2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct ext4_lazy_init
*ext4_li_info
;
59 static struct mutex ext4_li_mtx
;
60 static struct ratelimit_state ext4_mount_msg_ratelimit
;
62 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
63 unsigned long journal_devnum
);
64 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
65 static int ext4_commit_super(struct super_block
*sb
, int sync
);
66 static void ext4_mark_recovery_complete(struct super_block
*sb
,
67 struct ext4_super_block
*es
);
68 static void ext4_clear_journal_err(struct super_block
*sb
,
69 struct ext4_super_block
*es
);
70 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
71 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
72 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
73 static int ext4_unfreeze(struct super_block
*sb
);
74 static int ext4_freeze(struct super_block
*sb
);
75 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
76 const char *dev_name
, void *data
);
77 static inline int ext2_feature_set_ok(struct super_block
*sb
);
78 static inline int ext3_feature_set_ok(struct super_block
*sb
);
79 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
80 static void ext4_destroy_lazyinit_thread(void);
81 static void ext4_unregister_li_request(struct super_block
*sb
);
82 static void ext4_clear_request_list(void);
83 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
84 unsigned int journal_inum
);
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type
= {
118 .owner
= THIS_MODULE
,
121 .kill_sb
= kill_block_super
,
122 .fs_flags
= FS_REQUIRES_DEV
,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type
= {
133 .owner
= THIS_MODULE
,
136 .kill_sb
= kill_block_super
,
137 .fs_flags
= FS_REQUIRES_DEV
,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static int ext4_verify_csum_type(struct super_block
*sb
,
144 struct ext4_super_block
*es
)
146 if (!ext4_has_feature_metadata_csum(sb
))
149 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
152 static __le32
ext4_superblock_csum(struct super_block
*sb
,
153 struct ext4_super_block
*es
)
155 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
156 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
159 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
161 return cpu_to_le32(csum
);
164 static int ext4_superblock_csum_verify(struct super_block
*sb
,
165 struct ext4_super_block
*es
)
167 if (!ext4_has_metadata_csum(sb
))
170 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
173 void ext4_superblock_csum_set(struct super_block
*sb
)
175 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
177 if (!ext4_has_metadata_csum(sb
))
180 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
183 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
187 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
189 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
193 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
197 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
199 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
203 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
204 struct ext4_group_desc
*bg
)
206 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
207 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
208 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
211 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
212 struct ext4_group_desc
*bg
)
214 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
215 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
216 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
219 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
220 struct ext4_group_desc
*bg
)
222 return le32_to_cpu(bg
->bg_inode_table_lo
) |
223 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
224 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
227 __u32
ext4_free_group_clusters(struct super_block
*sb
,
228 struct ext4_group_desc
*bg
)
230 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
231 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
232 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
235 __u32
ext4_free_inodes_count(struct super_block
*sb
,
236 struct ext4_group_desc
*bg
)
238 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
239 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
240 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
243 __u32
ext4_used_dirs_count(struct super_block
*sb
,
244 struct ext4_group_desc
*bg
)
246 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
247 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
248 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
251 __u32
ext4_itable_unused_count(struct super_block
*sb
,
252 struct ext4_group_desc
*bg
)
254 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
255 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
256 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
259 void ext4_block_bitmap_set(struct super_block
*sb
,
260 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
262 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
263 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
264 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
267 void ext4_inode_bitmap_set(struct super_block
*sb
,
268 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
270 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
271 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
272 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
275 void ext4_inode_table_set(struct super_block
*sb
,
276 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
278 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
279 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
280 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
283 void ext4_free_group_clusters_set(struct super_block
*sb
,
284 struct ext4_group_desc
*bg
, __u32 count
)
286 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
287 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
288 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
291 void ext4_free_inodes_set(struct super_block
*sb
,
292 struct ext4_group_desc
*bg
, __u32 count
)
294 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
295 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
296 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
299 void ext4_used_dirs_set(struct super_block
*sb
,
300 struct ext4_group_desc
*bg
, __u32 count
)
302 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
303 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
304 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
307 void ext4_itable_unused_set(struct super_block
*sb
,
308 struct ext4_group_desc
*bg
, __u32 count
)
310 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
311 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
312 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
316 static void __save_error_info(struct super_block
*sb
, const char *func
,
319 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
321 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
322 if (bdev_read_only(sb
->s_bdev
))
324 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
325 es
->s_last_error_time
= cpu_to_le32(get_seconds());
326 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
327 es
->s_last_error_line
= cpu_to_le32(line
);
328 if (!es
->s_first_error_time
) {
329 es
->s_first_error_time
= es
->s_last_error_time
;
330 strncpy(es
->s_first_error_func
, func
,
331 sizeof(es
->s_first_error_func
));
332 es
->s_first_error_line
= cpu_to_le32(line
);
333 es
->s_first_error_ino
= es
->s_last_error_ino
;
334 es
->s_first_error_block
= es
->s_last_error_block
;
337 * Start the daily error reporting function if it hasn't been
340 if (!es
->s_error_count
)
341 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
342 le32_add_cpu(&es
->s_error_count
, 1);
345 static void save_error_info(struct super_block
*sb
, const char *func
,
348 __save_error_info(sb
, func
, line
);
349 ext4_commit_super(sb
, 1);
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
360 static int block_device_ejected(struct super_block
*sb
)
362 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
363 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
365 return bdi
->dev
== NULL
;
368 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
370 struct super_block
*sb
= journal
->j_private
;
371 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
372 int error
= is_journal_aborted(journal
);
373 struct ext4_journal_cb_entry
*jce
;
375 BUG_ON(txn
->t_state
== T_FINISHED
);
377 ext4_process_freed_data(sb
, txn
->t_tid
);
379 spin_lock(&sbi
->s_md_lock
);
380 while (!list_empty(&txn
->t_private_list
)) {
381 jce
= list_entry(txn
->t_private_list
.next
,
382 struct ext4_journal_cb_entry
, jce_list
);
383 list_del_init(&jce
->jce_list
);
384 spin_unlock(&sbi
->s_md_lock
);
385 jce
->jce_func(sb
, jce
, error
);
386 spin_lock(&sbi
->s_md_lock
);
388 spin_unlock(&sbi
->s_md_lock
);
391 /* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
394 * On ext2, we can store the error state of the filesystem in the
395 * superblock. That is not possible on ext4, because we may have other
396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
401 * We'll just use the jbd2_journal_abort() error code to record an error in
402 * the journal instead. On recovery, the journal will complain about
403 * that error until we've noted it down and cleared it.
406 static void ext4_handle_error(struct super_block
*sb
)
411 if (!test_opt(sb
, ERRORS_CONT
)) {
412 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
414 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
416 jbd2_journal_abort(journal
, -EIO
);
418 if (test_opt(sb
, ERRORS_RO
)) {
419 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
425 sb
->s_flags
|= MS_RDONLY
;
427 if (test_opt(sb
, ERRORS_PANIC
)) {
428 if (EXT4_SB(sb
)->s_journal
&&
429 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
431 panic("EXT4-fs (device %s): panic forced after error\n",
436 #define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
440 void __ext4_error(struct super_block
*sb
, const char *function
,
441 unsigned int line
, const char *fmt
, ...)
443 struct va_format vaf
;
446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
449 if (ext4_error_ratelimit(sb
)) {
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
458 save_error_info(sb
, function
, line
);
459 ext4_handle_error(sb
);
462 void __ext4_error_inode(struct inode
*inode
, const char *function
,
463 unsigned int line
, ext4_fsblk_t block
,
464 const char *fmt
, ...)
467 struct va_format vaf
;
468 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
473 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
474 es
->s_last_error_block
= cpu_to_le64(block
);
475 if (ext4_error_ratelimit(inode
->i_sb
)) {
480 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
483 block
, current
->comm
, &vaf
);
485 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
488 current
->comm
, &vaf
);
491 save_error_info(inode
->i_sb
, function
, line
);
492 ext4_handle_error(inode
->i_sb
);
495 void __ext4_error_file(struct file
*file
, const char *function
,
496 unsigned int line
, ext4_fsblk_t block
,
497 const char *fmt
, ...)
500 struct va_format vaf
;
501 struct ext4_super_block
*es
;
502 struct inode
*inode
= file_inode(file
);
503 char pathname
[80], *path
;
505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
508 es
= EXT4_SB(inode
->i_sb
)->s_es
;
509 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
510 if (ext4_error_ratelimit(inode
->i_sb
)) {
511 path
= file_path(file
, pathname
, sizeof(pathname
));
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
522 block
, current
->comm
, path
, &vaf
);
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
528 current
->comm
, path
, &vaf
);
531 save_error_info(inode
->i_sb
, function
, line
);
532 ext4_handle_error(inode
->i_sb
);
535 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
542 errstr
= "Corrupt filesystem";
545 errstr
= "Filesystem failed CRC";
548 errstr
= "IO failure";
551 errstr
= "Out of memory";
554 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
555 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
556 errstr
= "Journal has aborted";
558 errstr
= "Readonly filesystem";
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
578 void __ext4_std_error(struct super_block
*sb
, const char *function
,
579 unsigned int line
, int errno
)
584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
590 if (errno
== -EROFS
&& journal_current_handle() == NULL
&& sb_rdonly(sb
))
593 if (ext4_error_ratelimit(sb
)) {
594 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
595 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
596 sb
->s_id
, function
, line
, errstr
);
599 save_error_info(sb
, function
, line
);
600 ext4_handle_error(sb
);
604 * ext4_abort is a much stronger failure handler than ext4_error. The
605 * abort function may be used to deal with unrecoverable failures such
606 * as journal IO errors or ENOMEM at a critical moment in log management.
608 * We unconditionally force the filesystem into an ABORT|READONLY state,
609 * unless the error response on the fs has been set to panic in which
610 * case we take the easy way out and panic immediately.
613 void __ext4_abort(struct super_block
*sb
, const char *function
,
614 unsigned int line
, const char *fmt
, ...)
616 struct va_format vaf
;
619 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
622 save_error_info(sb
, function
, line
);
626 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
627 sb
->s_id
, function
, line
, &vaf
);
630 if (sb_rdonly(sb
) == 0) {
631 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
632 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
634 * Make sure updated value of ->s_mount_flags will be visible
635 * before ->s_flags update
638 sb
->s_flags
|= MS_RDONLY
;
639 if (EXT4_SB(sb
)->s_journal
)
640 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
641 save_error_info(sb
, function
, line
);
643 if (test_opt(sb
, ERRORS_PANIC
)) {
644 if (EXT4_SB(sb
)->s_journal
&&
645 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
647 panic("EXT4-fs panic from previous error\n");
651 void __ext4_msg(struct super_block
*sb
,
652 const char *prefix
, const char *fmt
, ...)
654 struct va_format vaf
;
657 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
663 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
667 #define ext4_warning_ratelimit(sb) \
668 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
671 void __ext4_warning(struct super_block
*sb
, const char *function
,
672 unsigned int line
, const char *fmt
, ...)
674 struct va_format vaf
;
677 if (!ext4_warning_ratelimit(sb
))
683 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
684 sb
->s_id
, function
, line
, &vaf
);
688 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
689 unsigned int line
, const char *fmt
, ...)
691 struct va_format vaf
;
694 if (!ext4_warning_ratelimit(inode
->i_sb
))
700 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
701 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
702 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
706 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
707 struct super_block
*sb
, ext4_group_t grp
,
708 unsigned long ino
, ext4_fsblk_t block
,
709 const char *fmt
, ...)
713 struct va_format vaf
;
715 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
717 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
720 es
->s_last_error_ino
= cpu_to_le32(ino
);
721 es
->s_last_error_block
= cpu_to_le64(block
);
722 __save_error_info(sb
, function
, line
);
724 if (ext4_error_ratelimit(sb
)) {
728 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
729 sb
->s_id
, function
, line
, grp
);
731 printk(KERN_CONT
"inode %lu: ", ino
);
733 printk(KERN_CONT
"block %llu:",
734 (unsigned long long) block
);
735 printk(KERN_CONT
"%pV\n", &vaf
);
739 if (test_opt(sb
, ERRORS_CONT
)) {
740 ext4_commit_super(sb
, 0);
744 ext4_unlock_group(sb
, grp
);
745 ext4_commit_super(sb
, 1);
746 ext4_handle_error(sb
);
748 * We only get here in the ERRORS_RO case; relocking the group
749 * may be dangerous, but nothing bad will happen since the
750 * filesystem will have already been marked read/only and the
751 * journal has been aborted. We return 1 as a hint to callers
752 * who might what to use the return value from
753 * ext4_grp_locked_error() to distinguish between the
754 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
755 * aggressively from the ext4 function in question, with a
756 * more appropriate error code.
758 ext4_lock_group(sb
, grp
);
762 void ext4_update_dynamic_rev(struct super_block
*sb
)
764 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
766 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
770 "updating to rev %d because of new feature flag, "
771 "running e2fsck is recommended",
774 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
775 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
776 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
777 /* leave es->s_feature_*compat flags alone */
778 /* es->s_uuid will be set by e2fsck if empty */
781 * The rest of the superblock fields should be zero, and if not it
782 * means they are likely already in use, so leave them alone. We
783 * can leave it up to e2fsck to clean up any inconsistencies there.
788 * Open the external journal device
790 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
792 struct block_device
*bdev
;
793 char b
[BDEVNAME_SIZE
];
795 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
801 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
802 __bdevname(dev
, b
), PTR_ERR(bdev
));
807 * Release the journal device
809 static void ext4_blkdev_put(struct block_device
*bdev
)
811 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
814 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
816 struct block_device
*bdev
;
817 bdev
= sbi
->journal_bdev
;
819 ext4_blkdev_put(bdev
);
820 sbi
->journal_bdev
= NULL
;
824 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
826 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
829 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
833 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
834 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
836 printk(KERN_ERR
"sb_info orphan list:\n");
837 list_for_each(l
, &sbi
->s_orphan
) {
838 struct inode
*inode
= orphan_list_entry(l
);
840 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
841 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
842 inode
->i_mode
, inode
->i_nlink
,
848 static int ext4_quota_off(struct super_block
*sb
, int type
);
850 static inline void ext4_quota_off_umount(struct super_block
*sb
)
854 /* Use our quota_off function to clear inode flags etc. */
855 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++)
856 ext4_quota_off(sb
, type
);
860 * This is a helper function which is used in the mount/remount
861 * codepaths (which holds s_umount) to fetch the quota file name.
863 static inline char *get_qf_name(struct super_block
*sb
,
864 struct ext4_sb_info
*sbi
,
867 return rcu_dereference_protected(sbi
->s_qf_names
[type
],
868 lockdep_is_held(&sb
->s_umount
));
871 static inline void ext4_quota_off_umount(struct super_block
*sb
)
876 static void ext4_put_super(struct super_block
*sb
)
878 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
879 struct ext4_super_block
*es
= sbi
->s_es
;
883 ext4_unregister_li_request(sb
);
884 ext4_quota_off_umount(sb
);
886 flush_workqueue(sbi
->rsv_conversion_wq
);
887 destroy_workqueue(sbi
->rsv_conversion_wq
);
889 if (sbi
->s_journal
) {
890 aborted
= is_journal_aborted(sbi
->s_journal
);
891 err
= jbd2_journal_destroy(sbi
->s_journal
);
892 sbi
->s_journal
= NULL
;
893 if ((err
< 0) && !aborted
)
894 ext4_abort(sb
, "Couldn't clean up the journal");
897 ext4_unregister_sysfs(sb
);
898 ext4_es_unregister_shrinker(sbi
);
899 del_timer_sync(&sbi
->s_err_report
);
900 ext4_release_system_zone(sb
);
902 ext4_ext_release(sb
);
904 if (!sb_rdonly(sb
) && !aborted
) {
905 ext4_clear_feature_journal_needs_recovery(sb
);
906 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
909 ext4_commit_super(sb
, 1);
911 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
912 brelse(sbi
->s_group_desc
[i
]);
913 kvfree(sbi
->s_group_desc
);
914 kvfree(sbi
->s_flex_groups
);
915 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
916 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
917 percpu_counter_destroy(&sbi
->s_dirs_counter
);
918 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
919 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
921 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
922 kfree(get_qf_name(sb
, sbi
, i
));
925 /* Debugging code just in case the in-memory inode orphan list
926 * isn't empty. The on-disk one can be non-empty if we've
927 * detected an error and taken the fs readonly, but the
928 * in-memory list had better be clean by this point. */
929 if (!list_empty(&sbi
->s_orphan
))
930 dump_orphan_list(sb
, sbi
);
931 J_ASSERT(list_empty(&sbi
->s_orphan
));
933 sync_blockdev(sb
->s_bdev
);
934 invalidate_bdev(sb
->s_bdev
);
935 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
937 * Invalidate the journal device's buffers. We don't want them
938 * floating about in memory - the physical journal device may
939 * hotswapped, and it breaks the `ro-after' testing code.
