features(sb)))
4335 ext4_msg(sb, KERN_WARNING,
4336 "feature flags set on rev 0 fs, "
4337 "running e2fsck is recommended");
4339 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4340 set_opt2(sb, HURD_COMPAT);
4341 if (ext4_has_feature_64bit(sb)) {
4342 ext4_msg(sb, KERN_ERR,
4343 "The Hurd can't support 64-bit file systems");
4348 * ea_inode feature uses l_i_version field which is not
4349 * available in HURD_COMPAT mode.
4351 if (ext4_has_feature_ea_inode(sb)) {
4352 ext4_msg(sb, KERN_ERR,
4353 "ea_inode feature is not supported for Hurd");
4358 if (IS_EXT2_SB(sb)) {
4359 if (ext2_feature_set_ok(sb))
4360 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4361 "using the ext4 subsystem");
4364 * If we're probing be silent, if this looks like
4365 * it's actually an ext[34] filesystem.
4367 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4369 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4370 "to feature incompatibilities");
4375 if (IS_EXT3_SB(sb)) {
4376 if (ext3_feature_set_ok(sb))
4377 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4378 "using the ext4 subsystem");
4381 * If we're probing be silent, if this looks like
4382 * it's actually an ext4 filesystem.
4384 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4386 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4387 "to feature incompatibilities");
4393 * Check feature flags regardless of the revision level, since we
4394 * previously didn't change the revision level when setting the flags,
4395 * so there is a chance incompat flags are set on a rev 0 filesystem.
4397 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4400 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4401 ext4_msg(sb, KERN_ERR,
4402 "Number of reserved GDT blocks insanely large: %d",
4403 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4407 if (bdev_dax_supported(sb->s_bdev, blocksize))
4408 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4410 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4411 if (ext4_has_feature_inline_data(sb)) {
4412 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4413 " that may contain inline data");
4416 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4417 ext4_msg(sb, KERN_ERR,
4418 "DAX unsupported by block device.");
4423 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4424 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4425 es->s_encryption_level);
4429 if (sb->s_blocksize != blocksize) {
4430 /* Validate the filesystem blocksize */
4431 if (!sb_set_blocksize(sb, blocksize)) {
4432 ext4_msg(sb, KERN_ERR, "bad block size %d",
4438 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4439 offset = do_div(logical_sb_block, blocksize);
4440 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4442 ext4_msg(sb, KERN_ERR,
4443 "Can't read superblock on 2nd try");
4448 es = (struct ext4_super_block *)(bh->b_data + offset);
4450 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4451 ext4_msg(sb, KERN_ERR,
4452 "Magic mismatch, very weird!");
4457 has_huge_files = ext4_has_feature_huge_file(sb);
4458 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4460 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4462 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4463 if (ext4_has_feature_64bit(sb)) {
4464 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4465 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4466 !is_power_of_2(sbi->s_desc_size)) {
4467 ext4_msg(sb, KERN_ERR,
4468 "unsupported descriptor size %lu",
4473 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4475 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4476 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4478 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4479 if (sbi->s_inodes_per_block == 0)
4481 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4482 sbi->s_inodes_per_group > blocksize * 8) {
4483 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4484 sbi->s_inodes_per_group);
4487 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4488 sbi->s_inodes_per_block;
4489 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4491 sbi->s_mount_state = le16_to_cpu(es->s_state);
4492 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4493 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4495 for (i = 0; i < 4; i++)
4496 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4497 sbi->s_def_hash_version = es->s_def_hash_version;
4498 if (ext4_has_feature_dir_index(sb)) {
4499 i = le32_to_cpu(es->s_flags);
4500 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4501 sbi->s_hash_unsigned = 3;
4502 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4503 #ifdef __CHAR_UNSIGNED__
4506 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4507 sbi->s_hash_unsigned = 3;
4511 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4516 /* Handle clustersize */
4517 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4518 if (ext4_has_feature_bigalloc(sb)) {
4519 if (clustersize < blocksize) {
4520 ext4_msg(sb, KERN_ERR,
4521 "cluster size (%d) smaller than "
4522 "block size (%d)", clustersize, blocksize);
4525 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4526 le32_to_cpu(es->s_log_block_size);
4527 sbi->s_clusters_per_group =
4528 le32_to_cpu(es->s_clusters_per_group);
4529 if (sbi->s_clusters_per_group > blocksize * 8) {
4530 ext4_msg(sb, KERN_ERR,
4531 "#clusters per group too big: %lu",
4532 sbi->s_clusters_per_group);
4535 if (sbi->s_blocks_per_group !=
4536 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4537 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4538 "clusters per group (%lu) inconsistent",
4539 sbi->s_blocks_per_group,
4540 sbi->s_clusters_per_group);
4544 if (clustersize != blocksize) {
4545 ext4_msg(sb, KERN_ERR,
4546 "fragment/cluster size (%d) != "
4547 "block size (%d)", clustersize, blocksize);
4550 if (sbi->s_blocks_per_group > blocksize * 8) {
4551 ext4_msg(sb, KERN_ERR,
4552 "#blocks per group too big: %lu",
4553 sbi->s_blocks_per_group);
4556 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4557 sbi->s_cluster_bits = 0;
4559 sbi->s_cluster_ratio = clustersize / blocksize;
4561 /* Do we have standard group size of clustersize * 8 blocks ? */
4562 if (sbi->s_blocks_per_group == clustersize << 3)
4563 set_opt2(sb, STD_GROUP_SIZE);
4566 * Test whether we have more sectors than will fit in sector_t,
4567 * and whether the max offset is addressable by the page cache.
4569 err = generic_check_addressable(sb->s_blocksize_bits,
4570 ext4_blocks_count(es));
4572 ext4_msg(sb, KERN_ERR, "filesystem"
4573 " too large to mount safely on this system");
4577 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4580 /* check blocks count against device size */
4581 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4582 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4583 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4584 "exceeds size of device (%llu blocks)",
4585 ext4_blocks_count(es), blocks_count);
4590 * It makes no sense for the first data block to be beyond the end
4591 * of the filesystem.
