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 DEFINE_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 void ext4_update_super(struct super_block
*sb
);
69 static int ext4_commit_super(struct super_block
*sb
);
70 static int ext4_mark_recovery_complete(struct super_block
*sb
,
71 struct ext4_super_block
*es
);
72 static int ext4_clear_journal_err(struct super_block
*sb
,
73 struct ext4_super_block
*es
);
74 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
75 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
76 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
77 static int ext4_unfreeze(struct super_block
*sb
);
78 static int ext4_freeze(struct super_block
*sb
);
79 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
80 const char *dev_name
, void *data
);
81 static inline int ext2_feature_set_ok(struct super_block
*sb
);
82 static inline int ext3_feature_set_ok(struct super_block
*sb
);
83 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block
*sb
);
86 static void ext4_clear_request_list(void);
87 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
88 unsigned int journal_inum
);
93 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
94 * i_mmap_rwsem (inode->i_mmap_rwsem)!
97 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
98 * page lock -> i_data_sem (rw)
100 * buffered write path:
101 * sb_start_write -> i_mutex -> mmap_lock
102 * sb_start_write -> i_mutex -> transaction start -> page lock ->
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
107 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
111 * sb_start_write -> i_mutex -> mmap_lock
112 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
115 * transaction start -> page lock(s) -> i_data_sem (rw)
118 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
119 static struct file_system_type ext2_fs_type
= {
120 .owner
= THIS_MODULE
,
123 .kill_sb
= kill_block_super
,
124 .fs_flags
= FS_REQUIRES_DEV
,
126 MODULE_ALIAS_FS("ext2");
127 MODULE_ALIAS("ext2");
128 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
130 #define IS_EXT2_SB(sb) (0)
134 static struct file_system_type ext3_fs_type
= {
135 .owner
= THIS_MODULE
,
138 .kill_sb
= kill_block_super
,
139 .fs_flags
= FS_REQUIRES_DEV
,
141 MODULE_ALIAS_FS("ext3");
142 MODULE_ALIAS("ext3");
143 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
146 static inline void __ext4_read_bh(struct buffer_head
*bh
, int op_flags
,
150 * buffer's verified bit is no longer valid after reading from
151 * disk again due to write out error, clear it to make sure we
152 * recheck the buffer contents.
154 clear_buffer_verified(bh
);
156 bh
->b_end_io
= end_io
? end_io
: end_buffer_read_sync
;
158 submit_bh(REQ_OP_READ
, op_flags
, bh
);
161 void ext4_read_bh_nowait(struct buffer_head
*bh
, int op_flags
,
164 BUG_ON(!buffer_locked(bh
));
166 if (ext4_buffer_uptodate(bh
)) {
170 __ext4_read_bh(bh
, op_flags
, end_io
);
173 int ext4_read_bh(struct buffer_head
*bh
, int op_flags
, bh_end_io_t
*end_io
)
175 BUG_ON(!buffer_locked(bh
));
177 if (ext4_buffer_uptodate(bh
)) {
182 __ext4_read_bh(bh
, op_flags
, end_io
);
185 if (buffer_uptodate(bh
))
190 int ext4_read_bh_lock(struct buffer_head
*bh
, int op_flags
, bool wait
)
192 if (trylock_buffer(bh
)) {
194 return ext4_read_bh(bh
, op_flags
, NULL
);
195 ext4_read_bh_nowait(bh
, op_flags
, NULL
);
200 if (buffer_uptodate(bh
))
208 * This works like __bread_gfp() except it uses ERR_PTR for error
209 * returns. Currently with sb_bread it's impossible to distinguish
210 * between ENOMEM and EIO situations (since both result in a NULL
213 static struct buffer_head
*__ext4_sb_bread_gfp(struct super_block
*sb
,
214 sector_t block
, int op_flags
,
217 struct buffer_head
*bh
;
220 bh
= sb_getblk_gfp(sb
, block
, gfp
);
222 return ERR_PTR(-ENOMEM
);
223 if (ext4_buffer_uptodate(bh
))
226 ret
= ext4_read_bh_lock(bh
, REQ_META
| op_flags
, true);
234 struct buffer_head
*ext4_sb_bread(struct super_block
*sb
, sector_t block
,
237 return __ext4_sb_bread_gfp(sb
, block
, op_flags
, __GFP_MOVABLE
);
240 struct buffer_head
*ext4_sb_bread_unmovable(struct super_block
*sb
,
243 return __ext4_sb_bread_gfp(sb
, block
, 0, 0);
246 void ext4_sb_breadahead_unmovable(struct super_block
*sb
, sector_t block
)
248 struct buffer_head
*bh
= sb_getblk_gfp(sb
, block
, 0);
251 ext4_read_bh_lock(bh
, REQ_RAHEAD
, false);
256 static int ext4_verify_csum_type(struct super_block
*sb
,
257 struct ext4_super_block
*es
)
259 if (!ext4_has_feature_metadata_csum(sb
))
262 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
265 static __le32
ext4_superblock_csum(struct super_block
*sb
,
266 struct ext4_super_block
*es
)
268 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
269 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
272 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
274 return cpu_to_le32(csum
);
277 static int ext4_superblock_csum_verify(struct super_block
*sb
,
278 struct ext4_super_block
*es
)
280 if (!ext4_has_metadata_csum(sb
))
283 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
286 void ext4_superblock_csum_set(struct super_block
*sb
)
288 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
290 if (!ext4_has_metadata_csum(sb
))
293 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
296 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
297 struct ext4_group_desc
*bg
)
299 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
300 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
301 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
304 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
305 struct ext4_group_desc
*bg
)
307 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
308 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
309 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
312 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
313 struct ext4_group_desc
*bg
)
315 return le32_to_cpu(bg
->bg_inode_table_lo
) |
316 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
317 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
320 __u32
ext4_free_group_clusters(struct super_block
*sb
,
321 struct ext4_group_desc
*bg
)
323 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
324 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
325 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
328 __u32
ext4_free_inodes_count(struct super_block
*sb
,
329 struct ext4_group_desc
*bg
)
331 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
332 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
333 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
336 __u32
ext4_used_dirs_count(struct super_block
*sb
,
337 struct ext4_group_desc
*bg
)
339 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
340 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
341 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
344 __u32
ext4_itable_unused_count(struct super_block
*sb
,
345 struct ext4_group_desc
*bg
)
347 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
348 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
349 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
352 void ext4_block_bitmap_set(struct super_block
*sb
,
353 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
355 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
356 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
357 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
360 void ext4_inode_bitmap_set(struct super_block
*sb
,
361 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
363 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
364 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
365 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
368 void ext4_inode_table_set(struct super_block
*sb
,
369 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
371 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
372 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
373 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
376 void ext4_free_group_clusters_set(struct super_block
*sb
,
377 struct ext4_group_desc
*bg
, __u32 count
)
379 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
380 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
381 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
384 void ext4_free_inodes_set(struct super_block
*sb
,
385 struct ext4_group_desc
*bg
, __u32 count
)
387 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
388 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
389 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
392 void ext4_used_dirs_set(struct super_block
*sb
,
393 struct ext4_group_desc
*bg
, __u32 count
)
395 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
396 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
397 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
400 void ext4_itable_unused_set(struct super_block
*sb
,
401 struct ext4_group_desc
*bg
, __u32 count
)
403 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
404 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
405 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
408 static void __ext4_update_tstamp(__le32
*lo
, __u8
*hi
, time64_t now
)
410 now
= clamp_val(now
, 0, (1ull << 40) - 1);
412 *lo
= cpu_to_le32(lower_32_bits(now
));
413 *hi
= upper_32_bits(now
);
416 static time64_t
__ext4_get_tstamp(__le32
*lo
, __u8
*hi
)
418 return ((time64_t
)(*hi
) << 32) + le32_to_cpu(*lo
);
420 #define ext4_update_tstamp(es, tstamp) \
421 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
422 ktime_get_real_seconds())
423 #define ext4_get_tstamp(es, tstamp) \
424 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
427 * The del_gendisk() function uninitializes the disk-specific data
428 * structures, including the bdi structure, without telling anyone
429 * else. Once this happens, any attempt to call mark_buffer_dirty()
430 * (for example, by ext4_commit_super), will cause a kernel OOPS.
431 * This is a kludge to prevent these oops until we can put in a proper
432 * hook in del_gendisk() to inform the VFS and file system layers.
434 static int block_device_ejected(struct super_block
*sb
)
436 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
437 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
439 return bdi
->dev
== NULL
;
442 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
444 struct super_block
*sb
= journal
->j_private
;
445 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
446 int error
= is_journal_aborted(journal
);
447 struct ext4_journal_cb_entry
*jce
;
449 BUG_ON(txn
->t_state
== T_FINISHED
);
451 ext4_process_freed_data(sb
, txn
->t_tid
);
453 spin_lock(&sbi
->s_md_lock
);
454 while (!list_empty(&txn
->t_private_list
)) {
455 jce
= list_entry(txn
->t_private_list
.next
,
456 struct ext4_journal_cb_entry
, jce_list
);
457 list_del_init(&jce
->jce_list
);
458 spin_unlock(&sbi
->s_md_lock
);
459 jce
->jce_func(sb
, jce
, error
);
460 spin_lock(&sbi
->s_md_lock
);
462 spin_unlock(&sbi
->s_md_lock
);
466 * This writepage callback for write_cache_pages()
467 * takes care of a few cases after page cleaning.
469 * write_cache_pages() already checks for dirty pages
470 * and calls clear_page_dirty_for_io(), which we want,
471 * to write protect the pages.
473 * However, we may have to redirty a page (see below.)
475 static int ext4_journalled_writepage_callback(struct page
*page
,
476 struct writeback_control
*wbc
,
479 transaction_t
*transaction
= (transaction_t
*) data
;
480 struct buffer_head
*bh
, *head
;
481 struct journal_head
*jh
;
483 bh
= head
= page_buffers(page
);
486 * We have to redirty a page in these cases:
487 * 1) If buffer is dirty, it means the page was dirty because it
488 * contains a buffer that needs checkpointing. So the dirty bit
489 * needs to be preserved so that checkpointing writes the buffer
491 * 2) If buffer is not part of the committing transaction
492 * (we may have just accidentally come across this buffer because
493 * inode range tracking is not exact) or if the currently running
494 * transaction already contains this buffer as well, dirty bit
495 * needs to be preserved so that the buffer gets writeprotected
496 * properly on running transaction's commit.
499 if (buffer_dirty(bh
) ||
500 (jh
&& (jh
->b_transaction
!= transaction
||
501 jh
->b_next_transaction
))) {
502 redirty_page_for_writepage(wbc
, page
);
505 } while ((bh
= bh
->b_this_page
) != head
);
508 return AOP_WRITEPAGE_ACTIVATE
;
511 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode
*jinode
)
513 struct address_space
*mapping
= jinode
->i_vfs_inode
->i_mapping
;
514 struct writeback_control wbc
= {
515 .sync_mode
= WB_SYNC_ALL
,
516 .nr_to_write
= LONG_MAX
,
517 .range_start
= jinode
->i_dirty_start
,
518 .range_end
= jinode
->i_dirty_end
,
521 return write_cache_pages(mapping
, &wbc
,
522 ext4_journalled_writepage_callback
,
523 jinode
->i_transaction
);
526 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode
*jinode
)
530 if (ext4_should_journal_data(jinode
->i_vfs_inode
))
531 ret
= ext4_journalled_submit_inode_data_buffers(jinode
);
533 ret
= jbd2_journal_submit_inode_data_buffers(jinode
);
538 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode
*jinode
)
542 if (!ext4_should_journal_data(jinode
->i_vfs_inode
))
543 ret
= jbd2_journal_finish_inode_data_buffers(jinode
);
548 static bool system_going_down(void)
550 return system_state
== SYSTEM_HALT
|| system_state
== SYSTEM_POWER_OFF
551 || system_state
== SYSTEM_RESTART
;
554 struct ext4_err_translation
{
559 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
561 static struct ext4_err_translation err_translation
[] = {
562 EXT4_ERR_TRANSLATE(EIO
),
563 EXT4_ERR_TRANSLATE(ENOMEM
),
564 EXT4_ERR_TRANSLATE(EFSBADCRC
),
565 EXT4_ERR_TRANSLATE(EFSCORRUPTED
),
566 EXT4_ERR_TRANSLATE(ENOSPC
),
567 EXT4_ERR_TRANSLATE(ENOKEY
),
568 EXT4_ERR_TRANSLATE(EROFS
),
569 EXT4_ERR_TRANSLATE(EFBIG
),
570 EXT4_ERR_TRANSLATE(EEXIST
),
571 EXT4_ERR_TRANSLATE(ERANGE
),
572 EXT4_ERR_TRANSLATE(EOVERFLOW
),
573 EXT4_ERR_TRANSLATE(EBUSY
),
574 EXT4_ERR_TRANSLATE(ENOTDIR
),
575 EXT4_ERR_TRANSLATE(ENOTEMPTY
),
576 EXT4_ERR_TRANSLATE(ESHUTDOWN
),
577 EXT4_ERR_TRANSLATE(EFAULT
),
580 static int ext4_errno_to_code(int errno
)
584 for (i
= 0; i
< ARRAY_SIZE(err_translation
); i
++)
585 if (err_translation
[i
].errno
== errno
)
586 return err_translation
[i
].code
;
587 return EXT4_ERR_UNKNOWN
;
590 static void save_error_info(struct super_block
*sb
, int error
,
591 __u32 ino
, __u64 block
,
592 const char *func
, unsigned int line
)
594 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
596 /* We default to EFSCORRUPTED error... */
598 error
= EFSCORRUPTED
;
600 spin_lock(&sbi
->s_error_lock
);
601 sbi
->s_add_error_count
++;
602 sbi
->s_last_error_code
= error
;
603 sbi
->s_last_error_line
= line
;
604 sbi
->s_last_error_ino
= ino
;
605 sbi
->s_last_error_block
= block
;
606 sbi
->s_last_error_func
= func
;
607 sbi
->s_last_error_time
= ktime_get_real_seconds();
608 if (!sbi
->s_first_error_time
) {
609 sbi
->s_first_error_code
= error
;
610 sbi
->s_first_error_line
= line
;
611 sbi
->s_first_error_ino
= ino
;
612 sbi
->s_first_error_block
= block
;
613 sbi
->s_first_error_func
= func
;
614 sbi
->s_first_error_time
= sbi
->s_last_error_time
;
616 spin_unlock(&sbi
->s_error_lock
);
619 /* Deal with the reporting of failure conditions on a filesystem such as
620 * inconsistencies detected or read IO failures.
622 * On ext2, we can store the error state of the filesystem in the
623 * superblock. That is not possible on ext4, because we may have other
624 * write ordering constraints on the superblock which prevent us from
625 * writing it out straight away; and given that the journal is about to
626 * be aborted, we can't rely on the current, or future, transactions to
627 * write out the superblock safely.
629 * We'll just use the jbd2_journal_abort() error code to record an error in
630 * the journal instead. On recovery, the journal will complain about
631 * that error until we've noted it down and cleared it.
633 * If force_ro is set, we unconditionally force the filesystem into an
634 * ABORT|READONLY state, unless the error response on the fs has been set to
635 * panic in which case we take the easy way out and panic immediately. This is
636 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
637 * at a critical moment in log management.
639 static void ext4_handle_error(struct super_block
*sb
, bool force_ro
, int error
,
640 __u32 ino
, __u64 block
,
641 const char *func
, unsigned int line
)
643 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
644 bool continue_fs
= !force_ro
&& test_opt(sb
, ERRORS_CONT
);
646 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
647 if (test_opt(sb
, WARN_ON_ERROR
))
650 if (!continue_fs
&& !sb_rdonly(sb
)) {
651 ext4_set_mount_flag(sb
, EXT4_MF_FS_ABORTED
);
653 jbd2_journal_abort(journal
, -EIO
);
656 if (!bdev_read_only(sb
->s_bdev
)) {
657 save_error_info(sb
, error
, ino
, block
, func
, line
);
659 * In case the fs should keep running, we need to writeout
660 * superblock through the journal. Due to lock ordering
661 * constraints, it may not be safe to do it right here so we
662 * defer superblock flushing to a workqueue.
665 schedule_work(&EXT4_SB(sb
)->s_error_work
);
667 ext4_commit_super(sb
);
670 if (sb_rdonly(sb
) || continue_fs
)
674 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
675 * could panic during 'reboot -f' as the underlying device got already
678 if (test_opt(sb
, ERRORS_PANIC
) && !system_going_down()) {
679 panic("EXT4-fs (device %s): panic forced after error\n",
682 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
684 * Make sure updated value of ->s_mount_flags will be visible before
688 sb
->s_flags
|= SB_RDONLY
;
691 static void flush_stashed_error_work(struct work_struct
*work
)
693 struct ext4_sb_info
*sbi
= container_of(work
, struct ext4_sb_info
,
695 journal_t
*journal
= sbi
->s_journal
;
699 * If the journal is still running, we have to write out superblock
700 * through the journal to avoid collisions of other journalled sb
703 * We use directly jbd2 functions here to avoid recursing back into
704 * ext4 error handling code during handling of previous errors.
706 if (!sb_rdonly(sbi
->s_sb
) && journal
) {
707 handle
= jbd2_journal_start(journal
, 1);
710 if (jbd2_journal_get_write_access(handle
, sbi
->s_sbh
)) {
711 jbd2_journal_stop(handle
);
714 ext4_update_super(sbi
->s_sb
);
715 if (jbd2_journal_dirty_metadata(handle
, sbi
->s_sbh
)) {
716 jbd2_journal_stop(handle
);
719 jbd2_journal_stop(handle
);
724 * Write through journal failed. Write sb directly to get error info
725 * out and hope for the best.
727 ext4_commit_super(sbi
->s_sb
);
730 #define ext4_error_ratelimit(sb) \
731 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
734 void __ext4_error(struct super_block
*sb
, const char *function
,
735 unsigned int line
, bool force_ro
, int error
, __u64 block
,
736 const char *fmt
, ...)
738 struct va_format vaf
;
741 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
744 trace_ext4_error(sb
, function
, line
);
745 if (ext4_error_ratelimit(sb
)) {
750 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
751 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
754 ext4_handle_error(sb
, force_ro
, error
, 0, block
, function
, line
);
757 void __ext4_error_inode(struct inode
*inode
, const char *function
,
758 unsigned int line
, ext4_fsblk_t block
, int error
,
759 const char *fmt
, ...)
762 struct va_format vaf
;
764 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
767 trace_ext4_error(inode
->i_sb
, function
, line
);
768 if (ext4_error_ratelimit(inode
->i_sb
)) {
773 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
774 "inode #%lu: block %llu: comm %s: %pV\n",
775 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
776 block
, current
->comm
, &vaf
);
778 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
779 "inode #%lu: comm %s: %pV\n",
780 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
781 current
->comm
, &vaf
);
784 ext4_handle_error(inode
->i_sb
, false, error
, inode
->i_ino
, block
,
788 void __ext4_error_file(struct file
*file
, const char *function
,
789 unsigned int line
, ext4_fsblk_t block
,
790 const char *fmt
, ...)
793 struct va_format vaf
;
794 struct inode
*inode
= file_inode(file
);
795 char pathname
[80], *path
;
797 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
800 trace_ext4_error(inode
->i_sb
, function
, line
);
801 if (ext4_error_ratelimit(inode
->i_sb
)) {
802 path
= file_path(file
, pathname
, sizeof(pathname
));
810 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
811 "block %llu: comm %s: path %s: %pV\n",
812 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
813 block
, current
->comm
, path
, &vaf
);
816 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
817 "comm %s: path %s: %pV\n",
818 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
819 current
->comm
, path
, &vaf
);
822 ext4_handle_error(inode
->i_sb
, false, EFSCORRUPTED
, inode
->i_ino
, block
,
826 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
833 errstr
= "Corrupt filesystem";
836 errstr
= "Filesystem failed CRC";
839 errstr
= "IO failure";
842 errstr
= "Out of memory";
845 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
846 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
847 errstr
= "Journal has aborted";
849 errstr
= "Readonly filesystem";
852 /* If the caller passed in an extra buffer for unknown
853 * errors, textualise them now. Else we just return
856 /* Check for truncated error codes... */
857 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
866 /* __ext4_std_error decodes expected errors from journaling functions
867 * automatically and invokes the appropriate error response. */
869 void __ext4_std_error(struct super_block
*sb
, const char *function
,
870 unsigned int line
, int errno
)
875 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
878 /* Special case: if the error is EROFS, and we're not already
879 * inside a transaction, then there's really no point in logging
881 if (errno
== -EROFS
&& journal_current_handle() == NULL
&& sb_rdonly(sb
))
884 if (ext4_error_ratelimit(sb
)) {
885 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
886 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
887 sb
->s_id
, function
, line
, errstr
);
890 ext4_handle_error(sb
, false, -errno
, 0, 0, function
, line
);
893 void __ext4_msg(struct super_block
*sb
,
894 const char *prefix
, const char *fmt
, ...)
896 struct va_format vaf
;
899 atomic_inc(&EXT4_SB(sb
)->s_msg_count
);
900 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
906 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
910 static int ext4_warning_ratelimit(struct super_block
*sb
)
912 atomic_inc(&EXT4_SB(sb
)->s_warning_count
);
913 return ___ratelimit(&(EXT4_SB(sb
)->s_warning_ratelimit_state
),
917 void __ext4_warning(struct super_block
*sb
, const char *function
,
918 unsigned int line
, const char *fmt
, ...)
920 struct va_format vaf
;
923 if (!ext4_warning_ratelimit(sb
))
929 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
930 sb
->s_id
, function
, line
, &vaf
);
934 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
935 unsigned int line
, const char *fmt
, ...)
937 struct va_format vaf
;
940 if (!ext4_warning_ratelimit(inode
->i_sb
))
946 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
947 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
948 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
952 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
953 struct super_block
*sb
, ext4_group_t grp
,
954 unsigned long ino
, ext4_fsblk_t block
,
955 const char *fmt
, ...)