941 sync_blockdev(sbi
->journal_bdev
);
942 invalidate_bdev(sbi
->journal_bdev
);
943 ext4_blkdev_remove(sbi
);
945 if (sbi
->s_ea_inode_cache
) {
946 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
947 sbi
->s_ea_inode_cache
= NULL
;
949 if (sbi
->s_ea_block_cache
) {
950 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
951 sbi
->s_ea_block_cache
= NULL
;
954 kthread_stop(sbi
->s_mmp_tsk
);
956 sb
->s_fs_info
= NULL
;
958 * Now that we are completely done shutting down the
959 * superblock, we need to actually destroy the kobject.
961 kobject_put(&sbi
->s_kobj
);
962 wait_for_completion(&sbi
->s_kobj_unregister
);
963 if (sbi
->s_chksum_driver
)
964 crypto_free_shash(sbi
->s_chksum_driver
);
965 kfree(sbi
->s_blockgroup_lock
);
966 fs_put_dax(sbi
->s_daxdev
);
970 static struct kmem_cache
*ext4_inode_cachep
;
973 * Called inside transaction, so use GFP_NOFS
975 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
977 struct ext4_inode_info
*ei
;
979 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
983 ei
->vfs_inode
.i_version
= 1;
984 spin_lock_init(&ei
->i_raw_lock
);
985 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
986 spin_lock_init(&ei
->i_prealloc_lock
);
987 ext4_es_init_tree(&ei
->i_es_tree
);
988 rwlock_init(&ei
->i_es_lock
);
989 INIT_LIST_HEAD(&ei
->i_es_list
);
992 ei
->i_es_shrink_lblk
= 0;
993 ei
->i_reserved_data_blocks
= 0;
994 ei
->i_da_metadata_calc_len
= 0;
995 ei
->i_da_metadata_calc_last_lblock
= 0;
996 spin_lock_init(&(ei
->i_block_reservation_lock
));
998 ei
->i_reserved_quota
= 0;
999 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
1002 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
1003 spin_lock_init(&ei
->i_completed_io_lock
);
1005 ei
->i_datasync_tid
= 0;
1006 atomic_set(&ei
->i_unwritten
, 0);
1007 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
1008 return &ei
->vfs_inode
;
1011 static int ext4_drop_inode(struct inode
*inode
)
1013 int drop
= generic_drop_inode(inode
);
1015 trace_ext4_drop_inode(inode
, drop
);
1019 static void ext4_i_callback(struct rcu_head
*head
)
1021 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1022 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
1025 static void ext4_destroy_inode(struct inode
*inode
)
1027 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
1028 ext4_msg(inode
->i_sb
, KERN_ERR
,
1029 "Inode %lu (%p): orphan list check failed!",
1030 inode
->i_ino
, EXT4_I(inode
));
1031 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
1032 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
1036 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
1039 static void init_once(void *foo
)
1041 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
1043 INIT_LIST_HEAD(&ei
->i_orphan
);
1044 init_rwsem(&ei
->xattr_sem
);
1045 init_rwsem(&ei
->i_data_sem
);
1046 init_rwsem(&ei
->i_mmap_sem
);
1047 inode_init_once(&ei
->vfs_inode
);
1050 static int __init
init_inodecache(void)
1052 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
1053 sizeof(struct ext4_inode_info
),
1054 0, (SLAB_RECLAIM_ACCOUNT
|
1055 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1057 if (ext4_inode_cachep
== NULL
)
1062 static void destroy_inodecache(void)
1065 * Make sure all delayed rcu free inodes are flushed before we
1069 kmem_cache_destroy(ext4_inode_cachep
);
1072 void ext4_clear_inode(struct inode
*inode
)
1074 invalidate_inode_buffers(inode
);
1077 ext4_discard_preallocations(inode
);
1078 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1079 if (EXT4_I(inode
)->jinode
) {
1080 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1081 EXT4_I(inode
)->jinode
);
1082 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1083 EXT4_I(inode
)->jinode
= NULL
;
1085 fscrypt_put_encryption_info(inode
);
1088 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1089 u64 ino
, u32 generation
)
1091 struct inode
*inode
;
1093 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1094 return ERR_PTR(-ESTALE
);
1095 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1096 return ERR_PTR(-ESTALE
);
1098 /* iget isn't really right if the inode is currently unallocated!!
1100 * ext4_read_inode will return a bad_inode if the inode had been
1101 * deleted, so we should be safe.
1103 * Currently we don't know the generation for parent directory, so
1104 * a generation of 0 means "accept any"
1106 inode
= ext4_iget_normal(sb
, ino
);
1108 return ERR_CAST(inode
);
1109 if (generation
&& inode
->i_generation
!= generation
) {
1111 return ERR_PTR(-ESTALE
);
1117 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1118 int fh_len
, int fh_type
)
1120 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1121 ext4_nfs_get_inode
);
1124 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1125 int fh_len
, int fh_type
)
1127 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1128 ext4_nfs_get_inode
);
1132 * Try to release metadata pages (indirect blocks, directories) which are
1133 * mapped via the block device. Since these pages could have journal heads
1134 * which would prevent try_to_free_buffers() from freeing them, we must use
1135 * jbd2 layer's try_to_free_buffers() function to release them.
1137 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1140 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1142 WARN_ON(PageChecked(page
));
1143 if (!page_has_buffers(page
))
1146 return jbd2_journal_try_to_free_buffers(journal
, page
,
1147 wait
& ~__GFP_DIRECT_RECLAIM
);
1148 return try_to_free_buffers(page
);
1151 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1152 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1154 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1155 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1158 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1161 handle_t
*handle
= fs_data
;
1162 int res
, res2
, credits
, retries
= 0;
1165 * Encrypting the root directory is not allowed because e2fsck expects
1166 * lost+found to exist and be unencrypted, and encrypting the root
1167 * directory would imply encrypting the lost+found directory as well as
1168 * the filename "lost+found" itself.
1170 if (inode
->i_ino
== EXT4_ROOT_INO
)
1173 res
= ext4_convert_inline_data(inode
);
1178 * If a journal handle was specified, then the encryption context is
1179 * being set on a new inode via inheritance and is part of a larger
1180 * transaction to create the inode. Otherwise the encryption context is
1181 * being set on an existing inode in its own transaction. Only in the
1182 * latter case should the "retry on ENOSPC" logic be used.
1186 res
= ext4_xattr_set_handle(handle
, inode
,
1187 EXT4_XATTR_INDEX_ENCRYPTION
,
1188 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1191 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1192 ext4_clear_inode_state(inode
,
1193 EXT4_STATE_MAY_INLINE_DATA
);
1195 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1196 * S_DAX may be disabled
1198 ext4_set_inode_flags(inode
);
1203 res
= dquot_initialize(inode
);
1207 res
= ext4_xattr_set_credits(inode
, len
, false /* is_create */,
1212 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
, credits
);
1214 return PTR_ERR(handle
);
1216 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1217 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1220 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1222 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1223 * S_DAX may be disabled
1225 ext4_set_inode_flags(inode
);
1226 res
= ext4_mark_inode_dirty(handle
, inode
);
1228 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1230 res2
= ext4_journal_stop(handle
);
1232 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1239 static bool ext4_dummy_context(struct inode
*inode
)
1241 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1244 static unsigned ext4_max_namelen(struct inode
*inode
)
1246 return S_ISLNK(inode
->i_mode
) ? inode
->i_sb
->s_blocksize
:
1250 static const struct fscrypt_operations ext4_cryptops
= {
1251 .key_prefix
= "ext4:",
1252 .get_context
= ext4_get_context
,
1253 .set_context
= ext4_set_context
,
1254 .dummy_context
= ext4_dummy_context
,
1255 .empty_dir
= ext4_empty_dir
,
1256 .max_namelen
= ext4_max_namelen
,
1261 static const char * const quotatypes
[] = INITQFNAMES
;
1262 #define QTYPE2NAME(t) (quotatypes[t])
1264 static int ext4_write_dquot(struct dquot
*dquot
);
1265 static int ext4_acquire_dquot(struct dquot
*dquot
);
1266 static int ext4_release_dquot(struct dquot
*dquot
);
1267 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1268 static int ext4_write_info(struct super_block
*sb
, int type
);
1269 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1270 const struct path
*path
);
1271 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1272 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1273 size_t len
, loff_t off
);
1274 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1275 const char *data
, size_t len
, loff_t off
);
1276 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1277 unsigned int flags
);
1278 static int ext4_enable_quotas(struct super_block
*sb
);
1279 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1281 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1283 return EXT4_I(inode
)->i_dquot
;
1286 static const struct dquot_operations ext4_quota_operations
= {
1287 .get_reserved_space
= ext4_get_reserved_space
,
1288 .write_dquot
= ext4_write_dquot
,
1289 .acquire_dquot
= ext4_acquire_dquot
,
1290 .release_dquot
= ext4_release_dquot
,
1291 .mark_dirty
= ext4_mark_dquot_dirty
,
1292 .write_info
= ext4_write_info
,
1293 .alloc_dquot
= dquot_alloc
,
1294 .destroy_dquot
= dquot_destroy
,
1295 .get_projid
= ext4_get_projid
,
1296 .get_inode_usage
= ext4_get_inode_usage
,
1297 .get_next_id
= ext4_get_next_id
,
1300 static const struct quotactl_ops ext4_qctl_operations
= {
1301 .quota_on
= ext4_quota_on
,
1302 .quota_off
= ext4_quota_off
,
1303 .quota_sync
= dquot_quota_sync
,
1304 .get_state
= dquot_get_state
,
1305 .set_info
= dquot_set_dqinfo
,
1306 .get_dqblk
= dquot_get_dqblk
,
1307 .set_dqblk
= dquot_set_dqblk
,
1308 .get_nextdqblk
= dquot_get_next_dqblk
,
1312 static const struct super_operations ext4_sops
= {
1313 .alloc_inode
= ext4_alloc_inode
,
1314 .destroy_inode
= ext4_destroy_inode
,
1315 .write_inode
= ext4_write_inode
,
1316 .dirty_inode
= ext4_dirty_inode
,
1317 .drop_inode
= ext4_drop_inode
,
1318 .evict_inode
= ext4_evict_inode
,
1319 .put_super
= ext4_put_super
,
1320 .sync_fs
= ext4_sync_fs
,
1321 .freeze_fs
= ext4_freeze
,
1322 .unfreeze_fs
= ext4_unfreeze
,
1323 .statfs
= ext4_statfs
,
1324 .remount_fs
= ext4_remount
,
1325 .show_options
= ext4_show_options
,
1327 .quota_read
= ext4_quota_read
,
1328 .quota_write
= ext4_quota_write
,
1329 .get_dquots
= ext4_get_dquots
,
1331 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1334 static const struct export_operations ext4_export_ops
= {
1335 .fh_to_dentry
= ext4_fh_to_dentry
,
1336 .fh_to_parent
= ext4_fh_to_parent
,
1337 .get_parent
= ext4_get_parent
,
1341 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1342 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1343 Opt_nouid32
, Opt_debug
, Opt_removed
,
1344 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1345 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1346 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1347 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1348 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1349 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1350 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1351 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1352 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1353 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1354 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1355 Opt_lazytime
, Opt_nolazytime
, Opt_debug_want_extra_isize
,
1356 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1357 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1358 Opt_dioread_nolock
, Opt_dioread_lock
,
1359 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1360 Opt_max_dir_size_kb
, Opt_nojournal_checksum
, Opt_nombcache
,
1363 static const match_table_t tokens
= {
1364 {Opt_bsd_df
, "bsddf"},
1365 {Opt_minix_df
, "minixdf"},
1366 {Opt_grpid
, "grpid"},
1367 {Opt_grpid
, "bsdgroups"},
1368 {Opt_nogrpid
, "nogrpid"},
1369 {Opt_nogrpid
, "sysvgroups"},
1370 {Opt_resgid
, "resgid=%u"},
1371 {Opt_resuid
, "resuid=%u"},
1373 {Opt_err_cont
, "errors=continue"},
1374 {Opt_err_panic
, "errors=panic"},
1375 {Opt_err_ro
, "errors=remount-ro"},
1376 {Opt_nouid32
, "nouid32"},
1377 {Opt_debug
, "debug"},
1378 {Opt_removed
, "oldalloc"},
1379 {Opt_removed
, "orlov"},
1380 {Opt_user_xattr
, "user_xattr"},
1381 {Opt_nouser_xattr
, "nouser_xattr"},
1383 {Opt_noacl
, "noacl"},
1384 {Opt_noload
, "norecovery"},
1385 {Opt_noload
, "noload"},
1386 {Opt_removed
, "nobh"},
1387 {Opt_removed
, "bh"},
1388 {Opt_commit
, "commit=%u"},
1389 {Opt_min_batch_time
, "min_batch_time=%u"},
1390 {Opt_max_batch_time
, "max_batch_time=%u"},
1391 {Opt_journal_dev
, "journal_dev=%u"},
1392 {Opt_journal_path
, "journal_path=%s"},
1393 {Opt_journal_checksum
, "journal_checksum"},
1394 {Opt_nojournal_checksum
, "nojournal_checksum"},
1395 {Opt_journal_async_commit
, "journal_async_commit"},
1396 {Opt_abort
, "abort"},
1397 {Opt_data_journal
, "data=journal"},
1398 {Opt_data_ordered
, "data=ordered"},
1399 {Opt_data_writeback
, "data=writeback"},
1400 {Opt_data_err_abort
, "data_err=abort"},
1401 {Opt_data_err_ignore
, "data_err=ignore"},
1402 {Opt_offusrjquota
, "usrjquota="},
1403 {Opt_usrjquota
, "usrjquota=%s"},
1404 {Opt_offgrpjquota
, "grpjquota="},
1405 {Opt_grpjquota
, "grpjquota=%s"},
1406 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1407 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1408 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1409 {Opt_grpquota
, "grpquota"},
1410 {Opt_noquota
, "noquota"},
1411 {Opt_quota
, "quota"},
1412 {Opt_usrquota
, "usrquota"},
1413 {Opt_prjquota
, "prjquota"},
1414 {Opt_barrier
, "barrier=%u"},
1415 {Opt_barrier
, "barrier"},
1416 {Opt_nobarrier
, "nobarrier"},
1417 {Opt_i_version
, "i_version"},
1419 {Opt_stripe
, "stripe=%u"},
1420 {Opt_delalloc
, "delalloc"},
1421 {Opt_lazytime
, "lazytime"},
1422 {Opt_nolazytime
, "nolazytime"},
1423 {Opt_debug_want_extra_isize
, "debug_want_extra_isize=%u"},
1424 {Opt_nodelalloc
, "nodelalloc"},
1425 {Opt_removed
, "mblk_io_submit"},
1426 {Opt_removed
, "nomblk_io_submit"},
1427 {Opt_block_validity
, "block_validity"},
1428 {Opt_noblock_validity
, "noblock_validity"},
1429 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1430 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1431 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1432 {Opt_auto_da_alloc
, "auto_da_alloc"},
1433 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1434 {Opt_dioread_nolock
, "dioread_nolock"},
1435 {Opt_dioread_lock
, "dioread_lock"},
1436 {Opt_discard
, "discard"},
1437 {Opt_nodiscard
, "nodiscard"},
1438 {Opt_init_itable
, "init_itable=%u"},
1439 {Opt_init_itable
, "init_itable"},
1440 {Opt_noinit_itable
, "noinit_itable"},
1441 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1442 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1443 {Opt_nombcache
, "nombcache"},
1444 {Opt_nombcache
, "no_mbcache"}, /* for backward compatibility */
1445 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1446 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1447 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1448 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1449 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1453 static ext4_fsblk_t
get_sb_block(void **data
)
1455 ext4_fsblk_t sb_block
;
1456 char *options
= (char *) *data
;
1458 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1459 return 1; /* Default location */
1462 /* TODO: use simple_strtoll with >32bit ext4 */
1463 sb_block
= simple_strtoul(options
, &options
, 0);
1464 if (*options
&& *options
!= ',') {
1465 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1469 if (*options
== ',')
1471 *data
= (void *) options
;
1476 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1477 static const char deprecated_msg
[] =
1478 "Mount option \"%s\" will be removed by %s\n"
1479 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1482 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1484 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1485 char *qname
, *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1488 if (sb_any_quota_loaded(sb
) && !old_qname
) {
1489 ext4_msg(sb
, KERN_ERR
,
1490 "Cannot change journaled "
1491 "quota options when quota turned on");
1494 if (ext4_has_feature_quota(sb
)) {
1495 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1496 "ignored when QUOTA feature is enabled");
1499 qname
= match_strdup(args
);
1501 ext4_msg(sb
, KERN_ERR
,
1502 "Not enough memory for storing quotafile name");
1506 if (strcmp(old_qname
, qname
) == 0)
1509 ext4_msg(sb
, KERN_ERR
,
1510 "%s quota file already specified",
1514 if (strchr(qname
, '/')) {
1515 ext4_msg(sb
, KERN_ERR
,
1516 "quotafile must be on filesystem root");
1519 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], qname
);
1527 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1530 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1531 char *old_qname
= get_qf_name(sb
, sbi
, qtype
);
1533 if (sb_any_quota_loaded(sb
) && old_qname
) {
1534 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1535 " when quota turned on");
1538 rcu_assign_pointer(sbi
->s_qf_names
[qtype
], NULL
);
1545 #define MOPT_SET 0x0001
1546 #define MOPT_CLEAR 0x0002
1547 #define MOPT_NOSUPPORT 0x0004
1548 #define MOPT_EXPLICIT 0x0008
1549 #define MOPT_CLEAR_ERR 0x0010
1550 #define MOPT_GTE0 0x0020
1553 #define MOPT_QFMT 0x0040
1555 #define MOPT_Q MOPT_NOSUPPORT
1556 #define MOPT_QFMT MOPT_NOSUPPORT
1558 #define MOPT_DATAJ 0x0080
1559 #define MOPT_NO_EXT2 0x0100
1560 #define MOPT_NO_EXT3 0x0200
1561 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1562 #define MOPT_STRING 0x0400
1564 static const struct mount_opts
{
1568 } ext4_mount_opts
[] = {
1569 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1570 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1571 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1572 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1573 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1574 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1575 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1576 MOPT_EXT4_ONLY
| MOPT_SET
},
1577 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1578 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1579 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1580 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1581 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1582 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1583 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1584 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1585 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1586 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1587 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1588 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1589 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1590 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1591 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1592 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1593 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1594 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1595 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1596 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1598 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1600 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1601 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1602 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1603 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1604 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1605 {Opt_commit
, 0, MOPT_GTE0
},
1606 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1607 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1608 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1609 {Opt_init_itable
, 0, MOPT_GTE0
},
1610 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1611 {Opt_stripe
, 0, MOPT_GTE0
},
1612 {Opt_resuid
, 0, MOPT_GTE0
},
1613 {Opt_resgid
, 0, MOPT_GTE0
},
1614 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1615 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1616 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1617 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1618 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1619 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1620 MOPT_NO_EXT2
| MOPT_DATAJ
},
1621 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1622 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1623 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1624 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1625 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1627 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1628 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1630 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1631 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1632 {Opt_debug_want_extra_isize
, 0, MOPT_GTE0
},
1633 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1634 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1636 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1638 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1640 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1641 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1642 MOPT_CLEAR
| MOPT_Q
},
1643 {Opt_usrjquota
, 0, MOPT_Q
},
1644 {Opt_grpjquota
, 0, MOPT_Q
},
1645 {Opt_offusrjquota
, 0, MOPT_Q
},
1646 {Opt_offgrpjquota
, 0, MOPT_Q
},
1647 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1648 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1649 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1650 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1651 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1652 {Opt_nombcache