4593 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4594 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4595 "block %u is beyond end of filesystem (%llu)",
4596 le32_to_cpu(es->s_first_data_block),
4597 ext4_blocks_count(es));
4600 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4601 (sbi->s_cluster_ratio == 1)) {
4602 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4603 "block is 0 with a 1k block and cluster size");
4607 blocks_count = (ext4_blocks_count(es) -
4608 le32_to_cpu(es->s_first_data_block) +
4609 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4610 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4611 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4612 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4613 "(block count %llu, first data block %u, "
4614 "blocks per group %lu)", blocks_count,
4615 ext4_blocks_count(es),
4616 le32_to_cpu(es->s_first_data_block),
4617 EXT4_BLOCKS_PER_GROUP(sb));
4620 sbi->s_groups_count = blocks_count;
4621 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4622 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4623 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4624 le32_to_cpu(es->s_inodes_count)) {
4625 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4626 le32_to_cpu(es->s_inodes_count),
4627 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4631 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4632 EXT4_DESC_PER_BLOCK(sb);
4633 if (ext4_has_feature_meta_bg(sb)) {
4634 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4635 ext4_msg(sb, KERN_WARNING,
4636 "first meta block group too large: %u "
4637 "(group descriptor block count %u)",
4638 le32_to_cpu(es->s_first_meta_bg), db_count);
4642 rcu_assign_pointer(sbi->s_group_desc,
4643 kvmalloc_array(db_count,
4644 sizeof(struct buffer_head *),
4646 if (sbi->s_group_desc == NULL) {
4647 ext4_msg(sb, KERN_ERR, "not enough memory");
4652 bgl_lock_init(sbi->s_blockgroup_lock);
4654 /* Pre-read the descriptors into the buffer cache */
4655 for (i = 0; i < db_count; i++) {
4656 block = descriptor_loc(sb, logical_sb_block, i);
4657 ext4_sb_breadahead_unmovable(sb, block);
4660 for (i = 0; i < db_count; i++) {
4661 struct buffer_head *bh;
4663 block = descriptor_loc(sb, logical_sb_block, i);
4664 bh = ext4_sb_bread_unmovable(sb, block);
4666 ext4_msg(sb, KERN_ERR,
4667 "can't read group descriptor %d", i);
4673 rcu_dereference(sbi->s_group_desc)[i] = bh;
4676 sbi->s_gdb_count = db_count;
4677 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4678 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4679 ret = -EFSCORRUPTED;
4683 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4685 /* Register extent status tree shrinker */
4686 if (ext4_es_register_shrinker(sbi))
4689 sbi->s_stripe = ext4_get_stripe_size(sbi);
4690 sbi->s_extent_max_zeroout_kb = 32;
4693 * set up enough so that it can read an inode
4695 sb->s_op = &ext4_sops;
4696 sb->s_export_op = &ext4_export_ops;
4697 sb->s_xattr = ext4_xattr_handlers;
4698 #ifdef CONFIG_FS_ENCRYPTION
4699 sb->s_cop = &ext4_cryptops;
4701 #ifdef CONFIG_FS_VERITY
4702 sb->s_vop = &ext4_verityops;
4705 sb->dq_op = &ext4_quota_operations;
4706 if (ext4_has_feature_quota(sb))
4707 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4709 sb->s_qcop = &ext4_qctl_operations;
4710 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4712 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4714 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4715 mutex_init(&sbi->s_orphan_lock);
4717 /* Initialize fast commit stuff */
4718 atomic_set(&sbi->s_fc_subtid, 0);
4719 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4720 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4721 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4722 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4723 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4724 sbi->s_fc_bytes = 0;
4725 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4726 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4727 spin_lock_init(&sbi->s_fc_lock);
4728 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4729 sbi->s_fc_replay_state.fc_regions = NULL;
4730 sbi->s_fc_replay_state.fc_regions_size = 0;
4731 sbi->s_fc_replay_state.fc_regions_used = 0;
4732 sbi->s_fc_replay_state.fc_regions_valid = 0;
4733 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4734 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4735 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4739 needs_recovery = (es->s_last_orphan != 0 ||
4740 ext4_has_feature_journal_needs_recovery(sb));
4742 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4743 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4744 goto failed_mount3a;
4747 * The first inode we look at is the journal inode. Don't try
4748 * root first: it may be modified in the journal!
4750 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4751 err = ext4_load_journal(sb, es, journal_devnum);
4753 goto failed_mount3a;
4754 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4755 ext4_has_feature_journal_needs_recovery(sb)) {
4756 ext4_msg(sb, KERN_ERR, "required journal recovery "
4757 "suppressed and not mounted read-only");
4758 goto failed_mount_wq;
4760 /* Nojournal mode, all journal mount options are illegal */
4761 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4762 ext4_msg(sb, KERN_ERR, "can't mount with "
4763 "journal_checksum, fs mounted w/o journal");
4764 goto failed_mount_wq;
4766 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4767 ext4_msg(sb, KERN_ERR, "can't mount with "
4768 "journal_async_commit, fs mounted w/o journal");
4769 goto failed_mount_wq;
4771 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4772 ext4_msg(sb, KERN_ERR, "can't mount with "
4773 "commit=%lu, fs mounted w/o journal",
4774 sbi->s_commit_interval / HZ);
4775 goto failed_mount_wq;
4777 if (EXT4_MOUNT_DATA_FLAGS &
4778 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4779 ext4_msg(sb, KERN_ERR, "can't mount with "
4780 "data=, fs mounted w/o journal");
4781 goto failed_mount_wq;
4783 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4784 clear_opt(sb, JOURNAL_CHECKSUM);
4785 clear_opt(sb, DATA_FLAGS);
4786 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4787 sbi->s_journal = NULL;
4792 if (ext4_has_feature_64bit(sb) &&
4793 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4794 JBD2_FEATURE_INCOMPAT_64BIT)) {
4795 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4796 goto failed_mount_wq;
4799 if (!set_journal_csum_feature_set(sb)) {
4800 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4802 goto failed_mount_wq;
4805 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4806 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4807 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4808 ext4_msg(sb, KERN_ERR,
4809 "Failed to set fast commit journal feature");
4810 goto failed_mount_wq;
4813 /* We have now updated the journal if required, so we can
4814 * validate the data journaling mode. */
4815 switch (test_opt(sb, DATA_FLAGS)) {
4817 /* No mode set, assume a default based on the journal
4818 * capabilities: ORDERED_DATA if the journal can
4819 * cope, else JOURNAL_DATA
4821 if (jbd2_journal_check_available_features
4822 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4823 set_opt(sb, ORDERED_DATA);
4824 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4826 set_opt(sb, JOURNAL_DATA);
4827 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4831 case EXT4_MOUNT_ORDERED_DATA:
4832 case EXT4_MOUNT_WRITEBACK_DATA:
4833 if (!jbd2_journal_check_available_features
4834 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4835 ext4_msg(sb, KERN_ERR, "Journal does not support "
4836 "requested data journaling mode");
4837 goto failed_mount_wq;
4844 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4845 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4846 ext4_msg(sb, KERN_ERR, "can't mount with "
4847 "journal_async_commit in data=ordered mode");
4848 goto failed_mount_wq;
4851 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4853 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4854 sbi->s_journal->j_submit_inode_data_buffers =
4855 ext4_journal_submit_inode_data_buffers;
4856 sbi->s_journal->j_finish_inode_data_buffers =
4857 ext4_journal_finish_inode_data_buffers;
4860 if (!test_opt(sb, NO_MBCACHE)) {
4861 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4862 if (!sbi->s_ea_block_cache) {
4863 ext4_msg(sb, KERN_ERR,
4864 "Failed to create ea_block_cache");
4865 goto failed_mount_wq;
4868 if (ext4_has_feature_ea_inode(sb)) {
4869 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4870 if (!sbi->s_ea_inode_cache) {
4871 ext4_msg(sb, KERN_ERR,
4872 "Failed to create ea_inode_cache");
4873 goto failed_mount_wq;
4878 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4879 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4880 goto failed_mount_wq;
4883 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4884 !ext4_has_feature_encrypt(sb)) {
4885 ext4_set_feature_encrypt(sb);
4886 ext4_commit_super(sb, 1);
4890 * Get the # of file system overhead blocks from the
4891 * superblock if present.