959 struct va_format vaf
;
962 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
965 trace_ext4_error(sb
, function
, line
);
966 if (ext4_error_ratelimit(sb
)) {
970 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
971 sb
->s_id
, function
, line
, grp
);
973 printk(KERN_CONT
"inode %lu: ", ino
);
975 printk(KERN_CONT
"block %llu:",
976 (unsigned long long) block
);
977 printk(KERN_CONT
"%pV\n", &vaf
);
981 if (test_opt(sb
, ERRORS_CONT
)) {
982 if (test_opt(sb
, WARN_ON_ERROR
))
984 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
985 if (!bdev_read_only(sb
->s_bdev
)) {
986 save_error_info(sb
, EFSCORRUPTED
, ino
, block
, function
,
988 schedule_work(&EXT4_SB(sb
)->s_error_work
);
992 ext4_unlock_group(sb
, grp
);
993 ext4_handle_error(sb
, false, EFSCORRUPTED
, ino
, block
, function
, line
);
995 * We only get here in the ERRORS_RO case; relocking the group
996 * may be dangerous, but nothing bad will happen since the
997 * filesystem will have already been marked read/only and the
998 * journal has been aborted. We return 1 as a hint to callers
999 * who might what to use the return value from
1000 * ext4_grp_locked_error() to distinguish between the
1001 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1002 * aggressively from the ext4 function in question, with a
1003 * more appropriate error code.
1005 ext4_lock_group(sb
, grp
);
1009 void ext4_mark_group_bitmap_corrupted(struct super_block
*sb
,
1013 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1014 struct ext4_group_info
*grp
= ext4_get_group_info(sb
, group
);
1015 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, group
, NULL
);
1018 if (flags
& EXT4_GROUP_INFO_BBITMAP_CORRUPT
) {
1019 ret
= ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT
,
1022 percpu_counter_sub(&sbi
->s_freeclusters_counter
,
1026 if (flags
& EXT4_GROUP_INFO_IBITMAP_CORRUPT
) {
1027 ret
= ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT
,
1032 count
= ext4_free_inodes_count(sb
, gdp
);
1033 percpu_counter_sub(&sbi
->s_freeinodes_counter
,
1039 void ext4_update_dynamic_rev(struct super_block
*sb
)
1041 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
1043 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
1047 "updating to rev %d because of new feature flag, "
1048 "running e2fsck is recommended",
1051 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
1052 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
1053 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
1054 /* leave es->s_feature_*compat flags alone */
1055 /* es->s_uuid will be set by e2fsck if empty */
1058 * The rest of the superblock fields should be zero, and if not it
1059 * means they are likely already in use, so leave them alone. We
1060 * can leave it up to e2fsck to clean up any inconsistencies there.
1065 * Open the external journal device
1067 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
1069 struct block_device
*bdev
;
1071 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
1077 ext4_msg(sb
, KERN_ERR
,
1078 "failed to open journal device unknown-block(%u,%u) %ld",
1079 MAJOR(dev
), MINOR(dev
), PTR_ERR(bdev
));
1084 * Release the journal device
1086 static void ext4_blkdev_put(struct block_device
*bdev
)
1088 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
1091 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
1093 struct block_device
*bdev
;
1094 bdev
= sbi
->s_journal_bdev
;
1096 ext4_blkdev_put(bdev
);
1097 sbi
->s_journal_bdev
= NULL
;
1101 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
1103 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
1106 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
1108 struct list_head
*l
;
1110 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
1111 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
1113 printk(KERN_ERR
"sb_info orphan list:\n");
1114 list_for_each(l
, &sbi
->s_orphan
) {
1115 struct inode
*inode
= orphan_list_entry(l
);
1117 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1118 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
1119 inode
->i_mode
, inode
->i_nlink
,
1120 NEXT_ORPHAN(inode
));
1125 static int ext4_quota_off(struct super_block
*sb
, int type
);
1127 static inline void ext4_quota_off_umount(struct super_block
*sb
)
1131 /* Use our quota_off function to clear inode flags etc. */
1132 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++)
1133 ext4_quota_off(sb
, type
);
1137 * This is a helper function which is used in the mount/remount
1138 * codepaths (which holds s_umount) to fetch the quota file name.
1140 static inline char *get_qf_name(struct super_block
*sb
,
1141 struct ext4_sb_info
*sbi
,
1144 return rcu_dereference_protected(sbi
->s_qf_names
[type
],
1145 lockdep_is_held(&sb
->s_umount
));
1148 static inline void ext4_quota_off_umount(struct super_block
*sb
)
1153 static void ext4_put_super(struct super_block
*sb
)
1155 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1156 struct ext4_super_block
*es
= sbi
->s_es
;
1157 struct buffer_head
**group_desc
;
1158 struct flex_groups
**flex_groups
;
1162 ext4_unregister_li_request(sb
);
1163 ext4_quota_off_umount(sb
);
1165 flush_work(&sbi
->s_error_work
);
1166 destroy_workqueue(sbi
->rsv_conversion_wq
);
1169 * Unregister sysfs before destroying jbd2 journal.
1170 * Since we could still access attr_journal_task attribute via sysfs
1171 * path which could have sbi->s_journal->j_task as NULL
1173 ext4_unregister_sysfs(sb
);
1175 if (sbi
->s_journal
) {
1176 aborted
= is_journal_aborted(sbi
->s_journal
);
1177 err
= jbd2_journal_destroy(sbi
->s_journal
);
1178 sbi
->s_journal
= NULL
;
1179 if ((err
< 0) && !aborted
) {
1180 ext4_abort(sb
, -err
, "Couldn't clean up the journal");
1184 ext4_es_unregister_shrinker(sbi
);
1185 del_timer_sync(&sbi
->s_err_report
);
1186 ext4_release_system_zone(sb
);
1187 ext4_mb_release(sb
);
1188 ext4_ext_release(sb
);
1190 if (!sb_rdonly(sb
) && !aborted
) {
1191 ext4_clear_feature_journal_needs_recovery(sb
);
1192 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
1195 ext4_commit_super(sb
);
1198 group_desc
= rcu_dereference(sbi
->s_group_desc
);
1199 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
1200 brelse(group_desc
[i
]);
1202 flex_groups
= rcu_dereference(sbi
->s_flex_groups
);
1204 for (i
= 0; i
< sbi
->s_flex_groups_allocated
; i
++)
1205 kvfree(flex_groups
[i
]);
1206 kvfree(flex_groups
);
1209 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
1210 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
1211 percpu_counter_destroy(&sbi
->s_dirs_counter
);
1212 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
1213 percpu_counter_destroy(&sbi
->s_sra_exceeded_retry_limit
);
1214 percpu_free_rwsem(&sbi
->s_writepages_rwsem
);
1216 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
1217 kfree(get_qf_name(sb
, sbi
, i
));
1220 /* Debugging code just in case the in-memory inode orphan list
1221 * isn't empty. The on-disk one can be non-empty if we've
1222 * detected an error and taken the fs readonly, but the
1223 * in-memory list had better be clean by this point. */
1224 if (!list_empty(&sbi
->s_orphan
))
1225 dump_orphan_list(sb
, sbi
);
1226 ASSERT(list_empty(&sbi
->s_orphan
));
1228 sync_blockdev(sb
->s_bdev
);
1229 invalidate_bdev(sb
->s_bdev
);
1230 if (sbi
->s_journal_bdev
&& sbi
->s_journal_bdev
!= sb
->s_bdev
) {
1232 * Invalidate the journal device's buffers. We don't want them
1233 * floating about in memory - the physical journal device may
1234 * hotswapped, and it breaks the `ro-after' testing code.
1236 sync_blockdev(sbi
->s_journal_bdev
);
1237 invalidate_bdev(sbi
->s_journal_bdev
);
1238 ext4_blkdev_remove(sbi
);
1241 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
1242 sbi
->s_ea_inode_cache
= NULL
;
1244 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
1245 sbi
->s_ea_block_cache
= NULL
;
1248 kthread_stop(sbi
->s_mmp_tsk
);
1250 sb
->s_fs_info
= NULL
;
1252 * Now that we are completely done shutting down the
1253 * superblock, we need to actually destroy the kobject.
1255 kobject_put(&sbi
->s_kobj
);
1256 wait_for_completion(&sbi
->s_kobj_unregister
);
1257 if (sbi
->s_chksum_driver
)
1258 crypto_free_shash(sbi
->s_chksum_driver
);
1259 kfree(sbi
->s_blockgroup_lock
);
1260 fs_put_dax(sbi
->s_daxdev
);
1261 fscrypt_free_dummy_policy(&sbi
->s_dummy_enc_policy
);
1262 #ifdef CONFIG_UNICODE
1263 utf8_unload(sb
->s_encoding
);
1268 static struct kmem_cache
*ext4_inode_cachep
;
1271 * Called inside transaction, so use GFP_NOFS
1273 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
1275 struct ext4_inode_info
*ei
;
1277 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
1281 inode_set_iversion(&ei
->vfs_inode
, 1);
1282 spin_lock_init(&ei
->i_raw_lock
);
1283 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
1284 atomic_set(&ei
->i_prealloc_active
, 0);
1285 spin_lock_init(&ei
->i_prealloc_lock
);
1286 ext4_es_init_tree(&ei
->i_es_tree
);
1287 rwlock_init(&ei
->i_es_lock
);
1288 INIT_LIST_HEAD(&ei
->i_es_list
);
1289 ei
->i_es_all_nr
= 0;
1290 ei
->i_es_shk_nr
= 0;
1291 ei
->i_es_shrink_lblk
= 0;
1292 ei
->i_reserved_data_blocks
= 0;
1293 spin_lock_init(&(ei
->i_block_reservation_lock
));
1294 ext4_init_pending_tree(&ei
->i_pending_tree
);
1296 ei
->i_reserved_quota
= 0;
1297 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
1300 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
1301 spin_lock_init(&ei
->i_completed_io_lock
);
1303 ei
->i_datasync_tid
= 0;
1304 atomic_set(&ei
->i_unwritten
, 0);
1305 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
1306 ext4_fc_init_inode(&ei
->vfs_inode
);
1307 mutex_init(&ei
->i_fc_lock
);
1308 return &ei
->vfs_inode
;
1311 static int ext4_drop_inode(struct inode
*inode
)
1313 int drop
= generic_drop_inode(inode
);
1316 drop
= fscrypt_drop_inode(inode
);
1318 trace_ext4_drop_inode(inode
, drop
);
1322 static void ext4_free_in_core_inode(struct inode
*inode
)
1324 fscrypt_free_inode(inode
);
1325 if (!list_empty(&(EXT4_I(inode
)->i_fc_list
))) {
1326 pr_warn("%s: inode %ld still in fc list",
1327 __func__
, inode
->i_ino
);
1329 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
1332 static void ext4_destroy_inode(struct inode
*inode
)
1334 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
1335 ext4_msg(inode
->i_sb
, KERN_ERR
,
1336 "Inode %lu (%p): orphan list check failed!",
1337 inode
->i_ino
, EXT4_I(inode
));
1338 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
1339 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
1345 static void init_once(void *foo
)
1347 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
1349 INIT_LIST_HEAD(&ei
->i_orphan
);
1350 init_rwsem(&ei
->xattr_sem
);
1351 init_rwsem(&ei
->i_data_sem
);
1352 init_rwsem(&ei
->i_mmap_sem
);
1353 inode_init_once(&ei
->vfs_inode
);
1354 ext4_fc_init_inode(&ei
->vfs_inode
);
1357 static int __init
init_inodecache(void)
1359 ext4_inode_cachep
= kmem_cache_create_usercopy("ext4_inode_cache",
1360 sizeof(struct ext4_inode_info
), 0,
1361 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
|
1363 offsetof(struct ext4_inode_info
, i_data
),
1364 sizeof_field(struct ext4_inode_info
, i_data
),
1366 if (ext4_inode_cachep
== NULL
)
1371 static void destroy_inodecache(void)
1374 * Make sure all delayed rcu free inodes are flushed before we
1378 kmem_cache_destroy(ext4_inode_cachep
);
1381 void ext4_clear_inode(struct inode
*inode
)
1384 invalidate_inode_buffers(inode
);
1386 ext4_discard_preallocations(inode
, 0);
1387 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1389 if (EXT4_I(inode
)->jinode
) {
1390 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1391 EXT4_I(inode
)->jinode
);
1392 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1393 EXT4_I(inode
)->jinode
= NULL
;
1395 fscrypt_put_encryption_info(inode
);
1396 fsverity_cleanup_inode(inode
);
1399 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1400 u64 ino
, u32 generation
)
1402 struct inode
*inode
;
1405 * Currently we don't know the generation for parent directory, so
1406 * a generation of 0 means "accept any"
1408 inode
= ext4_iget(sb
, ino
, EXT4_IGET_HANDLE
);
1410 return ERR_CAST(inode
);
1411 if (generation
&& inode
->i_generation
!= generation
) {
1413 return ERR_PTR(-ESTALE
);
1419 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1420 int fh_len
, int fh_type
)
1422 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1423 ext4_nfs_get_inode
);
1426 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1427 int fh_len
, int fh_type
)
1429 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1430 ext4_nfs_get_inode
);
1433 static int ext4_nfs_commit_metadata(struct inode
*inode
)
1435 struct writeback_control wbc
= {
1436 .sync_mode
= WB_SYNC_ALL
1439 trace_ext4_nfs_commit_metadata(inode
);
1440 return ext4_write_inode(inode
, &wbc
);
1444 * Try to release metadata pages (indirect blocks, directories) which are
1445 * mapped via the block device. Since these pages could have journal heads
1446 * which would prevent try_to_free_buffers() from freeing them, we must use
1447 * jbd2 layer's try_to_free_buffers() function to release them.
1449 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1452 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1454 WARN_ON(PageChecked(page
));
1455 if (!page_has_buffers(page
))
1458 return jbd2_journal_try_to_free_buffers(journal
, page
);
1460 return try_to_free_buffers(page
);
1463 #ifdef CONFIG_FS_ENCRYPTION
1464 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1466 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1467 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1470 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1473 handle_t
*handle
= fs_data
;
1474 int res
, res2
, credits
, retries
= 0;
1477 * Encrypting the root directory is not allowed because e2fsck expects
1478 * lost+found to exist and be unencrypted, and encrypting the root
1479 * directory would imply encrypting the lost+found directory as well as
1480 * the filename "lost+found" itself.
1482 if (inode
->i_ino
== EXT4_ROOT_INO
)
1485 if (WARN_ON_ONCE(IS_DAX(inode
) && i_size_read(inode
)))
1488 if (ext4_test_inode_flag(inode
, EXT4_INODE_DAX
))
1491 res
= ext4_convert_inline_data(inode
);
1496 * If a journal handle was specified, then the encryption context is
1497 * being set on a new inode via inheritance and is part of a larger
1498 * transaction to create the inode. Otherwise the encryption context is
1499 * being set on an existing inode in its own transaction. Only in the
1500 * latter case should the "retry on ENOSPC" logic be used.
1504 res
= ext4_xattr_set_handle(handle
, inode
,
1505 EXT4_XATTR_INDEX_ENCRYPTION
,
1506 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1509 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1510 ext4_clear_inode_state(inode
,
1511 EXT4_STATE_MAY_INLINE_DATA
);
1513 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1514 * S_DAX may be disabled
1516 ext4_set_inode_flags(inode
, false);
1521 res
= dquot_initialize(inode
);
1525 res
= ext4_xattr_set_credits(inode
, len
, false /* is_create */,
1530 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
, credits
);
1532 return PTR_ERR(handle
);
1534 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1535 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1538 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1540 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1541 * S_DAX may be disabled
1543 ext4_set_inode_flags(inode
, false);
1544 res
= ext4_mark_inode_dirty(handle
, inode
);
1546 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1548 res2
= ext4_journal_stop(handle
);
1550 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1557 static const union fscrypt_policy
*ext4_get_dummy_policy(struct super_block
*sb
)
1559 return EXT4_SB(sb
)->s_dummy_enc_policy
.policy
;
1562 static bool ext4_has_stable_inodes(struct super_block
*sb
)
1564 return ext4_has_feature_stable_inodes(sb
);
1567 static void ext4_get_ino_and_lblk_bits(struct super_block
*sb
,
1568 int *ino_bits_ret
, int *lblk_bits_ret
)
1570 *ino_bits_ret
= 8 * sizeof(EXT4_SB(sb
)->s_es
->s_inodes_count
);
1571 *lblk_bits_ret
= 8 * sizeof(ext4_lblk_t
);
1574 static const struct fscrypt_operations ext4_cryptops
= {
1575 .key_prefix
= "ext4:",
1576 .get_context
= ext4_get_context
,
1577 .set_context
= ext4_set_context
,
1578 .get_dummy_policy
= ext4_get_dummy_policy
,
1579 .empty_dir
= ext4_empty_dir
,
1580 .max_namelen
= EXT4_NAME_LEN
,
1581 .has_stable_inodes
= ext4_has_stable_inodes
,
1582 .get_ino_and_lblk_bits
= ext4_get_ino_and_lblk_bits
,
1587 static const char * const quotatypes
[] = INITQFNAMES
;
1588 #define QTYPE2NAME(t) (quotatypes[t])
1590 static int ext4_write_dquot(struct dquot
*dquot
);
1591 static int ext4_acquire_dquot(struct dquot
*dquot
);
1592 static int ext4_release_dquot(struct dquot
*dquot
);
1593 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1594 static int ext4_write_info(struct super_block
*sb
, int type
);
1595 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1596 const struct path
*path
);
1597 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1598 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1599 size_t len
, loff_t off
);
1600 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1601 const char *data
, size_t len
, loff_t off
);
1602 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1603 unsigned int flags
);
1604 static int ext4_enable_quotas(struct super_block
*sb
);
1606 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1608 return EXT4_I(inode
)->i_dquot
;
1611 static const struct dquot_operations ext4_quota_operations
= {
1612 .get_reserved_space
= ext4_get_reserved_space
,
1613 .write_dquot
= ext4_write_dquot
,
1614 .acquire_dquot
= ext4_acquire_dquot
,
1615 .release_dquot
= ext4_release_dquot
,
1616 .mark_dirty
= ext4_mark_dquot_dirty
,
1617 .write_info
= ext4_write_info
,
1618 .alloc_dquot
= dquot_alloc
,
1619 .destroy_dquot
= dquot_destroy
,
1620 .get_projid
= ext4_get_projid
,
1621 .get_inode_usage
= ext4_get_inode_usage
,
1622 .get_next_id
= dquot_get_next_id
,
1625 static const struct quotactl_ops ext4_qctl_operations
= {
1626 .quota_on
= ext4_quota_on
,
1627 .quota_off
= ext4_quota_off
,
1628 .quota_sync
= dquot_quota_sync
,
1629 .get_state
= dquot_get_state
,
1630 .set_info
= dquot_set_dqinfo
,
1631 .get_dqblk
= dquot_get_dqblk
,
1632 .set_dqblk
= dquot_set_dqblk
,
1633 .get_nextdqblk
= dquot_get_next_dqblk
,
1637 static const struct super_operations ext4_sops
= {
1638 .alloc_inode
= ext4_alloc_inode
,
1639 .free_inode
= ext4_free_in_core_inode
,
1640 .destroy_inode
= ext4_destroy_inode
,
1641 .write_inode
= ext4_write_inode
,
1642 .dirty_inode
= ext4_dirty_inode
,
1643 .drop_inode
= ext4_drop_inode
,
1644 .evict_inode
= ext4_evict_inode
,
1645 .put_super
= ext4_put_super
,
1646 .sync_fs
= ext4_sync_fs
,
1647 .freeze_fs
= ext4_freeze
,
1648 .unfreeze_fs
= ext4_unfreeze
,
1649 .statfs
= ext4_statfs
,
1650 .remount_fs
= ext4_remount
,
1651 .show_options
= ext4_show_options
,
1653 .quota_read
= ext4_quota_read
,
1654 .quota_write
= ext4_quota_write
,
1655 .get_dquots
= ext4_get_dquots
,
1657 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1660 static const struct export_operations ext4_export_ops
= {
1661 .fh_to_dentry
= ext4_fh_to_dentry
,
1662 .fh_to_parent
= ext4_fh_to_parent
,
1663 .get_parent
= ext4_get_parent
,
1664 .commit_metadata
= ext4_nfs_commit_metadata
,
1668 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1669 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1670 Opt_nouid32
, Opt_debug
, Opt_removed
,
1671 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1672 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1673 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1674 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1675 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1676 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1678 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1679 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1680 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1681 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
,
1682 Opt_dax
, Opt_dax_always
, Opt_dax_inode
, Opt_dax_never
,
1683 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_warn_on_error
,
1684 Opt_nowarn_on_error
, Opt_mblk_io_submit
,
1685 Opt_lazytime
, Opt_nolazytime
, Opt_debug_want_extra_isize
,
1686 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1687 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1688 Opt_dioread_nolock
, Opt_dioread_lock
,
1689 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1690 Opt_max_dir_size_kb
, Opt_nojournal_checksum
, Opt_nombcache
,
1691 Opt_prefetch_block_bitmaps
,
1692 #ifdef CONFIG_EXT4_DEBUG
1693 Opt_fc_debug_max_replay
, Opt_fc_debug_force
1697 static const match_table_t tokens
= {
1698 {Opt_bsd_df
, "bsddf"},
1699 {Opt_minix_df
, "minixdf"},
1700 {Opt_grpid
, "grpid"},
1701 {Opt_grpid
, "bsdgroups"},
1702 {Opt_nogrpid
, "nogrpid"},
1703 {Opt_nogrpid
, "sysvgroups"},
1704 {Opt_resgid
, "resgid=%u"},
1705 {Opt_resuid
, "resuid=%u"},
1707 {Opt_err_cont
, "errors=continue"},
1708 {Opt_err_panic
, "errors=panic"},
1709 {Opt_err_ro
, "errors=remount-ro"},
1710 {Opt_nouid32
, "nouid32"},
1711 {Opt_debug
, "debug"},
1712 {Opt_removed
, "oldalloc"},
1713 {Opt_removed
, "orlov"},
1714 {Opt_user_xattr
, "user_xattr"},
1715 {Opt_nouser_xattr
, "nouser_xattr"},
1717 {Opt_noacl
, "noacl"},
1718 {Opt_noload
, "norecovery"},
1719 {Opt_noload
, "noload"},
1720 {Opt_removed
, "nobh"},
1721 {Opt_removed
, "bh"},
1722 {Opt_commit
, "commit=%u"},
1723 {Opt_min_batch_time
, "min_batch_time=%u"},
1724 {Opt_max_batch_time