, EXT4_MOUNT_NO_MBCACHE
, MOPT_SET
},
1656 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1657 substring_t
*args
, unsigned long *journal_devnum
,
1658 unsigned int *journal_ioprio
, int is_remount
)
1660 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1661 const struct mount_opts
*m
;
1667 if (token
== Opt_usrjquota
)
1668 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1669 else if (token
== Opt_grpjquota
)
1670 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1671 else if (token
== Opt_offusrjquota
)
1672 return clear_qf_name(sb
, USRQUOTA
);
1673 else if (token
== Opt_offgrpjquota
)
1674 return clear_qf_name(sb
, GRPQUOTA
);
1678 case Opt_nouser_xattr
:
1679 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1682 return 1; /* handled by get_sb_block() */
1684 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1687 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1690 sb
->s_flags
|= SB_I_VERSION
;
1693 sb
->s_flags
|= MS_LAZYTIME
;
1695 case Opt_nolazytime
:
1696 sb
->s_flags
&= ~MS_LAZYTIME
;
1700 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1701 if (token
== m
->token
)
1704 if (m
->token
== Opt_err
) {
1705 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1706 "or missing value", opt
);
1710 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1711 ext4_msg(sb
, KERN_ERR
,
1712 "Mount option \"%s\" incompatible with ext2", opt
);
1715 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1716 ext4_msg(sb
, KERN_ERR
,
1717 "Mount option \"%s\" incompatible with ext3", opt
);
1721 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1723 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1725 if (m
->flags
& MOPT_EXPLICIT
) {
1726 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1727 set_opt2(sb
, EXPLICIT_DELALLOC
);
1728 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1729 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1733 if (m
->flags
& MOPT_CLEAR_ERR
)
1734 clear_opt(sb
, ERRORS_MASK
);
1735 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1736 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1737 "options when quota turned on");
1741 if (m
->flags
& MOPT_NOSUPPORT
) {
1742 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1743 } else if (token
== Opt_commit
) {
1745 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1746 sbi
->s_commit_interval
= HZ
* arg
;
1747 } else if (token
== Opt_debug_want_extra_isize
) {
1748 sbi
->s_want_extra_isize
= arg
;
1749 } else if (token
== Opt_max_batch_time
) {
1750 sbi
->s_max_batch_time
= arg
;
1751 } else if (token
== Opt_min_batch_time
) {
1752 sbi
->s_min_batch_time
= arg
;
1753 } else if (token
== Opt_inode_readahead_blks
) {
1754 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1755 ext4_msg(sb
, KERN_ERR
,
1756 "EXT4-fs: inode_readahead_blks must be "
1757 "0 or a power of 2 smaller than 2^31");
1760 sbi
->s_inode_readahead_blks
= arg
;
1761 } else if (token
== Opt_init_itable
) {
1762 set_opt(sb
, INIT_INODE_TABLE
);
1764 arg
= EXT4_DEF_LI_WAIT_MULT
;
1765 sbi
->s_li_wait_mult
= arg
;
1766 } else if (token
== Opt_max_dir_size_kb
) {
1767 sbi
->s_max_dir_size_kb
= arg
;
1768 } else if (token
== Opt_stripe
) {
1769 sbi
->s_stripe
= arg
;
1770 } else if (token
== Opt_resuid
) {
1771 uid
= make_kuid(current_user_ns(), arg
);
1772 if (!uid_valid(uid
)) {
1773 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1776 sbi
->s_resuid
= uid
;
1777 } else if (token
== Opt_resgid
) {
1778 gid
= make_kgid(current_user_ns(), arg
);
1779 if (!gid_valid(gid
)) {
1780 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1783 sbi
->s_resgid
= gid
;
1784 } else if (token
== Opt_journal_dev
) {
1786 ext4_msg(sb
, KERN_ERR
,
1787 "Cannot specify journal on remount");
1790 *journal_devnum
= arg
;
1791 } else if (token
== Opt_journal_path
) {
1793 struct inode
*journal_inode
;
1798 ext4_msg(sb
, KERN_ERR
,
1799 "Cannot specify journal on remount");
1802 journal_path
= match_strdup(&args
[0]);
1803 if (!journal_path
) {
1804 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1805 "journal device string");
1809 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1811 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1812 "journal device path: error %d", error
);
1813 kfree(journal_path
);
1817 journal_inode
= d_inode(path
.dentry
);
1818 if (!S_ISBLK(journal_inode
->i_mode
)) {
1819 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1820 "is not a block device", journal_path
);
1822 kfree(journal_path
);
1826 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1828 kfree(journal_path
);
1829 } else if (token
== Opt_journal_ioprio
) {
1831 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1836 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1837 } else if (token
== Opt_test_dummy_encryption
) {
1838 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1839 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1840 ext4_msg(sb
, KERN_WARNING
,
1841 "Test dummy encryption mode enabled");
1843 ext4_msg(sb
, KERN_WARNING
,
1844 "Test dummy encryption mount option ignored");
1846 } else if (m
->flags
& MOPT_DATAJ
) {
1848 if (!sbi
->s_journal
)
1849 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1850 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1851 ext4_msg(sb
, KERN_ERR
,
1852 "Cannot change data mode on remount");
1856 clear_opt(sb
, DATA_FLAGS
);
1857 sbi
->s_mount_opt
|= m
->mount_opt
;
1860 } else if (m
->flags
& MOPT_QFMT
) {
1861 if (sb_any_quota_loaded(sb
) &&
1862 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1863 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1864 "quota options when quota turned on");
1867 if (ext4_has_feature_quota(sb
)) {
1868 ext4_msg(sb
, KERN_INFO
,
1869 "Quota format mount options ignored "
1870 "when QUOTA feature is enabled");
1873 sbi
->s_jquota_fmt
= m
->mount_opt
;
1875 } else if (token
== Opt_dax
) {
1876 #ifdef CONFIG_FS_DAX
1877 ext4_msg(sb
, KERN_WARNING
,
1878 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1879 sbi
->s_mount_opt
|= m
->mount_opt
;
1881 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1884 } else if (token
== Opt_data_err_abort
) {
1885 sbi
->s_mount_opt
|= m
->mount_opt
;
1886 } else if (token
== Opt_data_err_ignore
) {
1887 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1891 if (m
->flags
& MOPT_CLEAR
)
1893 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1894 ext4_msg(sb
, KERN_WARNING
,
1895 "buggy handling of option %s", opt
);
1900 sbi
->s_mount_opt
|= m
->mount_opt
;
1902 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1907 static int parse_options(char *options
, struct super_block
*sb
,
1908 unsigned long *journal_devnum
,
1909 unsigned int *journal_ioprio
,
1912 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1913 char *p
, __maybe_unused
*usr_qf_name
, __maybe_unused
*grp_qf_name
;
1914 substring_t args
[MAX_OPT_ARGS
];
1920 while ((p
= strsep(&options
, ",")) != NULL
) {
1924 * Initialize args struct so we know whether arg was
1925 * found; some options take optional arguments.
1927 args
[0].to
= args
[0].from
= NULL
;
1928 token
= match_token(p
, tokens
, args
);
1929 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1930 journal_ioprio
, is_remount
) < 0)
1935 * We do the test below only for project quotas. 'usrquota' and
1936 * 'grpquota' mount options are allowed even without quota feature
1937 * to support legacy quotas in quota files.
1939 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
1940 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
1941 "Cannot enable project quota enforcement.");
1944 usr_qf_name
= get_qf_name(sb
, sbi
, USRQUOTA
);
1945 grp_qf_name
= get_qf_name(sb
, sbi
, GRPQUOTA
);
1946 if (usr_qf_name
|| grp_qf_name
) {
1947 if (test_opt(sb
, USRQUOTA
) && usr_qf_name
)
1948 clear_opt(sb
, USRQUOTA
);
1950 if (test_opt(sb
, GRPQUOTA
) && grp_qf_name
)
1951 clear_opt(sb
, GRPQUOTA
);
1953 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1954 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1959 if (!sbi
->s_jquota_fmt
) {
1960 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1966 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1968 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1970 if (blocksize
< PAGE_SIZE
) {
1971 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1972 "dioread_nolock if block size != PAGE_SIZE");
1979 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1980 struct super_block
*sb
)
1982 #if defined(CONFIG_QUOTA)
1983 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1984 char *usr_qf_name
, *grp_qf_name
;
1986 if (sbi
->s_jquota_fmt
) {
1989 switch (sbi
->s_jquota_fmt
) {
2000 seq_printf(seq
, ",jqfmt=%s", fmtname
);
2004 usr_qf_name
= rcu_dereference(sbi
->s_qf_names
[USRQUOTA
]);
2005 grp_qf_name
= rcu_dereference(sbi
->s_qf_names
[GRPQUOTA
]);
2007 seq_show_option(seq
, "usrjquota", usr_qf_name
);
2009 seq_show_option(seq
, "grpjquota", grp_qf_name
);
2014 static const char *token2str(int token
)
2016 const struct match_token
*t
;
2018 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
2019 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
2026 * - it's set to a non-default value OR
2027 * - if the per-sb default is different from the global default
2029 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
2032 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2033 struct ext4_super_block
*es
= sbi
->s_es
;
2034 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
2035 const struct mount_opts
*m
;
2036 char sep
= nodefs
? '\n' : ',';
2038 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2039 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2041 if (sbi
->s_sb_block
!= 1)
2042 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
2044 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
2045 int want_set
= m
->flags
& MOPT_SET
;
2046 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
2047 (m
->flags
& MOPT_CLEAR_ERR
))
2049 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2050 continue; /* skip if same as the default */
2052 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2053 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2054 continue; /* select Opt_noFoo vs Opt_Foo */
2055 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2058 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(&init_user_ns
, EXT4_DEF_RESUID
)) ||
2059 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2060 SEQ_OPTS_PRINT("resuid=%u",
2061 from_kuid_munged(&init_user_ns
, sbi
->s_resuid
));
2062 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(&init_user_ns
, EXT4_DEF_RESGID
)) ||
2063 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2064 SEQ_OPTS_PRINT("resgid=%u",
2065 from_kgid_munged(&init_user_ns
, sbi
->s_resgid
));
2066 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2067 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2068 SEQ_OPTS_PUTS("errors=remount-ro");
2069 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2070 SEQ_OPTS_PUTS("errors=continue");
2071 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2072 SEQ_OPTS_PUTS("errors=panic");
2073 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2074 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2075 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2076 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2077 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2078 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2079 if (sb
->s_flags
& SB_I_VERSION
)
2080 SEQ_OPTS_PUTS("i_version");
2081 if (nodefs
|| sbi
->s_stripe
)
2082 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2083 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
2084 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2085 SEQ_OPTS_PUTS("data=journal");
2086 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2087 SEQ_OPTS_PUTS("data=ordered");
2088 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2089 SEQ_OPTS_PUTS("data=writeback");
2092 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2093 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2094 sbi
->s_inode_readahead_blks
);
2096 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
2097 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2098 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2099 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2100 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2101 if (test_opt(sb
, DATA_ERR_ABORT
))
2102 SEQ_OPTS_PUTS("data_err=abort");
2103 if (DUMMY_ENCRYPTION_ENABLED(sbi
))
2104 SEQ_OPTS_PUTS("test_dummy_encryption");
2106 ext4_show_quota_options(seq
, sb
);
2110 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2112 return _ext4_show_options(seq
, root
->d_sb
, 0);
2115 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2117 struct super_block
*sb
= seq
->private;
2120 seq_puts(seq
, sb_rdonly(sb
) ? "ro" : "rw");
2121 rc
= _ext4_show_options(seq
, sb
, 1);
2122 seq_puts(seq
, "\n");
2126 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2129 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2132 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2133 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2134 "forcing read-only mode");
2139 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2140 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2141 "running e2fsck is recommended");
2142 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2143 ext4_msg(sb
, KERN_WARNING
,
2144 "warning: mounting fs with errors, "
2145 "running e2fsck is recommended");
2146 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2147 le16_to_cpu(es
->s_mnt_count
) >=
2148 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2149 ext4_msg(sb
, KERN_WARNING
,
2150 "warning: maximal mount count reached, "
2151 "running e2fsck is recommended");
2152 else if (le32_to_cpu(es
->s_checkinterval
) &&
2153 (le32_to_cpu(es
->s_lastcheck
) +
2154 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
2155 ext4_msg(sb
, KERN_WARNING
,
2156 "warning: checktime reached, "
2157 "running e2fsck is recommended");
2158 if (!sbi
->s_journal
)
2159 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2160 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2161 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2162 le16_add_cpu(&es
->s_mnt_count
, 1);
2163 es
->s_mtime
= cpu_to_le32(get_seconds());
2164 ext4_update_dynamic_rev(sb
);
2166 ext4_set_feature_journal_needs_recovery(sb
);
2168 ext4_commit_super(sb
, 1);
2170 if (test_opt(sb
, DEBUG
))
2171 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2172 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2174 sbi
->s_groups_count
,
2175 EXT4_BLOCKS_PER_GROUP(sb
),
2176 EXT4_INODES_PER_GROUP(sb
),
2177 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2179 cleancache_init_fs(sb
);
2183 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2185 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2186 struct flex_groups
*new_groups
;
2189 if (!sbi
->s_log_groups_per_flex
)
2192 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2193 if (size
<= sbi
->s_flex_groups_allocated
)
2196 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2197 new_groups
= kvzalloc(size
, GFP_KERNEL
);
2199 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2200 size
/ (int) sizeof(struct flex_groups
));
2204 if (sbi
->s_flex_groups
) {
2205 memcpy(new_groups
, sbi
->s_flex_groups
,
2206 (sbi
->s_flex_groups_allocated
*
2207 sizeof(struct flex_groups
)));
2208 kvfree(sbi
->s_flex_groups
);
2210 sbi
->s_flex_groups
= new_groups
;
2211 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2215 static int ext4_fill_flex_info(struct super_block
*sb
)
2217 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2218 struct ext4_group_desc
*gdp
= NULL
;
2219 ext4_group_t flex_group
;
2222 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2223 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2224 sbi
->s_log_groups_per_flex
= 0;
2228 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2232 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2233 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2235 flex_group
= ext4_flex_group(sbi
, i
);
2236 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2237 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2238 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2239 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2240 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2241 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2249 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2250 struct ext4_group_desc
*gdp
)
2252 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2254 __le32 le_group
= cpu_to_le32(block_group
);
2255 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2257 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2258 /* Use new metadata_csum algorithm */
2260 __u16 dummy_csum
= 0;
2262 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2264 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2265 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2266 sizeof(dummy_csum
));
2267 offset
+= sizeof(dummy_csum
);
2268 if (offset
< sbi
->s_desc_size
)
2269 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2270 sbi
->s_desc_size
- offset
);
2272 crc
= csum32
& 0xFFFF;
2276 /* old crc16 code */
2277 if (!ext4_has_feature_gdt_csum(sb
))
2280 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2281 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2282 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2283 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2284 /* for checksum of struct ext4_group_desc do the rest...*/
2285 if (ext4_has_feature_64bit(sb
) &&
2286 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2287 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2288 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2292 return cpu_to_le16(crc
);
2295 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2296 struct ext4_group_desc
*gdp
)
2298 if (ext4_has_group_desc_csum(sb
) &&
2299 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2305 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2306 struct ext4_group_desc
*gdp
)
2308 if (!ext4_has_group_desc_csum(sb
))
2310 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2313 /* Called at mount-time, super-block is locked */
2314 static int ext4_check_descriptors(struct super_block
*sb
,
2315 ext4_fsblk_t sb_block
,
2316 ext4_group_t
*first_not_zeroed
)
2318 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2319 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2320 ext4_fsblk_t last_block
;
2321 ext4_fsblk_t last_bg_block
= sb_block
+ ext4_bg_num_gdb(sb
, 0);
2322 ext4_fsblk_t block_bitmap
;
2323 ext4_fsblk_t inode_bitmap
;
2324 ext4_fsblk_t inode_table
;
2325 int flexbg_flag
= 0;
2326 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2328 if (ext4_has_feature_flex_bg(sb
))
2331 ext4_debug("Checking group descriptors");
2333 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2334 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2336 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2337 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2339 last_block
= first_block
+
2340 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2342 if ((grp
== sbi
->s_groups_count
) &&
2343 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2346 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2347 if (block_bitmap
== sb_block
) {
2348 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2349 "Block bitmap for group %u overlaps "
2354 if (block_bitmap
>= sb_block
+ 1 &&
2355 block_bitmap
<= last_bg_block
) {
2356 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2357 "Block bitmap for group %u overlaps "
2358 "block group descriptors", i
);
2362 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2363 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2364 "Block bitmap for group %u not in group "
2365 "(block %llu)!", i
, block_bitmap
);
2368 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2369 if (inode_bitmap
== sb_block
) {
2370 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2371 "Inode bitmap for group %u overlaps "
2376 if (inode_bitmap
>= sb_block
+ 1 &&
2377 inode_bitmap
<= last_bg_block
) {
2378 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2379 "Inode bitmap for group %u overlaps "
2380 "block group descriptors", i
);
2384 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2385 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2386 "Inode bitmap for group %u not in group "
2387 "(block %llu)!", i
, inode_bitmap
);
2390 inode_table
= ext4_inode_table(sb
, gdp
);
2391 if (inode_table
== sb_block
) {
2392 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2393 "Inode table for group %u overlaps "
2398 if (inode_table
>= sb_block
+ 1 &&
2399 inode_table
<= last_bg_block
) {
2400 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2401 "Inode table for group %u overlaps "
2402 "block group descriptors", i
);
2406 if (inode_table
< first_block
||
2407 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2408 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2409 "Inode table for group %u not in group "
2410 "(block %llu)!", i
, inode_table
);
2413 ext4_lock_group(sb
, i
);
2414 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2415 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2416 "Checksum for group %u failed (%u!=%u)",
2417 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2418 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2419 if (!sb_rdonly(sb
)) {
2420 ext4_unlock_group(sb
, i
);
2424 ext4_unlock_group(sb
, i
);
2426 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2428 if (NULL
!= first_not_zeroed
)
2429 *first_not_zeroed
= grp
;
2433 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2434 * the superblock) which were deleted from all directories, but held open by
2435 * a process at the time of a crash. We walk the list and try to delete these
2436 * inodes at recovery time (only with a read-write filesystem).