4893 if (es->s_overhead_clusters)
4894 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4896 err = ext4_calculate_overhead(sb);
4898 goto failed_mount_wq;
4902 * The maximum number of concurrent works can be high and
4903 * concurrency isn't really necessary. Limit it to 1.
4905 EXT4_SB(sb)->rsv_conversion_wq =
4906 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4907 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4908 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4914 * The jbd2_journal_load will have done any necessary log recovery,
4915 * so we can safely mount the rest of the filesystem now.
4918 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4920 ext4_msg(sb, KERN_ERR, "get root inode failed");
4921 ret = PTR_ERR(root);
4925 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4926 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4931 #ifdef CONFIG_UNICODE
4933 sb->s_d_op = &ext4_dentry_ops;
4936 sb->s_root = d_make_root(root);
4938 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4943 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4944 if (ret == -EROFS) {
4945 sb->s_flags |= SB_RDONLY;
4948 goto failed_mount4a;
4950 ext4_set_resv_clusters(sb);
4952 if (test_opt(sb, BLOCK_VALIDITY)) {
4953 err = ext4_setup_system_zone(sb);
4955 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4957 goto failed_mount4a;
4960 ext4_fc_replay_cleanup(sb);
4963 err = ext4_mb_init(sb);
4965 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4970 block = ext4_count_free_clusters(sb);
4971 ext4_free_blocks_count_set(sbi->s_es,
4972 EXT4_C2B(sbi, block));
4973 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4976 unsigned long freei = ext4_count_free_inodes(sb);
4977 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4978 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4982 err = percpu_counter_init(&sbi->s_dirs_counter,
4983 ext4_count_dirs(sb), GFP_KERNEL);
4985 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4988 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4991 ext4_msg(sb, KERN_ERR, "insufficient memory");
4995 if (ext4_has_feature_flex_bg(sb))
4996 if (!ext4_fill_flex_info(sb)) {
4997 ext4_msg(sb, KERN_ERR,
4998 "unable to initialize "
4999 "flex_bg meta info!");
5003 err = ext4_register_li_request(sb, first_not_zeroed);
5007 err = ext4_register_sysfs(sb);
5012 /* Enable quota usage during mount. */
5013 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5014 err = ext4_enable_quotas(sb);
5018 #endif /* CONFIG_QUOTA */
5021 * Save the original bdev mapping's wb_err value which could be
5022 * used to detect the metadata async write error.
5024 spin_lock_init(&sbi->s_bdev_wb_lock);
5025 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5026 &sbi->s_bdev_wb_err);
5027 sb->s_bdev->bd_super = sb;
5028 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5029 ext4_orphan_cleanup(sb, es);
5030 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5031 if (needs_recovery) {
5032 ext4_msg(sb, KERN_INFO, "recovery complete");
5033 err = ext4_mark_recovery_complete(sb, es);
5037 if (EXT4_SB(sb)->s_journal) {
5038 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5039 descr = " journalled data mode";
5040 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5041 descr = " ordered data mode";
5043 descr = " writeback data mode";
5045 descr = "out journal";
5047 if (test_opt(sb, DISCARD)) {
5048 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5049 if (!blk_queue_discard(q))
5050 ext4_msg(sb, KERN_WARNING,
5051 "mounting with \"discard\" option, but "
5052 "the device does not support discard");
5055 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5056 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5057 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5058 (int) sizeof(sbi->s_es->s_mount_opts),
5059 sbi->s_es->s_mount_opts,
5060 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5061 ext4_quota_mode(sb));
5063 if (es->s_error_count)
5064 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5066 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5067 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5068 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5069 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5070 atomic_set(&sbi->s_warning_count, 0);
5071 atomic_set(&sbi->s_msg_count, 0);
5078 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5082 ext4_unregister_sysfs(sb);
5083 kobject_put(&sbi->s_kobj);
5085 ext4_unregister_li_request(sb);
5087 ext4_mb_release(sb);
5089 flex_groups = rcu_dereference(sbi->s_flex_groups);
5091 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5092 kvfree(flex_groups[i]);
5093 kvfree(flex_groups);
5096 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5097 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5098 percpu_counter_destroy(&sbi->s_dirs_counter);
5099 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5100 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5102 ext4_ext_release(sb);
5103 ext4_release_system_zone(sb);
5108 ext4_msg(sb, KERN_ERR, "mount failed");
5109 if (EXT4_SB(sb)->rsv_conversion_wq)
5110 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5112 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5113 sbi->s_ea_inode_cache = NULL;
5115 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5116 sbi->s_ea_block_cache = NULL;
5118 if (sbi->s_journal) {
5119 jbd2_journal_destroy(sbi->s_journal);
5120 sbi->s_journal = NULL;
5123 ext4_es_unregister_shrinker(sbi);
5125 del_timer_sync(&sbi->s_err_report);
5127 kthread_stop(sbi->s_mmp_tsk);
5130 group_desc = rcu_dereference(sbi->s_group_desc);
5131 for (i = 0; i < db_count; i++)
5132 brelse(group_desc[i]);
5136 if (sbi->s_chksum_driver)
5137 crypto_free_shash(sbi->s_chksum_driver);
5139 #ifdef CONFIG_UNICODE
5140 utf8_unload(sb->s_encoding);
5144 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5145 kfree(get_qf_name(sb, sbi, i));
5147 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5148 ext4_blkdev_remove(sbi);
5151 sb->s_fs_info = NULL;
5152 kfree(sbi->s_blockgroup_lock);
5156 fs_put_dax(dax_dev);
5157 return err ? err : ret;
5161 * Setup any per-fs journal parameters now. We'll do this both on
5162 * initial mount, once the journal has been initialised but before we've
5163 * done any recovery; and again on any subsequent remount.