, "max_batch_time=%u"},
1725 {Opt_journal_dev
, "journal_dev=%u"},
1726 {Opt_journal_path
, "journal_path=%s"},
1727 {Opt_journal_checksum
, "journal_checksum"},
1728 {Opt_nojournal_checksum
, "nojournal_checksum"},
1729 {Opt_journal_async_commit
, "journal_async_commit"},
1730 {Opt_abort
, "abort"},
1731 {Opt_data_journal
, "data=journal"},
1732 {Opt_data_ordered
, "data=ordered"},
1733 {Opt_data_writeback
, "data=writeback"},
1734 {Opt_data_err_abort
, "data_err=abort"},
1735 {Opt_data_err_ignore
, "data_err=ignore"},
1736 {Opt_offusrjquota
, "usrjquota="},
1737 {Opt_usrjquota
, "usrjquota=%s"},
1738 {Opt_offgrpjquota
, "grpjquota="},
1739 {Opt_grpjquota
, "grpjquota=%s"},
1740 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1741 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1742 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1743 {Opt_grpquota
, "grpquota"},
1744 {Opt_noquota
, "noquota"},
1745 {Opt_quota
, "quota"},
1746 {Opt_usrquota
, "usrquota"},
1747 {Opt_prjquota
, "prjquota"},
1748 {Opt_barrier
, "barrier=%u"},
1749 {Opt_barrier
, "barrier"},
1750 {Opt_nobarrier
, "nobarrier"},
1751 {Opt_i_version
, "i_version"},
1753 {Opt_dax_always
, "dax=always"},
1754 {Opt_dax_inode
, "dax=inode"},
1755 {Opt_dax_never
, "dax=never"},
1756 {Opt_stripe
, "stripe=%u"},
1757 {Opt_delalloc
, "delalloc"},
1758 {Opt_warn_on_error
, "warn_on_error"},
1759 {Opt_nowarn_on_error
, "nowarn_on_error"},
1760 {Opt_lazytime
, "lazytime"},
1761 {Opt_nolazytime
, "nolazytime"},
1762 {Opt_debug_want_extra_isize
, "debug_want_extra_isize=%u"},
1763 {Opt_nodelalloc
, "nodelalloc"},
1764 {Opt_removed
, "mblk_io_submit"},
1765 {Opt_removed
, "nomblk_io_submit"},
1766 {Opt_block_validity
, "block_validity"},
1767 {Opt_noblock_validity
, "noblock_validity"},
1768 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1769 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1770 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1771 {Opt_auto_da_alloc
, "auto_da_alloc"},
1772 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1773 {Opt_dioread_nolock
, "dioread_nolock"},
1774 {Opt_dioread_lock
, "nodioread_nolock"},
1775 {Opt_dioread_lock
, "dioread_lock"},
1776 {Opt_discard
, "discard"},
1777 {Opt_nodiscard
, "nodiscard"},
1778 {Opt_init_itable
, "init_itable=%u"},
1779 {Opt_init_itable
, "init_itable"},
1780 {Opt_noinit_itable
, "noinit_itable"},
1781 #ifdef CONFIG_EXT4_DEBUG
1782 {Opt_fc_debug_force
, "fc_debug_force"},
1783 {Opt_fc_debug_max_replay
, "fc_debug_max_replay=%u"},
1785 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1786 {Opt_test_dummy_encryption
, "test_dummy_encryption=%s"},
1787 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1788 {Opt_inlinecrypt
, "inlinecrypt"},
1789 {Opt_nombcache
, "nombcache"},
1790 {Opt_nombcache
, "no_mbcache"}, /* for backward compatibility */
1791 {Opt_prefetch_block_bitmaps
, "prefetch_block_bitmaps"},
1792 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1793 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1794 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1795 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1796 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1800 static ext4_fsblk_t
get_sb_block(void **data
)
1802 ext4_fsblk_t sb_block
;
1803 char *options
= (char *) *data
;
1805 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1806 return 1; /* Default location */
1809 /* TODO: use simple_strtoll with >32bit ext4 */
1810 sb_block
= simple_strtoul(options
, &options
, 0);
1811 if (*options
&& *options
!= ',') {
1812 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1816 if (*options
== ',')
1818 *data
= (void *) options
;
1823 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1824 static const char deprecated_msg
[] =
1825 "Mount option \"%s\" will be removed by %s\n"
1826 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1829 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1831 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1832 char *qname
, *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1835 if (sb_any_quota_loaded(sb
) && !old_qname
) {
1836 ext4_msg(sb
, KERN_ERR
,
1837 "Cannot change journaled "
1838 "quota options when quota turned on");
1841 if (ext4_has_feature_quota(sb
)) {
1842 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1843 "ignored when QUOTA feature is enabled");
1846 qname
= match_strdup(args
);
1848 ext4_msg(sb
, KERN_ERR
,
1849 "Not enough memory for storing quotafile name");
1853 if (strcmp(old_qname
, qname
) == 0)
1856 ext4_msg(sb
, KERN_ERR
,
1857 "%s quota file already specified",
1861 if (strchr(qname
, '/')) {
1862 ext4_msg(sb
, KERN_ERR
,
1863 "quotafile must be on filesystem root");
1866 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], qname
);
1874 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1877 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1878 char *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1880 if (sb_any_quota_loaded(sb
) && old_qname
) {
1881 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1882 " when quota turned on");
1885 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], NULL
);
1892 #define MOPT_SET 0x0001
1893 #define MOPT_CLEAR 0x0002
1894 #define MOPT_NOSUPPORT 0x0004
1895 #define MOPT_EXPLICIT 0x0008
1896 #define MOPT_CLEAR_ERR 0x0010
1897 #define MOPT_GTE0 0x0020
1900 #define MOPT_QFMT 0x0040
1902 #define MOPT_Q MOPT_NOSUPPORT
1903 #define MOPT_QFMT MOPT_NOSUPPORT
1905 #define MOPT_DATAJ 0x0080
1906 #define MOPT_NO_EXT2 0x0100
1907 #define MOPT_NO_EXT3 0x0200
1908 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1909 #define MOPT_STRING 0x0400
1910 #define MOPT_SKIP 0x0800
1911 #define MOPT_2 0x1000
1913 static const struct mount_opts
{
1917 } ext4_mount_opts
[] = {
1918 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1919 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1920 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1921 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1922 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1923 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1924 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1925 MOPT_EXT4_ONLY
| MOPT_SET
},
1926 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1927 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1928 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1929 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1930 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1931 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1932 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1933 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1934 {Opt_warn_on_error
, EXT4_MOUNT_WARN_ON_ERROR
, MOPT_SET
},
1935 {Opt_nowarn_on_error
, EXT4_MOUNT_WARN_ON_ERROR
, MOPT_CLEAR
},
1936 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1937 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1938 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1939 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1940 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1941 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1942 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1943 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1944 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1945 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1946 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1947 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1949 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1951 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1952 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1953 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1954 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1955 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1956 {Opt_commit
, 0, MOPT_GTE0
},
1957 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1958 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1959 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1960 {Opt_init_itable
, 0, MOPT_GTE0
},
1961 {Opt_dax
, EXT4_MOUNT_DAX_ALWAYS
, MOPT_SET
| MOPT_SKIP
},
1962 {Opt_dax_always
, EXT4_MOUNT_DAX_ALWAYS
,
1963 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_SKIP
},
1964 {Opt_dax_inode
, EXT4_MOUNT2_DAX_INODE
,
1965 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_SKIP
},
1966 {Opt_dax_never
, EXT4_MOUNT2_DAX_NEVER
,
1967 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_SKIP
},
1968 {Opt_stripe
, 0, MOPT_GTE0
},
1969 {Opt_resuid
, 0, MOPT_GTE0
},
1970 {Opt_resgid
, 0, MOPT_GTE0
},
1971 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1972 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1973 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1974 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1975 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1976 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1977 MOPT_NO_EXT2
| MOPT_DATAJ
},
1978 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1979 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1980 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1981 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1982 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1984 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1985 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1987 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1988 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1989 {Opt_debug_want_extra_isize
, 0, MOPT_GTE0
},
1990 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1991 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1993 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1995 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1997 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1998 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1999 MOPT_CLEAR
| MOPT_Q
},
2000 {Opt_usrjquota
, 0, MOPT_Q
| MOPT_STRING
},
2001 {Opt_grpjquota
, 0, MOPT_Q
| MOPT_STRING
},
2002 {Opt_offusrjquota
, 0, MOPT_Q
},
2003 {Opt_offgrpjquota
, 0, MOPT_Q
},
2004 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
2005 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
2006 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
2007 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
2008 {Opt_test_dummy_encryption
, 0, MOPT_STRING
},
2009 {Opt_nombcache
, EXT4_MOUNT_NO_MBCACHE
, MOPT_SET
},
2010 {Opt_prefetch_block_bitmaps
, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS
,
2012 #ifdef CONFIG_EXT4_DEBUG
2013 {Opt_fc_debug_force
, EXT4_MOUNT2_JOURNAL_FAST_COMMIT
,
2014 MOPT_SET
| MOPT_2
| MOPT_EXT4_ONLY
},
2015 {Opt_fc_debug_max_replay
, 0, MOPT_GTE0
},
2020 #ifdef CONFIG_UNICODE
2021 static const struct ext4_sb_encodings
{
2025 } ext4_sb_encoding_map
[] = {
2026 {EXT4_ENC_UTF8_12_1
, "utf8", "12.1.0"},
2029 static int ext4_sb_read_encoding(const struct ext4_super_block
*es
,
2030 const struct ext4_sb_encodings
**encoding
,
2033 __u16 magic
= le16_to_cpu(es
->s_encoding
);
2036 for (i
= 0; i
< ARRAY_SIZE(ext4_sb_encoding_map
); i
++)
2037 if (magic
== ext4_sb_encoding_map
[i
].magic
)
2040 if (i
>= ARRAY_SIZE(ext4_sb_encoding_map
))
2043 *encoding
= &ext4_sb_encoding_map
[i
];
2044 *flags
= le16_to_cpu(es
->s_encoding_flags
);
2050 static int ext4_set_test_dummy_encryption(struct super_block
*sb
,
2052 const substring_t
*arg
,
2055 #ifdef CONFIG_FS_ENCRYPTION
2056 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2060 * This mount option is just for testing, and it's not worthwhile to
2061 * implement the extra complexity (e.g. RCU protection) that would be
2062 * needed to allow it to be set or changed during remount. We do allow
2063 * it to be specified during remount, but only if there is no change.
2065 if (is_remount
&& !sbi
->s_dummy_enc_policy
.policy
) {
2066 ext4_msg(sb
, KERN_WARNING
,
2067 "Can't set test_dummy_encryption on remount");
2070 err
= fscrypt_set_test_dummy_encryption(sb
, arg
->from
,
2071 &sbi
->s_dummy_enc_policy
);
2074 ext4_msg(sb
, KERN_WARNING
,
2075 "Can't change test_dummy_encryption on remount");
2076 else if (err
== -EINVAL
)
2077 ext4_msg(sb
, KERN_WARNING
,
2078 "Value of option \"%s\" is unrecognized", opt
);
2080 ext4_msg(sb
, KERN_WARNING
,
2081 "Error processing option \"%s\" [%d]",
2085 ext4_msg(sb
, KERN_WARNING
, "Test dummy encryption mode enabled");
2087 ext4_msg(sb
, KERN_WARNING
,
2088 "Test dummy encryption mount option ignored");
2093 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
2094 substring_t
*args
, unsigned long *journal_devnum
,
2095 unsigned int *journal_ioprio
, int is_remount
)
2097 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2098 const struct mount_opts
*m
;
2104 if (token
== Opt_usrjquota
)
2105 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
2106 else if (token
== Opt_grpjquota
)
2107 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
2108 else if (token
== Opt_offusrjquota
)
2109 return clear_qf_name(sb
, USRQUOTA
);
2110 else if (token
== Opt_offgrpjquota
)
2111 return clear_qf_name(sb
, GRPQUOTA
);
2115 case Opt_nouser_xattr
:
2116 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
2119 return 1; /* handled by get_sb_block() */
2121 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
2124 ext4_set_mount_flag(sb
, EXT4_MF_FS_ABORTED
);
2127 sb
->s_flags
|= SB_I_VERSION
;
2130 sb
->s_flags
|= SB_LAZYTIME
;
2132 case Opt_nolazytime
:
2133 sb
->s_flags
&= ~SB_LAZYTIME
;
2135 case Opt_inlinecrypt
:
2136 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2137 sb
->s_flags
|= SB_INLINECRYPT
;
2139 ext4_msg(sb
, KERN_ERR
, "inline encryption not supported");
2144 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
2145 if (token
== m
->token
)
2148 if (m
->token
== Opt_err
) {
2149 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
2150 "or missing value", opt
);
2154 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
2155 ext4_msg(sb
, KERN_ERR
,
2156 "Mount option \"%s\" incompatible with ext2", opt
);
2159 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
2160 ext4_msg(sb
, KERN_ERR
,
2161 "Mount option \"%s\" incompatible with ext3", opt
);
2165 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
2167 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
2169 if (m
->flags
& MOPT_EXPLICIT
) {
2170 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
2171 set_opt2(sb
, EXPLICIT_DELALLOC
);
2172 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
2173 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
2177 if (m
->flags
& MOPT_CLEAR_ERR
)
2178 clear_opt(sb
, ERRORS_MASK
);
2179 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
2180 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
2181 "options when quota turned on");
2185 if (m
->flags
& MOPT_NOSUPPORT
) {
2186 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
2187 } else if (token
== Opt_commit
) {
2189 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
2190 else if (arg
> INT_MAX
/ HZ
) {
2191 ext4_msg(sb
, KERN_ERR
,
2192 "Invalid commit interval %d, "
2193 "must be smaller than %d",
2197 sbi
->s_commit_interval
= HZ
* arg
;
2198 } else if (token
== Opt_debug_want_extra_isize
) {
2201 (arg
> (sbi
->s_inode_size
- EXT4_GOOD_OLD_INODE_SIZE
))) {
2202 ext4_msg(sb
, KERN_ERR
,
2203 "Invalid want_extra_isize %d", arg
);
2206 sbi
->s_want_extra_isize
= arg
;
2207 } else if (token
== Opt_max_batch_time
) {
2208 sbi
->s_max_batch_time
= arg
;
2209 } else if (token
== Opt_min_batch_time
) {
2210 sbi
->s_min_batch_time
= arg
;
2211 } else if (token
== Opt_inode_readahead_blks
) {
2212 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
2213 ext4_msg(sb
, KERN_ERR
,
2214 "EXT4-fs: inode_readahead_blks must be "
2215 "0 or a power of 2 smaller than 2^31");
2218 sbi
->s_inode_readahead_blks
= arg
;
2219 } else if (token
== Opt_init_itable
) {
2220 set_opt(sb
, INIT_INODE_TABLE
);
2222 arg
= EXT4_DEF_LI_WAIT_MULT
;
2223 sbi
->s_li_wait_mult
= arg
;
2224 } else if (token
== Opt_max_dir_size_kb
) {
2225 sbi
->s_max_dir_size_kb
= arg
;
2226 #ifdef CONFIG_EXT4_DEBUG
2227 } else if (token
== Opt_fc_debug_max_replay
) {
2228 sbi
->s_fc_debug_max_replay
= arg
;
2230 } else if (token
== Opt_stripe
) {
2231 sbi
->s_stripe
= arg
;
2232 } else if (token
== Opt_resuid
) {
2233 uid
= make_kuid(current_user_ns(), arg
);
2234 if (!uid_valid(uid
)) {
2235 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
2238 sbi
->s_resuid
= uid
;
2239 } else if (token
== Opt_resgid
) {
2240 gid
= make_kgid(current_user_ns(), arg
);
2241 if (!gid_valid(gid
)) {
2242 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
2245 sbi
->s_resgid
= gid
;
2246 } else if (token
== Opt_journal_dev
) {
2248 ext4_msg(sb
, KERN_ERR
,
2249 "Cannot specify journal on remount");
2252 *journal_devnum
= arg
;
2253 } else if (token
== Opt_journal_path
) {
2255 struct inode
*journal_inode
;
2260 ext4_msg(sb
, KERN_ERR
,
2261 "Cannot specify journal on remount");
2264 journal_path
= match_strdup(&args
[0]);
2265 if (!journal_path
) {
2266 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
2267 "journal device string");
2271 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
2273 ext4_msg(sb
, KERN_ERR
, "error: could not find "
2274 "journal device path: error %d", error
);
2275 kfree(journal_path
);
2279 journal_inode
= d_inode(path
.dentry
);
2280 if (!S_ISBLK(journal_inode
->i_mode
)) {
2281 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
2282 "is not a block device", journal_path
);
2284 kfree(journal_path
);
2288 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
2290 kfree(journal_path
);
2291 } else if (token
== Opt_journal_ioprio
) {
2293 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
2298 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
2299 } else if (token
== Opt_test_dummy_encryption
) {
2300 return ext4_set_test_dummy_encryption(sb
, opt
, &args
[0],
2302 } else if (m
->flags
& MOPT_DATAJ
) {
2304 if (!sbi
->s_journal
)
2305 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
2306 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
2307 ext4_msg(sb
, KERN_ERR
,
2308 "Cannot change data mode on remount");
2312 clear_opt(sb
, DATA_FLAGS
);
2313 sbi
->s_mount_opt
|= m
->mount_opt
;
2316 } else if (m
->flags
& MOPT_QFMT
) {
2317 if (sb_any_quota_loaded(sb
) &&
2318 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
2319 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
2320 "quota options when quota turned on");
2323 if (ext4_has_feature_quota(sb
)) {
2324 ext4_msg(sb
, KERN_INFO
,
2325 "Quota format mount options ignored "
2326 "when QUOTA feature is enabled");
2329 sbi
->s_jquota_fmt
= m
->mount_opt
;
2331 } else if (token
== Opt_dax
|| token
== Opt_dax_always
||
2332 token
== Opt_dax_inode
|| token
== Opt_dax_never
) {
2333 #ifdef CONFIG_FS_DAX
2336 case Opt_dax_always
:
2338 (!(sbi
->s_mount_opt
& EXT4_MOUNT_DAX_ALWAYS
) ||
2339 (sbi
->s_mount_opt2
& EXT4_MOUNT2_DAX_NEVER
))) {
2340 fail_dax_change_remount
:
2341 ext4_msg(sb
, KERN_ERR
, "can't change "
2342 "dax mount option while remounting");
2346 (test_opt(sb
, DATA_FLAGS
) ==
2347 EXT4_MOUNT_JOURNAL_DATA
)) {
2348 ext4_msg(sb
, KERN_ERR
, "can't mount with "
2349 "both data=journal and dax");
2352 ext4_msg(sb
, KERN_WARNING
,
2353 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2354 sbi
->s_mount_opt
|= EXT4_MOUNT_DAX_ALWAYS
;
2355 sbi
->s_mount_opt2
&= ~EXT4_MOUNT2_DAX_NEVER
;
2359 (!(sbi
->s_mount_opt2
& EXT4_MOUNT2_DAX_NEVER
) ||
2360 (sbi
->s_mount_opt
& EXT4_MOUNT_DAX_ALWAYS
)))
2361 goto fail_dax_change_remount
;
2362 sbi
->s_mount_opt2
|= EXT4_MOUNT2_DAX_NEVER
;
2363 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DAX_ALWAYS
;
2367 ((sbi
->s_mount_opt
& EXT4_MOUNT_DAX_ALWAYS
) ||
2368 (sbi
->s_mount_opt2
& EXT4_MOUNT2_DAX_NEVER
) ||
2369 !(sbi
->s_mount_opt2
& EXT4_MOUNT2_DAX_INODE
)))
2370 goto fail_dax_change_remount
;
2371 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DAX_ALWAYS
;
2372 sbi
->s_mount_opt2
&= ~EXT4_MOUNT2_DAX_NEVER
;
2373 /* Strictly for printing options */
2374 sbi
->s_mount_opt2
|= EXT4_MOUNT2_DAX_INODE
;
2378 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
2379 sbi
->s_mount_opt2
|= EXT4_MOUNT2_DAX_NEVER
;
2380 sbi
->s_mount_opt
&= ~EXT4_MOUNT_DAX_ALWAYS
;
2383 } else if (token
== Opt_data_err_abort
) {
2384 sbi
->s_mount_opt
|= m
->mount_opt
;
2385 } else if (token
== Opt_data_err_ignore
) {
2386 sbi
->s_mount_opt
&= ~m
->mount_opt
;
2390 if (m
->flags
& MOPT_CLEAR
)
2392 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
2393 ext4_msg(sb
, KERN_WARNING
,
2394 "buggy handling of option %s", opt
);
2398 if (m
->flags
& MOPT_2
) {
2400 sbi
->s_mount_opt2
|= m
->mount_opt
;
2402 sbi
->s_mount_opt2
&= ~m
->mount_opt
;
2405 sbi
->s_mount_opt
|= m
->mount_opt
;
2407 sbi
->s_mount_opt
&= ~m
->mount_opt
;
2413 static int parse_options(char *options
, struct super_block
*sb
,
2414 unsigned long *journal_devnum
,
2415 unsigned int *journal_ioprio
,
2418 struct ext4_sb_info __maybe_unused
*sbi
= EXT4_SB(sb
);
2419 char *p
, __maybe_unused
*usr_qf_name
, __maybe_unused
*grp_qf_name
;
2420 substring_t args
[MAX_OPT_ARGS
];
2426 while ((p
= strsep(&options
, ",")) != NULL
) {
2430 * Initialize args struct so we know whether arg was
2431 * found; some options take optional arguments.
2433 args
[0].to
= args
[0].from
= NULL
;
2434 token
= match_token(p
, tokens
, args
);
2435 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
2436 journal_ioprio
, is_remount
) < 0)
2441 * We do the test below only for project quotas. 'usrquota' and
2442 * 'grpquota' mount options are allowed even without quota feature
2443 * to support legacy quotas in quota files.