2438 * In order to keep the orphan inode chain consistent during traversal (in
2439 * case of crash during recovery), we link each inode into the superblock
2440 * orphan list_head and handle it the same way as an inode deletion during
2441 * normal operation (which journals the operations for us).
2443 * We only do an iget() and an iput() on each inode, which is very safe if we
2444 * accidentally point at an in-use or already deleted inode. The worst that
2445 * can happen in this case is that we get a "bit already cleared" message from
2446 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2447 * e2fsck was run on this filesystem, and it must have already done the orphan
2448 * inode cleanup for us, so we can safely abort without any further action.
2450 static void ext4_orphan_cleanup(struct super_block
*sb
,
2451 struct ext4_super_block
*es
)
2453 unsigned int s_flags
= sb
->s_flags
;
2454 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2456 int quota_update
= 0;
2459 if (!es
->s_last_orphan
) {
2460 jbd_debug(4, "no orphan inodes to clean up\n");
2464 if (bdev_read_only(sb
->s_bdev
)) {
2465 ext4_msg(sb
, KERN_ERR
, "write access "
2466 "unavailable, skipping orphan cleanup");
2470 /* Check if feature set would not allow a r/w mount */
2471 if (!ext4_feature_set_ok(sb
, 0)) {
2472 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2473 "unknown ROCOMPAT features");
2477 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2478 /* don't clear list on RO mount w/ errors */
2479 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2480 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2481 "clearing orphan list.\n");
2482 es
->s_last_orphan
= 0;
2484 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2488 if (s_flags
& MS_RDONLY
) {
2489 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2490 sb
->s_flags
&= ~MS_RDONLY
;
2493 /* Needed for iput() to work correctly and not trash data */
2494 sb
->s_flags
|= MS_ACTIVE
;
2497 * Turn on quotas which were not enabled for read-only mounts if
2498 * filesystem has quota feature, so that they are updated correctly.
2500 if (ext4_has_feature_quota(sb
) && (s_flags
& MS_RDONLY
)) {
2501 int ret
= ext4_enable_quotas(sb
);
2506 ext4_msg(sb
, KERN_ERR
,
2507 "Cannot turn on quotas: error %d", ret
);
2510 /* Turn on journaled quotas used for old sytle */
2511 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2512 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2513 int ret
= ext4_quota_on_mount(sb
, i
);
2518 ext4_msg(sb
, KERN_ERR
,
2519 "Cannot turn on journaled "
2520 "quota: type %d: error %d", i
, ret
);
2525 while (es
->s_last_orphan
) {
2526 struct inode
*inode
;
2529 * We may have encountered an error during cleanup; if
2530 * so, skip the rest.
2532 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2533 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2534 es
->s_last_orphan
= 0;
2538 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2539 if (IS_ERR(inode
)) {
2540 es
->s_last_orphan
= 0;
2544 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2545 dquot_initialize(inode
);
2546 if (inode
->i_nlink
) {
2547 if (test_opt(sb
, DEBUG
))
2548 ext4_msg(sb
, KERN_DEBUG
,
2549 "%s: truncating inode %lu to %lld bytes",
2550 __func__
, inode
->i_ino
, inode
->i_size
);
2551 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2552 inode
->i_ino
, inode
->i_size
);
2554 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2555 ret
= ext4_truncate(inode
);
2557 ext4_std_error(inode
->i_sb
, ret
);
2558 inode_unlock(inode
);
2561 if (test_opt(sb
, DEBUG
))
2562 ext4_msg(sb
, KERN_DEBUG
,
2563 "%s: deleting unreferenced inode %lu",
2564 __func__
, inode
->i_ino
);
2565 jbd_debug(2, "deleting unreferenced inode %lu\n",
2569 iput(inode
); /* The delete magic happens here! */
2572 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2575 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2576 PLURAL(nr_orphans
));
2578 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2579 PLURAL(nr_truncates
));
2581 /* Turn off quotas if they were enabled for orphan cleanup */
2583 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2584 if (sb_dqopt(sb
)->files
[i
])
2585 dquot_quota_off(sb
, i
);
2589 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2593 * Maximal extent format file size.
2594 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2595 * extent format containers, within a sector_t, and within i_blocks
2596 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2597 * so that won't be a limiting factor.
2599 * However there is other limiting factor. We do store extents in the form
2600 * of starting block and length, hence the resulting length of the extent
2601 * covering maximum file size must fit into on-disk format containers as
2602 * well. Given that length is always by 1 unit bigger than max unit (because
2603 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2605 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2607 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2610 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2612 /* small i_blocks in vfs inode? */
2613 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2615 * CONFIG_LBDAF is not enabled implies the inode
2616 * i_block represent total blocks in 512 bytes
2617 * 32 == size of vfs inode i_blocks * 8
2619 upper_limit
= (1LL << 32) - 1;
2621 /* total blocks in file system block size */
2622 upper_limit
>>= (blkbits
- 9);
2623 upper_limit
<<= blkbits
;
2627 * 32-bit extent-start container, ee_block. We lower the maxbytes
2628 * by one fs block, so ee_len can cover the extent of maximum file
2631 res
= (1LL << 32) - 1;
2634 /* Sanity check against vm- & vfs- imposed limits */
2635 if (res
> upper_limit
)
2642 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2643 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2644 * We need to be 1 filesystem block less than the 2^48 sector limit.
2646 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2648 loff_t res
= EXT4_NDIR_BLOCKS
;
2651 /* This is calculated to be the largest file size for a dense, block
2652 * mapped file such that the file's total number of 512-byte sectors,
2653 * including data and all indirect blocks, does not exceed (2^48 - 1).
2655 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2656 * number of 512-byte sectors of the file.
2659 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2661 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2662 * the inode i_block field represents total file blocks in
2663 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2665 upper_limit
= (1LL << 32) - 1;
2667 /* total blocks in file system block size */
2668 upper_limit
>>= (bits
- 9);
2672 * We use 48 bit ext4_inode i_blocks
2673 * With EXT4_HUGE_FILE_FL set the i_blocks
2674 * represent total number of blocks in
2675 * file system block size
2677 upper_limit
= (1LL << 48) - 1;
2681 /* indirect blocks */
2683 /* double indirect blocks */
2684 meta_blocks
+= 1 + (1LL << (bits
-2));
2685 /* tripple indirect blocks */
2686 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2688 upper_limit
-= meta_blocks
;
2689 upper_limit
<<= bits
;
2691 res
+= 1LL << (bits
-2);
2692 res
+= 1LL << (2*(bits
-2));
2693 res
+= 1LL << (3*(bits
-2));
2695 if (res
> upper_limit
)
2698 if (res
> MAX_LFS_FILESIZE
)
2699 res
= MAX_LFS_FILESIZE
;
2704 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2705 ext4_fsblk_t logical_sb_block
, int nr
)
2707 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2708 ext4_group_t bg
, first_meta_bg
;
2711 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2713 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2714 return logical_sb_block
+ nr
+ 1;
2715 bg
= sbi
->s_desc_per_block
* nr
;
2716 if (ext4_bg_has_super(sb
, bg
))
2720 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2721 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2722 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2725 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2726 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2729 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2733 * ext4_get_stripe_size: Get the stripe size.
2734 * @sbi: In memory super block info
2736 * If we have specified it via mount option, then
2737 * use the mount option value. If the value specified at mount time is
2738 * greater than the blocks per group use the super block value.
2739 * If the super block value is greater than blocks per group return 0.
2740 * Allocator needs it be less than blocks per group.
2743 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2745 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2746 unsigned long stripe_width
=
2747 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2750 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2751 ret
= sbi
->s_stripe
;
2752 else if (stripe_width
&& stripe_width
<= sbi
->s_blocks_per_group
)
2754 else if (stride
&& stride
<= sbi
->s_blocks_per_group
)
2760 * If the stripe width is 1, this makes no sense and
2761 * we set it to 0 to turn off stripe handling code.
2770 * Check whether this filesystem can be mounted based on
2771 * the features present and the RDONLY/RDWR mount requested.
2772 * Returns 1 if this filesystem can be mounted as requested,
2773 * 0 if it cannot be.
2775 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2777 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2778 ext4_msg(sb
, KERN_ERR
,
2779 "Couldn't mount because of "
2780 "unsupported optional features (%x)",
2781 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2782 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2789 if (ext4_has_feature_readonly(sb
)) {
2790 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2791 sb
->s_flags
|= MS_RDONLY
;
2795 /* Check that feature set is OK for a read-write mount */
2796 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2797 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2798 "unsupported optional features (%x)",
2799 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2800 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2804 * Large file size enabled file system can only be mounted
2805 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2807 if (ext4_has_feature_huge_file(sb
)) {
2808 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2809 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2810 "cannot be mounted RDWR without "
2815 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2816 ext4_msg(sb
, KERN_ERR
,
2817 "Can't support bigalloc feature without "
2818 "extents feature\n");
2822 #ifndef CONFIG_QUOTA
2823 if (ext4_has_feature_quota(sb
) && !readonly
) {
2824 ext4_msg(sb
, KERN_ERR
,
2825 "Filesystem with quota feature cannot be mounted RDWR "
2826 "without CONFIG_QUOTA");
2829 if (ext4_has_feature_project(sb
) && !readonly
) {
2830 ext4_msg(sb
, KERN_ERR
,
2831 "Filesystem with project quota feature cannot be mounted RDWR "
2832 "without CONFIG_QUOTA");
2835 #endif /* CONFIG_QUOTA */
2840 * This function is called once a day if we have errors logged
2841 * on the file system
2843 static void print_daily_error_info(unsigned long arg
)
2845 struct super_block
*sb
= (struct super_block
*) arg
;
2846 struct ext4_sb_info
*sbi
;
2847 struct ext4_super_block
*es
;
2852 if (es
->s_error_count
)
2853 /* fsck newer than v1.41.13 is needed to clean this condition. */
2854 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2855 le32_to_cpu(es
->s_error_count
));
2856 if (es
->s_first_error_time
) {
2857 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2858 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2859 (int) sizeof(es
->s_first_error_func
),
2860 es
->s_first_error_func
,
2861 le32_to_cpu(es
->s_first_error_line
));
2862 if (es
->s_first_error_ino
)
2863 printk(KERN_CONT
": inode %u",
2864 le32_to_cpu(es
->s_first_error_ino
));
2865 if (es
->s_first_error_block
)
2866 printk(KERN_CONT
": block %llu", (unsigned long long)
2867 le64_to_cpu(es
->s_first_error_block
));
2868 printk(KERN_CONT
"\n");
2870 if (es
->s_last_error_time
) {
2871 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2872 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2873 (int) sizeof(es
->s_last_error_func
),
2874 es
->s_last_error_func
,
2875 le32_to_cpu(es
->s_last_error_line
));
2876 if (es
->s_last_error_ino
)
2877 printk(KERN_CONT
": inode %u",
2878 le32_to_cpu(es
->s_last_error_ino
));
2879 if (es
->s_last_error_block
)
2880 printk(KERN_CONT
": block %llu", (unsigned long long)
2881 le64_to_cpu(es
->s_last_error_block
));
2882 printk(KERN_CONT
"\n");
2884 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2887 /* Find next suitable group and run ext4_init_inode_table */
2888 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2890 struct ext4_group_desc
*gdp
= NULL
;
2891 ext4_group_t group
, ngroups
;
2892 struct super_block
*sb
;
2893 unsigned long timeout
= 0;
2897 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2899 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2900 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2906 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2910 if (group
>= ngroups
)
2915 ret
= ext4_init_inode_table(sb
, group
,
2916 elr
->lr_timeout
? 0 : 1);
2917 if (elr
->lr_timeout
== 0) {
2918 timeout
= (jiffies
- timeout
) *
2919 elr
->lr_sbi
->s_li_wait_mult
;
2920 elr
->lr_timeout
= timeout
;
2922 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2923 elr
->lr_next_group
= group
+ 1;
2929 * Remove lr_request from the list_request and free the
2930 * request structure. Should be called with li_list_mtx held
2932 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2934 struct ext4_sb_info
*sbi
;
2941 list_del(&elr
->lr_request
);
2942 sbi
->s_li_request
= NULL
;
2946 static void ext4_unregister_li_request(struct super_block
*sb
)
2948 mutex_lock(&ext4_li_mtx
);
2949 if (!ext4_li_info
) {
2950 mutex_unlock(&ext4_li_mtx
);
2954 mutex_lock(&ext4_li_info
->li_list_mtx
);
2955 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2956 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2957 mutex_unlock(&ext4_li_mtx
);
2960 static struct task_struct
*ext4_lazyinit_task
;
2963 * This is the function where ext4lazyinit thread lives. It walks
2964 * through the request list searching for next scheduled filesystem.