5165 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5167 struct ext4_sb_info *sbi = EXT4_SB(sb);
5169 journal->j_commit_interval = sbi->s_commit_interval;
5170 journal->j_min_batch_time = sbi->s_min_batch_time;
5171 journal->j_max_batch_time = sbi->s_max_batch_time;
5172 ext4_fc_init(sb, journal);
5174 write_lock(&journal->j_state_lock);
5175 if (test_opt(sb, BARRIER))
5176 journal->j_flags |= JBD2_BARRIER;
5178 journal->j_flags &= ~JBD2_BARRIER;
5179 if (test_opt(sb, DATA_ERR_ABORT))
5180 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5182 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5183 write_unlock(&journal->j_state_lock);
5186 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5187 unsigned int journal_inum)
5189 struct inode *journal_inode;
5192 * Test for the existence of a valid inode on disk. Bad things
5193 * happen if we iget() an unused inode, as the subsequent iput()
5194 * will try to delete it.
5196 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5197 if (IS_ERR(journal_inode)) {
5198 ext4_msg(sb, KERN_ERR, "no journal found");
5201 if (!journal_inode->i_nlink) {
5202 make_bad_inode(journal_inode);
5203 iput(journal_inode);
5204 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5208 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5209 journal_inode, journal_inode->i_size);
5210 if (!S_ISREG(journal_inode->i_mode)) {
5211 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5212 iput(journal_inode);
5215 return journal_inode;
5218 static journal_t *ext4_get_journal(struct super_block *sb,
5219 unsigned int journal_inum)
5221 struct inode *journal_inode;
5224 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5227 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5231 journal = jbd2_journal_init_inode(journal_inode);
5233 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5234 iput(journal_inode);
5237 journal->j_private = sb;
5238 ext4_init_journal_params(sb, journal);
5242 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5245 struct buffer_head *bh;
5249 int hblock, blocksize;
5250 ext4_fsblk_t sb_block;
5251 unsigned long offset;
5252 struct ext4_super_block *es;
5253 struct block_device *bdev;
5255 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5258 bdev = ext4_blkdev_get(j_dev, sb);
5262 blocksize = sb->s_blocksize;
5263 hblock = bdev_logical_block_size(bdev);
5264 if (blocksize < hblock) {
5265 ext4_msg(sb, KERN_ERR,
5266 "blocksize too small for journal device");
5270 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5271 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5272 set_blocksize(bdev, blocksize);
5273 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5274 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5275 "external journal");
5279 es = (struct ext4_super_block *) (bh->b_data + offset);
5280 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5281 !(le32_to_cpu(es->s_feature_incompat) &
5282 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5283 ext4_msg(sb, KERN_ERR, "external journal has "
5289 if ((le32_to_cpu(es->s_feature_ro_compat) &
5290 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5291 es->s_checksum != ext4_superblock_csum(sb, es)) {
5292 ext4_msg(sb, KERN_ERR, "external journal has "
5293 "corrupt superblock");
5298 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5299 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5304 len = ext4_blocks_count(es);
5305 start = sb_block + 1;
5306 brelse(bh); /* we're done with the superblock */
5308 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5309 start, len, blocksize);
5311 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5314 journal->j_private = sb;
5315 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5316 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5319 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5320 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5321 "user (unsupported) - %d",
5322 be32_to_cpu(journal->j_superblock->s_nr_users));
5325 EXT4_SB(sb)->s_journal_bdev = bdev;
5326 ext4_init_journal_params(sb, journal);
5330 jbd2_journal_destroy(journal);
5332 ext4_blkdev_put(bdev);
5336 static int ext4_load_journal(struct super_block *sb,
5337 struct ext4_super_block *es,
5338 unsigned long journal_devnum)
5341 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5344 int really_read_only;
5347 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5348 return -EFSCORRUPTED;
5350 if (journal_devnum &&
5351 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5352 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5353 "numbers have changed");
5354 journal_dev = new_decode_dev(journal_devnum);
5356 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5358 if (journal_inum && journal_dev) {
5359 ext4_msg(sb, KERN_ERR,
5360 "filesystem has both journal inode and journal device!");
5365 journal = ext4_get_journal(sb, journal_inum);
5369 journal = ext4_get_dev_journal(sb, journal_dev);
5374 journal_dev_ro = bdev_read_only(journal->j_dev);
5375 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5377 if (journal_dev_ro && !sb_rdonly(sb)) {
5378 ext4_msg(sb, KERN_ERR,
5379 "journal device read-only, try mounting with '-o ro'");
5385 * Are we loading a blank journal or performing recovery after a
5386 * crash? For recovery, we need to check in advance whether we
5387 * can get read-write access to the device.
5389 if (ext4_has_feature_journal_needs_recovery(sb)) {
5390 if (sb_rdonly(sb)) {
5391 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5392 "required on readonly filesystem");
5393 if (really_read_only) {
5394 ext4_msg(sb, KERN_ERR, "write access "
5395 "unavailable, cannot proceed "
5396 "(try mounting with noload)");
5400 ext4_msg(sb, KERN_INFO, "write access will "
5401 "be enabled during recovery");
5405 if (!(journal->j_flags & JBD2_BARRIER))
5406 ext4_msg(sb, KERN_INFO, "barriers disabled");
5408 if (!ext4_has_feature_journal_needs_recovery(sb))
5409 err = jbd2_journal_wipe(journal, !really_read_only);
5411 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5413 memcpy(save, ((char *) es) +
5414 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5415 err = jbd2_journal_load(journal);
5417 memcpy(((char *) es) + EXT4_S_ERR_START,
5418 save, EXT4_S_ERR_LEN);
5423 ext4_msg(sb, KERN_ERR, "error loading journal");
5427 EXT4_SB(sb)->s_journal = journal;
5428 err = ext4_clear_journal_err(sb, es);
5430 EXT4_SB(sb)->s_journal = NULL;
5431 jbd2_journal_destroy(journal);
5435 if (!really_read_only && journal_devnum &&
5436 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5437 es->s_journal_dev = cpu_to_le32(journal_devnum);
5439 /* Make sure we flush the recovery flag to disk. */
5440 ext4_commit_super(sb, 1);
5446 jbd2_journal_destroy(journal);
5450 static int ext4_commit_super(struct super_block *sb, int sync)
5452 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5453 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5456 if (!sbh || block_device_ejected(sb))
5460 * If the file system is mounted read-only, don't update the
5461 * superblock write time. This avoids updating the superblock
5462 * write time when we are mounting the root file system
5463 * read/only but we need to replay the journal; at that point,
5464 * for people who are east of GMT and who make their clock
5465 * tick in localtime for Windows bug-for-bug compatibility,
5466 * the clock is set in the future, and this will cause e2fsck
5467 * to complain and force a full file system check.