2445 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
2446 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
2447 "Cannot enable project quota enforcement.");
2450 usr_qf_name
= get_qf_name(sb
, sbi
, USRQUOTA
);
2451 grp_qf_name
= get_qf_name(sb
, sbi
, GRPQUOTA
);
2452 if (usr_qf_name
|| grp_qf_name
) {
2453 if (test_opt(sb
, USRQUOTA
) && usr_qf_name
)
2454 clear_opt(sb
, USRQUOTA
);
2456 if (test_opt(sb
, GRPQUOTA
) && grp_qf_name
)
2457 clear_opt(sb
, GRPQUOTA
);
2459 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
2460 ext4_msg(sb
, KERN_ERR
, "old and new quota "
2465 if (!sbi
->s_jquota_fmt
) {
2466 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
2472 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
2474 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
2475 if (blocksize
< PAGE_SIZE
)
2476 ext4_msg(sb
, KERN_WARNING
, "Warning: mounting with an "
2477 "experimental mount option 'dioread_nolock' "
2478 "for blocksize < PAGE_SIZE");
2483 static inline void ext4_show_quota_options(struct seq_file
*seq
,
2484 struct super_block
*sb
)
2486 #if defined(CONFIG_QUOTA)
2487 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2488 char *usr_qf_name
, *grp_qf_name
;
2490 if (sbi
->s_jquota_fmt
) {
2493 switch (sbi
->s_jquota_fmt
) {
2504 seq_printf(seq
, ",jqfmt=%s", fmtname
);
2508 usr_qf_name
= rcu_dereference(sbi
->s_qf_names
[USRQUOTA
]);
2509 grp_qf_name
= rcu_dereference(sbi
->s_qf_names
[GRPQUOTA
]);
2511 seq_show_option(seq
, "usrjquota", usr_qf_name
);
2513 seq_show_option(seq
, "grpjquota", grp_qf_name
);
2518 static const char *token2str(int token
)
2520 const struct match_token
*t
;
2522 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
2523 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
2530 * - it's set to a non-default value OR
2531 * - if the per-sb default is different from the global default
2533 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
2536 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2537 struct ext4_super_block
*es
= sbi
->s_es
;
2538 int def_errors
, def_mount_opt
= sbi
->s_def_mount_opt
;
2539 const struct mount_opts
*m
;
2540 char sep
= nodefs
? '\n' : ',';
2542 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2543 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2545 if (sbi
->s_sb_block
!= 1)
2546 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
2548 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
2549 int want_set
= m
->flags
& MOPT_SET
;
2550 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
2551 (m
->flags
& MOPT_CLEAR_ERR
) || m
->flags
& MOPT_SKIP
)
2553 if (!nodefs
&& !(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2554 continue; /* skip if same as the default */
2556 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2557 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2558 continue; /* select Opt_noFoo vs Opt_Foo */
2559 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2562 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
2563 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2564 SEQ_OPTS_PRINT("resuid=%u",
2565 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
2566 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
2567 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2568 SEQ_OPTS_PRINT("resgid=%u",
2569 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
2570 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2571 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2572 SEQ_OPTS_PUTS("errors=remount-ro");
2573 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2574 SEQ_OPTS_PUTS("errors=continue");
2575 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2576 SEQ_OPTS_PUTS("errors=panic");
2577 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2578 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2579 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2580 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2581 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2582 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2583 if (sb
->s_flags
& SB_I_VERSION
)
2584 SEQ_OPTS_PUTS("i_version");
2585 if (nodefs
|| sbi
->s_stripe
)
2586 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2587 if (nodefs
|| EXT4_MOUNT_DATA_FLAGS
&
2588 (sbi
->s_mount_opt
^ def_mount_opt
)) {
2589 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2590 SEQ_OPTS_PUTS("data=journal");
2591 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2592 SEQ_OPTS_PUTS("data=ordered");
2593 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2594 SEQ_OPTS_PUTS("data=writeback");
2597 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2598 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2599 sbi
->s_inode_readahead_blks
);
2601 if (test_opt(sb
, INIT_INODE_TABLE
) && (nodefs
||
2602 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2603 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2604 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2605 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2606 if (test_opt(sb
, DATA_ERR_ABORT
))
2607 SEQ_OPTS_PUTS("data_err=abort");
2609 fscrypt_show_test_dummy_encryption(seq
, sep
, sb
);
2611 if (sb
->s_flags
& SB_INLINECRYPT
)
2612 SEQ_OPTS_PUTS("inlinecrypt");
2614 if (test_opt(sb
, DAX_ALWAYS
)) {
2616 SEQ_OPTS_PUTS("dax");
2618 SEQ_OPTS_PUTS("dax=always");
2619 } else if (test_opt2(sb
, DAX_NEVER
)) {
2620 SEQ_OPTS_PUTS("dax=never");
2621 } else if (test_opt2(sb
, DAX_INODE
)) {
2622 SEQ_OPTS_PUTS("dax=inode");
2624 ext4_show_quota_options(seq
, sb
);
2628 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2630 return _ext4_show_options(seq
, root
->d_sb
, 0);
2633 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2635 struct super_block
*sb
= seq
->private;
2638 seq_puts(seq
, sb_rdonly(sb
) ? "ro" : "rw");
2639 rc
= _ext4_show_options(seq
, sb
, 1);
2640 seq_puts(seq
, "\n");
2644 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2647 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2650 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2651 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2652 "forcing read-only mode");
2658 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2659 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2660 "running e2fsck is recommended");
2661 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2662 ext4_msg(sb
, KERN_WARNING
,
2663 "warning: mounting fs with errors, "
2664 "running e2fsck is recommended");
2665 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2666 le16_to_cpu(es
->s_mnt_count
) >=
2667 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2668 ext4_msg(sb
, KERN_WARNING
,
2669 "warning: maximal mount count reached, "
2670 "running e2fsck is recommended");
2671 else if (le32_to_cpu(es
->s_checkinterval
) &&
2672 (ext4_get_tstamp(es
, s_lastcheck
) +
2673 le32_to_cpu(es
->s_checkinterval
) <= ktime_get_real_seconds()))
2674 ext4_msg(sb
, KERN_WARNING
,
2675 "warning: checktime reached, "
2676 "running e2fsck is recommended");
2677 if (!sbi
->s_journal
)
2678 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2679 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2680 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2681 le16_add_cpu(&es
->s_mnt_count
, 1);
2682 ext4_update_tstamp(es
, s_mtime
);
2684 ext4_set_feature_journal_needs_recovery(sb
);
2686 err
= ext4_commit_super(sb
);
2688 if (test_opt(sb
, DEBUG
))
2689 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2690 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2692 sbi
->s_groups_count
,
2693 EXT4_BLOCKS_PER_GROUP(sb
),
2694 EXT4_INODES_PER_GROUP(sb
),
2695 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2697 cleancache_init_fs(sb
);
2701 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2703 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2704 struct flex_groups
**old_groups
, **new_groups
;
2707 if (!sbi
->s_log_groups_per_flex
)
2710 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2711 if (size
<= sbi
->s_flex_groups_allocated
)
2714 new_groups
= kvzalloc(roundup_pow_of_two(size
*
2715 sizeof(*sbi
->s_flex_groups
)), GFP_KERNEL
);
2717 ext4_msg(sb
, KERN_ERR
,
2718 "not enough memory for %d flex group pointers", size
);
2721 for (i
= sbi
->s_flex_groups_allocated
; i
< size
; i
++) {
2722 new_groups
[i
] = kvzalloc(roundup_pow_of_two(
2723 sizeof(struct flex_groups
)),
2725 if (!new_groups
[i
]) {
2726 for (j
= sbi
->s_flex_groups_allocated
; j
< i
; j
++)
2727 kvfree(new_groups
[j
]);
2729 ext4_msg(sb
, KERN_ERR
,
2730 "not enough memory for %d flex groups", size
);
2735 old_groups
= rcu_dereference(sbi
->s_flex_groups
);
2737 memcpy(new_groups
, old_groups
,
2738 (sbi
->s_flex_groups_allocated
*
2739 sizeof(struct flex_groups
*)));
2741 rcu_assign_pointer(sbi
->s_flex_groups
, new_groups
);
2742 sbi
->s_flex_groups_allocated
= size
;
2744 ext4_kvfree_array_rcu(old_groups
);
2748 static int ext4_fill_flex_info(struct super_block
*sb
)
2750 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2751 struct ext4_group_desc
*gdp
= NULL
;
2752 struct flex_groups
*fg
;
2753 ext4_group_t flex_group
;
2756 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2757 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2758 sbi
->s_log_groups_per_flex
= 0;
2762 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2766 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2767 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2769 flex_group
= ext4_flex_group(sbi
, i
);
2770 fg
= sbi_array_rcu_deref(sbi
, s_flex_groups
, flex_group
);
2771 atomic_add(ext4_free_inodes_count(sb
, gdp
), &fg
->free_inodes
);
2772 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2773 &fg
->free_clusters
);
2774 atomic_add(ext4_used_dirs_count(sb
, gdp
), &fg
->used_dirs
);
2782 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2783 struct ext4_group_desc
*gdp
)
2785 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2787 __le32 le_group
= cpu_to_le32(block_group
);
2788 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2790 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2791 /* Use new metadata_csum algorithm */
2793 __u16 dummy_csum
= 0;
2795 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2797 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2798 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2799 sizeof(dummy_csum
));
2800 offset
+= sizeof(dummy_csum
);
2801 if (offset
< sbi
->s_desc_size
)
2802 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2803 sbi
->s_desc_size
- offset
);
2805 crc
= csum32
& 0xFFFF;
2809 /* old crc16 code */
2810 if (!ext4_has_feature_gdt_csum(sb
))
2813 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2814 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2815 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2816 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2817 /* for checksum of struct ext4_group_desc do the rest...*/
2818 if (ext4_has_feature_64bit(sb
) &&
2819 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2820 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2821 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2825 return cpu_to_le16(crc
);
2828 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2829 struct ext4_group_desc
*gdp
)
2831 if (ext4_has_group_desc_csum(sb
) &&
2832 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2838 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2839 struct ext4_group_desc
*gdp
)
2841 if (!ext4_has_group_desc_csum(sb
))
2843 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2846 /* Called at mount-time, super-block is locked */
2847 static int ext4_check_descriptors(struct super_block
*sb
,
2848 ext4_fsblk_t sb_block
,
2849 ext4_group_t
*first_not_zeroed
)
2851 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2852 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2853 ext4_fsblk_t last_block
;
2854 ext4_fsblk_t last_bg_block
= sb_block
+ ext4_bg_num_gdb(sb
, 0);
2855 ext4_fsblk_t block_bitmap
;
2856 ext4_fsblk_t inode_bitmap
;
2857 ext4_fsblk_t inode_table
;
2858 int flexbg_flag
= 0;
2859 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2861 if (ext4_has_feature_flex_bg(sb
))
2864 ext4_debug("Checking group descriptors");
2866 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2867 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2869 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2870 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2872 last_block
= first_block
+
2873 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2875 if ((grp
== sbi
->s_groups_count
) &&
2876 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2879 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2880 if (block_bitmap
== sb_block
) {
2881 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2882 "Block bitmap for group %u overlaps "
2887 if (block_bitmap
>= sb_block
+ 1 &&
2888 block_bitmap
<= last_bg_block
) {
2889 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2890 "Block bitmap for group %u overlaps "
2891 "block group descriptors", i
);
2895 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2896 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2897 "Block bitmap for group %u not in group "
2898 "(block %llu)!", i
, block_bitmap
);
2901 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2902 if (inode_bitmap
== sb_block
) {
2903 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2904 "Inode bitmap for group %u overlaps "
2909 if (inode_bitmap
>= sb_block
+ 1 &&
2910 inode_bitmap
<= last_bg_block
) {
2911 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2912 "Inode bitmap for group %u overlaps "
2913 "block group descriptors", i
);
2917 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2918 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2919 "Inode bitmap for group %u not in group "
2920 "(block %llu)!", i
, inode_bitmap
);
2923 inode_table
= ext4_inode_table(sb
, gdp
);
2924 if (inode_table
== sb_block
) {
2925 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2926 "Inode table for group %u overlaps "
2931 if (inode_table
>= sb_block
+ 1 &&
2932 inode_table
<= last_bg_block
) {
2933 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2934 "Inode table for group %u overlaps "
2935 "block group descriptors", i
);
2939 if (inode_table
< first_block
||
2940 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2941 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2942 "Inode table for group %u not in group "
2943 "(block %llu)!", i
, inode_table
);
2946 ext4_lock_group(sb
, i
);
2947 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2948 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2949 "Checksum for group %u failed (%u!=%u)",
2950 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2951 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2952 if (!sb_rdonly(sb
)) {
2953 ext4_unlock_group(sb
, i
);
2957 ext4_unlock_group(sb
, i
);
2959 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2961 if (NULL
!= first_not_zeroed
)
2962 *first_not_zeroed
= grp
;
2966 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2967 * the superblock) which were deleted from all directories, but held open by
2968 * a process at the time of a crash. We walk the list and try to delete these
2969 * inodes at recovery time (only with a read-write filesystem).
2971 * In order to keep the orphan inode chain consistent during traversal (in
2972 * case of crash during recovery), we link each inode into the superblock
2973 * orphan list_head and handle it the same way as an inode deletion during
2974 * normal operation (which journals the operations for us).
2976 * We only do an iget() and an iput() on each inode, which is very safe if we
2977 * accidentally point at an in-use or already deleted inode. The worst that
2978 * can happen in this case is that we get a "bit already cleared" message from
2979 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2980 * e2fsck was run on this filesystem, and it must have already done the orphan
2981 * inode cleanup for us, so we can safely abort without any further action.
2983 static void ext4_orphan_cleanup(struct super_block
*sb
,
2984 struct ext4_super_block
*es
)
2986 unsigned int s_flags
= sb
->s_flags
;
2987 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2989 int quota_update
= 0;
2992 if (!es
->s_last_orphan
) {
2993 jbd_debug(4, "no orphan inodes to clean up\n");
2997 if (bdev_read_only(sb
->s_bdev
)) {
2998 ext4_msg(sb
, KERN_ERR
, "write access "
2999 "unavailable, skipping orphan cleanup");
3003 /* Check if feature set would not allow a r/w mount */
3004 if (!ext4_feature_set_ok(sb
, 0)) {
3005 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
3006 "unknown ROCOMPAT features");
3010 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
3011 /* don't clear list on RO mount w/ errors */
3012 if (es
->s_last_orphan
&& !(s_flags
& SB_RDONLY
)) {
3013 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
3014 "clearing orphan list.\n");
3015 es
->s_last_orphan
= 0;
3017 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3021 if (s_flags
& SB_RDONLY
) {
3022 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
3023 sb
->s_flags
&= ~SB_RDONLY
;
3026 /* Needed for iput() to work correctly and not trash data */
3027 sb
->s_flags
|= SB_ACTIVE
;
3030 * Turn on quotas which were not enabled for read-only mounts if
3031 * filesystem has quota feature, so that they are updated correctly.
3033 if (ext4_has_feature_quota(sb
) && (s_flags
& SB_RDONLY
)) {
3034 int ret
= ext4_enable_quotas(sb
);
3039 ext4_msg(sb
, KERN_ERR
,
3040 "Cannot turn on quotas: error %d", ret
);
3043 /* Turn on journaled quotas used for old sytle */
3044 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
3045 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
3046 int ret
= ext4_quota_on_mount(sb
, i
);
3051 ext4_msg(sb
, KERN_ERR
,
3052 "Cannot turn on journaled "
3053 "quota: type %d: error %d", i
, ret
);
3058 while (es
->s_last_orphan
) {
3059 struct inode
*inode
;
3062 * We may have encountered an error during cleanup; if
3063 * so, skip the rest.
3065 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
3066 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3067 es
->s_last_orphan
= 0;
3071 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
3072 if (IS_ERR(inode
)) {
3073 es
->s_last_orphan
= 0;
3077 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
3078 dquot_initialize(inode
);
3079 if (inode
->i_nlink
) {
3080 if (test_opt(sb
, DEBUG
))
3081 ext4_msg(sb
, KERN_DEBUG
,
3082 "%s: truncating inode %lu to %lld bytes",
3083 __func__
, inode
->i_ino
, inode
->i_size
);
3084 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3085 inode
->i_ino
, inode
->i_size
);
3087 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
3088 ret
= ext4_truncate(inode
);
3090 ext4_std_error(inode
->i_sb
, ret
);
3091 inode_unlock(inode
);
3094 if (test_opt(sb
, DEBUG
))
3095 ext4_msg(sb
, KERN_DEBUG
,
3096 "%s: deleting unreferenced inode %lu",
3097 __func__
, inode
->i_ino
);
3098 jbd_debug(2, "deleting unreferenced inode %lu\n",
3102 iput(inode
); /* The delete magic happens here! */
3105 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3108 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
3109 PLURAL(nr_orphans
));
3111 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
3112 PLURAL(nr_truncates
));
3114 /* Turn off quotas if they were enabled for orphan cleanup */
3116 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
3117 if (sb_dqopt(sb
)->files
[i
])
3118 dquot_quota_off(sb
, i
);
3122 sb
->s_flags
= s_flags
; /* Restore SB_RDONLY status */
3126 * Maximal extent format file size.
3127 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3128 * extent format containers, within a sector_t, and within i_blocks
3129 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3130 * so that won't be a limiting factor.
3132 * However there is other limiting factor. We do store extents in the form
3133 * of starting block and length, hence the resulting length of the extent
3134 * covering maximum file size must fit into on-disk format containers as
3135 * well. Given that length is always by 1 unit bigger than max unit (because
3136 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3138 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3140 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
3143 loff_t upper_limit
= MAX_LFS_FILESIZE
;
3145 BUILD_BUG_ON(sizeof(blkcnt_t
) < sizeof(u64
));
3147 if (!has_huge_files
) {
3148 upper_limit
= (1LL << 32) - 1;
3150 /* total blocks in file system block size */
3151 upper_limit
>>= (blkbits
- 9);
3152 upper_limit
<<= blkbits
;
3156 * 32-bit extent-start container, ee_block. We lower the maxbytes
3157 * by one fs block, so ee_len can cover the extent of maximum file
3160 res
= (1LL << 32) - 1;
3163 /* Sanity check against vm- & vfs- imposed limits */
3164 if (res
> upper_limit
)
3171 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3172 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3173 * We need to be 1 filesystem block less than the 2^48 sector limit.
3175 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
3177 loff_t res
= EXT4_NDIR_BLOCKS
;
3180 /* This is calculated to be the largest file size for a dense, block
3181 * mapped file such that the file's total number of 512-byte sectors,
3182 * including data and all indirect blocks, does not exceed (2^48 - 1).
3184 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3185 * number of 512-byte sectors of the file.
3188 if (!has_huge_files
) {
3190 * !has_huge_files or implies that the inode i_block field
3191 * represents total file blocks in 2^32 512-byte sectors ==
3192 * size of vfs inode i_blocks * 8
3194 upper_limit
= (1LL << 32) - 1;
3196 /* total blocks in file system block size */
3197 upper_limit
>>= (bits
- 9);
3201 * We use 48 bit ext4_inode i_blocks
3202 * With EXT4_HUGE_FILE_FL set the i_blocks
3203 * represent total number of blocks in
3204 * file system block size
3206 upper_limit
= (1LL << 48) - 1;
3210 /* indirect blocks */
3212 /* double indirect blocks */
3213 meta_blocks
+= 1 + (1LL << (bits
-2));
3214 /* tripple indirect blocks */
3215 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
3217 upper_limit
-= meta_blocks
;
3218 upper_limit
<<= bits
;
3220 res
+= 1LL << (bits
-2);
3221 res
+= 1LL << (2*(bits
-2));
3222 res
+= 1LL << (3*(bits
-2));
3224 if (res
> upper_limit
)
3227 if (res
> MAX_LFS_FILESIZE
)
3228 res
= MAX_LFS_FILESIZE
;
3233 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
3234 ext4_fsblk_t logical_sb_block
, int nr
)
3236 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3237 ext4_group_t bg
, first_meta_bg
;
3240 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
3242 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
3243 return logical_sb_block
+ nr
+ 1;
3244 bg
= sbi
->s_desc_per_block
* nr
;
3245 if (ext4_bg_has_super(sb
, bg
))
3249 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3250 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3251 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3254 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
3255 le32_to_cpu(sbi
->s_es
->s_first_data_block
) == 0)
3258 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
3262 * ext4_get_stripe_size: Get the stripe size.
3263 * @sbi: In memory super block info
3265 * If we have specified it via mount option, then
3266 * use the mount option value. If the value specified at mount time is
3267 * greater than the blocks per group use the super block value.
3268 * If the super block value is greater than blocks per group return 0.
3269 * Allocator needs it be less than blocks per group.
3272 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
3274 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
3275 unsigned long stripe_width
=
3276 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
3279 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
3280 ret
= sbi
->s_stripe
;
3281 else if (stripe_width
&& stripe_width
<= sbi
->s_blocks_per_group
)
3283 else if (stride
&& stride
<= sbi
->s_blocks_per_group
)
3289 * If the stripe width is 1, this makes no sense and
3290 * we set it to 0 to turn off stripe handling code.
3299 * Check whether this filesystem can be mounted based on
3300 * the features present and the RDONLY/RDWR mount requested.
3301 * Returns 1 if this filesystem can be mounted as requested,
3302 * 0 if it cannot be.