2965 * When such a fs is found, run the lazy initialization request
2966 * (ext4_rn_li_request) and keep track of the time spend in this
2967 * function. Based on that time we compute next schedule time of
2968 * the request. When walking through the list is complete, compute
2969 * next waking time and put itself into sleep.
2971 static int ext4_lazyinit_thread(void *arg
)
2973 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2974 struct list_head
*pos
, *n
;
2975 struct ext4_li_request
*elr
;
2976 unsigned long next_wakeup
, cur
;
2978 BUG_ON(NULL
== eli
);
2982 next_wakeup
= MAX_JIFFY_OFFSET
;
2984 mutex_lock(&eli
->li_list_mtx
);
2985 if (list_empty(&eli
->li_request_list
)) {
2986 mutex_unlock(&eli
->li_list_mtx
);
2989 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2992 elr
= list_entry(pos
, struct ext4_li_request
,
2995 if (time_before(jiffies
, elr
->lr_next_sched
)) {
2996 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2997 next_wakeup
= elr
->lr_next_sched
;
3000 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
3001 if (sb_start_write_trylock(elr
->lr_super
)) {
3004 * We hold sb->s_umount, sb can not
3005 * be removed from the list, it is
3006 * now safe to drop li_list_mtx
3008 mutex_unlock(&eli
->li_list_mtx
);
3009 err
= ext4_run_li_request(elr
);
3010 sb_end_write(elr
->lr_super
);
3011 mutex_lock(&eli
->li_list_mtx
);
3014 up_read((&elr
->lr_super
->s_umount
));
3016 /* error, remove the lazy_init job */
3018 ext4_remove_li_request(elr
);
3022 elr
->lr_next_sched
= jiffies
+
3024 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3026 if (time_before(elr
->lr_next_sched
, next_wakeup
))
3027 next_wakeup
= elr
->lr_next_sched
;
3029 mutex_unlock(&eli
->li_list_mtx
);
3034 if ((time_after_eq(cur
, next_wakeup
)) ||
3035 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
3040 schedule_timeout_interruptible(next_wakeup
- cur
);
3042 if (kthread_should_stop()) {
3043 ext4_clear_request_list();
3050 * It looks like the request list is empty, but we need
3051 * to check it under the li_list_mtx lock, to prevent any
3052 * additions into it, and of course we should lock ext4_li_mtx
3053 * to atomically free the list and ext4_li_info, because at
3054 * this point another ext4 filesystem could be registering
3057 mutex_lock(&ext4_li_mtx
);
3058 mutex_lock(&eli
->li_list_mtx
);
3059 if (!list_empty(&eli
->li_request_list
)) {
3060 mutex_unlock(&eli
->li_list_mtx
);
3061 mutex_unlock(&ext4_li_mtx
);
3064 mutex_unlock(&eli
->li_list_mtx
);
3065 kfree(ext4_li_info
);
3066 ext4_li_info
= NULL
;
3067 mutex_unlock(&ext4_li_mtx
);
3072 static void ext4_clear_request_list(void)
3074 struct list_head
*pos
, *n
;
3075 struct ext4_li_request
*elr
;
3077 mutex_lock(&ext4_li_info
->li_list_mtx
);
3078 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3079 elr
= list_entry(pos
, struct ext4_li_request
,
3081 ext4_remove_li_request(elr
);
3083 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3086 static int ext4_run_lazyinit_thread(void)
3088 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3089 ext4_li_info
, "ext4lazyinit");
3090 if (IS_ERR(ext4_lazyinit_task
)) {
3091 int err
= PTR_ERR(ext4_lazyinit_task
);
3092 ext4_clear_request_list();
3093 kfree(ext4_li_info
);
3094 ext4_li_info
= NULL
;
3095 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3096 "initialization thread\n",
3100 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3105 * Check whether it make sense to run itable init. thread or not.
3106 * If there is at least one uninitialized inode table, return
3107 * corresponding group number, else the loop goes through all
3108 * groups and return total number of groups.
3110 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3112 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3113 struct ext4_group_desc
*gdp
= NULL
;
3115 if (!ext4_has_group_desc_csum(sb
))
3118 for (group
= 0; group
< ngroups
; group
++) {
3119 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3123 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3130 static int ext4_li_info_new(void)
3132 struct ext4_lazy_init
*eli
= NULL
;
3134 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3138 INIT_LIST_HEAD(&eli
->li_request_list
);
3139 mutex_init(&eli
->li_list_mtx
);
3141 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3148 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3151 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3152 struct ext4_li_request
*elr
;
3154 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3160 elr
->lr_next_group
= start
;
3163 * Randomize first schedule time of the request to
3164 * spread the inode table initialization requests
3167 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3168 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3172 int ext4_register_li_request(struct super_block
*sb
,
3173 ext4_group_t first_not_zeroed
)
3175 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3176 struct ext4_li_request
*elr
= NULL
;
3177 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3180 mutex_lock(&ext4_li_mtx
);
3181 if (sbi
->s_li_request
!= NULL
) {
3183 * Reset timeout so it can be computed again, because
3184 * s_li_wait_mult might have changed.
3186 sbi
->s_li_request
->lr_timeout
= 0;
3190 if (first_not_zeroed
== ngroups
|| sb_rdonly(sb
) ||
3191 !test_opt(sb
, INIT_INODE_TABLE
))
3194 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3200 if (NULL
== ext4_li_info
) {
3201 ret
= ext4_li_info_new();
3206 mutex_lock(&ext4_li_info
->li_list_mtx
);
3207 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3208 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3210 sbi
->s_li_request
= elr
;
3212 * set elr to NULL here since it has been inserted to
3213 * the request_list and the removal and free of it is
3214 * handled by ext4_clear_request_list from now on.
3218 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3219 ret
= ext4_run_lazyinit_thread();
3224 mutex_unlock(&ext4_li_mtx
);
3231 * We do not need to lock anything since this is called on
3234 static void ext4_destroy_lazyinit_thread(void)
3237 * If thread exited earlier
3238 * there's nothing to be done.
3240 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3243 kthread_stop(ext4_lazyinit_task
);
3246 static int set_journal_csum_feature_set(struct super_block
*sb
)
3249 int compat
, incompat
;
3250 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3252 if (ext4_has_metadata_csum(sb
)) {
3253 /* journal checksum v3 */
3255 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3257 /* journal checksum v1 */
3258 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3262 jbd2_journal_clear_features(sbi
->s_journal
,
3263 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3264 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3265 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3266 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3267 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3269 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3271 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3272 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3275 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3276 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3278 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3279 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3286 * Note: calculating the overhead so we can be compatible with
3287 * historical BSD practice is quite difficult in the face of
3288 * clusters/bigalloc. This is because multiple metadata blocks from
3289 * different block group can end up in the same allocation cluster.
3290 * Calculating the exact overhead in the face of clustered allocation
3291 * requires either O(all block bitmaps) in memory or O(number of block
3292 * groups**2) in time. We will still calculate the superblock for
3293 * older file systems --- and if we come across with a bigalloc file
3294 * system with zero in s_overhead_clusters the estimate will be close to
3295 * correct especially for very large cluster sizes --- but for newer
3296 * file systems, it's better to calculate this figure once at mkfs
3297 * time, and store it in the superblock. If the superblock value is
3298 * present (even for non-bigalloc file systems), we will use it.
3300 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3303 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3304 struct ext4_group_desc
*gdp
;
3305 ext4_fsblk_t first_block
, last_block
, b
;
3306 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3307 int s
, j
, count
= 0;
3309 if (!ext4_has_feature_bigalloc(sb
))
3310 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3311 sbi
->s_itb_per_group
+ 2);
3313 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3314 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3315 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3316 for (i
= 0; i
< ngroups
; i
++) {
3317 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3318 b
= ext4_block_bitmap(sb
, gdp
);
3319 if (b
>= first_block
&& b
<= last_block
) {
3320 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3323 b
= ext4_inode_bitmap(sb
, gdp
);
3324 if (b
>= first_block
&& b
<= last_block
) {
3325 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3328 b
= ext4_inode_table(sb
, gdp
);
3329 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3330 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3331 int c
= EXT4_B2C(sbi
, b
- first_block
);
3332 ext4_set_bit(c
, buf
);
3338 if (ext4_bg_has_super(sb
, grp
)) {
3339 ext4_set_bit(s
++, buf
);
3342 j
= ext4_bg_num_gdb(sb
, grp
);
3343 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3344 ext4_error(sb
, "Invalid number of block group "
3345 "descriptor blocks: %d", j
);
3346 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3350 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3354 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3355 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3359 * Compute the overhead and stash it in sbi->s_overhead
3361 int ext4_calculate_overhead(struct super_block
*sb
)
3363 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3364 struct ext4_super_block
*es
= sbi
->s_es
;
3365 struct inode
*j_inode
;
3366 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3367 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3368 ext4_fsblk_t overhead
= 0;
3369 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3375 * Compute the overhead (FS structures). This is constant
3376 * for a given filesystem unless the number of block groups
3377 * changes so we cache the previous value until it does.
3381 * All of the blocks before first_data_block are overhead
3383 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3386 * Add the overhead found in each block group
3388 for (i
= 0; i
< ngroups
; i
++) {
3391 blks
= count_overhead(sb
, i
, buf
);
3394 memset(buf
, 0, PAGE_SIZE
);
3399 * Add the internal journal blocks whether the journal has been
3402 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3403 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3404 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3405 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3407 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3408 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3411 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3414 sbi
->s_overhead
= overhead
;
3416 free_page((unsigned long) buf
);
3420 static void ext4_set_resv_clusters(struct super_block
*sb
)
3422 ext4_fsblk_t resv_clusters
;
3423 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3426 * There's no need to reserve anything when we aren't using extents.
3427 * The space estimates are exact, there are no unwritten extents,
3428 * hole punching doesn't need new metadata... This is needed especially
3429 * to keep ext2/3 backward compatibility.
3431 if (!ext4_has_feature_extents(sb
))
3434 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3435 * This should cover the situations where we can not afford to run
3436 * out of space like for example punch hole, or converting
3437 * unwritten extents in delalloc path. In most cases such
3438 * allocation would require 1, or 2 blocks, higher numbers are
3441 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3442 sbi
->s_cluster_bits
);
3444 do_div(resv_clusters
, 50);
3445 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3447 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3450 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3452 struct dax_device
*dax_dev
= fs_dax_get_by_bdev(sb
->s_bdev
);
3453 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3454 struct buffer_head
*bh
;
3455 struct ext4_super_block
*es
= NULL
;
3456 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3458 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3459 ext4_fsblk_t logical_sb_block
;
3460 unsigned long offset
= 0;
3461 unsigned long journal_devnum
= 0;
3462 unsigned long def_mount_opts
;
3466 int blocksize
, clustersize
;
3467 unsigned int db_count
;
3469 int needs_recovery
, has_huge_files
, has_bigalloc
;
3472 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3473 ext4_group_t first_not_zeroed
;
3475 if ((data
&& !orig_data
) || !sbi
)
3478 sbi
->s_daxdev
= dax_dev
;
3479 sbi
->s_blockgroup_lock
=
3480 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3481 if (!sbi
->s_blockgroup_lock
)
3484 sb
->s_fs_info
= sbi
;
3486 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3487 sbi
->s_sb_block
= sb_block
;
3488 if (sb
->s_bdev
->bd_part
)
3489 sbi
->s_sectors_written_start
=
3490 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3492 /* Cleanup superblock name */
3493 strreplace(sb
->s_id
, '/', '!');
3495 /* -EINVAL is default */
3497 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3499 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3504 * The ext4 superblock will not be buffer aligned for other than 1kB
3505 * block sizes. We need to calculate the offset from buffer start.
3507 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3508 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3509 offset
= do_div(logical_sb_block
, blocksize
);
3511 logical_sb_block
= sb_block
;
3514 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3515 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3519 * Note: s_es must be initialized as soon as possible because
3520 * some ext4 macro-instructions depend on its value
3522 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3524 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3525 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3527 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3529 /* Warn if metadata_csum and gdt_csum are both set. */
3530 if (ext4_has_feature_metadata_csum(sb
) &&
3531 ext4_has_feature_gdt_csum(sb
))
3532 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3533 "redundant flags; please run fsck.");
3535 /* Check for a known checksum algorithm */
3536 if (!ext4_verify_csum_type(sb
, es
)) {
3537 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3538 "unknown checksum algorithm.");
3543 /* Load the checksum driver */
3544 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3545 if (IS_ERR(sbi
->s_chksum_driver
)) {
3546 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3547 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3548 sbi
->s_chksum_driver
= NULL
;
3552 /* Check superblock checksum */
3553 if (!ext4_superblock_csum_verify(sb
, es
)) {
3554 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3555 "invalid superblock checksum. Run e2fsck?");
3561 /* Precompute checksum seed for all metadata */
3562 if (ext4_has_feature_csum_seed(sb
))
3563 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3564 else if (ext4_has_metadata_csum(sb
) || ext4_has_feature_ea_inode(sb
))
3565 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3566 sizeof(es
->s_uuid
));
3568 /* Set defaults before we parse the mount options */
3569 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3570 set_opt(sb
, INIT_INODE_TABLE
);
3571 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3573 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3575 if (def_mount_opts
& EXT4_DEFM_UID16
)
3576 set_opt(sb
, NO_UID32
);
3577 /* xattr user namespace & acls are now defaulted on */
3578 set_opt(sb
, XATTR_USER
);
3579 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3580 set_opt(sb
, POSIX_ACL
);
3582 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3583 if (ext4_has_metadata_csum(sb
))
3584 set_opt(sb
, JOURNAL_CHECKSUM
);
3586 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3587 set_opt(sb
, JOURNAL_DATA
);
3588 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3589 set_opt(sb
, ORDERED_DATA
);
3590 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3591 set_opt(sb
, WRITEBACK_DATA
);
3593 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3594 set_opt(sb
, ERRORS_PANIC
);
3595 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3596 set_opt(sb
, ERRORS_CONT
);
3598 set_opt(sb
, ERRORS_RO
);
3599 /* block_validity enabled by default; disable with noblock_validity */
3600 set_opt(sb
, BLOCK_VALIDITY
);
3601 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3602 set_opt(sb
, DISCARD
);
3604 sbi
->s_resuid
= make_kuid(&init_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3605 sbi
->s_resgid
= make_kgid(&init_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3606 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3607 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3608 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3610 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3611 set_opt(sb
, BARRIER
);
3614 * enable delayed allocation by default
3615 * Use -o nodelalloc to turn it off
3617 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3618 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3619 set_opt(sb
, DELALLOC
);
3622 * set default s_li_wait_mult for lazyinit, for the case there is
3623 * no mount option specified.
3625 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3627 if (sbi
->s_es
->s_mount_opts
[0]) {
3628 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3629 sizeof(sbi
->s_es
->s_mount_opts
),
3633 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3634 &journal_ioprio
, 0)) {
3635 ext4_msg(sb
, KERN_WARNING
,
3636 "failed to parse options in superblock: %s",
3639 kfree(s_mount_opts
);
3641 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3642 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3643 &journal_ioprio
, 0))
3646 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3647 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3648 "with data=journal disables delayed "
3649 "allocation and O_DIRECT support!\n");
3650 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3651 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3652 "both data=journal and delalloc");
3655 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3656 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3657 "both data=journal and dioread_nolock");
3660 if (test_opt(sb
, DAX
)) {
3661 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3662 "both data=journal and dax");
3665 if (ext4_has_feature_encrypt(sb
)) {
3666 ext4_msg(sb
, KERN_WARNING
,
3667 "encrypted files will use data=ordered "
3668 "instead of data journaling mode");
3670 if (test_opt(sb
, DELALLOC
))
3671 clear_opt(sb
, DELALLOC
);
3673 sb
->s_iflags
|= SB_I_CGROUPWB
;
3676 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3677 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3679 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3680 (ext4_has_compat_features(sb
) ||
3681 ext4_has_ro_compat_features(sb
) ||
3682 ext4_has_incompat_features(sb
)))
3683 ext4_msg(sb
, KERN_WARNING
,
3684 "feature flags set on rev 0 fs, "
3685 "running e2fsck is recommended");
3687 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3688 set_opt2(sb
, HURD_COMPAT
);
3689 if (ext4_has_feature_64bit(sb
)) {
3690 ext4_msg(sb
, KERN_ERR
,
3691 "The Hurd can't support 64-bit file systems");
3696 * ea_inode feature uses l_i_version field which is not
3697 * available in HURD_COMPAT mode.