5469 if (!(sb->s_flags & SB_RDONLY))
5470 ext4_update_tstamp(es, s_wtime);
5471 if (sb->s_bdev->bd_part)
5472 es->s_kbytes_written =
5473 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5474 ((part_stat_read(sb->s_bdev->bd_part,
5475 sectors[STAT_WRITE]) -
5476 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5478 es->s_kbytes_written =
5479 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5480 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5481 ext4_free_blocks_count_set(es,
5482 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5483 &EXT4_SB(sb)->s_freeclusters_counter)));
5484 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5485 es->s_free_inodes_count =
5486 cpu_to_le32(percpu_counter_sum_positive(
5487 &EXT4_SB(sb)->s_freeinodes_counter));
5488 BUFFER_TRACE(sbh, "marking dirty");
5489 ext4_superblock_csum_set(sb);
5492 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5494 * Oh, dear. A previous attempt to write the
5495 * superblock failed. This could happen because the
5496 * USB device was yanked out. Or it could happen to
5497 * be a transient write error and maybe the block will
5498 * be remapped. Nothing we can do but to retry the
5499 * write and hope for the best.
5501 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5502 "superblock detected");
5503 clear_buffer_write_io_error(sbh);
5504 set_buffer_uptodate(sbh);
5506 mark_buffer_dirty(sbh);
5509 error = __sync_dirty_buffer(sbh,
5510 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5511 if (buffer_write_io_error(sbh)) {
5512 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5514 clear_buffer_write_io_error(sbh);
5515 set_buffer_uptodate(sbh);
5522 * Have we just finished recovery? If so, and if we are mounting (or
5523 * remounting) the filesystem readonly, then we will end up with a
5524 * consistent fs on disk. Record that fact.
5526 static int ext4_mark_recovery_complete(struct super_block *sb,
5527 struct ext4_super_block *es)
5530 journal_t *journal = EXT4_SB(sb)->s_journal;
5532 if (!ext4_has_feature_journal(sb)) {
5533 if (journal != NULL) {
5534 ext4_error(sb, "Journal got removed while the fs was "
5536 return -EFSCORRUPTED;
5540 jbd2_journal_lock_updates(journal);
5541 err = jbd2_journal_flush(journal);
5545 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5546 ext4_clear_feature_journal_needs_recovery(sb);
5547 ext4_commit_super(sb, 1);
5550 jbd2_journal_unlock_updates(journal);
5555 * If we are mounting (or read-write remounting) a filesystem whose journal
5556 * has recorded an error from a previous lifetime, move that error to the
5557 * main filesystem now.
5559 static int ext4_clear_journal_err(struct super_block *sb,
5560 struct ext4_super_block *es)
5566 if (!ext4_has_feature_journal(sb)) {
5567 ext4_error(sb, "Journal got removed while the fs was mounted!");
5568 return -EFSCORRUPTED;
5571 journal = EXT4_SB(sb)->s_journal;
5574 * Now check for any error status which may have been recorded in the
5575 * journal by a prior ext4_error() or ext4_abort()
5578 j_errno = jbd2_journal_errno(journal);
5582 errstr = ext4_decode_error(sb, j_errno, nbuf);
5583 ext4_warning(sb, "Filesystem error recorded "
5584 "from previous mount: %s", errstr);
5585 ext4_warning(sb, "Marking fs in need of filesystem check.");
5587 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5588 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5589 ext4_commit_super(sb, 1);
5591 jbd2_journal_clear_err(journal);
5592 jbd2_journal_update_sb_errno(journal);
5598 * Force the running and committing transactions to commit,
5599 * and wait on the commit.
5601 int ext4_force_commit(struct super_block *sb)
5608 journal = EXT4_SB(sb)->s_journal;
5609 return ext4_journal_force_commit(journal);
5612 static int ext4_sync_fs(struct super_block *sb, int wait)
5616 bool needs_barrier = false;
5617 struct ext4_sb_info *sbi = EXT4_SB(sb);
5619 if (unlikely(ext4_forced_shutdown(sbi)))
5622 trace_ext4_sync_fs(sb, wait);
5623 flush_workqueue(sbi->rsv_conversion_wq);
5625 * Writeback quota in non-journalled quota case - journalled quota has
5628 dquot_writeback_dquots(sb, -1);
5630 * Data writeback is possible w/o journal transaction, so barrier must
5631 * being sent at the end of the function. But we can skip it if
5632 * transaction_commit will do it for us.
5634 if (sbi->s_journal) {
5635 target = jbd2_get_latest_transaction(sbi->s_journal);
5636 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5637 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5638 needs_barrier = true;
5640 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5642 ret = jbd2_log_wait_commit(sbi->s_journal,
5645 } else if (wait && test_opt(sb, BARRIER))
5646 needs_barrier = true;
5647 if (needs_barrier) {
5649 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5658 * LVM calls this function before a (read-only) snapshot is created. This
5659 * gives us a chance to flush the journal completely and mark the fs clean.
5661 * Note that only this function cannot bring a filesystem to be in a clean
5662 * state independently. It relies on upper layer to stop all data & metadata
5665 static int ext4_freeze(struct super_block *sb)
5673 journal = EXT4_SB(sb)->s_journal;
5676 /* Now we set up the journal barrier. */
5677 jbd2_journal_lock_updates(journal);
5680 * Don't clear the needs_recovery flag if we failed to
5681 * flush the journal.
5683 error = jbd2_journal_flush(journal);
5687 /* Journal blocked and flushed, clear needs_recovery flag. */
5688 ext4_clear_feature_journal_needs_recovery(sb);
5691 error = ext4_commit_super(sb, 1);
5694 /* we rely on upper layer to stop further updates */
5695 jbd2_journal_unlock_updates(journal);
5700 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5701 * flag here, even though the filesystem is not technically dirty yet.