3304 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
3306 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
3307 ext4_msg(sb
, KERN_ERR
,
3308 "Couldn't mount because of "
3309 "unsupported optional features (%x)",
3310 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
3311 ~EXT4_FEATURE_INCOMPAT_SUPP
));
3315 #ifndef CONFIG_UNICODE
3316 if (ext4_has_feature_casefold(sb
)) {
3317 ext4_msg(sb
, KERN_ERR
,
3318 "Filesystem with casefold feature cannot be "
3319 "mounted without CONFIG_UNICODE");
3327 if (ext4_has_feature_readonly(sb
)) {
3328 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
3329 sb
->s_flags
|= SB_RDONLY
;
3333 /* Check that feature set is OK for a read-write mount */
3334 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
3335 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
3336 "unsupported optional features (%x)",
3337 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
3338 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
3341 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
3342 ext4_msg(sb
, KERN_ERR
,
3343 "Can't support bigalloc feature without "
3344 "extents feature\n");
3348 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3349 if (!readonly
&& (ext4_has_feature_quota(sb
) ||
3350 ext4_has_feature_project(sb
))) {
3351 ext4_msg(sb
, KERN_ERR
,
3352 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3355 #endif /* CONFIG_QUOTA */
3360 * This function is called once a day if we have errors logged
3361 * on the file system
3363 static void print_daily_error_info(struct timer_list
*t
)
3365 struct ext4_sb_info
*sbi
= from_timer(sbi
, t
, s_err_report
);
3366 struct super_block
*sb
= sbi
->s_sb
;
3367 struct ext4_super_block
*es
= sbi
->s_es
;
3369 if (es
->s_error_count
)
3370 /* fsck newer than v1.41.13 is needed to clean this condition. */
3371 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
3372 le32_to_cpu(es
->s_error_count
));
3373 if (es
->s_first_error_time
) {
3374 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3376 ext4_get_tstamp(es
, s_first_error_time
),
3377 (int) sizeof(es
->s_first_error_func
),
3378 es
->s_first_error_func
,
3379 le32_to_cpu(es
->s_first_error_line
));
3380 if (es
->s_first_error_ino
)
3381 printk(KERN_CONT
": inode %u",
3382 le32_to_cpu(es
->s_first_error_ino
));
3383 if (es
->s_first_error_block
)
3384 printk(KERN_CONT
": block %llu", (unsigned long long)
3385 le64_to_cpu(es
->s_first_error_block
));
3386 printk(KERN_CONT
"\n");
3388 if (es
->s_last_error_time
) {
3389 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %llu: %.*s:%d",
3391 ext4_get_tstamp(es
, s_last_error_time
),
3392 (int) sizeof(es
->s_last_error_func
),
3393 es
->s_last_error_func
,
3394 le32_to_cpu(es
->s_last_error_line
));
3395 if (es
->s_last_error_ino
)
3396 printk(KERN_CONT
": inode %u",
3397 le32_to_cpu(es
->s_last_error_ino
));
3398 if (es
->s_last_error_block
)
3399 printk(KERN_CONT
": block %llu", (unsigned long long)
3400 le64_to_cpu(es
->s_last_error_block
));
3401 printk(KERN_CONT
"\n");
3403 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
3406 /* Find next suitable group and run ext4_init_inode_table */
3407 static int ext4_run_li_request(struct ext4_li_request
*elr
)
3409 struct ext4_group_desc
*gdp
= NULL
;
3410 struct super_block
*sb
= elr
->lr_super
;
3411 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3412 ext4_group_t group
= elr
->lr_next_group
;
3413 unsigned long timeout
= 0;
3414 unsigned int prefetch_ios
= 0;
3417 if (elr
->lr_mode
== EXT4_LI_MODE_PREFETCH_BBITMAP
) {
3418 elr
->lr_next_group
= ext4_mb_prefetch(sb
, group
,
3419 EXT4_SB(sb
)->s_mb_prefetch
, &prefetch_ios
);
3421 ext4_mb_prefetch_fini(sb
, elr
->lr_next_group
,
3423 trace_ext4_prefetch_bitmaps(sb
, group
, elr
->lr_next_group
,
3425 if (group
>= elr
->lr_next_group
) {
3427 if (elr
->lr_first_not_zeroed
!= ngroups
&&
3428 !sb_rdonly(sb
) && test_opt(sb
, INIT_INODE_TABLE
)) {
3429 elr
->lr_next_group
= elr
->lr_first_not_zeroed
;
3430 elr
->lr_mode
= EXT4_LI_MODE_ITABLE
;
3437 for (; group
< ngroups
; group
++) {
3438 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3444 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3448 if (group
>= ngroups
)
3453 ret
= ext4_init_inode_table(sb
, group
,
3454 elr
->lr_timeout
? 0 : 1);
3455 trace_ext4_lazy_itable_init(sb
, group
);
3456 if (elr
->lr_timeout
== 0) {
3457 timeout
= (jiffies
- timeout
) *
3458 EXT4_SB(elr
->lr_super
)->s_li_wait_mult
;
3459 elr
->lr_timeout
= timeout
;
3461 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
3462 elr
->lr_next_group
= group
+ 1;
3468 * Remove lr_request from the list_request and free the
3469 * request structure. Should be called with li_list_mtx held
3471 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
3476 list_del(&elr
->lr_request
);
3477 EXT4_SB(elr
->lr_super
)->s_li_request
= NULL
;
3481 static void ext4_unregister_li_request(struct super_block
*sb
)
3483 mutex_lock(&ext4_li_mtx
);
3484 if (!ext4_li_info
) {
3485 mutex_unlock(&ext4_li_mtx
);
3489 mutex_lock(&ext4_li_info
->li_list_mtx
);
3490 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
3491 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3492 mutex_unlock(&ext4_li_mtx
);
3495 static struct task_struct
*ext4_lazyinit_task
;
3498 * This is the function where ext4lazyinit thread lives. It walks
3499 * through the request list searching for next scheduled filesystem.
3500 * When such a fs is found, run the lazy initialization request
3501 * (ext4_rn_li_request) and keep track of the time spend in this
3502 * function. Based on that time we compute next schedule time of
3503 * the request. When walking through the list is complete, compute
3504 * next waking time and put itself into sleep.
3506 static int ext4_lazyinit_thread(void *arg
)
3508 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
3509 struct list_head
*pos
, *n
;
3510 struct ext4_li_request
*elr
;
3511 unsigned long next_wakeup
, cur
;
3513 BUG_ON(NULL
== eli
);
3517 next_wakeup
= MAX_JIFFY_OFFSET
;
3519 mutex_lock(&eli
->li_list_mtx
);
3520 if (list_empty(&eli
->li_request_list
)) {
3521 mutex_unlock(&eli
->li_list_mtx
);
3524 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
3527 elr
= list_entry(pos
, struct ext4_li_request
,
3530 if (time_before(jiffies
, elr
->lr_next_sched
)) {
3531 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3532 next_wakeup
= elr
->lr_next_sched
;
3535 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
3536 if (sb_start_write_trylock(elr
->lr_super
)) {
3539 * We hold sb->s_umount, sb can not
3540 * be removed from the list, it is
3541 * now safe to drop li_list_mtx
3543 mutex_unlock(&eli
->li_list_mtx
);
3544 err
= ext4_run_li_request(elr
);
3545 sb_end_write(elr
->lr_super
);
3546 mutex_lock(&eli
->li_list_mtx
);
3549 up_read((&elr
->lr_super
->s_umount
));
3551 /* error, remove the lazy_init job */
3553 ext4_remove_li_request(elr
);
3557 elr
->lr_next_sched
= jiffies
+
3559 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3561 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3562 next_wakeup
= elr
->lr_next_sched
;
3564 mutex_unlock(&eli
->li_list_mtx
);
3569 if ((time_after_eq(cur
, next_wakeup
)) ||
3570 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3575 schedule_timeout_interruptible(next_wakeup
- cur
);
3577 if (kthread_should_stop()) {
3578 ext4_clear_request_list();
3585 * It looks like the request list is empty, but we need
3586 * to check it under the li_list_mtx lock, to prevent any
3587 * additions into it, and of course we should lock ext4_li_mtx
3588 * to atomically free the list and ext4_li_info, because at
3589 * this point another ext4 filesystem could be registering
3592 mutex_lock(&ext4_li_mtx
);
3593 mutex_lock(&eli
->li_list_mtx
);
3594 if (!list_empty(&eli
->li_request_list
)) {
3595 mutex_unlock(&eli
->li_list_mtx
);
3596 mutex_unlock(&ext4_li_mtx
);
3599 mutex_unlock(&eli
->li_list_mtx
);
3600 kfree(ext4_li_info
);
3601 ext4_li_info
= NULL
;
3602 mutex_unlock(&ext4_li_mtx
);
3607 static void ext4_clear_request_list(void)
3609 struct list_head
*pos
, *n
;
3610 struct ext4_li_request
*elr
;
3612 mutex_lock(&ext4_li_info
->li_list_mtx
);
3613 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3614 elr
= list_entry(pos
, struct ext4_li_request
,
3616 ext4_remove_li_request(elr
);
3618 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3621 static int ext4_run_lazyinit_thread(void)
3623 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3624 ext4_li_info
, "ext4lazyinit");
3625 if (IS_ERR(ext4_lazyinit_task
)) {
3626 int err
= PTR_ERR(ext4_lazyinit_task
);
3627 ext4_clear_request_list();
3628 kfree(ext4_li_info
);
3629 ext4_li_info
= NULL
;
3630 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3631 "initialization thread\n",
3635 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3640 * Check whether it make sense to run itable init. thread or not.
3641 * If there is at least one uninitialized inode table, return
3642 * corresponding group number, else the loop goes through all
3643 * groups and return total number of groups.
3645 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3647 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3648 struct ext4_group_desc
*gdp
= NULL
;
3650 if (!ext4_has_group_desc_csum(sb
))
3653 for (group
= 0; group
< ngroups
; group
++) {
3654 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3658 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3665 static int ext4_li_info_new(void)
3667 struct ext4_lazy_init
*eli
= NULL
;
3669 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3673 INIT_LIST_HEAD(&eli
->li_request_list
);
3674 mutex_init(&eli
->li_list_mtx
);
3676 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3683 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3686 struct ext4_li_request
*elr
;
3688 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3693 elr
->lr_first_not_zeroed
= start
;
3694 if (test_opt(sb
, PREFETCH_BLOCK_BITMAPS
))
3695 elr
->lr_mode
= EXT4_LI_MODE_PREFETCH_BBITMAP
;
3697 elr
->lr_mode
= EXT4_LI_MODE_ITABLE
;
3698 elr
->lr_next_group
= start
;
3702 * Randomize first schedule time of the request to
3703 * spread the inode table initialization requests
3706 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3707 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3711 int ext4_register_li_request(struct super_block
*sb
,
3712 ext4_group_t first_not_zeroed
)
3714 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3715 struct ext4_li_request
*elr
= NULL
;
3716 ext4_group_t ngroups
= sbi
->s_groups_count
;
3719 mutex_lock(&ext4_li_mtx
);
3720 if (sbi
->s_li_request
!= NULL
) {
3722 * Reset timeout so it can be computed again, because
3723 * s_li_wait_mult might have changed.
3725 sbi
->s_li_request
->lr_timeout
= 0;
3729 if (!test_opt(sb
, PREFETCH_BLOCK_BITMAPS
) &&
3730 (first_not_zeroed
== ngroups
|| sb_rdonly(sb
) ||
3731 !test_opt(sb
, INIT_INODE_TABLE
)))
3734 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3740 if (NULL
== ext4_li_info
) {
3741 ret
= ext4_li_info_new();
3746 mutex_lock(&ext4_li_info
->li_list_mtx
);
3747 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3748 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3750 sbi
->s_li_request
= elr
;
3752 * set elr to NULL here since it has been inserted to
3753 * the request_list and the removal and free of it is
3754 * handled by ext4_clear_request_list from now on.
3758 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3759 ret
= ext4_run_lazyinit_thread();
3764 mutex_unlock(&ext4_li_mtx
);
3771 * We do not need to lock anything since this is called on
3774 static void ext4_destroy_lazyinit_thread(void)
3777 * If thread exited earlier
3778 * there's nothing to be done.
3780 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3783 kthread_stop(ext4_lazyinit_task
);
3786 static int set_journal_csum_feature_set(struct super_block
*sb
)
3789 int compat
, incompat
;
3790 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3792 if (ext4_has_metadata_csum(sb
)) {
3793 /* journal checksum v3 */
3795 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3797 /* journal checksum v1 */
3798 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3802 jbd2_journal_clear_features(sbi
->s_journal
,
3803 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3804 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3805 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3806 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3807 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3809 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3811 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3812 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3815 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3816 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3818 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3819 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3826 * Note: calculating the overhead so we can be compatible with
3827 * historical BSD practice is quite difficult in the face of
3828 * clusters/bigalloc. This is because multiple metadata blocks from
3829 * different block group can end up in the same allocation cluster.
3830 * Calculating the exact overhead in the face of clustered allocation
3831 * requires either O(all block bitmaps) in memory or O(number of block
3832 * groups**2) in time. We will still calculate the superblock for
3833 * older file systems --- and if we come across with a bigalloc file
3834 * system with zero in s_overhead_clusters the estimate will be close to
3835 * correct especially for very large cluster sizes --- but for newer
3836 * file systems, it's better to calculate this figure once at mkfs
3837 * time, and store it in the superblock. If the superblock value is
3838 * present (even for non-bigalloc file systems), we will use it.
3840 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3843 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3844 struct ext4_group_desc
*gdp
;
3845 ext4_fsblk_t first_block
, last_block
, b
;
3846 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3847 int s
, j
, count
= 0;
3849 if (!ext4_has_feature_bigalloc(sb
))
3850 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3851 sbi
->s_itb_per_group
+ 2);
3853 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3854 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3855 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3856 for (i
= 0; i
< ngroups
; i
++) {
3857 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3858 b
= ext4_block_bitmap(sb
, gdp
);
3859 if (b
>= first_block
&& b
<= last_block
) {
3860 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3863 b
= ext4_inode_bitmap(sb
, gdp
);
3864 if (b
>= first_block
&& b
<= last_block
) {
3865 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3868 b
= ext4_inode_table(sb
, gdp
);
3869 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3870 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3871 int c
= EXT4_B2C(sbi
, b
- first_block
);
3872 ext4_set_bit(c
, buf
);
3878 if (ext4_bg_has_super(sb
, grp
)) {
3879 ext4_set_bit(s
++, buf
);
3882 j
= ext4_bg_num_gdb(sb
, grp
);
3883 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3884 ext4_error(sb
, "Invalid number of block group "
3885 "descriptor blocks: %d", j
);
3886 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3890 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3894 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3895 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3899 * Compute the overhead and stash it in sbi->s_overhead
3901 int ext4_calculate_overhead(struct super_block
*sb
)
3903 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3904 struct ext4_super_block
*es
= sbi
->s_es
;
3905 struct inode
*j_inode
;
3906 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3907 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3908 ext4_fsblk_t overhead
= 0;
3909 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3915 * Compute the overhead (FS structures). This is constant
3916 * for a given filesystem unless the number of block groups
3917 * changes so we cache the previous value until it does.
3921 * All of the blocks before first_data_block are overhead
3923 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3926 * Add the overhead found in each block group
3928 for (i
= 0; i
< ngroups
; i
++) {
3931 blks
= count_overhead(sb
, i
, buf
);
3934 memset(buf
, 0, PAGE_SIZE
);
3939 * Add the internal journal blocks whether the journal has been
3942 if (sbi
->s_journal
&& !sbi
->s_journal_bdev
)
3943 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_total_len
);
3944 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
&& j_inum
) {
3945 /* j_inum for internal journal is non-zero */
3946 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3948 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3949 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3952 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3955 sbi
->s_overhead
= overhead
;
3957 free_page((unsigned long) buf
);
3961 static void ext4_set_resv_clusters(struct super_block
*sb
)
3963 ext4_fsblk_t resv_clusters
;
3964 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3967 * There's no need to reserve anything when we aren't using extents.
3968 * The space estimates are exact, there are no unwritten extents,
3969 * hole punching doesn't need new metadata... This is needed especially
3970 * to keep ext2/3 backward compatibility.
3972 if (!ext4_has_feature_extents(sb
))
3975 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3976 * This should cover the situations where we can not afford to run
3977 * out of space like for example punch hole, or converting
3978 * unwritten extents in delalloc path. In most cases such
3979 * allocation would require 1, or 2 blocks, higher numbers are
3982 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3983 sbi
->s_cluster_bits
);
3985 do_div(resv_clusters
, 50);
3986 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3988 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3991 static const char *ext4_quota_mode(struct super_block
*sb
)
3994 if (!ext4_quota_capable(sb
))
3997 if (EXT4_SB(sb
)->s_journal
&& ext4_is_quota_journalled(sb
))
3998 return "journalled";
4006 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
4008 struct dax_device
*dax_dev
= fs_dax_get_by_bdev(sb
->s_bdev
);
4009 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4010 struct buffer_head
*bh
, **group_desc
;
4011 struct ext4_super_block
*es
= NULL
;
4012 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
4013 struct flex_groups
**flex_groups
;
4015 ext4_fsblk_t sb_block
= get_sb_block(&data
);
4016 ext4_fsblk_t logical_sb_block
;
4017 unsigned long offset
= 0;
4018 unsigned long journal_devnum
= 0;
4019 unsigned long def_mount_opts
;
4023 int blocksize
, clustersize
;
4024 unsigned int db_count
;
4026 int needs_recovery
, has_huge_files
;
4029 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4030 ext4_group_t first_not_zeroed
;
4032 if ((data
&& !orig_data
) || !sbi
)
4035 sbi
->s_daxdev
= dax_dev
;
4036 sbi
->s_blockgroup_lock
=
4037 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
4038 if (!sbi
->s_blockgroup_lock
)
4041 sb
->s_fs_info
= sbi
;
4043 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
4044 sbi
->s_sb_block
= sb_block
;
4045 sbi
->s_sectors_written_start
=
4046 part_stat_read(sb
->s_bdev
, sectors
[STAT_WRITE
]);
4048 /* Cleanup superblock name */
4049 strreplace(sb
->s_id
, '/', '!');
4051 /* -EINVAL is default */
4053 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
4055 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
4060 * The ext4 superblock will not be buffer aligned for other than 1kB
4061 * block sizes. We need to calculate the offset from buffer start.
4063 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
4064 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
4065 offset
= do_div(logical_sb_block
, blocksize
);
4067 logical_sb_block
= sb_block
;
4070 bh
= ext4_sb_bread_unmovable(sb
, logical_sb_block
);
4072 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
4077 * Note: s_es must be initialized as soon as possible because
4078 * some ext4 macro-instructions depend on its value
4080 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4082 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
4083 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
4085 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
4087 /* Warn if metadata_csum and gdt_csum are both set. */
4088 if (ext4_has_feature_metadata_csum(sb
) &&
4089 ext4_has_feature_gdt_csum(sb
))
4090 ext4_warning(sb
, "metadata_csum and uninit_bg are "
4091 "redundant flags; please run fsck.");
4093 /* Check for a known checksum algorithm */
4094 if (!ext4_verify_csum_type(sb
, es
)) {
4095 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
4096 "unknown checksum algorithm.");
4101 /* Load the checksum driver */
4102 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
4103 if (IS_ERR(sbi
->s_chksum_driver
)) {
4104 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
4105 ret
= PTR_ERR(sbi
->s_chksum_driver
);
4106 sbi
->s_chksum_driver
= NULL
;
4110 /* Check superblock checksum */
4111 if (!ext4_superblock_csum_verify(sb
, es
)) {
4112 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
4113 "invalid superblock checksum. Run e2fsck?");
4119 /* Precompute checksum seed for all metadata */
4120 if (ext4_has_feature_csum_seed(sb
))
4121 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
4122 else if (ext4_has_metadata_csum(sb
) || ext4_has_feature_ea_inode(sb
))
4123 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
4124 sizeof(es
->s_uuid
));
4126 /* Set defaults before we parse the mount options */
4127 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
4128 set_opt(sb
, INIT_INODE_TABLE
);
4129 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
4131 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
4133 if (def_mount_opts
& EXT4_DEFM_UID16
)
4134 set_opt(sb
, NO_UID32
);
4135 /* xattr user namespace & acls are now defaulted on */
4136 set_opt(sb
, XATTR_USER
);
4137 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4138 set_opt(sb
, POSIX_ACL
);
4140 if (ext4_has_feature_fast_commit(sb
))
4141 set_opt2(sb
, JOURNAL_FAST_COMMIT
);
4142 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4143 if (ext4_has_metadata_csum(sb
))
4144 set_opt(sb
, JOURNAL_CHECKSUM
);
4146 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
4147 set_opt(sb
, JOURNAL_DATA
);
4148 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
4149 set_opt(sb
, ORDERED_DATA
);
4150 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
4151 set_opt(sb
, WRITEBACK_DATA
);
4153 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
4154 set_opt(sb
, ERRORS_PANIC
);
4155 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
4156 set_opt(sb
, ERRORS_CONT
);
4158 set_opt(sb
, ERRORS_RO
);
4159 /* block_validity enabled by default; disable with noblock_validity */
4160 set_opt(sb
, BLOCK_VALIDITY
);
4161 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
4162 set_opt(sb
, DISCARD
);
4164 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
4165 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
4166 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
4167 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
4168 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
4170 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
4171 set_opt(sb
, BARRIER
);
4174 * enable delayed allocation by default
4175 * Use -o nodelalloc to turn it off
4177 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
4178 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
4179 set_opt(sb
, DELALLOC
);
4182 * set default s_li_wait_mult for lazyinit, for the case there is
4183 * no mount option specified.