3699 if (ext4_has_feature_ea_inode(sb
)) {
3700 ext4_msg(sb
, KERN_ERR
,
3701 "ea_inode feature is not supported for Hurd");
3706 if (IS_EXT2_SB(sb
)) {
3707 if (ext2_feature_set_ok(sb
))
3708 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3709 "using the ext4 subsystem");
3712 * If we're probing be silent, if this looks like
3713 * it's actually an ext[34] filesystem.
3715 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3717 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3718 "to feature incompatibilities");
3723 if (IS_EXT3_SB(sb
)) {
3724 if (ext3_feature_set_ok(sb
))
3725 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3726 "using the ext4 subsystem");
3729 * If we're probing be silent, if this looks like
3730 * it's actually an ext4 filesystem.
3732 if (silent
&& ext4_feature_set_ok(sb
, sb_rdonly(sb
)))
3734 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3735 "to feature incompatibilities");
3741 * Check feature flags regardless of the revision level, since we
3742 * previously didn't change the revision level when setting the flags,
3743 * so there is a chance incompat flags are set on a rev 0 filesystem.
3745 if (!ext4_feature_set_ok(sb
, (sb_rdonly(sb
))))
3748 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3749 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3750 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3751 ext4_msg(sb
, KERN_ERR
,
3752 "Unsupported filesystem blocksize %d (%d log_block_size)",
3753 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3756 if (le32_to_cpu(es
->s_log_block_size
) >
3757 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3758 ext4_msg(sb
, KERN_ERR
,
3759 "Invalid log block size: %u",
3760 le32_to_cpu(es
->s_log_block_size
));
3763 if (le32_to_cpu(es
->s_log_cluster_size
) >
3764 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3765 ext4_msg(sb
, KERN_ERR
,
3766 "Invalid log cluster size: %u",
3767 le32_to_cpu(es
->s_log_cluster_size
));
3771 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3772 ext4_msg(sb
, KERN_ERR
,
3773 "Number of reserved GDT blocks insanely large: %d",
3774 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3778 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3779 if (ext4_has_feature_inline_data(sb
)) {
3780 ext4_msg(sb
, KERN_ERR
, "Cannot use DAX on a filesystem"
3781 " that may contain inline data");
3784 if (!bdev_dax_supported(sb
->s_bdev
, blocksize
))
3788 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3789 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3790 es
->s_encryption_level
);
3794 if (sb
->s_blocksize
!= blocksize
) {
3795 /* Validate the filesystem blocksize */
3796 if (!sb_set_blocksize(sb
, blocksize
)) {
3797 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3803 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3804 offset
= do_div(logical_sb_block
, blocksize
);
3805 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3807 ext4_msg(sb
, KERN_ERR
,
3808 "Can't read superblock on 2nd try");
3811 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3813 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3814 ext4_msg(sb
, KERN_ERR
,
3815 "Magic mismatch, very weird!");
3820 has_huge_files
= ext4_has_feature_huge_file(sb
);
3821 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3823 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3825 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3826 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3827 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3829 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3830 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3831 if (sbi
->s_first_ino
< EXT4_GOOD_OLD_FIRST_INO
) {
3832 ext4_msg(sb
, KERN_ERR
, "invalid first ino: %u",
3836 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3837 (!is_power_of_2(sbi
->s_inode_size
)) ||
3838 (sbi
->s_inode_size
> blocksize
)) {
3839 ext4_msg(sb
, KERN_ERR
,
3840 "unsupported inode size: %d",
3844 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3845 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3848 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3849 if (ext4_has_feature_64bit(sb
)) {
3850 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3851 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3852 !is_power_of_2(sbi
->s_desc_size
)) {
3853 ext4_msg(sb
, KERN_ERR
,
3854 "unsupported descriptor size %lu",
3859 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3861 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3862 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3864 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3865 if (sbi
->s_inodes_per_block
== 0)
3867 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3868 sbi
->s_inodes_per_group
> blocksize
* 8) {
3869 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3870 sbi
->s_blocks_per_group
);
3873 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3874 sbi
->s_inodes_per_block
;
3875 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3877 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3878 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3879 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3881 for (i
= 0; i
< 4; i
++)
3882 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3883 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3884 if (ext4_has_feature_dir_index(sb
)) {
3885 i
= le32_to_cpu(es
->s_flags
);
3886 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3887 sbi
->s_hash_unsigned
= 3;
3888 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3889 #ifdef __CHAR_UNSIGNED__
3892 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3893 sbi
->s_hash_unsigned
= 3;
3897 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3902 /* Handle clustersize */
3903 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3904 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
3906 if (clustersize
< blocksize
) {
3907 ext4_msg(sb
, KERN_ERR
,
3908 "cluster size (%d) smaller than "
3909 "block size (%d)", clustersize
, blocksize
);
3912 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3913 le32_to_cpu(es
->s_log_block_size
);
3914 sbi
->s_clusters_per_group
=
3915 le32_to_cpu(es
->s_clusters_per_group
);
3916 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3917 ext4_msg(sb
, KERN_ERR
,
3918 "#clusters per group too big: %lu",
3919 sbi
->s_clusters_per_group
);
3922 if (sbi
->s_blocks_per_group
!=
3923 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3924 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3925 "clusters per group (%lu) inconsistent",
3926 sbi
->s_blocks_per_group
,
3927 sbi
->s_clusters_per_group
);
3931 if (clustersize
!= blocksize
) {
3932 ext4_msg(sb
, KERN_ERR
,
3933 "fragment/cluster size (%d) != "
3934 "block size (%d)", clustersize
, blocksize
);
3937 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3938 ext4_msg(sb
, KERN_ERR
,
3939 "#blocks per group too big: %lu",
3940 sbi
->s_blocks_per_group
);
3943 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3944 sbi
->s_cluster_bits
= 0;
3946 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3948 /* Do we have standard group size of clustersize * 8 blocks ? */
3949 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3950 set_opt2(sb
, STD_GROUP_SIZE
);
3953 * Test whether we have more sectors than will fit in sector_t,
3954 * and whether the max offset is addressable by the page cache.
3956 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3957 ext4_blocks_count(es
));
3959 ext4_msg(sb
, KERN_ERR
, "filesystem"
3960 " too large to mount safely on this system");
3961 if (sizeof(sector_t
) < 8)
3962 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3966 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3969 /* check blocks count against device size */
3970 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3971 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3972 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3973 "exceeds size of device (%llu blocks)",
3974 ext4_blocks_count(es
), blocks_count
);
3979 * It makes no sense for the first data block to be beyond the end
3980 * of the filesystem.
3982 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3983 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3984 "block %u is beyond end of filesystem (%llu)",
3985 le32_to_cpu(es
->s_first_data_block
),
3986 ext4_blocks_count(es
));
3989 if ((es
->s_first_data_block
== 0) && (es
->s_log_block_size
== 0) &&
3990 (sbi
->s_cluster_ratio
== 1)) {
3991 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3992 "block is 0 with a 1k block and cluster size");
3996 blocks_count
= (ext4_blocks_count(es
) -
3997 le32_to_cpu(es
->s_first_data_block
) +
3998 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3999 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
4000 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
4001 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
4002 "(block count %llu, first data block %u, "
4003 "blocks per group %lu)", sbi
->s_groups_count
,
4004 ext4_blocks_count(es
),
4005 le32_to_cpu(es
->s_first_data_block
),
4006 EXT4_BLOCKS_PER_GROUP(sb
));
4009 sbi
->s_groups_count
= blocks_count
;
4010 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
4011 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
4012 if (((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
) !=
4013 le32_to_cpu(es
->s_inodes_count
)) {
4014 ext4_msg(sb
, KERN_ERR
, "inodes count not valid: %u vs %llu",
4015 le32_to_cpu(es
->s_inodes_count
),
4016 ((u64
)sbi
->s_groups_count
* sbi
->s_inodes_per_group
));
4020 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
4021 EXT4_DESC_PER_BLOCK(sb
);
4022 if (ext4_has_feature_meta_bg(sb
)) {
4023 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
4024 ext4_msg(sb
, KERN_WARNING
,
4025 "first meta block group too large: %u "
4026 "(group descriptor block count %u)",
4027 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
4031 sbi
->s_group_desc
= kvmalloc(db_count
*
4032 sizeof(struct buffer_head
*),
4034 if (sbi
->s_group_desc
== NULL
) {
4035 ext4_msg(sb
, KERN_ERR
, "not enough memory");
4040 bgl_lock_init(sbi
->s_blockgroup_lock
);
4042 /* Pre-read the descriptors into the buffer cache */
4043 for (i
= 0; i
< db_count
; i
++) {
4044 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4045 sb_breadahead(sb
, block
);
4048 for (i
= 0; i
< db_count
; i
++) {
4049 block
= descriptor_loc(sb
, logical_sb_block
, i
);
4050 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
4051 if (!sbi
->s_group_desc
[i
]) {
4052 ext4_msg(sb
, KERN_ERR
,
4053 "can't read group descriptor %d", i
);
4058 sbi
->s_gdb_count
= db_count
;
4059 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
4060 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
4061 ret
= -EFSCORRUPTED
;
4065 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
4066 spin_lock_init(&sbi
->s_next_gen_lock
);
4068 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
4069 (unsigned long) sb
);
4071 /* Register extent status tree shrinker */
4072 if (ext4_es_register_shrinker(sbi
))
4075 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
4076 sbi
->s_extent_max_zeroout_kb
= 32;
4079 * set up enough so that it can read an inode
4081 sb
->s_op
= &ext4_sops
;
4082 sb
->s_export_op
= &ext4_export_ops
;
4083 sb
->s_xattr
= ext4_xattr_handlers
;
4084 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4085 sb
->s_cop
= &ext4_cryptops
;
4088 sb
->dq_op
= &ext4_quota_operations
;
4089 if (ext4_has_feature_quota(sb
))
4090 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
4092 sb
->s_qcop
= &ext4_qctl_operations
;
4093 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
4095 memcpy(&sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
4097 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
4098 mutex_init(&sbi
->s_orphan_lock
);
4102 needs_recovery
= (es
->s_last_orphan
!= 0 ||
4103 ext4_has_feature_journal_needs_recovery(sb
));
4105 if (ext4_has_feature_mmp(sb
) && !sb_rdonly(sb
))
4106 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
4107 goto failed_mount3a
;
4110 * The first inode we look at is the journal inode. Don't try
4111 * root first: it may be modified in the journal!
4113 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
4114 err
= ext4_load_journal(sb
, es
, journal_devnum
);
4116 goto failed_mount3a
;
4117 } else if (test_opt(sb
, NOLOAD
) && !sb_rdonly(sb
) &&
4118 ext4_has_feature_journal_needs_recovery(sb
)) {
4119 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
4120 "suppressed and not mounted read-only");
4121 goto failed_mount_wq
;
4123 /* Nojournal mode, all journal mount options are illegal */
4124 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
4125 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4126 "journal_checksum, fs mounted w/o journal");
4127 goto failed_mount_wq
;
4129 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4130 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4131 "journal_async_commit, fs mounted w/o journal");
4132 goto failed_mount_wq
;
4134 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
4135 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4136 "commit=%lu, fs mounted w/o journal",
4137 sbi
->s_commit_interval
/ HZ
);
4138 goto failed_mount_wq
;
4140 if (EXT4_MOUNT_DATA_FLAGS
&
4141 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
4142 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4143 "data=, fs mounted w/o journal");
4144 goto failed_mount_wq
;
4146 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4147 clear_opt(sb
, JOURNAL_CHECKSUM
);
4148 clear_opt(sb
, DATA_FLAGS
);
4149 sbi
->s_journal
= NULL
;
4154 if (ext4_has_feature_64bit(sb
) &&
4155 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4156 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4157 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4158 goto failed_mount_wq
;
4161 if (!set_journal_csum_feature_set(sb
)) {
4162 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4164 goto failed_mount_wq
;
4167 /* We have now updated the journal if required, so we can
4168 * validate the data journaling mode. */
4169 switch (test_opt(sb
, DATA_FLAGS
)) {
4171 /* No mode set, assume a default based on the journal
4172 * capabilities: ORDERED_DATA if the journal can
4173 * cope, else JOURNAL_DATA
4175 if (jbd2_journal_check_available_features
4176 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4177 set_opt(sb
, ORDERED_DATA
);
4179 set_opt(sb
, JOURNAL_DATA
);
4182 case EXT4_MOUNT_ORDERED_DATA
:
4183 case EXT4_MOUNT_WRITEBACK_DATA
:
4184 if (!jbd2_journal_check_available_features
4185 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4186 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4187 "requested data journaling mode");
4188 goto failed_mount_wq
;
4194 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4195 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4196 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4197 "journal_async_commit in data=ordered mode");
4198 goto failed_mount_wq
;
4201 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4203 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4206 if (!test_opt(sb
, NO_MBCACHE
)) {
4207 sbi
->s_ea_block_cache
= ext4_xattr_create_cache();
4208 if (!sbi
->s_ea_block_cache
) {
4209 ext4_msg(sb
, KERN_ERR
,
4210 "Failed to create ea_block_cache");
4211 goto failed_mount_wq
;
4214 if (ext4_has_feature_ea_inode(sb
)) {
4215 sbi
->s_ea_inode_cache
= ext4_xattr_create_cache();
4216 if (!sbi
->s_ea_inode_cache
) {
4217 ext4_msg(sb
, KERN_ERR
,
4218 "Failed to create ea_inode_cache");
4219 goto failed_mount_wq
;
4224 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4225 (blocksize
!= PAGE_SIZE
)) {
4226 ext4_msg(sb
, KERN_ERR
,
4227 "Unsupported blocksize for fs encryption");
4228 goto failed_mount_wq
;
4231 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !sb_rdonly(sb
) &&
4232 !ext4_has_feature_encrypt(sb
)) {
4233 ext4_set_feature_encrypt(sb
);
4234 ext4_commit_super(sb
, 1);
4238 * Get the # of file system overhead blocks from the
4239 * superblock if present.
4241 if (es
->s_overhead_clusters
)
4242 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4244 err
= ext4_calculate_overhead(sb
);
4246 goto failed_mount_wq
;
4250 * The maximum number of concurrent works can be high and
4251 * concurrency isn't really necessary. Limit it to 1.
4253 EXT4_SB(sb
)->rsv_conversion_wq
=
4254 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4255 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4256 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4262 * The jbd2_journal_load will have done any necessary log recovery,
4263 * so we can safely mount the rest of the filesystem now.