5703 static int ext4_unfreeze(struct super_block *sb)
5705 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5708 if (EXT4_SB(sb)->s_journal) {
5709 /* Reset the needs_recovery flag before the fs is unlocked. */
5710 ext4_set_feature_journal_needs_recovery(sb);
5713 ext4_commit_super(sb, 1);
5718 * Structure to save mount options for ext4_remount's benefit
5720 struct ext4_mount_options {
5721 unsigned long s_mount_opt;
5722 unsigned long s_mount_opt2;
5725 unsigned long s_commit_interval;
5726 u32 s_min_batch_time, s_max_batch_time;
5729 char *s_qf_names[EXT4_MAXQUOTAS];
5733 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5735 struct ext4_super_block *es;
5736 struct ext4_sb_info *sbi = EXT4_SB(sb);
5737 unsigned long old_sb_flags, vfs_flags;
5738 struct ext4_mount_options old_opts;
5739 int enable_quota = 0;
5741 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5745 char *to_free[EXT4_MAXQUOTAS];
5747 char *orig_data = kstrdup(data, GFP_KERNEL);
5749 if (data && !orig_data)
5752 /* Store the original options */
5753 old_sb_flags = sb->s_flags;
5754 old_opts.s_mount_opt = sbi->s_mount_opt;
5755 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5756 old_opts.s_resuid = sbi->s_resuid;
5757 old_opts.s_resgid = sbi->s_resgid;
5758 old_opts.s_commit_interval = sbi->s_commit_interval;
5759 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5760 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5762 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5763 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5764 if (sbi->s_qf_names[i]) {
5765 char *qf_name = get_qf_name(sb, sbi, i);
5767 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5768 if (!old_opts.s_qf_names[i]) {
5769 for (j = 0; j < i; j++)
5770 kfree(old_opts.s_qf_names[j]);
5775 old_opts.s_qf_names[i] = NULL;
5777 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5778 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5781 * Some options can be enabled by ext4 and/or by VFS mount flag
5782 * either way we need to make sure it matches in both *flags and
5783 * s_flags. Copy those selected flags from *flags to s_flags
5785 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5786 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5788 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5793 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5794 test_opt(sb, JOURNAL_CHECKSUM)) {
5795 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5796 "during remount not supported; ignoring");
5797 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5800 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5801 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5802 ext4_msg(sb, KERN_ERR, "can't mount with "
5803 "both data=journal and delalloc");
5807 if (test_opt(sb, DIOREAD_NOLOCK)) {
5808 ext4_msg(sb, KERN_ERR, "can't mount with "
5809 "both data=journal and dioread_nolock");
5813 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5814 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5815 ext4_msg(sb, KERN_ERR, "can't mount with "
5816 "journal_async_commit in data=ordered mode");
5822 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5823 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5828 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5829 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5831 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5832 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5836 if (sbi->s_journal) {
5837 ext4_init_journal_params(sb, sbi->s_journal);
5838 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5841 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5842 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5847 if (*flags & SB_RDONLY) {
5848 err = sync_filesystem(sb);
5851 err = dquot_suspend(sb, -1);
5856 * First of all, the unconditional stuff we have to do
5857 * to disable replay of the journal when we next remount
5859 sb->s_flags |= SB_RDONLY;
5862 * OK, test if we are remounting a valid rw partition
5863 * readonly, and if so set the rdonly flag and then
5864 * mark the partition as valid again.
5866 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5867 (sbi->s_mount_state & EXT4_VALID_FS))
5868 es->s_state = cpu_to_le16(sbi->s_mount_state);
5870 if (sbi->s_journal) {
5872 * We let remount-ro finish even if marking fs
5873 * as clean failed...
5875 ext4_mark_recovery_complete(sb, es);
5878 kthread_stop(sbi->s_mmp_tsk);
5880 /* Make sure we can mount this feature set readwrite */
5881 if (ext4_has_feature_readonly(sb) ||
5882 !ext4_feature_set_ok(sb, 0)) {
5887 * Make sure the group descriptor checksums
5888 * are sane. If they aren't, refuse to remount r/w.
5890 for (g = 0; g < sbi->s_groups_count; g++) {
5891 struct ext4_group_desc *gdp =
5892 ext4_get_group_desc(sb, g, NULL);
5894 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5895 ext4_msg(sb, KERN_ERR,
5896 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5897 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5898 le16_to_cpu(gdp->bg_checksum));
5905 * If we have an unprocessed orphan list hanging
5906 * around from a previously readonly bdev mount,
5907 * require a full umount/remount for now.
5909 if (es->s_last_orphan) {
5910 ext4_msg(sb, KERN_WARNING, "Couldn't "
5911 "remount RDWR because of unprocessed "
5912 "orphan inode list. Please "
5913 "umount/remount instead");
5919 * Mounting a RDONLY partition read-write, so reread
5920 * and store the current valid flag. (It may have
5921 * been changed by e2fsck since we originally mounted
5924 if (sbi->s_journal) {
5925 err = ext4_clear_journal_err(sb, es);
5929 sbi->s_mount_state = le16_to_cpu(es->s_state);
5931 err = ext4_setup_super(sb, es, 0);
5935 sb->s_flags &= ~SB_RDONLY;
5936 if (ext4_has_feature_mmp(sb))
5937 if (ext4_multi_mount_protect(sb,
5938 le64_to_cpu(es->s_mmp_block))) {
5947 * Reinitialize lazy itable initialization thread based on
5950 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5951 ext4_unregister_li_request(sb);
5953 ext4_group_t first_not_zeroed;
5954 first_not_zeroed = ext4_has_uninit_itable(sb);
5955 ext4_register_li_request(sb, first_not_zeroed);
5959 * Handle creation of system zone data early because it can fail.
5960 * Releasing of existing data is done when we are sure remount will
5963 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5964 err = ext4_setup_system_zone(sb);
5969 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5970 err = ext4_commit_super(sb, 1);
5976 /* Release old quota file names */
5977 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5978 kfree(old_opts.s_qf_names[i]);
5980 if (sb_any_quota_suspended(sb))
5981 dquot_resume(sb, -1);
5982 else if (ext4_has_feature_quota(sb)) {
5983 err = ext4_enable_quotas(sb);
5989 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5990 ext4_release_system_zone(sb);
5993 * Some options can be enabled by ext4 and/or by VFS mount flag
5994 * either way we need to make sure it matches in both *flags and
5995 * s_flags. Copy those selected flags from s_flags to *flags
5997 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5999 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6000 orig_data, ext4_quota_mode(sb));
6005 sb->s_flags = old_sb_flags;
6006 sbi->s_mount_opt = old_opts.s_mount_opt;
6007 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6008 sbi->s_resuid = old_opts.s_resuid;
6009 sbi->s_resgid = old_opts.s_resgid;
6010 sbi->s_commit_interval = old_opts.s_commit_interval;
6011 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6012 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6013 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6014 ext4_release_system_zone(sb);
6016 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6017 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6018 to_free[i] = get_qf_name(sb, sbi, i);
6019 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6022 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6030 static int ext4_statfs_project(struct super_block *sb,
6031 kprojid_t projid, struct kstatfs *buf)
6034 struct dquot *dquot;
6038 qid = make_kqid_projid(projid);
6039 dquot = dqget(sb, qid);
6041 return PTR_ERR(dquot);
6042 spin_lock(&dquot->dq_dqb_lock);
6044 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6045 dquot->dq_dqb.dqb_bhardlimit);
6046 limit >>= sb->s_blocksize_bits;
6048 if (limit && buf->f_blocks > limit) {
6049 curblock = (dquot->dq_dqb.dqb_curspace +
6050 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6051 buf->f_blocks = limit;
6052 buf->f_bfree = buf->f_bavail =
6053 (buf->f_blocks > curblock) ?