4185 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
4187 if (le32_to_cpu(es
->s_log_block_size
) >
4188 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
4189 ext4_msg(sb
, KERN_ERR
,
4190 "Invalid log block size: %u",
4191 le32_to_cpu(es
->s_log_block_size
));
4194 if (le32_to_cpu(es
->s_log_cluster_size
) >
4195 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
4196 ext4_msg(sb
, KERN_ERR
,
4197 "Invalid log cluster size: %u",
4198 le32_to_cpu(es
->s_log_cluster_size
));
4202 blocksize
= EXT4_MIN_BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
4204 if (blocksize
== PAGE_SIZE
)
4205 set_opt(sb
, DIOREAD_NOLOCK
);
4207 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
4208 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
4209 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
4211 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
4212 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
4213 if (sbi
->s_first_ino
< EXT4_GOOD_OLD_FIRST_INO
) {
4214 ext4_msg(sb
, KERN_ERR
, "invalid first ino: %u",
4218 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
4219 (!is_power_of_2(sbi
->s_inode_size
)) ||
4220 (sbi
->s_inode_size
> blocksize
)) {
4221 ext4_msg(sb
, KERN_ERR
,
4222 "unsupported inode size: %d",
4224 ext4_msg(sb
, KERN_ERR
, "blocksize: %d", blocksize
);
4228 * i_atime_extra is the last extra field available for
4229 * [acm]times in struct ext4_inode. Checking for that
4230 * field should suffice to ensure we have extra space
4233 if (sbi
->s_inode_size
>= offsetof(struct ext4_inode
, i_atime_extra
) +
4234 sizeof(((struct ext4_inode
*)0)->i_atime_extra
)) {
4235 sb
->s_time_gran
= 1;
4236 sb
->s_time_max
= EXT4_EXTRA_TIMESTAMP_MAX
;
4238 sb
->s_time_gran
= NSEC_PER_SEC
;
4239 sb
->s_time_max
= EXT4_NON_EXTRA_TIMESTAMP_MAX
;
4241 sb
->s_time_min
= EXT4_TIMESTAMP_MIN
;
4243 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4244 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4245 EXT4_GOOD_OLD_INODE_SIZE
;
4246 if (ext4_has_feature_extra_isize(sb
)) {
4247 unsigned v
, max
= (sbi
->s_inode_size
-
4248 EXT4_GOOD_OLD_INODE_SIZE
);
4250 v
= le16_to_cpu(es
->s_want_extra_isize
);
4252 ext4_msg(sb
, KERN_ERR
,
4253 "bad s_want_extra_isize: %d", v
);
4256 if (sbi
->s_want_extra_isize
< v
)
4257 sbi
->s_want_extra_isize
= v
;
4259 v
= le16_to_cpu(es
->s_min_extra_isize
);
4261 ext4_msg(sb
, KERN_ERR
,
4262 "bad s_min_extra_isize: %d", v
);
4265 if (sbi
->s_want_extra_isize
< v
)
4266 sbi
->s_want_extra_isize
= v
;
4270 if (sbi
->s_es
->s_mount_opts
[0]) {
4271 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
4272 sizeof(sbi
->s_es
->s_mount_opts
),
4276 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
4277 &journal_ioprio
, 0)) {
4278 ext4_msg(sb
, KERN_WARNING
,
4279 "failed to parse options in superblock: %s",
4282 kfree(s_mount_opts
);
4284 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
4285 if (!parse_options((char *) data
, sb
, &journal_devnum
,
4286 &journal_ioprio
, 0))
4289 #ifdef CONFIG_UNICODE
4290 if (ext4_has_feature_casefold(sb
) && !sb
->s_encoding
) {
4291 const struct ext4_sb_encodings
*encoding_info
;
4292 struct unicode_map
*encoding
;
4293 __u16 encoding_flags
;
4295 if (ext4_has_feature_encrypt(sb
)) {
4296 ext4_msg(sb
, KERN_ERR
,
4297 "Can't mount with encoding and encryption");
4301 if (ext4_sb_read_encoding(es
, &encoding_info
,
4303 ext4_msg(sb
, KERN_ERR
,
4304 "Encoding requested by superblock is unknown");
4308 encoding
= utf8_load(encoding_info
->version
);
4309 if (IS_ERR(encoding
)) {
4310 ext4_msg(sb
, KERN_ERR
,
4311 "can't mount with superblock charset: %s-%s "
4312 "not supported by the kernel. flags: 0x%x.",
4313 encoding_info
->name
, encoding_info
->version
,
4317 ext4_msg(sb
, KERN_INFO
,"Using encoding defined by superblock: "
4318 "%s-%s with flags 0x%hx", encoding_info
->name
,
4319 encoding_info
->version
?:"\b", encoding_flags
);
4321 sb
->s_encoding
= encoding
;
4322 sb
->s_encoding_flags
= encoding_flags
;
4326 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4327 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4328 /* can't mount with both data=journal and dioread_nolock. */
4329 clear_opt(sb
, DIOREAD_NOLOCK
);
4330 clear_opt2(sb
, JOURNAL_FAST_COMMIT
);
4331 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4332 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4333 "both data=journal and delalloc");
4336 if (test_opt(sb
, DAX_ALWAYS
)) {
4337 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4338 "both data=journal and dax");
4341 if (ext4_has_feature_encrypt(sb
)) {
4342 ext4_msg(sb
, KERN_WARNING
,
4343 "encrypted files will use data=ordered "
4344 "instead of data journaling mode");
4346 if (test_opt(sb
, DELALLOC
))
4347 clear_opt(sb
, DELALLOC
);
4349 sb
->s_iflags
|= SB_I_CGROUPWB
;
4352 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
4353 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
4355 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
4356 (ext4_has_compat_features(sb
) ||
4357 ext4_has_ro_compat_features(sb
) ||
4358 ext4_has_incompat_features(sb
)))
4359 ext4_msg(sb
, KERN_WARNING
,
4360 "feature flags set on rev 0 fs, "
4361 "running e2fsck is recommended");
4363 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
4364 set_opt2(sb
, HURD_COMPAT
);
4365 if (ext4_has_feature_64bit(sb
)) {
4366 ext4_msg(sb
, KERN_ERR
,
4367 "The Hurd can't support 64-bit file systems");
4372 * ea_inode feature uses l_i_version field which is not
4373 * available in HURD_COMPAT mode.
4375 if (ext4_has_feature_ea_inode(sb
)) {
4376 ext4_msg(sb
, KERN_ERR
,
4377 "ea_inode feature is not supported for Hurd");
4382 if (IS_EXT2_SB(sb
)) {
4383 if (ext2_feature_set_ok(sb
))
4384 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
4385 "using the ext4 subsystem");
4388 * If we're probing be silent, if this looks like
4389 * it's actually an ext[34] filesystem.
4391 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
4393 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
4394 "to feature incompatibilities");
4399 if (IS_EXT3_SB(sb
)) {
4400 if (ext3_feature_set_ok(sb
))
4401 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
4402 "using the ext4 subsystem");
4405 * If we're probing be silent, if this looks like
4406 * it's actually an ext4 filesystem.
4408 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
4410 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
4411 "to feature incompatibilities");
4417 * Check feature flags regardless of the revision level, since we
4418 * previously didn't change the revision level when setting the flags,
4419 * so there is a chance incompat flags are set on a rev 0 filesystem.
4421 if (!ext4_feature_set_ok(sb
, (sb_rdonly(sb
))))
4424 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
4425 ext4_msg(sb
, KERN_ERR
,
4426 "Number of reserved GDT blocks insanely large: %d",
4427 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
4431 if (bdev_dax_supported(sb
->s_bdev
, blocksize
))
4432 set_bit(EXT4_FLAGS_BDEV_IS_DAX
, &sbi
->s_ext4_flags
);
4434 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX_ALWAYS
) {
4435 if (ext4_has_feature_inline_data(sb
)) {
4436 ext4_msg(sb
, KERN_ERR
, "Cannot use DAX on a filesystem"
4437 " that may contain inline data");
4440 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX
, &sbi
->s_ext4_flags
)) {
4441 ext4_msg(sb
, KERN_ERR
,
4442 "DAX unsupported by block device.");
4447 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
4448 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
4449 es
->s_encryption_level
);
4453 if (sb
->s_blocksize
!= blocksize
) {
4454 /* Validate the filesystem blocksize */
4455 if (!sb_set_blocksize(sb
, blocksize
)) {
4456 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
4462 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
4463 offset
= do_div(logical_sb_block
, blocksize
);
4464 bh
= ext4_sb_bread_unmovable(sb
, logical_sb_block
);
4466 ext4_msg(sb
, KERN_ERR
,
4467 "Can't read superblock on 2nd try");
4472 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
4474 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
4475 ext4_msg(sb
, KERN_ERR
,
4476 "Magic mismatch, very weird!");
4481 has_huge_files
= ext4_has_feature_huge_file(sb
);
4482 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
4484 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
4486 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
4487 if (ext4_has_feature_64bit(sb
)) {
4488 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
4489 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
4490 !is_power_of_2(sbi
->s_desc_size
)) {
4491 ext4_msg(sb
, KERN_ERR
,
4492 "unsupported descriptor size %lu",
4497 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
4499 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
4500 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
4502 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
4503 if (sbi
->s_inodes_per_block
== 0)
4505 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
4506 sbi
->s_inodes_per_group
> blocksize
* 8) {
4507 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
4508 sbi
->s_inodes_per_group
);
4511 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
4512 sbi
->s_inodes_per_block
;
4513 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
4515 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4516 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
4517 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
4519 for (i
= 0; i
< 4; i
++)
4520 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
4521 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
4522 if (ext4_has_feature_dir_index(sb
)) {
4523 i
= le32_to_cpu(es
->s_flags
);
4524 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
4525 sbi
->s_hash_unsigned
= 3;
4526 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
4527 #ifdef __CHAR_UNSIGNED__
4530 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
4531 sbi
->s_hash_unsigned
= 3;
4535 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
4540 /* Handle clustersize */
4541 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
4542 if (ext4_has_feature_bigalloc(sb
)) {
4543 if (clustersize
< blocksize
) {
4544 ext4_msg(sb
, KERN_ERR
,
4545 "cluster size (%d) smaller than "
4546 "block size (%d)", clustersize
, blocksize
);
4549 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
4550 le32_to_cpu(es
->s_log_block_size
);
4551 sbi
->s_clusters_per_group
=
4552 le32_to_cpu(es
->s_clusters_per_group
);
4553 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
4554 ext4_msg(sb
, KERN_ERR
,
4555 "#clusters per group too big: %lu",
4556 sbi
->s_clusters_per_group
);
4559 if (sbi
->s_blocks_per_group
!=
4560 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
4561 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
4562 "clusters per group (%lu) inconsistent",
4563 sbi
->s_blocks_per_group
,
4564 sbi
->s_clusters_per_group
);
4568 if (clustersize
!= blocksize
) {
4569 ext4_msg(sb
, KERN_ERR
,
4570 "fragment/cluster size (%d) != "
4571 "block size (%d)", clustersize
, blocksize
);
4574 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
4575 ext4_msg(sb
, KERN_ERR
,
4576 "#blocks per group too big: %lu",
4577 sbi
->s_blocks_per_group
);
4580 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
4581 sbi
->s_cluster_bits
= 0;
4583 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
4585 /* Do we have standard group size of clustersize * 8 blocks ? */
4586 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
4587 set_opt2(sb
, STD_GROUP_SIZE
);
4590 * Test whether we have more sectors than will fit in sector_t,
4591 * and whether the max offset is addressable by the page cache.
4593 err
= generic_check_addressable(sb
->s_blocksize_bits
,
4594 ext4_blocks_count(es
));
4596 ext4_msg(sb
, KERN_ERR
, "filesystem"
4597 " too large to mount safely on this system");
4601 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
4604 /* check blocks count against device size */
4605 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
4606 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
4607 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
4608 "exceeds size of device (%llu blocks)",
4609 ext4_blocks_count(es
), blocks_count
);
4614 * It makes no sense for the first data block to be beyond the end
4615 * of the filesystem.
4617 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
4618 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4619 "block %u is beyond end of filesystem (%llu)",
4620 le32_to_cpu(es
->s_first_data_block
),
4621 ext4_blocks_count(es
));
4624 if ((es
->s_first_data_block
== 0) && (es
->s_log_block_size
== 0) &&
4625 (sbi
->s_cluster_ratio
== 1)) {
4626 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
4627 "block is 0 with a 1k block and cluster size");
4631 blocks_count
= (ext4_blocks_count(es
) -
4632 le32_to_cpu(es
->s_first_data_block
) +
4633 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
4634 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
4635 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
4636 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %llu "
4637 "(block count %llu, first data block %u, "
4638 "blocks per group %lu)", blocks_count
,
4639 ext4_blocks_count(es
),
4640 le32_to_cpu(es
->s_first_data_block
),
4641 EXT4_BLOCKS_PER_GROUP(sb
));
4644 sbi
->s_groups_count
= blocks_count
;
4645 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
4646 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
4647 if (((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
) !=
4648 le32_to_cpu(es
->s_inodes_count
)) {
4649 ext4_msg(sb
, KERN_ERR
, "inodes count not valid: %u vs %llu",
4650 le32_to_cpu(es
->s_inodes_count
),
4651 ((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
));
4655 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
4656 EXT4_DESC_PER_BLOCK(sb
);
4657 if (ext4_has_feature_meta_bg(sb
)) {
4658 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
4659 ext4_msg(sb
, KERN_WARNING
,
4660 "first meta block group too large: %u "
4661 "(group descriptor block count %u)",
4662 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
4666 rcu_assign_pointer(sbi
->s_group_desc
,
4667 kvmalloc_array(db_count
,
4668 sizeof(struct buffer_head
*),
4670 if (sbi
->s_group_desc
== NULL
) {
4671 ext4_msg(sb
, KERN_ERR
, "not enough memory");
4676 bgl_lock_init(sbi
->s_blockgroup_lock
);
4678 /* Pre-read the descriptors into the buffer cache */
4679 for (i
= 0; i
< db_count
; i
++) {
4680 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4681 ext4_sb_breadahead_unmovable(sb
, block
);
4684 for (i
= 0; i
< db_count
; i
++) {
4685 struct buffer_head
*bh
;
4687 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4688 bh
= ext4_sb_bread_unmovable(sb
, block
);
4690 ext4_msg(sb
, KERN_ERR
,
4691 "can't read group descriptor %d", i
);
4697 rcu_dereference(sbi
->s_group_desc
)[i
] = bh
;
4700 sbi
->s_gdb_count
= db_count
;
4701 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
4702 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
4703 ret
= -EFSCORRUPTED
;
4707 timer_setup(&sbi
->s_err_report
, print_daily_error_info
, 0);
4708 spin_lock_init(&sbi
->s_error_lock
);
4709 INIT_WORK(&sbi
->s_error_work
, flush_stashed_error_work
);
4711 /* Register extent status tree shrinker */
4712 if (ext4_es_register_shrinker(sbi
))
4715 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
4716 sbi
->s_extent_max_zeroout_kb
= 32;
4719 * set up enough so that it can read an inode
4721 sb
->s_op
= &ext4_sops
;
4722 sb
->s_export_op
= &ext4_export_ops
;
4723 sb
->s_xattr
= ext4_xattr_handlers
;
4724 #ifdef CONFIG_FS_ENCRYPTION
4725 sb
->s_cop
= &ext4_cryptops
;
4727 #ifdef CONFIG_FS_VERITY
4728 sb
->s_vop
= &ext4_verityops
;
4731 sb
->dq_op
= &ext4_quota_operations
;
4732 if (ext4_has_feature_quota(sb
))
4733 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
4735 sb
->s_qcop
= &ext4_qctl_operations
;
4736 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
4738 memcpy(&sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
4740 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
4741 mutex_init(&sbi
->s_orphan_lock
);
4743 /* Initialize fast commit stuff */
4744 atomic_set(&sbi
->s_fc_subtid
, 0);
4745 atomic_set(&sbi
->s_fc_ineligible_updates
, 0);
4746 INIT_LIST_HEAD(&sbi
->s_fc_q
[FC_Q_MAIN
]);
4747 INIT_LIST_HEAD(&sbi
->s_fc_q
[FC_Q_STAGING
]);
4748 INIT_LIST_HEAD(&sbi
->s_fc_dentry_q
[FC_Q_MAIN
]);
4749 INIT_LIST_HEAD(&sbi
->s_fc_dentry_q
[FC_Q_STAGING
]);
4750 sbi
->s_fc_bytes
= 0;
4751 ext4_clear_mount_flag(sb
, EXT4_MF_FC_INELIGIBLE
);
4752 ext4_clear_mount_flag(sb
, EXT4_MF_FC_COMMITTING
);
4753 spin_lock_init(&sbi
->s_fc_lock
);
4754 memset(&sbi
->s_fc_stats
, 0, sizeof(sbi
->s_fc_stats
));
4755 sbi
->s_fc_replay_state
.fc_regions
= NULL
;
4756 sbi
->s_fc_replay_state
.fc_regions_size
= 0;
4757 sbi
->s_fc_replay_state
.fc_regions_used
= 0;
4758 sbi
->s_fc_replay_state
.fc_regions_valid
= 0;
4759 sbi
->s_fc_replay_state
.fc_modified_inodes
= NULL
;
4760 sbi
->s_fc_replay_state
.fc_modified_inodes_size
= 0;
4761 sbi
->s_fc_replay_state
.fc_modified_inodes_used
= 0;
4765 needs_recovery
= (es
->s_last_orphan
!= 0 ||
4766 ext4_has_feature_journal_needs_recovery(sb
));
4768 if (ext4_has_feature_mmp(sb
) && !sb_rdonly(sb
))
4769 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
4770 goto failed_mount3a
;
4773 * The first inode we look at is the journal inode. Don't try
4774 * root first: it may be modified in the journal!
4776 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
4777 err
= ext4_load_journal(sb
, es
, journal_devnum
);
4779 goto failed_mount3a
;
4780 } else if (test_opt(sb
, NOLOAD
) && !sb_rdonly(sb
) &&
4781 ext4_has_feature_journal_needs_recovery(sb
)) {
4782 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
4783 "suppressed and not mounted read-only");
4784 goto failed_mount_wq
;
4786 /* Nojournal mode, all journal mount options are illegal */
4787 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
4788 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4789 "journal_checksum, fs mounted w/o journal");
4790 goto failed_mount_wq
;
4792 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4793 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4794 "journal_async_commit, fs mounted w/o journal");
4795 goto failed_mount_wq
;
4797 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
4798 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4799 "commit=%lu, fs mounted w/o journal",
4800 sbi
->s_commit_interval
/ HZ
);
4801 goto failed_mount_wq
;
4803 if (EXT4_MOUNT_DATA_FLAGS
&
4804 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
4805 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4806 "data=, fs mounted w/o journal");
4807 goto failed_mount_wq
;
4809 sbi
->s_def_mount_opt
&= ~EXT4_MOUNT_JOURNAL_CHECKSUM
;
4810 clear_opt(sb
, JOURNAL_CHECKSUM
);
4811 clear_opt(sb
, DATA_FLAGS
);
4812 clear_opt2(sb
, JOURNAL_FAST_COMMIT
);
4813 sbi
->s_journal
= NULL
;
4818 if (ext4_has_feature_64bit(sb
) &&
4819 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4820 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4821 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4822 goto failed_mount_wq
;
4825 if (!set_journal_csum_feature_set(sb
)) {
4826 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4828 goto failed_mount_wq
;
4831 if (test_opt2(sb
, JOURNAL_FAST_COMMIT
) &&
4832 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4833 JBD2_FEATURE_INCOMPAT_FAST_COMMIT
)) {
4834 ext4_msg(sb
, KERN_ERR
,
4835 "Failed to set fast commit journal feature");
4836 goto failed_mount_wq
;
4839 /* We have now updated the journal if required, so we can
4840 * validate the data journaling mode. */
4841 switch (test_opt(sb
, DATA_FLAGS
)) {
4843 /* No mode set, assume a default based on the journal
4844 * capabilities: ORDERED_DATA if the journal can
4845 * cope, else JOURNAL_DATA
4847 if (jbd2_journal_check_available_features
4848 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4849 set_opt(sb
, ORDERED_DATA
);
4850 sbi
->s_def_mount_opt
|= EXT4_MOUNT_ORDERED_DATA
;
4852 set_opt(sb
, JOURNAL_DATA
);
4853 sbi
->s_def_mount_opt
|= EXT4_MOUNT_JOURNAL_DATA
;
4857 case EXT4_MOUNT_ORDERED_DATA
:
4858 case EXT4_MOUNT_WRITEBACK_DATA
:
4859 if (!jbd2_journal_check_available_features
4860 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4861 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4862 "requested data journaling mode");
4863 goto failed_mount_wq
;
4870 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4871 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4872 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4873 "journal_async_commit in data=ordered mode");
4874 goto failed_mount_wq
;
4877 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4879 sbi
->s_journal
->j_submit_inode_data_buffers
=
4880 ext4_journal_submit_inode_data_buffers
;
4881 sbi
->s_journal
->j_finish_inode_data_buffers
=
4882 ext4_journal_finish_inode_data_buffers
;
4885 if (!test_opt(sb
, NO_MBCACHE
)) {
4886 sbi
->s_ea_block_cache
= ext4_xattr_create_cache();
4887 if (!sbi
->s_ea_block_cache
) {
4888 ext4_msg(sb
, KERN_ERR
,
4889 "Failed to create ea_block_cache");
4890 goto failed_mount_wq
;
4893 if (ext4_has_feature_ea_inode(sb
)) {
4894 sbi
->s_ea_inode_cache
= ext4_xattr_create_cache();
4895 if (!sbi
->s_ea_inode_cache
) {
4896 ext4_msg(sb
, KERN_ERR
,
4897 "Failed to create ea_inode_cache");
4898 goto failed_mount_wq
;
4903 if (ext4_has_feature_verity(sb
) && blocksize
!= PAGE_SIZE
) {
4904 ext4_msg(sb
, KERN_ERR
, "Unsupported blocksize for fs-verity");
4905 goto failed_mount_wq
;
4908 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !sb_rdonly(sb
) &&
4909 !ext4_has_feature_encrypt(sb
)) {
4910 ext4_set_feature_encrypt(sb
);
4911 ext4_commit_super(sb
);
4915 * Get the # of file system overhead blocks from the
4916 * superblock if present.
4918 if (es
->s_overhead_clusters
)
4919 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4921 err
= ext4_calculate_overhead(sb
);
4923 goto failed_mount_wq
;
4927 * The maximum number of concurrent works can be high and
4928 * concurrency isn't really necessary. Limit it to 1.
4930 EXT4_SB(sb
)->rsv_conversion_wq
=
4931 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4932 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4933 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4939 * The jbd2_journal_load will have done any necessary log recovery,
4940 * so we can safely mount the rest of the filesystem now.