4266 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4268 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4269 ret
= PTR_ERR(root
);
4273 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4274 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4278 sb
->s_root
= d_make_root(root
);
4280 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4285 if (ext4_setup_super(sb
, es
, sb_rdonly(sb
)))
4286 sb
->s_flags
|= MS_RDONLY
;
4288 /* determine the minimum size of new large inodes, if present */
4289 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
&&
4290 sbi
->s_want_extra_isize
== 0) {
4291 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4292 EXT4_GOOD_OLD_INODE_SIZE
;
4293 if (ext4_has_feature_extra_isize(sb
)) {
4294 if (sbi
->s_want_extra_isize
<
4295 le16_to_cpu(es
->s_want_extra_isize
))
4296 sbi
->s_want_extra_isize
=
4297 le16_to_cpu(es
->s_want_extra_isize
);
4298 if (sbi
->s_want_extra_isize
<
4299 le16_to_cpu(es
->s_min_extra_isize
))
4300 sbi
->s_want_extra_isize
=
4301 le16_to_cpu(es
->s_min_extra_isize
);
4304 /* Check if enough inode space is available */
4305 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4306 sbi
->s_inode_size
) {
4307 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4308 EXT4_GOOD_OLD_INODE_SIZE
;
4309 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4313 ext4_set_resv_clusters(sb
);
4315 err
= ext4_setup_system_zone(sb
);
4317 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4319 goto failed_mount4a
;
4323 err
= ext4_mb_init(sb
);
4325 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4330 block
= ext4_count_free_clusters(sb
);
4331 ext4_free_blocks_count_set(sbi
->s_es
,
4332 EXT4_C2B(sbi
, block
));
4333 ext4_superblock_csum_set(sb
);
4334 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4337 unsigned long freei
= ext4_count_free_inodes(sb
);
4338 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4339 ext4_superblock_csum_set(sb
);
4340 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4344 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4345 ext4_count_dirs(sb
), GFP_KERNEL
);
4347 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4350 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4353 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4357 if (ext4_has_feature_flex_bg(sb
))
4358 if (!ext4_fill_flex_info(sb
)) {
4359 ext4_msg(sb
, KERN_ERR
,
4360 "unable to initialize "
4361 "flex_bg meta info!");
4365 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4369 err
= ext4_register_sysfs(sb
);
4374 /* Enable quota usage during mount. */
4375 if (ext4_has_feature_quota(sb
) && !sb_rdonly(sb
)) {
4376 err
= ext4_enable_quotas(sb
);
4380 #endif /* CONFIG_QUOTA */
4382 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4383 ext4_orphan_cleanup(sb
, es
);
4384 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4385 if (needs_recovery
) {
4386 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4387 ext4_mark_recovery_complete(sb
, es
);
4389 if (EXT4_SB(sb
)->s_journal
) {
4390 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4391 descr
= " journalled data mode";
4392 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4393 descr
= " ordered data mode";
4395 descr
= " writeback data mode";
4397 descr
= "out journal";
4399 if (test_opt(sb
, DISCARD
)) {
4400 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4401 if (!blk_queue_discard(q
))
4402 ext4_msg(sb
, KERN_WARNING
,
4403 "mounting with \"discard\" option, but "
4404 "the device does not support discard");
4407 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4408 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4409 "Opts: %.*s%s%s", descr
,
4410 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4411 sbi
->s_es
->s_mount_opts
,
4412 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4414 if (es
->s_error_count
)
4415 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4417 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4418 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4419 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4420 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4427 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4432 ext4_unregister_sysfs(sb
);
4435 ext4_unregister_li_request(sb
);
4437 ext4_mb_release(sb
);
4438 if (sbi
->s_flex_groups
)
4439 kvfree(sbi
->s_flex_groups
);
4440 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4441 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4442 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4443 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4444 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
4446 ext4_ext_release(sb
);
4447 ext4_release_system_zone(sb
);
4452 ext4_msg(sb
, KERN_ERR
, "mount failed");
4453 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4454 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4456 if (sbi
->s_ea_inode_cache
) {
4457 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
4458 sbi
->s_ea_inode_cache
= NULL
;
4460 if (sbi
->s_ea_block_cache
) {
4461 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
4462 sbi
->s_ea_block_cache
= NULL
;
4464 if (sbi
->s_journal
) {
4465 jbd2_journal_destroy(sbi
->s_journal
);
4466 sbi
->s_journal
= NULL
;
4469 ext4_es_unregister_shrinker(sbi
);
4471 del_timer_sync(&sbi
->s_err_report
);
4473 kthread_stop(sbi
->s_mmp_tsk
);
4475 for (i
= 0; i
< db_count
; i
++)
4476 brelse(sbi
->s_group_desc
[i
]);
4477 kvfree(sbi
->s_group_desc
);
4479 if (sbi
->s_chksum_driver
)
4480 crypto_free_shash(sbi
->s_chksum_driver
);
4482 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4483 kfree(sbi
->s_qf_names
[i
]);
4485 ext4_blkdev_remove(sbi
);
4488 sb
->s_fs_info
= NULL
;
4489 kfree(sbi
->s_blockgroup_lock
);
4493 fs_put_dax(dax_dev
);
4494 return err
? err
: ret
;
4498 * Setup any per-fs journal parameters now. We'll do this both on
4499 * initial mount, once the journal has been initialised but before we've
4500 * done any recovery; and again on any subsequent remount.
4502 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4504 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4506 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4507 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4508 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4510 write_lock(&journal
->j_state_lock
);
4511 if (test_opt(sb
, BARRIER
))
4512 journal
->j_flags
|= JBD2_BARRIER
;
4514 journal
->j_flags
&= ~JBD2_BARRIER
;
4515 if (test_opt(sb
, DATA_ERR_ABORT
))
4516 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4518 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4519 write_unlock(&journal
->j_state_lock
);
4522 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4523 unsigned int journal_inum
)
4525 struct inode
*journal_inode
;
4528 * Test for the existence of a valid inode on disk. Bad things
4529 * happen if we iget() an unused inode, as the subsequent iput()
4530 * will try to delete it.
4532 journal_inode
= ext4_iget(sb
, journal_inum
);
4533 if (IS_ERR(journal_inode
)) {
4534 ext4_msg(sb
, KERN_ERR
, "no journal found");
4537 if (!journal_inode
->i_nlink
) {
4538 make_bad_inode(journal_inode
);
4539 iput(journal_inode
);
4540 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4544 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4545 journal_inode
, journal_inode
->i_size
);
4546 if (!S_ISREG(journal_inode
->i_mode
)) {
4547 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4548 iput(journal_inode
);
4551 return journal_inode
;
4554 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4555 unsigned int journal_inum
)
4557 struct inode
*journal_inode
;
4560 BUG_ON(!ext4_has_feature_journal(sb
));
4562 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4566 journal
= jbd2_journal_init_inode(journal_inode
);
4568 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4569 iput(journal_inode
);
4572 journal
->j_private
= sb
;
4573 ext4_init_journal_params(sb
, journal
);
4577 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4580 struct buffer_head
*bh
;
4584 int hblock
, blocksize
;
4585 ext4_fsblk_t sb_block
;
4586 unsigned long offset
;
4587 struct ext4_super_block
*es
;
4588 struct block_device
*bdev
;
4590 BUG_ON(!ext4_has_feature_journal(sb
));
4592 bdev
= ext4_blkdev_get(j_dev
, sb
);
4596 blocksize
= sb
->s_blocksize
;
4597 hblock
= bdev_logical_block_size(bdev
);
4598 if (blocksize
< hblock
) {
4599 ext4_msg(sb
, KERN_ERR
,
4600 "blocksize too small for journal device");
4604 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4605 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4606 set_blocksize(bdev
, blocksize
);
4607 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4608 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4609 "external journal");
4613 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4614 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4615 !(le32_to_cpu(es
->s_feature_incompat
) &
4616 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4617 ext4_msg(sb
, KERN_ERR
, "external journal has "
4623 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4624 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4625 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4626 ext4_msg(sb
, KERN_ERR
, "external journal has "
4627 "corrupt superblock");
4632 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4633 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4638 len
= ext4_blocks_count(es
);
4639 start
= sb_block
+ 1;
4640 brelse(bh
); /* we're done with the superblock */
4642 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4643 start
, len
, blocksize
);
4645 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4648 journal
->j_private
= sb
;
4649 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4650 wait_on_buffer(journal
->j_sb_buffer
);
4651 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4652 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4655 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4656 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4657 "user (unsupported) - %d",
4658 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4661 EXT4_SB(sb
)->journal_bdev
= bdev
;
4662 ext4_init_journal_params(sb
, journal
);
4666 jbd2_journal_destroy(journal
);
4668 ext4_blkdev_put(bdev
);
4672 static int ext4_load_journal(struct super_block
*sb
,
4673 struct ext4_super_block
*es
,
4674 unsigned long journal_devnum
)
4677 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4680 int really_read_only
;
4682 BUG_ON(!ext4_has_feature_journal(sb
));
4684 if (journal_devnum
&&
4685 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4686 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4687 "numbers have changed");
4688 journal_dev
= new_decode_dev(journal_devnum
);
4690 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4692 really_read_only
= bdev_read_only(sb
->s_bdev
);
4695 * Are we loading a blank journal or performing recovery after a
4696 * crash? For recovery, we need to check in advance whether we
4697 * can get read-write access to the device.
4699 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4700 if (sb_rdonly(sb
)) {
4701 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4702 "required on readonly filesystem");
4703 if (really_read_only
) {
4704 ext4_msg(sb
, KERN_ERR
, "write access "
4705 "unavailable, cannot proceed");
4708 ext4_msg(sb
, KERN_INFO
, "write access will "
4709 "be enabled during recovery");
4713 if (journal_inum
&& journal_dev
) {
4714 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4715 "and inode journals!");
4720 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4723 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4727 if (!(journal
->j_flags
& JBD2_BARRIER
))
4728 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4730 if (!ext4_has_feature_journal_needs_recovery(sb
))
4731 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4733 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4735 memcpy(save
, ((char *) es
) +
4736 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4737 err
= jbd2_journal_load(journal
);
4739 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4740 save
, EXT4_S_ERR_LEN
);
4745 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4746 jbd2_journal_destroy(journal
);
4750 EXT4_SB(sb
)->s_journal
= journal
;
4751 ext4_clear_journal_err(sb
, es
);
4753 if (!really_read_only
&& journal_devnum
&&
4754 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4755 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4757 /* Make sure we flush the recovery flag to disk. */
4758 ext4_commit_super(sb
, 1);
4764 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4766 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4767 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4770 if (!sbh
|| block_device_ejected(sb
))
4774 * The superblock bh should be mapped, but it might not be if the
4775 * device was hot-removed. Not much we can do but fail the I/O.
4777 if (!buffer_mapped(sbh
))
4781 * If the file system is mounted read-only, don't update the
4782 * superblock write time. This avoids updating the superblock
4783 * write time when we are mounting the root file system
4784 * read/only but we need to replay the journal; at that point,
4785 * for people who are east of GMT and who make their clock
4786 * tick in localtime for Windows bug-for-bug compatibility,
4787 * the clock is set in the future, and this will cause e2fsck
4788 * to complain and force a full file system check.
4790 if (!(sb
->s_flags
& MS_RDONLY
))
4791 es
->s_wtime
= cpu_to_le32(get_seconds());
4792 if (sb
->s_bdev
->bd_part
)
4793 es
->s_kbytes_written
=
4794 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4795 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4796 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4798 es
->s_kbytes_written
=
4799 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4800 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4801 ext4_free_blocks_count_set(es
,
4802 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4803 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4804 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4805 es
->s_free_inodes_count
=
4806 cpu_to_le32(percpu_counter_sum_positive(
4807 &EXT4_SB(sb
)->s_freeinodes_counter
));
4808 BUFFER_TRACE(sbh
, "marking dirty");
4809 ext4_superblock_csum_set(sb
);
4812 if (buffer_write_io_error(sbh
)) {
4814 * Oh, dear. A previous attempt to write the
4815 * superblock failed. This could happen because the
4816 * USB device was yanked out. Or it could happen to
4817 * be a transient write error and maybe the block will
4818 * be remapped. Nothing we can do but to retry the
4819 * write and hope for the best.
4821 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4822 "superblock detected");
4823 clear_buffer_write_io_error(sbh
);
4824 set_buffer_uptodate(sbh
);
4826 mark_buffer_dirty(sbh
);
4829 error
= __sync_dirty_buffer(sbh
,
4830 REQ_SYNC
| (test_opt(sb
, BARRIER
) ? REQ_FUA
: 0));
4834 error
= buffer_write_io_error(sbh
);
4836 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4838 clear_buffer_write_io_error(sbh
);
4839 set_buffer_uptodate(sbh
);
4846 * Have we just finished recovery? If so, and if we are mounting (or
4847 * remounting) the filesystem readonly, then we will end up with a
4848 * consistent fs on disk. Record that fact.
4850 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4851 struct ext4_super_block
*es
)
4853 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4855 if (!ext4_has_feature_journal(sb
)) {
4856 BUG_ON(journal
!= NULL
);
4859 jbd2_journal_lock_updates(journal
);
4860 if (jbd2_journal_flush(journal
) < 0)
4863 if (ext4_has_feature_journal_needs_recovery(sb
) && sb_rdonly(sb
)) {
4864 ext4_clear_feature_journal_needs_recovery(sb
);
4865 ext4_commit_super(sb
, 1);
4869 jbd2_journal_unlock_updates(journal
);
4873 * If we are mounting (or read-write remounting) a filesystem whose journal
4874 * has recorded an error from a previous lifetime, move that error to the
4875 * main filesystem now.
4877 static void ext4_clear_journal_err(struct super_block
*sb
,
4878 struct ext4_super_block
*es
)
4884 BUG_ON(!ext4_has_feature_journal(sb
));
4886 journal
= EXT4_SB(sb
)->s_journal
;
4889 * Now check for any error status which may have been recorded in the
4890 * journal by a prior ext4_error() or ext4_abort()
4893 j_errno
= jbd2_journal_errno(journal
);
4897 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4898 ext4_warning(sb
, "Filesystem error recorded "
4899 "from previous mount: %s", errstr
);
4900 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4902 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4903 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4904 ext4_commit_super(sb
, 1);
4906 jbd2_journal_clear_err(journal
);
4907 jbd2_journal_update_sb_errno(journal
);
4912 * Force the running and committing transactions to commit,
4913 * and wait on the commit.
4915 int ext4_force_commit(struct super_block
*sb
)
4922 journal
= EXT4_SB(sb
)->s_journal
;
4923 return ext4_journal_force_commit(journal
);
4926 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4930 bool needs_barrier
= false;
4931 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4933 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
4936 trace_ext4_sync_fs(sb
, wait
);
4937 flush_workqueue(sbi
->rsv_conversion_wq
);
4939 * Writeback quota in non-journalled quota case - journalled quota has
4942 dquot_writeback_dquots(sb
, -1);
4944 * Data writeback is possible w/o journal transaction, so barrier must
4945 * being sent at the end of the function. But we can skip it if
4946 * transaction_commit will do it for us.
4948 if (sbi
->s_journal
) {
4949 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4950 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4951 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4952 needs_barrier
= true;
4954 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4956 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4959 } else if (wait
&& test_opt(sb
, BARRIER
))
4960 needs_barrier
= true;
4961 if (needs_barrier
) {
4963 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4972 * LVM calls this function before a (read-only) snapshot is created. This
4973 * gives us a chance to flush the journal completely and mark the fs clean.
4975 * Note that only this function cannot bring a filesystem to be in a clean
4976 * state independently. It relies on upper layer to stop all data & metadata
4979 static int ext4_freeze(struct super_block
*sb
)
4987 journal
= EXT4_SB(sb
)->s_journal
;
4990 /* Now we set up the journal barrier. */
4991 jbd2_journal_lock_updates(journal
);
4994 * Don't clear the needs_recovery flag if we failed to
4995 * flush the journal.
4997 error
= jbd2_journal_flush(journal
);
5001 /* Journal blocked and flushed, clear needs_recovery flag. */
5002 ext4_clear_feature_journal_needs_recovery(sb
);
5005 error
= ext4_commit_super(sb
, 1);
5008 /* we rely on upper layer to stop further updates */
5009 jbd2_journal_unlock_updates(journal
);
5014 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5015 * flag here, even though the filesystem is not technically dirty yet.
5017 static int ext4_unfreeze(struct super_block
*sb
)
5019 if (sb_rdonly(sb
) || ext4_forced_shutdown(EXT4_SB(sb
)))
5022 if (EXT4_SB(sb
)->s_journal
) {
5023 /* Reset the needs_recovery flag before the fs is unlocked. */
5024 ext4_set_feature_journal_needs_recovery(sb
);
5027 ext4_commit_super(sb
, 1);
5032 * Structure to save mount options for ext4_remount's benefit
5034 struct ext4_mount_options
{
5035 unsigned long s_mount_opt
;
5036 unsigned long s_mount_opt2
;
5039 unsigned long s_commit_interval
;
5040 u32 s_min_batch_time
, s_max_batch_time
;
5043 char *s_qf_names
[EXT4_MAXQUOTAS
];
5047 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
5049 struct ext4_super_block
*es
;
5050 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5051 unsigned long old_sb_flags
;
5052 struct ext4_mount_options old_opts
;
5053 int enable_quota
= 0;
5055 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
5059 char *to_free
[EXT4_MAXQUOTAS
];
5061 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
5063 /* Store the original options */
5064 old_sb_flags
= sb
->s_flags
;
5065 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
5066 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
5067 old_opts
.s_resuid
= sbi
->s_resuid
;
5068 old_opts
.s_resgid
= sbi
->s_resgid
;
5069 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
5070 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
5071 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
5073 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
5074 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5075 if (sbi
->s_qf_names
[i
]) {
5076 char *qf_name
= get_qf_name(sb
, sbi
, i
);
5078 old_opts
.s_qf_names
[i
] = kstrdup(qf_name
, GFP_KERNEL
);
5079 if (!old_opts
.s_qf_names
[i
]) {
5080 for (j
= 0; j
< i
; j
++)
5081 kfree(old_opts
.s_qf_names
[j
]);
5086 old_opts
.s_qf_names
[i
] = NULL
;
5088 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
5089 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
5091 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
5096 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
5097 test_opt(sb
, JOURNAL_CHECKSUM
)) {
5098 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
5099 "during remount not supported; ignoring");
5100 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
5103 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
5104 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
5105 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5106 "both data=journal and delalloc");
5110 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
5111 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5112 "both data=journal and dioread_nolock");
5116 if (test_opt(sb
, DAX
)) {
5117 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5118 "both data=journal and dax");
5122 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
5123 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
5124 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5125 "journal_async_commit in data=ordered mode");
5131 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_NO_MBCACHE
) {
5132 ext4_msg(sb
, KERN_ERR
, "can't enable nombcache during remount");
5137 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
5138 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
5139 "dax flag with busy inodes while remounting");
5140 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
5143 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
5144 ext4_abort(sb
, "Abort forced by user");
5146 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
5147 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
5151 if (sbi
->s_journal
) {
5152 ext4_init_journal_params(sb
, sbi
->s_journal
);
5153 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
5156 if (*flags
& MS_LAZYTIME
)
5157 sb
->s_flags
|= MS_LAZYTIME
;
5159 if ((bool)(*flags
& MS_RDONLY
) != sb_rdonly(sb
)) {
5160 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
5165 if (*flags
& MS_RDONLY
) {
5166 err
= sync_filesystem(sb
);
5169 err
= dquot_suspend(sb
, -1);
5174 * First of all, the unconditional stuff we have to do
5175 * to disable replay of the journal when we next remount
5177 sb
->s_flags
|= MS_RDONLY
;
5180 * OK, test if we are remounting a valid rw partition
5181 * readonly, and if so set the rdonly flag and then
5182 * mark the partition as valid again.