6054 (buf->f_blocks - curblock) : 0;
6057 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6058 dquot->dq_dqb.dqb_ihardlimit);
6059 if (limit && buf->f_files > limit) {
6060 buf->f_files = limit;
6062 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6063 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6066 spin_unlock(&dquot->dq_dqb_lock);
6072 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6074 struct super_block *sb = dentry->d_sb;
6075 struct ext4_sb_info *sbi = EXT4_SB(sb);
6076 struct ext4_super_block *es = sbi->s_es;
6077 ext4_fsblk_t overhead = 0, resv_blocks;
6080 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6082 if (!test_opt(sb, MINIX_DF))
6083 overhead = sbi->s_overhead;
6085 buf->f_type = EXT4_SUPER_MAGIC;
6086 buf->f_bsize = sb->s_blocksize;
6087 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6088 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6089 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6090 /* prevent underflow in case that few free space is available */
6091 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6092 buf->f_bavail = buf->f_bfree -
6093 (ext4_r_blocks_count(es) + resv_blocks);
6094 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6096 buf->f_files = le32_to_cpu(es->s_inodes_count);
6097 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6098 buf->f_namelen = EXT4_NAME_LEN;
6099 fsid = le64_to_cpup((void *)es->s_uuid) ^
6100 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6101 buf->f_fsid = u64_to_fsid(fsid);
6104 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6105 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6106 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6115 * Helper functions so that transaction is started before we acquire dqio_sem
6116 * to keep correct lock ordering of transaction > dqio_sem
6118 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6120 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6123 static int ext4_write_dquot(struct dquot *dquot)
6127 struct inode *inode;
6129 inode = dquot_to_inode(dquot);
6130 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6131 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6133 return PTR_ERR(handle);
6134 ret = dquot_commit(dquot);
6135 err = ext4_journal_stop(handle);
6141 static int ext4_acquire_dquot(struct dquot *dquot)
6146 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6147 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6149 return PTR_ERR(handle);
6150 ret = dquot_acquire(dquot);
6151 err = ext4_journal_stop(handle);
6157 static int ext4_release_dquot(struct dquot *dquot)
6162 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6163 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6164 if (IS_ERR(handle)) {
6165 /* Release dquot anyway to avoid endless cycle in dqput() */
6166 dquot_release(dquot);
6167 return PTR_ERR(handle);
6169 ret = dquot_release(dquot);
6170 err = ext4_journal_stop(handle);
6176 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6178 struct super_block *sb = dquot->dq_sb;
6180 if (ext4_is_quota_journalled(sb)) {
6181 dquot_mark_dquot_dirty(dquot);
6182 return ext4_write_dquot(dquot);
6184 return dquot_mark_dquot_dirty(dquot);
6188 static int ext4_write_info(struct super_block *sb, int type)
6193 /* Data block + inode block */
6194 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6196 return PTR_ERR(handle);
6197 ret = dquot_commit_info(sb, type);
6198 err = ext4_journal_stop(handle);
6205 * Turn on quotas during mount time - we need to find
6206 * the quota file and such...
6208 static int ext4_quota_on_mount(struct super_block *sb, int type)
6210 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6211 EXT4_SB(sb)->s_jquota_fmt, type);
6214 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6216 struct ext4_inode_info *ei = EXT4_I(inode);
6218 /* The first argument of lockdep_set_subclass has to be
6219 * *exactly* the same as the argument to init_rwsem() --- in
6220 * this case, in init_once() --- or lockdep gets unhappy
6221 * because the name of the lock is set using the
6222 * stringification of the argument to init_rwsem().
6224 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6225 lockdep_set_subclass(&ei->i_data_sem, subclass);
6229 * Standard function to be called on quota_on
6231 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6232 const struct path *path)
6236 if (!test_opt(sb, QUOTA))
6239 /* Quotafile not on the same filesystem? */
6240 if (path->dentry->d_sb != sb)
6243 /* Quota already enabled for this file? */
6244 if (IS_NOQUOTA(d_inode(path->dentry)))
6247 /* Journaling quota? */
6248 if (EXT4_SB(sb)->s_qf_names[type]) {
6249 /* Quotafile not in fs root? */
6250 if (path->dentry->d_parent != sb->s_root)
6251 ext4_msg(sb, KERN_WARNING,
6252 "Quota file not on filesystem root. "
6253 "Journaled quota will not work");
6254 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6257 * Clear the flag just in case mount options changed since
6260 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6264 * When we journal data on quota file, we have to flush journal to see
6265 * all updates to the file when we bypass pagecache...
6267 if (EXT4_SB(sb)->s_journal &&
6268 ext4_should_journal_data(d_inode(path->dentry))) {
6270 * We don't need to lock updates but journal_flush() could
6271 * otherwise be livelocked...
6273 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6274 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6275 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6280 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6281 err = dquot_quota_on(sb, type, format_id, path);
6283 lockdep_set_quota_inode(path->dentry->d_inode,
6286 struct inode *inode = d_inode(path->dentry);
6290 * Set inode flags to prevent userspace from messing with quota
6291 * files. If this fails, we return success anyway since quotas
6292 * are already enabled and this is not a hard failure.
6295 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6298 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6299 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6300 S_NOATIME | S_IMMUTABLE);
6301 err = ext4_mark_inode_dirty(handle, inode);
6302 ext4_journal_stop(handle);
6304 inode_unlock(inode);
6309 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6313 struct inode *qf_inode;
6314 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6315 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6316 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6317 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6320 BUG_ON(!ext4_has_feature_quota(sb));
6322 if (!qf_inums[type])
6325 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6326 if (IS_ERR(qf_inode)) {
6327 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6328 return PTR_ERR(qf_inode);
6331 /* Don't account quota for quota files to avoid recursion */
6332 qf_inode->i_flags |= S_NOQUOTA;
6333 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6334 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6336 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6342 /* Enable usage tracking for all quota types. */
6343 static int ext4_enable_quotas(struct super_block *sb)
6346 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6347 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6348 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6349 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6351 bool quota_mopt[EXT4_MAXQUOTAS] = {
6352 test_opt(sb, USRQUOTA),
6353 test_opt(sb, GRPQUOTA),
6354 test_opt(sb, PRJQUOTA),
6357 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6358 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6359 if (qf_inums[type]) {
6360 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6361 DQUOT_USAGE_ENABLED |
6362 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6365 "Failed to enable quota tracking "
6366 "(type=%d, err=%d). Please run "
6367 "e2fsck to fix.", type, err);
6368 for (type--; type >= 0; type--)
6369 dquot_quota_off(sb, type);
6378 static int ext4_quota_off(struct super_block *sb, int type)
6380 struct inode *inode = sb_dqopt(sb)->files[type];
6384 /* Force all delayed allocation blocks to be allocated.