4943 root
= ext4_iget(sb
, EXT4_ROOT_INO
, EXT4_IGET_SPECIAL
);
4945 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4946 ret
= PTR_ERR(root
);
4950 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4951 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4956 sb
->s_root
= d_make_root(root
);
4958 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4963 ret
= ext4_setup_super(sb
, es
, sb_rdonly(sb
));
4964 if (ret
== -EROFS
) {
4965 sb
->s_flags
|= SB_RDONLY
;
4968 goto failed_mount4a
;
4970 ext4_set_resv_clusters(sb
);
4972 if (test_opt(sb
, BLOCK_VALIDITY
)) {
4973 err
= ext4_setup_system_zone(sb
);
4975 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4977 goto failed_mount4a
;
4980 ext4_fc_replay_cleanup(sb
);
4983 err
= ext4_mb_init(sb
);
4985 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4991 * We can only set up the journal commit callback once
4992 * mballoc is initialized
4995 sbi
->s_journal
->j_commit_callback
=
4996 ext4_journal_commit_callback
;
4998 block
= ext4_count_free_clusters(sb
);
4999 ext4_free_blocks_count_set(sbi
->s_es
,
5000 EXT4_C2B(sbi
, block
));
5001 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
5004 unsigned long freei
= ext4_count_free_inodes(sb
);
5005 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
5006 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
5010 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
5011 ext4_count_dirs(sb
), GFP_KERNEL
);
5013 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
5016 err
= percpu_counter_init(&sbi
->s_sra_exceeded_retry_limit
, 0,
5019 err
= percpu_init_rwsem(&sbi
->s_writepages_rwsem
);
5022 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
5026 if (ext4_has_feature_flex_bg(sb
))
5027 if (!ext4_fill_flex_info(sb
)) {
5028 ext4_msg(sb
, KERN_ERR
,
5029 "unable to initialize "
5030 "flex_bg meta info!");
5034 err
= ext4_register_li_request(sb
, first_not_zeroed
);
5038 err
= ext4_register_sysfs(sb
);
5043 /* Enable quota usage during mount. */
5044 if (ext4_has_feature_quota(sb
) && !sb_rdonly(sb
)) {
5045 err
= ext4_enable_quotas(sb
);
5049 #endif /* CONFIG_QUOTA */
5052 * Save the original bdev mapping's wb_err value which could be
5053 * used to detect the metadata async write error.
5055 spin_lock_init(&sbi
->s_bdev_wb_lock
);
5056 errseq_check_and_advance(&sb
->s_bdev
->bd_inode
->i_mapping
->wb_err
,
5057 &sbi
->s_bdev_wb_err
);
5058 sb
->s_bdev
->bd_super
= sb
;
5059 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
5060 ext4_orphan_cleanup(sb
, es
);
5061 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
5062 if (needs_recovery
) {
5063 ext4_msg(sb
, KERN_INFO
, "recovery complete");
5064 err
= ext4_mark_recovery_complete(sb
, es
);
5068 if (EXT4_SB(sb
)->s_journal
) {
5069 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
5070 descr
= " journalled data mode";
5071 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
5072 descr
= " ordered data mode";
5074 descr
= " writeback data mode";
5076 descr
= "out journal";
5078 if (test_opt(sb
, DISCARD
)) {
5079 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
5080 if (!blk_queue_discard(q
))
5081 ext4_msg(sb
, KERN_WARNING
,
5082 "mounting with \"discard\" option, but "
5083 "the device does not support discard");
5086 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
5087 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
5088 "Opts: %.*s%s%s. Quota mode: %s.", descr
,
5089 (int) sizeof(sbi
->s_es
->s_mount_opts
),
5090 sbi
->s_es
->s_mount_opts
,
5091 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
,
5092 ext4_quota_mode(sb
));
5094 if (es
->s_error_count
)
5095 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
5097 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5098 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
5099 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
5100 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
5101 atomic_set(&sbi
->s_warning_count
, 0);
5102 atomic_set(&sbi
->s_msg_count
, 0);
5109 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
5113 ext4_unregister_sysfs(sb
);
5114 kobject_put(&sbi
->s_kobj
);
5116 ext4_unregister_li_request(sb
);
5118 ext4_mb_release(sb
);
5120 flex_groups
= rcu_dereference(sbi
->s_flex_groups
);
5122 for (i
= 0; i
< sbi
->s_flex_groups_allocated
; i
++)
5123 kvfree(flex_groups
[i
]);
5124 kvfree(flex_groups
);
5127 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
5128 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
5129 percpu_counter_destroy(&sbi
->s_dirs_counter
);
5130 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
5131 percpu_counter_destroy(&sbi
->s_sra_exceeded_retry_limit
);
5132 percpu_free_rwsem(&sbi
->s_writepages_rwsem
);
5134 ext4_ext_release(sb
);
5135 ext4_release_system_zone(sb
);
5140 ext4_msg(sb
, KERN_ERR
, "mount failed");
5141 if (EXT4_SB(sb
)->rsv_conversion_wq
)
5142 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
5144 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
5145 sbi
->s_ea_inode_cache
= NULL
;
5147 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
5148 sbi
->s_ea_block_cache
= NULL
;
5150 if (sbi
->s_journal
) {
5151 jbd2_journal_destroy(sbi
->s_journal
);
5152 sbi
->s_journal
= NULL
;
5155 ext4_es_unregister_shrinker(sbi
);
5157 flush_work(&sbi
->s_error_work
);
5158 del_timer_sync(&sbi
->s_err_report
);
5160 kthread_stop(sbi
->s_mmp_tsk
);
5163 group_desc
= rcu_dereference(sbi
->s_group_desc
);
5164 for (i
= 0; i
< db_count
; i
++)
5165 brelse(group_desc
[i
]);
5169 if (sbi
->s_chksum_driver
)
5170 crypto_free_shash(sbi
->s_chksum_driver
);
5172 #ifdef CONFIG_UNICODE
5173 utf8_unload(sb
->s_encoding
);
5177 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5178 kfree(get_qf_name(sb
, sbi
, i
));
5180 fscrypt_free_dummy_policy(&sbi
->s_dummy_enc_policy
);
5181 ext4_blkdev_remove(sbi
);
5184 sb
->s_fs_info
= NULL
;
5185 kfree(sbi
->s_blockgroup_lock
);
5189 fs_put_dax(dax_dev
);
5190 return err
? err
: ret
;
5194 * Setup any per-fs journal parameters now. We'll do this both on
5195 * initial mount, once the journal has been initialised but before we've
5196 * done any recovery; and again on any subsequent remount.
5198 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
5200 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5202 journal
->j_commit_interval
= sbi
->s_commit_interval
;
5203 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
5204 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
5205 ext4_fc_init(sb
, journal
);
5207 write_lock(&journal
->j_state_lock
);
5208 if (test_opt(sb
, BARRIER
))
5209 journal
->j_flags
|= JBD2_BARRIER
;
5211 journal
->j_flags
&= ~JBD2_BARRIER
;
5212 if (test_opt(sb
, DATA_ERR_ABORT
))
5213 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
5215 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
5216 write_unlock(&journal
->j_state_lock
);
5219 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
5220 unsigned int journal_inum
)
5222 struct inode
*journal_inode
;
5225 * Test for the existence of a valid inode on disk. Bad things
5226 * happen if we iget() an unused inode, as the subsequent iput()
5227 * will try to delete it.
5229 journal_inode
= ext4_iget(sb
, journal_inum
, EXT4_IGET_SPECIAL
);
5230 if (IS_ERR(journal_inode
)) {
5231 ext4_msg(sb
, KERN_ERR
, "no journal found");
5234 if (!journal_inode
->i_nlink
) {
5235 make_bad_inode(journal_inode
);
5236 iput(journal_inode
);
5237 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
5241 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5242 journal_inode
, journal_inode
->i_size
);
5243 if (!S_ISREG(journal_inode
->i_mode
)) {
5244 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
5245 iput(journal_inode
);
5248 return journal_inode
;
5251 static journal_t
*ext4_get_journal(struct super_block
*sb
,
5252 unsigned int journal_inum
)
5254 struct inode
*journal_inode
;
5257 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb
)))
5260 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
5264 journal
= jbd2_journal_init_inode(journal_inode
);
5266 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
5267 iput(journal_inode
);
5270 journal
->j_private
= sb
;
5271 ext4_init_journal_params(sb
, journal
);
5275 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
5278 struct buffer_head
*bh
;
5282 int hblock
, blocksize
;
5283 ext4_fsblk_t sb_block
;
5284 unsigned long offset
;
5285 struct ext4_super_block
*es
;
5286 struct block_device
*bdev
;
5288 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb
)))
5291 bdev
= ext4_blkdev_get(j_dev
, sb
);
5295 blocksize
= sb
->s_blocksize
;
5296 hblock
= bdev_logical_block_size(bdev
);
5297 if (blocksize
< hblock
) {
5298 ext4_msg(sb
, KERN_ERR
,
5299 "blocksize too small for journal device");
5303 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
5304 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
5305 set_blocksize(bdev
, blocksize
);
5306 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
5307 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
5308 "external journal");
5312 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
5313 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
5314 !(le32_to_cpu(es
->s_feature_incompat
) &
5315 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
5316 ext4_msg(sb
, KERN_ERR
, "external journal has "
5322 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
5323 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
5324 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
5325 ext4_msg(sb
, KERN_ERR
, "external journal has "
5326 "corrupt superblock");
5331 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
5332 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
5337 len
= ext4_blocks_count(es
);
5338 start
= sb_block
+ 1;
5339 brelse(bh
); /* we're done with the superblock */
5341 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
5342 start
, len
, blocksize
);
5344 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
5347 journal
->j_private
= sb
;
5348 if (ext4_read_bh_lock(journal
->j_sb_buffer
, REQ_META
| REQ_PRIO
, true)) {
5349 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
5352 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
5353 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
5354 "user (unsupported) - %d",
5355 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
5358 EXT4_SB(sb
)->s_journal_bdev
= bdev
;
5359 ext4_init_journal_params(sb
, journal
);
5363 jbd2_journal_destroy(journal
);
5365 ext4_blkdev_put(bdev
);
5369 static int ext4_load_journal(struct super_block
*sb
,
5370 struct ext4_super_block
*es
,
5371 unsigned long journal_devnum
)
5374 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
5377 int really_read_only
;
5380 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb
)))
5381 return -EFSCORRUPTED
;
5383 if (journal_devnum
&&
5384 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
5385 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
5386 "numbers have changed");
5387 journal_dev
= new_decode_dev(journal_devnum
);
5389 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
5391 if (journal_inum
&& journal_dev
) {
5392 ext4_msg(sb
, KERN_ERR
,
5393 "filesystem has both journal inode and journal device!");
5398 journal
= ext4_get_journal(sb
, journal_inum
);
5402 journal
= ext4_get_dev_journal(sb
, journal_dev
);
5407 journal_dev_ro
= bdev_read_only(journal
->j_dev
);
5408 really_read_only
= bdev_read_only(sb
->s_bdev
) | journal_dev_ro
;
5410 if (journal_dev_ro
&& !sb_rdonly(sb
)) {
5411 ext4_msg(sb
, KERN_ERR
,
5412 "journal device read-only, try mounting with '-o ro'");
5418 * Are we loading a blank journal or performing recovery after a
5419 * crash? For recovery, we need to check in advance whether we
5420 * can get read-write access to the device.
5422 if (ext4_has_feature_journal_needs_recovery(sb
)) {
5423 if (sb_rdonly(sb
)) {
5424 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
5425 "required on readonly filesystem");
5426 if (really_read_only
) {
5427 ext4_msg(sb
, KERN_ERR
, "write access "
5428 "unavailable, cannot proceed "
5429 "(try mounting with noload)");
5433 ext4_msg(sb
, KERN_INFO
, "write access will "
5434 "be enabled during recovery");
5438 if (!(journal
->j_flags
& JBD2_BARRIER
))
5439 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
5441 if (!ext4_has_feature_journal_needs_recovery(sb
))
5442 err
= jbd2_journal_wipe(journal
, !really_read_only
);
5444 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
5446 memcpy(save
, ((char *) es
) +
5447 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
5448 err
= jbd2_journal_load(journal
);
5450 memcpy(((char *) es
) + EXT4_S_ERR_START
,
5451 save
, EXT4_S_ERR_LEN
);
5456 ext4_msg(sb
, KERN_ERR
, "error loading journal");
5460 EXT4_SB(sb
)->s_journal
= journal
;
5461 err
= ext4_clear_journal_err(sb
, es
);
5463 EXT4_SB(sb
)->s_journal
= NULL
;
5464 jbd2_journal_destroy(journal
);
5468 if (!really_read_only
&& journal_devnum
&&
5469 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
5470 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
5472 /* Make sure we flush the recovery flag to disk. */
5473 ext4_commit_super(sb
);
5479 jbd2_journal_destroy(journal
);
5483 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5484 static void ext4_update_super(struct super_block
*sb
)
5486 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5487 struct ext4_super_block
*es
= sbi
->s_es
;
5488 struct buffer_head
*sbh
= sbi
->s_sbh
;
5492 * If the file system is mounted read-only, don't update the
5493 * superblock write time. This avoids updating the superblock
5494 * write time when we are mounting the root file system
5495 * read/only but we need to replay the journal; at that point,
5496 * for people who are east of GMT and who make their clock
5497 * tick in localtime for Windows bug-for-bug compatibility,
5498 * the clock is set in the future, and this will cause e2fsck
5499 * to complain and force a full file system check.
5501 if (!(sb
->s_flags
& SB_RDONLY
))
5502 ext4_update_tstamp(es
, s_wtime
);
5503 es
->s_kbytes_written
=
5504 cpu_to_le64(sbi
->s_kbytes_written
+
5505 ((part_stat_read(sb
->s_bdev
, sectors
[STAT_WRITE
]) -
5506 sbi
->s_sectors_written_start
) >> 1));
5507 if (percpu_counter_initialized(&sbi
->s_freeclusters_counter
))
5508 ext4_free_blocks_count_set(es
,
5509 EXT4_C2B(sbi
, percpu_counter_sum_positive(
5510 &sbi
->s_freeclusters_counter
)));
5511 if (percpu_counter_initialized(&sbi
->s_freeinodes_counter
))
5512 es
->s_free_inodes_count
=
5513 cpu_to_le32(percpu_counter_sum_positive(
5514 &sbi
->s_freeinodes_counter
));
5515 /* Copy error information to the on-disk superblock */
5516 spin_lock(&sbi
->s_error_lock
);
5517 if (sbi
->s_add_error_count
> 0) {
5518 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
5519 if (!es
->s_first_error_time
&& !es
->s_first_error_time_hi
) {
5520 __ext4_update_tstamp(&es
->s_first_error_time
,
5521 &es
->s_first_error_time_hi
,
5522 sbi
->s_first_error_time
);
5523 strncpy(es
->s_first_error_func
, sbi
->s_first_error_func
,
5524 sizeof(es
->s_first_error_func
));
5525 es
->s_first_error_line
=
5526 cpu_to_le32(sbi
->s_first_error_line
);
5527 es
->s_first_error_ino
=
5528 cpu_to_le32(sbi
->s_first_error_ino
);
5529 es
->s_first_error_block
=
5530 cpu_to_le64(sbi
->s_first_error_block
);
5531 es
->s_first_error_errcode
=
5532 ext4_errno_to_code(sbi
->s_first_error_code
);
5534 __ext4_update_tstamp(&es
->s_last_error_time
,
5535 &es
->s_last_error_time_hi
,
5536 sbi
->s_last_error_time
);
5537 strncpy(es
->s_last_error_func
, sbi
->s_last_error_func
,
5538 sizeof(es
->s_last_error_func
));
5539 es
->s_last_error_line
= cpu_to_le32(sbi
->s_last_error_line
);
5540 es
->s_last_error_ino
= cpu_to_le32(sbi
->s_last_error_ino
);
5541 es
->s_last_error_block
= cpu_to_le64(sbi
->s_last_error_block
);
5542 es
->s_last_error_errcode
=
5543 ext4_errno_to_code(sbi
->s_last_error_code
);
5545 * Start the daily error reporting function if it hasn't been
5548 if (!es
->s_error_count
)
5549 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
);
5550 le32_add_cpu(&es
->s_error_count
, sbi
->s_add_error_count
);
5551 sbi
->s_add_error_count
= 0;
5553 spin_unlock(&sbi
->s_error_lock
);
5555 ext4_superblock_csum_set(sb
);
5559 static int ext4_commit_super(struct super_block
*sb
)
5561 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
5564 if (!sbh
|| block_device_ejected(sb
))
5567 ext4_update_super(sb
);
5569 if (buffer_write_io_error(sbh
) || !buffer_uptodate(sbh
)) {
5571 * Oh, dear. A previous attempt to write the
5572 * superblock failed. This could happen because the
5573 * USB device was yanked out. Or it could happen to
5574 * be a transient write error and maybe the block will
5575 * be remapped. Nothing we can do but to retry the
5576 * write and hope for the best.
5578 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
5579 "superblock detected");
5580 clear_buffer_write_io_error(sbh
);
5581 set_buffer_uptodate(sbh
);
5583 BUFFER_TRACE(sbh
, "marking dirty");
5584 mark_buffer_dirty(sbh
);
5585 error
= __sync_dirty_buffer(sbh
,
5586 REQ_SYNC
| (test_opt(sb
, BARRIER
) ? REQ_FUA
: 0));
5587 if (buffer_write_io_error(sbh
)) {
5588 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
5590 clear_buffer_write_io_error(sbh
);
5591 set_buffer_uptodate(sbh
);
5597 * Have we just finished recovery? If so, and if we are mounting (or
5598 * remounting) the filesystem readonly, then we will end up with a
5599 * consistent fs on disk. Record that fact.
5601 static int ext4_mark_recovery_complete(struct super_block
*sb
,
5602 struct ext4_super_block
*es
)
5605 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
5607 if (!ext4_has_feature_journal(sb
)) {
5608 if (journal
!= NULL
) {
5609 ext4_error(sb
, "Journal got removed while the fs was "
5611 return -EFSCORRUPTED
;
5615 jbd2_journal_lock_updates(journal
);
5616 err
= jbd2_journal_flush(journal
);
5620 if (ext4_has_feature_journal_needs_recovery(sb
) && sb_rdonly(sb
)) {
5621 ext4_clear_feature_journal_needs_recovery(sb
);
5622 ext4_commit_super(sb
);
5625 jbd2_journal_unlock_updates(journal
);
5630 * If we are mounting (or read-write remounting) a filesystem whose journal
5631 * has recorded an error from a previous lifetime, move that error to the
5632 * main filesystem now.
5634 static int ext4_clear_journal_err(struct super_block
*sb
,
5635 struct ext4_super_block
*es
)
5641 if (!ext4_has_feature_journal(sb
)) {
5642 ext4_error(sb
, "Journal got removed while the fs was mounted!");
5643 return -EFSCORRUPTED
;
5646 journal
= EXT4_SB(sb
)->s_journal
;
5649 * Now check for any error status which may have been recorded in the
5650 * journal by a prior ext4_error() or ext4_abort()
5653 j_errno
= jbd2_journal_errno(journal
);
5657 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
5658 ext4_warning(sb
, "Filesystem error recorded "
5659 "from previous mount: %s", errstr
);
5660 ext4_warning(sb
, "Marking fs in need of filesystem check.");
5662 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
5663 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
5664 ext4_commit_super(sb
);
5666 jbd2_journal_clear_err(journal
);
5667 jbd2_journal_update_sb_errno(journal
);
5673 * Force the running and committing transactions to commit,
5674 * and wait on the commit.
5676 int ext4_force_commit(struct super_block
*sb
)
5683 journal
= EXT4_SB(sb
)->s_journal
;
5684 return ext4_journal_force_commit(journal
);
5687 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
5691 bool needs_barrier
= false;
5692 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5694 if (unlikely(ext4_forced_shutdown(sbi
)))
5697 trace_ext4_sync_fs(sb
, wait
);
5698 flush_workqueue(sbi
->rsv_conversion_wq
);
5700 * Writeback quota in non-journalled quota case - journalled quota has
5703 dquot_writeback_dquots(sb
, -1);
5705 * Data writeback is possible w/o journal transaction, so barrier must
5706 * being sent at the end of the function. But we can skip it if
5707 * transaction_commit will do it for us.
5709 if (sbi
->s_journal
) {
5710 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
5711 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
5712 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
5713 needs_barrier
= true;
5715 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
5717 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
5720 } else if (wait
&& test_opt(sb
, BARRIER
))
5721 needs_barrier
= true;
5722 if (needs_barrier
) {
5724 err
= blkdev_issue_flush(sb
->s_bdev
);
5733 * LVM calls this function before a (read-only) snapshot is created. This
5734 * gives us a chance to flush the journal completely and mark the fs clean.
5736 * Note that only this function cannot bring a filesystem to be in a clean
5737 * state independently. It relies on upper layer to stop all data & metadata
5740 static int ext4_freeze(struct super_block
*sb
)
5748 journal
= EXT4_SB(sb
)->s_journal
;
5751 /* Now we set up the journal barrier. */
5752 jbd2_journal_lock_updates(journal
);
5755 * Don't clear the needs_recovery flag if we failed to
5756 * flush the journal.
5758 error
= jbd2_journal_flush(journal
);
5762 /* Journal blocked and flushed, clear needs_recovery flag. */
5763 ext4_clear_feature_journal_needs_recovery(sb
);
5766 error
= ext4_commit_super(sb
);
5769 /* we rely on upper layer to stop further updates */
5770 jbd2_journal_unlock_updates(journal
);
5775 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5776 * flag here, even though the filesystem is not technically dirty yet.