5184 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5185 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5186 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5189 ext4_mark_recovery_complete(sb
, es
);
5191 kthread_stop(sbi
->s_mmp_tsk
);
5193 /* Make sure we can mount this feature set readwrite */
5194 if (ext4_has_feature_readonly(sb
) ||
5195 !ext4_feature_set_ok(sb
, 0)) {
5200 * Make sure the group descriptor checksums
5201 * are sane. If they aren't, refuse to remount r/w.
5203 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5204 struct ext4_group_desc
*gdp
=
5205 ext4_get_group_desc(sb
, g
, NULL
);
5207 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5208 ext4_msg(sb
, KERN_ERR
,
5209 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5210 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5211 le16_to_cpu(gdp
->bg_checksum
));
5218 * If we have an unprocessed orphan list hanging
5219 * around from a previously readonly bdev mount,
5220 * require a full umount/remount for now.
5222 if (es
->s_last_orphan
) {
5223 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5224 "remount RDWR because of unprocessed "
5225 "orphan inode list. Please "
5226 "umount/remount instead");
5232 * Mounting a RDONLY partition read-write, so reread
5233 * and store the current valid flag. (It may have
5234 * been changed by e2fsck since we originally mounted
5238 ext4_clear_journal_err(sb
, es
);
5239 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5240 if (!ext4_setup_super(sb
, es
, 0))
5241 sb
->s_flags
&= ~MS_RDONLY
;
5242 if (ext4_has_feature_mmp(sb
))
5243 if (ext4_multi_mount_protect(sb
,
5244 le64_to_cpu(es
->s_mmp_block
))) {
5253 * Reinitialize lazy itable initialization thread based on
5256 if (sb_rdonly(sb
) || !test_opt(sb
, INIT_INODE_TABLE
))
5257 ext4_unregister_li_request(sb
);
5259 ext4_group_t first_not_zeroed
;
5260 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5261 ext4_register_li_request(sb
, first_not_zeroed
);
5264 ext4_setup_system_zone(sb
);
5265 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5266 ext4_commit_super(sb
, 1);
5269 /* Release old quota file names */
5270 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5271 kfree(old_opts
.s_qf_names
[i
]);
5273 if (sb_any_quota_suspended(sb
))
5274 dquot_resume(sb
, -1);
5275 else if (ext4_has_feature_quota(sb
)) {
5276 err
= ext4_enable_quotas(sb
);
5283 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5284 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5289 sb
->s_flags
= old_sb_flags
;
5290 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5291 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5292 sbi
->s_resuid
= old_opts
.s_resuid
;
5293 sbi
->s_resgid
= old_opts
.s_resgid
;
5294 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5295 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5296 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5298 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5299 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5300 to_free
[i
] = get_qf_name(sb
, sbi
, i
);
5301 rcu_assign_pointer(sbi
->s_qf_names
[i
], old_opts
.s_qf_names
[i
]);
5304 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5312 static int ext4_statfs_project(struct super_block
*sb
,
5313 kprojid_t projid
, struct kstatfs
*buf
)
5316 struct dquot
*dquot
;
5320 qid
= make_kqid_projid(projid
);
5321 dquot
= dqget(sb
, qid
);
5323 return PTR_ERR(dquot
);
5324 spin_lock(&dquot
->dq_dqb_lock
);
5326 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5327 dquot
->dq_dqb
.dqb_bsoftlimit
:
5328 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5329 if (limit
&& buf
->f_blocks
> limit
) {
5330 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
5331 buf
->f_blocks
= limit
;
5332 buf
->f_bfree
= buf
->f_bavail
=
5333 (buf
->f_blocks
> curblock
) ?
5334 (buf
->f_blocks
- curblock
) : 0;
5337 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5338 dquot
->dq_dqb
.dqb_isoftlimit
:
5339 dquot
->dq_dqb
.dqb_ihardlimit
;
5340 if (limit
&& buf
->f_files
> limit
) {
5341 buf
->f_files
= limit
;
5343 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5344 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5347 spin_unlock(&dquot
->dq_dqb_lock
);
5353 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5355 struct super_block
*sb
= dentry
->d_sb
;
5356 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5357 struct ext4_super_block
*es
= sbi
->s_es
;
5358 ext4_fsblk_t overhead
= 0, resv_blocks
;
5361 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5363 if (!test_opt(sb
, MINIX_DF
))
5364 overhead
= sbi
->s_overhead
;
5366 buf
->f_type
= EXT4_SUPER_MAGIC
;
5367 buf
->f_bsize
= sb
->s_blocksize
;
5368 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5369 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5370 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5371 /* prevent underflow in case that few free space is available */
5372 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5373 buf
->f_bavail
= buf
->f_bfree
-
5374 (ext4_r_blocks_count(es
) + resv_blocks
);
5375 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5377 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5378 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5379 buf
->f_namelen
= EXT4_NAME_LEN
;
5380 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5381 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5382 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5383 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5386 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5387 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5388 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5397 * Helper functions so that transaction is started before we acquire dqio_sem
5398 * to keep correct lock ordering of transaction > dqio_sem
5400 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5402 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5405 static int ext4_write_dquot(struct dquot
*dquot
)
5409 struct inode
*inode
;
5411 inode
= dquot_to_inode(dquot
);
5412 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5413 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5415 return PTR_ERR(handle
);
5416 ret
= dquot_commit(dquot
);
5417 err
= ext4_journal_stop(handle
);
5423 static int ext4_acquire_dquot(struct dquot
*dquot
)
5428 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5429 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5431 return PTR_ERR(handle
);
5432 ret
= dquot_acquire(dquot
);
5433 err
= ext4_journal_stop(handle
);
5439 static int ext4_release_dquot(struct dquot
*dquot
)
5444 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5445 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5446 if (IS_ERR(handle
)) {
5447 /* Release dquot anyway to avoid endless cycle in dqput() */
5448 dquot_release(dquot
);
5449 return PTR_ERR(handle
);
5451 ret
= dquot_release(dquot
);
5452 err
= ext4_journal_stop(handle
);
5458 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5460 struct super_block
*sb
= dquot
->dq_sb
;
5461 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5463 /* Are we journaling quotas? */
5464 if (ext4_has_feature_quota(sb
) ||
5465 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5466 dquot_mark_dquot_dirty(dquot
);
5467 return ext4_write_dquot(dquot
);
5469 return dquot_mark_dquot_dirty(dquot
);
5473 static int ext4_write_info(struct super_block
*sb
, int type
)
5478 /* Data block + inode block */
5479 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5481 return PTR_ERR(handle
);
5482 ret
= dquot_commit_info(sb
, type
);
5483 err
= ext4_journal_stop(handle
);
5490 * Turn on quotas during mount time - we need to find
5491 * the quota file and such...
5493 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5495 return dquot_quota_on_mount(sb
, get_qf_name(sb
, EXT4_SB(sb
), type
),
5496 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5499 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5501 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5503 /* The first argument of lockdep_set_subclass has to be
5504 * *exactly* the same as the argument to init_rwsem() --- in
5505 * this case, in init_once() --- or lockdep gets unhappy
5506 * because the name of the lock is set using the
5507 * stringification of the argument to init_rwsem().
5509 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5510 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5514 * Standard function to be called on quota_on
5516 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5517 const struct path
*path
)
5521 if (!test_opt(sb
, QUOTA
))
5524 /* Quotafile not on the same filesystem? */
5525 if (path
->dentry
->d_sb
!= sb
)
5527 /* Journaling quota? */
5528 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5529 /* Quotafile not in fs root? */
5530 if (path
->dentry
->d_parent
!= sb
->s_root
)
5531 ext4_msg(sb
, KERN_WARNING
,
5532 "Quota file not on filesystem root. "
5533 "Journaled quota will not work");
5534 sb_dqopt(sb
)->flags
|= DQUOT_NOLIST_DIRTY
;
5537 * Clear the flag just in case mount options changed since
5540 sb_dqopt(sb
)->flags
&= ~DQUOT_NOLIST_DIRTY
;
5544 * When we journal data on quota file, we have to flush journal to see
5545 * all updates to the file when we bypass pagecache...
5547 if (EXT4_SB(sb
)->s_journal
&&
5548 ext4_should_journal_data(d_inode(path
->dentry
))) {
5550 * We don't need to lock updates but journal_flush() could
5551 * otherwise be livelocked...
5553 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5554 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5555 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5560 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5561 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5563 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5566 struct inode
*inode
= d_inode(path
->dentry
);
5570 * Set inode flags to prevent userspace from messing with quota
5571 * files. If this fails, we return success anyway since quotas
5572 * are already enabled and this is not a hard failure.
5575 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5578 EXT4_I(inode
)->i_flags
|= EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
;
5579 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
5580 S_NOATIME
| S_IMMUTABLE
);
5581 ext4_mark_inode_dirty(handle
, inode
);
5582 ext4_journal_stop(handle
);
5584 inode_unlock(inode
);
5589 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5593 struct inode
*qf_inode
;
5594 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5595 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5596 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5597 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5600 BUG_ON(!ext4_has_feature_quota(sb
));
5602 if (!qf_inums
[type
])
5605 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5606 if (IS_ERR(qf_inode
)) {
5607 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5608 return PTR_ERR(qf_inode
);
5611 /* Don't account quota for quota files to avoid recursion */
5612 qf_inode
->i_flags
|= S_NOQUOTA
;
5613 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5614 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5617 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5622 /* Enable usage tracking for all quota types. */
5623 static int ext4_enable_quotas(struct super_block
*sb
)
5626 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5627 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5628 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5629 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5631 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5632 test_opt(sb
, USRQUOTA
),
5633 test_opt(sb
, GRPQUOTA
),
5634 test_opt(sb
, PRJQUOTA
),
5637 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
| DQUOT_NOLIST_DIRTY
;
5638 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5639 if (qf_inums
[type
]) {
5640 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5641 DQUOT_USAGE_ENABLED
|
5642 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5644 for (type
--; type
>= 0; type
--)
5645 dquot_quota_off(sb
, type
);
5648 "Failed to enable quota tracking "
5649 "(type=%d, err=%d). Please run "
5650 "e2fsck to fix.", type
, err
);
5658 static int ext4_quota_off(struct super_block
*sb
, int type
)
5660 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5664 /* Force all delayed allocation blocks to be allocated.
5665 * Caller already holds s_umount sem */
5666 if (test_opt(sb
, DELALLOC
))
5667 sync_filesystem(sb
);
5669 if (!inode
|| !igrab(inode
))
5672 err
= dquot_quota_off(sb
, type
);
5673 if (err
|| ext4_has_feature_quota(sb
))
5678 * Update modification times of quota files when userspace can
5679 * start looking at them. If we fail, we return success anyway since
5680 * this is not a hard failure and quotas are already disabled.
5682 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5685 EXT4_I(inode
)->i_flags
&= ~(EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
);
5686 inode_set_flags(inode
, 0, S_NOATIME
| S_IMMUTABLE
);
5687 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5688 ext4_mark_inode_dirty(handle
, inode
);
5689 ext4_journal_stop(handle
);
5691 inode_unlock(inode
);
5693 lockdep_set_quota_inode(inode
, I_DATA_SEM_NORMAL
);
5697 return dquot_quota_off(sb
, type
);
5700 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5701 * acquiring the locks... As quota files are never truncated and quota code
5702 * itself serializes the operations (and no one else should touch the files)
5703 * we don't have to be afraid of races */
5704 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5705 size_t len
, loff_t off
)
5707 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5708 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5709 int offset
= off
& (sb
->s_blocksize
- 1);
5712 struct buffer_head
*bh
;
5713 loff_t i_size
= i_size_read(inode
);
5717 if (off
+len
> i_size
)
5720 while (toread
> 0) {
5721 tocopy
= sb
->s_blocksize
- offset
< toread
?
5722 sb
->s_blocksize
- offset
: toread
;
5723 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5726 if (!bh
) /* A hole? */
5727 memset(data
, 0, tocopy
);
5729 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5739 /* Write to quotafile (we know the transaction is already started and has
5740 * enough credits) */
5741 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5742 const char *data
, size_t len
, loff_t off
)
5744 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5745 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5746 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5748 struct buffer_head
*bh
;
5749 handle_t
*handle
= journal_current_handle();
5751 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5752 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5753 " cancelled because transaction is not started",
5754 (unsigned long long)off
, (unsigned long long)len
);
5758 * Since we account only one data block in transaction credits,
5759 * then it is impossible to cross a block boundary.
5761 if (sb
->s_blocksize
- offset
< len
) {
5762 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5763 " cancelled because not block aligned",
5764 (unsigned long long)off
, (unsigned long long)len
);
5769 bh
= ext4_bread(handle
, inode
, blk
,
5770 EXT4_GET_BLOCKS_CREATE
|
5771 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5772 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5773 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5778 BUFFER_TRACE(bh
, "get write access");
5779 err
= ext4_journal_get_write_access(handle
, bh
);
5785 memcpy(bh
->b_data
+offset
, data
, len
);
5786 flush_dcache_page(bh
->b_page
);
5788 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5791 if (inode
->i_size
< off
+ len
) {
5792 i_size_write(inode
, off
+ len
);
5793 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5794 ext4_mark_inode_dirty(handle
, inode
);
5799 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5801 const struct quota_format_ops
*ops
;
5803 if (!sb_has_quota_loaded(sb
, qid
->type
))
5805 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5806 if (!ops
|| !ops
->get_next_id
)
5808 return dquot_get_next_id(sb
, qid
);
5812 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5813 const char *dev_name
, void *data
)
5815 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5818 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5819 static inline void register_as_ext2(void)
5821 int err
= register_filesystem(&ext2_fs_type
);
5824 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5827 static inline void unregister_as_ext2(void)
5829 unregister_filesystem(&ext2_fs_type
);
5832 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5834 if (ext4_has_unknown_ext2_incompat_features(sb
))
5838 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5843 static inline void register_as_ext2(void) { }
5844 static inline void unregister_as_ext2(void) { }
5845 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5848 static inline void register_as_ext3(void)
5850 int err
= register_filesystem(&ext3_fs_type
);
5853 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5856 static inline void unregister_as_ext3(void)
5858 unregister_filesystem(&ext3_fs_type
);
5861 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5863 if (ext4_has_unknown_ext3_incompat_features(sb
))
5865 if (!ext4_has_feature_journal(sb
))
5869 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5874 static struct file_system_type ext4_fs_type
= {
5875 .owner
= THIS_MODULE
,
5877 .mount
= ext4_mount
,
5878 .kill_sb
= kill_block_super
,
5879 .fs_flags
= FS_REQUIRES_DEV
,
5881 MODULE_ALIAS_FS("ext4");
5883 /* Shared across all ext4 file systems */
5884 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5886 static int __init
ext4_init_fs(void)
5890 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5891 ext4_li_info
= NULL
;
5892 mutex_init(&ext4_li_mtx
);
5894 /* Build-time check for flags consistency */
5895 ext4_check_flag_values();
5897 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5898 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5900 err
= ext4_init_es();
5904 err
= ext4_init_pageio();
5908 err
= ext4_init_system_zone();
5912 err
= ext4_init_sysfs();
5916 err
= ext4_init_mballoc();
5919 err
= init_inodecache();
5924 err
= register_filesystem(&ext4_fs_type
);
5930 unregister_as_ext2();
5931 unregister_as_ext3();
5932 destroy_inodecache();
5934 ext4_exit_mballoc();
5938 ext4_exit_system_zone();
5947 static void __exit
ext4_exit_fs(void)
5949 ext4_destroy_lazyinit_thread();
5950 unregister_as_ext2();
5951 unregister_as_ext3();
5952 unregister_filesystem(&ext4_fs_type
);
5953 destroy_inodecache();
5954 ext4_exit_mballoc();
5956 ext4_exit_system_zone();
5961 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5962 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5963 MODULE_LICENSE("GPL");
5964 MODULE_SOFTDEP("pre: crc32c");
5965 module_init(ext4_init_fs
)
5966 module_exit(ext4_exit_fs
)