6385 * Caller already holds s_umount sem */
6386 if (test_opt(sb, DELALLOC))
6387 sync_filesystem(sb);
6389 if (!inode || !igrab(inode))
6392 err = dquot_quota_off(sb, type);
6393 if (err || ext4_has_feature_quota(sb))
6398 * Update modification times of quota files when userspace can
6399 * start looking at them. If we fail, we return success anyway since
6400 * this is not a hard failure and quotas are already disabled.
6402 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6403 if (IS_ERR(handle)) {
6404 err = PTR_ERR(handle);
6407 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6408 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6409 inode->i_mtime = inode->i_ctime = current_time(inode);
6410 err = ext4_mark_inode_dirty(handle, inode);
6411 ext4_journal_stop(handle);
6413 inode_unlock(inode);
6415 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6419 return dquot_quota_off(sb, type);
6422 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6423 * acquiring the locks... As quota files are never truncated and quota code
6424 * itself serializes the operations (and no one else should touch the files)
6425 * we don't have to be afraid of races */
6426 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6427 size_t len, loff_t off)
6429 struct inode *inode = sb_dqopt(sb)->files[type];
6430 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6431 int offset = off & (sb->s_blocksize - 1);
6434 struct buffer_head *bh;
6435 loff_t i_size = i_size_read(inode);
6439 if (off+len > i_size)
6442 while (toread > 0) {
6443 tocopy = sb->s_blocksize - offset < toread ?
6444 sb->s_blocksize - offset : toread;
6445 bh = ext4_bread(NULL, inode, blk, 0);
6448 if (!bh) /* A hole? */
6449 memset(data, 0, tocopy);
6451 memcpy(data, bh->b_data+offset, tocopy);
6461 /* Write to quotafile (we know the transaction is already started and has
6462 * enough credits) */
6463 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6464 const char *data, size_t len, loff_t off)
6466 struct inode *inode = sb_dqopt(sb)->files[type];
6467 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6468 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6470 struct buffer_head *bh;
6471 handle_t *handle = journal_current_handle();
6473 if (EXT4_SB(sb)->s_journal && !handle) {
6474 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6475 " cancelled because transaction is not started",
6476 (unsigned long long)off, (unsigned long long)len);
6480 * Since we account only one data block in transaction credits,
6481 * then it is impossible to cross a block boundary.
6483 if (sb->s_blocksize - offset < len) {
6484 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6485 " cancelled because not block aligned",
6486 (unsigned long long)off, (unsigned long long)len);
6491 bh = ext4_bread(handle, inode, blk,
6492 EXT4_GET_BLOCKS_CREATE |
6493 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6494 } while (PTR_ERR(bh) == -ENOSPC &&
6495 ext4_should_retry_alloc(inode->i_sb, &retries));
6500 BUFFER_TRACE(bh, "get write access");
6501 err = ext4_journal_get_write_access(handle, bh);
6507 memcpy(bh->b_data+offset, data, len);
6508 flush_dcache_page(bh->b_page);
6510 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6513 if (inode->i_size < off + len) {
6514 i_size_write(inode, off + len);
6515 EXT4_I(inode)->i_disksize = inode->i_size;
6516 err2 = ext4_mark_inode_dirty(handle, inode);
6517 if (unlikely(err2 && !err))
6520 return err ? err : len;
6524 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6525 const char *dev_name, void *data)
6527 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6530 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6531 static inline void register_as_ext2(void)
6533 int err = register_filesystem(&ext2_fs_type);
6536 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6539 static inline void unregister_as_ext2(void)
6541 unregister_filesystem(&ext2_fs_type);
6544 static inline int ext2_feature_set_ok(struct super_block *sb)
6546 if (ext4_has_unknown_ext2_incompat_features(sb))
6550 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6555 static inline void register_as_ext2(void) { }
6556 static inline void unregister_as_ext2(void) { }
6557 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6560 static inline void register_as_ext3(void)
6562 int err = register_filesystem(&ext3_fs_type);
6565 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6568 static inline void unregister_as_ext3(void)
6570 unregister_filesystem(&ext3_fs_type);
6573 static inline int ext3_feature_set_ok(struct super_block *sb)
6575 if (ext4_has_unknown_ext3_incompat_features(sb))
6577 if (!ext4_has_feature_journal(sb))
6581 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6586 static struct file_system_type ext4_fs_type = {
6587 .owner = THIS_MODULE,
6589 .mount = ext4_mount,
6590 .kill_sb = kill_block_super,
6591 .fs_flags = FS_REQUIRES_DEV,
6593 MODULE_ALIAS_FS("ext4");
6595 /* Shared across all ext4 file systems */
6596 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6598 static int __init ext4_init_fs(void)
6602 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6603 ext4_li_info = NULL;
6604 mutex_init(&ext4_li_mtx);
6606 /* Build-time check for flags consistency */
6607 ext4_check_flag_values();
6609 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6610 init_waitqueue_head(&ext4__ioend_wq[i]);
6612 err = ext4_init_es();
6616 err = ext4_init_pending();
6620 err = ext4_init_post_read_processing();
6624 err = ext4_init_pageio();
6628 err = ext4_init_system_zone();
6632 err = ext4_init_sysfs();
6636 err = ext4_init_mballoc();
6639 err = init_inodecache();
6643 err = ext4_fc_init_dentry_cache();
6649 err = register_filesystem(&ext4_fs_type);
6655 unregister_as_ext2();
6656 unregister_as_ext3();
6658 destroy_inodecache();
6660 ext4_exit_mballoc();
6664 ext4_exit_system_zone();
6668 ext4_exit_post_read_processing();
6670 ext4_exit_pending();
6677 static void __exit ext4_exit_fs(void)
6679 ext4_destroy_lazyinit_thread();
6680 unregister_as_ext2();
6681 unregister_as_ext3();
6682 unregister_filesystem(&ext4_fs_type);
6683 destroy_inodecache();
6684 ext4_exit_mballoc();
6686 ext4_exit_system_zone();
6688 ext4_exit_post_read_processing();
6690 ext4_exit_pending();
6693 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6694 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6695 MODULE_LICENSE("GPL");
6696 MODULE_SOFTDEP("pre: crc32c");
6697 module_init(ext4_init_fs)
6698 module_exit(ext4_exit_fs)