5778 static int ext4_unfreeze(struct super_block
*sb
)
5780 if (sb_rdonly(sb
) || ext4_forced_shutdown(EXT4_SB(sb
)))
5783 if (EXT4_SB(sb
)->s_journal
) {
5784 /* Reset the needs_recovery flag before the fs is unlocked. */
5785 ext4_set_feature_journal_needs_recovery(sb
);
5788 ext4_commit_super(sb
);
5793 * Structure to save mount options for ext4_remount's benefit
5795 struct ext4_mount_options
{
5796 unsigned long s_mount_opt
;
5797 unsigned long s_mount_opt2
;
5800 unsigned long s_commit_interval
;
5801 u32 s_min_batch_time
, s_max_batch_time
;
5804 char *s_qf_names
[EXT4_MAXQUOTAS
];
5808 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
5810 struct ext4_super_block
*es
;
5811 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5812 unsigned long old_sb_flags
, vfs_flags
;
5813 struct ext4_mount_options old_opts
;
5814 int enable_quota
= 0;
5816 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
5820 char *to_free
[EXT4_MAXQUOTAS
];
5822 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
5824 if (data
&& !orig_data
)
5827 /* Store the original options */
5828 old_sb_flags
= sb
->s_flags
;
5829 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
5830 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
5831 old_opts
.s_resuid
= sbi
->s_resuid
;
5832 old_opts
.s_resgid
= sbi
->s_resgid
;
5833 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
5834 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
5835 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
5837 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
5838 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5839 if (sbi
->s_qf_names
[i
]) {
5840 char *qf_name
= get_qf_name(sb
, sbi
, i
);
5842 old_opts
.s_qf_names
[i
] = kstrdup(qf_name
, GFP_KERNEL
);
5843 if (!old_opts
.s_qf_names
[i
]) {
5844 for (j
= 0; j
< i
; j
++)
5845 kfree(old_opts
.s_qf_names
[j
]);
5850 old_opts
.s_qf_names
[i
] = NULL
;
5852 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
5853 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
5856 * Some options can be enabled by ext4 and/or by VFS mount flag
5857 * either way we need to make sure it matches in both *flags and
5858 * s_flags. Copy those selected flags from *flags to s_flags
5860 vfs_flags
= SB_LAZYTIME
| SB_I_VERSION
;
5861 sb
->s_flags
= (sb
->s_flags
& ~vfs_flags
) | (*flags
& vfs_flags
);
5863 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
5868 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
5869 test_opt(sb
, JOURNAL_CHECKSUM
)) {
5870 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
5871 "during remount not supported; ignoring");
5872 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
5875 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
5876 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
5877 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5878 "both data=journal and delalloc");
5882 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
5883 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5884 "both data=journal and dioread_nolock");
5888 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
5889 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
5890 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5891 "journal_async_commit in data=ordered mode");
5897 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_NO_MBCACHE
) {
5898 ext4_msg(sb
, KERN_ERR
, "can't enable nombcache during remount");
5903 if (ext4_test_mount_flag(sb
, EXT4_MF_FS_ABORTED
))
5904 ext4_abort(sb
, EXT4_ERR_ESHUTDOWN
, "Abort forced by user");
5906 sb
->s_flags
= (sb
->s_flags
& ~SB_POSIXACL
) |
5907 (test_opt(sb
, POSIX_ACL
) ? SB_POSIXACL
: 0);
5911 if (sbi
->s_journal
) {
5912 ext4_init_journal_params(sb
, sbi
->s_journal
);
5913 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
5916 /* Flush outstanding errors before changing fs state */
5917 flush_work(&sbi
->s_error_work
);
5919 if ((bool)(*flags
& SB_RDONLY
) != sb_rdonly(sb
)) {
5920 if (ext4_test_mount_flag(sb
, EXT4_MF_FS_ABORTED
)) {
5925 if (*flags
& SB_RDONLY
) {
5926 err
= sync_filesystem(sb
);
5929 err
= dquot_suspend(sb
, -1);
5934 * First of all, the unconditional stuff we have to do
5935 * to disable replay of the journal when we next remount
5937 sb
->s_flags
|= SB_RDONLY
;
5940 * OK, test if we are remounting a valid rw partition
5941 * readonly, and if so set the rdonly flag and then
5942 * mark the partition as valid again.
5944 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5945 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5946 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5948 if (sbi
->s_journal
) {
5950 * We let remount-ro finish even if marking fs
5951 * as clean failed...
5953 ext4_mark_recovery_complete(sb
, es
);
5956 kthread_stop(sbi
->s_mmp_tsk
);
5958 /* Make sure we can mount this feature set readwrite */
5959 if (ext4_has_feature_readonly(sb
) ||
5960 !ext4_feature_set_ok(sb
, 0)) {
5965 * Make sure the group descriptor checksums
5966 * are sane. If they aren't, refuse to remount r/w.
5968 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5969 struct ext4_group_desc
*gdp
=
5970 ext4_get_group_desc(sb
, g
, NULL
);
5972 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5973 ext4_msg(sb
, KERN_ERR
,
5974 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5975 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5976 le16_to_cpu(gdp
->bg_checksum
));
5983 * If we have an unprocessed orphan list hanging
5984 * around from a previously readonly bdev mount,
5985 * require a full umount/remount for now.
5987 if (es
->s_last_orphan
) {
5988 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5989 "remount RDWR because of unprocessed "
5990 "orphan inode list. Please "
5991 "umount/remount instead");
5997 * Mounting a RDONLY partition read-write, so reread
5998 * and store the current valid flag. (It may have
5999 * been changed by e2fsck since we originally mounted
6002 if (sbi
->s_journal
) {
6003 err
= ext4_clear_journal_err(sb
, es
);
6007 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
6009 err
= ext4_setup_super(sb
, es
, 0);
6013 sb
->s_flags
&= ~SB_RDONLY
;
6014 if (ext4_has_feature_mmp(sb
))
6015 if (ext4_multi_mount_protect(sb
,
6016 le64_to_cpu(es
->s_mmp_block
))) {
6025 * Reinitialize lazy itable initialization thread based on
6028 if (sb_rdonly(sb
) || !test_opt(sb
, INIT_INODE_TABLE
))
6029 ext4_unregister_li_request(sb
);
6031 ext4_group_t first_not_zeroed
;
6032 first_not_zeroed
= ext4_has_uninit_itable(sb
);
6033 ext4_register_li_request(sb
, first_not_zeroed
);
6037 * Handle creation of system zone data early because it can fail.
6038 * Releasing of existing data is done when we are sure remount will
6041 if (test_opt(sb
, BLOCK_VALIDITY
) && !sbi
->s_system_blks
) {
6042 err
= ext4_setup_system_zone(sb
);
6047 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& SB_RDONLY
)) {
6048 err
= ext4_commit_super(sb
);
6054 /* Release old quota file names */
6055 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
6056 kfree(old_opts
.s_qf_names
[i
]);
6058 if (sb_any_quota_suspended(sb
))
6059 dquot_resume(sb
, -1);
6060 else if (ext4_has_feature_quota(sb
)) {
6061 err
= ext4_enable_quotas(sb
);
6067 if (!test_opt(sb
, BLOCK_VALIDITY
) && sbi
->s_system_blks
)
6068 ext4_release_system_zone(sb
);
6071 * Some options can be enabled by ext4 and/or by VFS mount flag
6072 * either way we need to make sure it matches in both *flags and
6073 * s_flags. Copy those selected flags from s_flags to *flags
6075 *flags
= (*flags
& ~vfs_flags
) | (sb
->s_flags
& vfs_flags
);
6077 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s. Quota mode: %s.",
6078 orig_data
, ext4_quota_mode(sb
));
6083 sb
->s_flags
= old_sb_flags
;
6084 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
6085 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
6086 sbi
->s_resuid
= old_opts
.s_resuid
;
6087 sbi
->s_resgid
= old_opts
.s_resgid
;
6088 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
6089 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
6090 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
6091 if (!test_opt(sb
, BLOCK_VALIDITY
) && sbi
->s_system_blks
)
6092 ext4_release_system_zone(sb
);
6094 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
6095 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
6096 to_free
[i
] = get_qf_name(sb
, sbi
, i
);
6097 rcu_assign_pointer(sbi
->s_qf_names
[i
], old_opts
.s_qf_names
[i
]);
6100 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
6108 static int ext4_statfs_project(struct super_block
*sb
,
6109 kprojid_t projid
, struct kstatfs
*buf
)
6112 struct dquot
*dquot
;
6116 qid
= make_kqid_projid(projid
);
6117 dquot
= dqget(sb
, qid
);
6119 return PTR_ERR(dquot
);
6120 spin_lock(&dquot
->dq_dqb_lock
);
6122 limit
= min_not_zero(dquot
->dq_dqb
.dqb_bsoftlimit
,
6123 dquot
->dq_dqb
.dqb_bhardlimit
);
6124 limit
>>= sb
->s_blocksize_bits
;
6126 if (limit
&& buf
->f_blocks
> limit
) {
6127 curblock
= (dquot
->dq_dqb
.dqb_curspace
+
6128 dquot
->dq_dqb
.dqb_rsvspace
) >> sb
->s_blocksize_bits
;
6129 buf
->f_blocks
= limit
;
6130 buf
->f_bfree
= buf
->f_bavail
=
6131 (buf
->f_blocks
> curblock
) ?
6132 (buf
->f_blocks
- curblock
) : 0;
6135 limit
= min_not_zero(dquot
->dq_dqb
.dqb_isoftlimit
,
6136 dquot
->dq_dqb
.dqb_ihardlimit
);
6137 if (limit
&& buf
->f_files
> limit
) {
6138 buf
->f_files
= limit
;
6140 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
6141 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
6144 spin_unlock(&dquot
->dq_dqb_lock
);
6150 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
6152 struct super_block
*sb
= dentry
->d_sb
;
6153 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
6154 struct ext4_super_block
*es
= sbi
->s_es
;
6155 ext4_fsblk_t overhead
= 0, resv_blocks
;
6158 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
6160 if (!test_opt(sb
, MINIX_DF
))
6161 overhead
= sbi
->s_overhead
;
6163 buf
->f_type
= EXT4_SUPER_MAGIC
;
6164 buf
->f_bsize
= sb
->s_blocksize
;
6165 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
6166 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
6167 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
6168 /* prevent underflow in case that few free space is available */
6169 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
6170 buf
->f_bavail
= buf
->f_bfree
-
6171 (ext4_r_blocks_count(es
) + resv_blocks
);
6172 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
6174 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
6175 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
6176 buf
->f_namelen
= EXT4_NAME_LEN
;
6177 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
6178 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
6179 buf
->f_fsid
= u64_to_fsid(fsid
);
6182 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
6183 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
6184 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
6193 * Helper functions so that transaction is started before we acquire dqio_sem
6194 * to keep correct lock ordering of transaction > dqio_sem
6196 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
6198 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
6201 static int ext4_write_dquot(struct dquot
*dquot
)
6205 struct inode
*inode
;
6207 inode
= dquot_to_inode(dquot
);
6208 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
6209 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
6211 return PTR_ERR(handle
);
6212 ret
= dquot_commit(dquot
);
6213 err
= ext4_journal_stop(handle
);
6219 static int ext4_acquire_dquot(struct dquot
*dquot
)
6224 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
6225 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
6227 return PTR_ERR(handle
);
6228 ret
= dquot_acquire(dquot
);
6229 err
= ext4_journal_stop(handle
);
6235 static int ext4_release_dquot(struct dquot
*dquot
)
6240 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
6241 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
6242 if (IS_ERR(handle
)) {
6243 /* Release dquot anyway to avoid endless cycle in dqput() */
6244 dquot_release(dquot
);
6245 return PTR_ERR(handle
);
6247 ret
= dquot_release(dquot
);
6248 err
= ext4_journal_stop(handle
);
6254 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
6256 struct super_block
*sb
= dquot
->dq_sb
;
6258 if (ext4_is_quota_journalled(sb
)) {
6259 dquot_mark_dquot_dirty(dquot
);
6260 return ext4_write_dquot(dquot
);
6262 return dquot_mark_dquot_dirty(dquot
);
6266 static int ext4_write_info(struct super_block
*sb
, int type
)
6271 /* Data block + inode block */
6272 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
6274 return PTR_ERR(handle
);
6275 ret
= dquot_commit_info(sb
, type
);
6276 err
= ext4_journal_stop(handle
);
6283 * Turn on quotas during mount time - we need to find
6284 * the quota file and such...
6286 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
6288 return dquot_quota_on_mount(sb
, get_qf_name(sb
, EXT4_SB(sb
), type
),
6289 EXT4_SB(sb
)->s_jquota_fmt
, type
);
6292 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
6294 struct ext4_inode_info
*ei
= EXT4_I(inode
);
6296 /* The first argument of lockdep_set_subclass has to be
6297 * *exactly* the same as the argument to init_rwsem() --- in
6298 * this case, in init_once() --- or lockdep gets unhappy
6299 * because the name of the lock is set using the
6300 * stringification of the argument to init_rwsem().
6302 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
6303 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
6307 * Standard function to be called on quota_on
6309 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
6310 const struct path
*path
)
6314 if (!test_opt(sb
, QUOTA
))
6317 /* Quotafile not on the same filesystem? */
6318 if (path
->dentry
->d_sb
!= sb
)
6321 /* Quota already enabled for this file? */
6322 if (IS_NOQUOTA(d_inode(path
->dentry
)))
6325 /* Journaling quota? */
6326 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
6327 /* Quotafile not in fs root? */
6328 if (path
->dentry
->d_parent
!= sb
->s_root
)
6329 ext4_msg(sb
, KERN_WARNING
,
6330 "Quota file not on filesystem root. "
6331 "Journaled quota will not work");
6332 sb_dqopt(sb
)->flags
|= DQUOT_NOLIST_DIRTY
;
6335 * Clear the flag just in case mount options changed since
6338 sb_dqopt(sb
)->flags
&= ~DQUOT_NOLIST_DIRTY
;
6342 * When we journal data on quota file, we have to flush journal to see
6343 * all updates to the file when we bypass pagecache...
6345 if (EXT4_SB(sb
)->s_journal
&&
6346 ext4_should_journal_data(d_inode(path
->dentry
))) {
6348 * We don't need to lock updates but journal_flush() could
6349 * otherwise be livelocked...
6351 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
6352 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
6353 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
6358 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
6359 err
= dquot_quota_on(sb
, type
, format_id
, path
);
6361 lockdep_set_quota_inode(path
->dentry
->d_inode
,
6364 struct inode
*inode
= d_inode(path
->dentry
);
6368 * Set inode flags to prevent userspace from messing with quota
6369 * files. If this fails, we return success anyway since quotas
6370 * are already enabled and this is not a hard failure.
6373 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
6376 EXT4_I(inode
)->i_flags
|= EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
;
6377 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
6378 S_NOATIME
| S_IMMUTABLE
);
6379 err
= ext4_mark_inode_dirty(handle
, inode
);
6380 ext4_journal_stop(handle
);
6382 inode_unlock(inode
);
6387 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
6391 struct inode
*qf_inode
;
6392 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
6393 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
6394 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
6395 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
6398 BUG_ON(!ext4_has_feature_quota(sb
));
6400 if (!qf_inums
[type
])
6403 qf_inode
= ext4_iget(sb
, qf_inums
[type
], EXT4_IGET_SPECIAL
);
6404 if (IS_ERR(qf_inode
)) {
6405 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
6406 return PTR_ERR(qf_inode
);
6409 /* Don't account quota for quota files to avoid recursion */
6410 qf_inode
->i_flags
|= S_NOQUOTA
;
6411 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
6412 err
= dquot_load_quota_inode(qf_inode
, type
, format_id
, flags
);
6414 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
6420 /* Enable usage tracking for all quota types. */
6421 static int ext4_enable_quotas(struct super_block
*sb
)
6424 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
6425 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
6426 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
6427 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
6429 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
6430 test_opt(sb
, USRQUOTA
),
6431 test_opt(sb
, GRPQUOTA
),
6432 test_opt(sb
, PRJQUOTA
),
6435 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NOLIST_DIRTY
;
6436 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
6437 if (qf_inums
[type
]) {
6438 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
6439 DQUOT_USAGE_ENABLED
|
6440 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
6443 "Failed to enable quota tracking "
6444 "(type=%d, err=%d). Please run "
6445 "e2fsck to fix.", type
, err
);
6446 for (type
--; type
>= 0; type
--)
6447 dquot_quota_off(sb
, type
);
6456 static int ext4_quota_off(struct super_block
*sb
, int type
)
6458 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
6462 /* Force all delayed allocation blocks to be allocated.
6463 * Caller already holds s_umount sem */
6464 if (test_opt(sb
, DELALLOC
))
6465 sync_filesystem(sb
);
6467 if (!inode
|| !igrab(inode
))
6470 err
= dquot_quota_off(sb
, type
);
6471 if (err
|| ext4_has_feature_quota(sb
))
6476 * Update modification times of quota files when userspace can
6477 * start looking at them. If we fail, we return success anyway since
6478 * this is not a hard failure and quotas are already disabled.
6480 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
6481 if (IS_ERR(handle
)) {
6482 err
= PTR_ERR(handle
);
6485 EXT4_I(inode
)->i_flags
&= ~(EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
);
6486 inode_set_flags(inode
, 0, S_NOATIME
| S_IMMUTABLE
);
6487 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
6488 err
= ext4_mark_inode_dirty(handle
, inode
);
6489 ext4_journal_stop(handle
);
6491 inode_unlock(inode
);
6493 lockdep_set_quota_inode(inode
, I_DATA_SEM_NORMAL
);
6497 return dquot_quota_off(sb
, type
);
6500 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6501 * acquiring the locks... As quota files are never truncated and quota code
6502 * itself serializes the operations (and no one else should touch the files)
6503 * we don't have to be afraid of races */
6504 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
6505 size_t len
, loff_t off
)
6507 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
6508 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
6509 int offset
= off
& (sb
->s_blocksize
- 1);
6512 struct buffer_head
*bh
;
6513 loff_t i_size
= i_size_read(inode
);
6517 if (off
+len
> i_size
)
6520 while (toread
> 0) {
6521 tocopy
= sb
->s_blocksize
- offset
< toread
?
6522 sb
->s_blocksize
- offset
: toread
;
6523 bh
= ext4_bread(NULL
, inode
, blk
, 0);
6526 if (!bh
) /* A hole? */
6527 memset(data
, 0, tocopy
);
6529 memcpy(data
, bh
->b_data
+offset
, tocopy
);
6539 /* Write to quotafile (we know the transaction is already started and has
6540 * enough credits) */
6541 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
6542 const char *data
, size_t len
, loff_t off
)
6544 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
6545 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
6546 int err
= 0, err2
= 0, offset
= off
& (sb
->s_blocksize
- 1);
6548 struct buffer_head
*bh
;
6549 handle_t
*handle
= journal_current_handle();
6551 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
6552 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
6553 " cancelled because transaction is not started",
6554 (unsigned long long)off
, (unsigned long long)len
);
6558 * Since we account only one data block in transaction credits,
6559 * then it is impossible to cross a block boundary.
6561 if (sb
->s_blocksize
- offset
< len
) {
6562 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
6563 " cancelled because not block aligned",
6564 (unsigned long long)off
, (unsigned long long)len
);
6569 bh
= ext4_bread(handle
, inode
, blk
,
6570 EXT4_GET_BLOCKS_CREATE
|
6571 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
6572 } while (PTR_ERR(bh
) == -ENOSPC
&&
6573 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
6578 BUFFER_TRACE(bh
, "get write access");
6579 err
= ext4_journal_get_write_access(handle
, bh
);
6585 memcpy(bh
->b_data
+offset
, data
, len
);
6586 flush_dcache_page(bh
->b_page
);
6588 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
6591 if (inode
->i_size
< off
+ len
) {
6592 i_size_write(inode
, off
+ len
);
6593 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
6594 err2
= ext4_mark_inode_dirty(handle
, inode
);
6595 if (unlikely(err2
&& !err
))
6598 return err
? err
: len
;
6602 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
6603 const char *dev_name
, void *data
)
6605 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
6608 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6609 static inline void register_as_ext2(void)
6611 int err
= register_filesystem(&ext2_fs_type
);
6614 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
6617 static inline void unregister_as_ext2(void)
6619 unregister_filesystem(&ext2_fs_type
);
6622 static inline int ext2_feature_set_ok(struct super_block
*sb
)
6624 if (ext4_has_unknown_ext2_incompat_features(sb
))
6628 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
6633 static inline void register_as_ext2(void) { }
6634 static inline void unregister_as_ext2(void) { }
6635 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
6638 static inline void register_as_ext3(void)
6640 int err
= register_filesystem(&ext3_fs_type
);
6643 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
6646 static inline void unregister_as_ext3(void)
6648 unregister_filesystem(&ext3_fs_type
);
6651 static inline int ext3_feature_set_ok(struct super_block
*sb
)
6653 if (ext4_has_unknown_ext3_incompat_features(sb
))
6655 if (!ext4_has_feature_journal(sb
))
6659 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
6664 static struct file_system_type ext4_fs_type
= {
6665 .owner
= THIS_MODULE
,
6667 .mount
= ext4_mount
,
6668 .kill_sb
= kill_block_super
,
6669 .fs_flags
= FS_REQUIRES_DEV
| FS_ALLOW_IDMAP
,
6671 MODULE_ALIAS_FS("ext4");
6673 /* Shared across all ext4 file systems */
6674 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
6676 static int __init
ext4_init_fs(void)
6680 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
6681 ext4_li_info
= NULL
;
6683 /* Build-time check for flags consistency */
6684 ext4_check_flag_values();
6686 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
6687 init_waitqueue_head(&ext4__ioend_wq
[i
]);
6689 err
= ext4_init_es();
6693 err
= ext4_init_pending();
6697 err
= ext4_init_post_read_processing();
6701 err
= ext4_init_pageio();
6705 err
= ext4_init_system_zone();
6709 err
= ext4_init_sysfs();
6713 err
= ext4_init_mballoc();
6716 err
= init_inodecache();
6720 err
= ext4_fc_init_dentry_cache();
6726 err
= register_filesystem(&ext4_fs_type
);
6732 unregister_as_ext2();
6733 unregister_as_ext3();
6735 destroy_inodecache();
6737 ext4_exit_mballoc();
6741 ext4_exit_system_zone();
6745 ext4_exit_post_read_processing();
6747 ext4_exit_pending();
6754 static void __exit
ext4_exit_fs(void)
6756 ext4_destroy_lazyinit_thread();
6757 unregister_as_ext2();
6758 unregister_as_ext3();
6759 unregister_filesystem(&ext4_fs_type
);
6760 destroy_inodecache();
6761 ext4_exit_mballoc();
6763 ext4_exit_system_zone();
6765 ext4_exit_post_read_processing();
6767 ext4_exit_pending();
6770 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6771 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6772 MODULE_LICENSE("GPL");
6773 MODULE_SOFTDEP("pre: crc32c");
6774 module_init(ext4_init_fs
)
6775 module_exit(ext4_exit_fs
)