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/cleancache.h>
41 #include <linux/uaccess.h>
42 #include <linux/user_namespace.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"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct ext4_lazy_init
*ext4_li_info
;
58 static struct mutex ext4_li_mtx
;
59 static struct ratelimit_state ext4_mount_msg_ratelimit
;
61 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
62 unsigned long journal_devnum
);
63 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
64 static int ext4_commit_super(struct super_block
*sb
, int sync
);
65 static void ext4_mark_recovery_complete(struct super_block
*sb
,
66 struct ext4_super_block
*es
);
67 static void ext4_clear_journal_err(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
70 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
71 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
72 static int ext4_unfreeze(struct super_block
*sb
);
73 static int ext4_freeze(struct super_block
*sb
);
74 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
75 const char *dev_name
, void *data
);
76 static inline int ext2_feature_set_ok(struct super_block
*sb
);
77 static inline int ext3_feature_set_ok(struct super_block
*sb
);
78 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block
*sb
);
81 static void ext4_clear_request_list(void);
82 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
83 unsigned int journal_inum
);
88 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
89 * i_mmap_rwsem (inode->i_mmap_rwsem)!
92 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
93 * page lock -> i_data_sem (rw)
95 * buffered write path:
96 * sb_start_write -> i_mutex -> mmap_sem
97 * sb_start_write -> i_mutex -> transaction start -> page lock ->
101 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
102 * i_mmap_rwsem (w) -> page lock
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * transaction start -> i_data_sem (rw)
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
109 * transaction start -> i_data_sem (rw)
112 * transaction start -> page lock(s) -> i_data_sem (rw)
115 static bool userns_mounts
= false;
116 module_param(userns_mounts
, bool, 0644);
117 MODULE_PARM_DESC(userns_mounts
, "Allow mounts from unprivileged user namespaces");
119 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
120 static struct file_system_type ext2_fs_type
= {
121 .owner
= THIS_MODULE
,
124 .kill_sb
= kill_block_super
,
125 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
127 MODULE_ALIAS_FS("ext2");
128 MODULE_ALIAS("ext2");
129 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
131 #define IS_EXT2_SB(sb) (0)
135 static struct file_system_type ext3_fs_type
= {
136 .owner
= THIS_MODULE
,
139 .kill_sb
= kill_block_super
,
140 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
142 MODULE_ALIAS_FS("ext3");
143 MODULE_ALIAS("ext3");
144 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
146 static int ext4_verify_csum_type(struct super_block
*sb
,
147 struct ext4_super_block
*es
)
149 if (!ext4_has_feature_metadata_csum(sb
))
152 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
155 static __le32
ext4_superblock_csum(struct super_block
*sb
,
156 struct ext4_super_block
*es
)
158 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
159 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
162 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
164 return cpu_to_le32(csum
);
167 static int ext4_superblock_csum_verify(struct super_block
*sb
,
168 struct ext4_super_block
*es
)
170 if (!ext4_has_metadata_csum(sb
))
173 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
176 void ext4_superblock_csum_set(struct super_block
*sb
)
178 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
180 if (!ext4_has_metadata_csum(sb
))
183 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
186 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
190 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
192 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
196 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
200 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
202 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
206 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
207 struct ext4_group_desc
*bg
)
209 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
210 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
211 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
214 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
215 struct ext4_group_desc
*bg
)
217 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
218 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
219 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
222 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
223 struct ext4_group_desc
*bg
)
225 return le32_to_cpu(bg
->bg_inode_table_lo
) |
226 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
227 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
230 __u32
ext4_free_group_clusters(struct super_block
*sb
,
231 struct ext4_group_desc
*bg
)
233 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
234 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
235 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
238 __u32
ext4_free_inodes_count(struct super_block
*sb
,
239 struct ext4_group_desc
*bg
)
241 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
242 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
243 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
246 __u32
ext4_used_dirs_count(struct super_block
*sb
,
247 struct ext4_group_desc
*bg
)
249 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
250 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
251 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
254 __u32
ext4_itable_unused_count(struct super_block
*sb
,
255 struct ext4_group_desc
*bg
)
257 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
258 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
259 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
262 void ext4_block_bitmap_set(struct super_block
*sb
,
263 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
265 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
266 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
267 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
270 void ext4_inode_bitmap_set(struct super_block
*sb
,
271 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
273 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
274 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
275 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
278 void ext4_inode_table_set(struct super_block
*sb
,
279 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
281 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
282 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
283 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
286 void ext4_free_group_clusters_set(struct super_block
*sb
,
287 struct ext4_group_desc
*bg
, __u32 count
)
289 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
290 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
291 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
294 void ext4_free_inodes_set(struct super_block
*sb
,
295 struct ext4_group_desc
*bg
, __u32 count
)
297 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
298 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
299 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
302 void ext4_used_dirs_set(struct super_block
*sb
,
303 struct ext4_group_desc
*bg
, __u32 count
)
305 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
306 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
307 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
310 void ext4_itable_unused_set(struct super_block
*sb
,
311 struct ext4_group_desc
*bg
, __u32 count
)
313 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
314 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
315 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
319 static void __save_error_info(struct super_block
*sb
, const char *func
,
322 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
324 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
325 if (bdev_read_only(sb
->s_bdev
))
327 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
328 es
->s_last_error_time
= cpu_to_le32(get_seconds());
329 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
330 es
->s_last_error_line
= cpu_to_le32(line
);
331 if (!es
->s_first_error_time
) {
332 es
->s_first_error_time
= es
->s_last_error_time
;
333 strncpy(es
->s_first_error_func
, func
,
334 sizeof(es
->s_first_error_func
));
335 es
->s_first_error_line
= cpu_to_le32(line
);
336 es
->s_first_error_ino
= es
->s_last_error_ino
;
337 es
->s_first_error_block
= es
->s_last_error_block
;
340 * Start the daily error reporting function if it hasn't been
343 if (!es
->s_error_count
)
344 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
345 le32_add_cpu(&es
->s_error_count
, 1);
348 static void save_error_info(struct super_block
*sb
, const char *func
,
351 __save_error_info(sb
, func
, line
);
352 ext4_commit_super(sb
, 1);
356 * The del_gendisk() function uninitializes the disk-specific data
357 * structures, including the bdi structure, without telling anyone
358 * else. Once this happens, any attempt to call mark_buffer_dirty()
359 * (for example, by ext4_commit_super), will cause a kernel OOPS.
360 * This is a kludge to prevent these oops until we can put in a proper
361 * hook in del_gendisk() to inform the VFS and file system layers.
363 static int block_device_ejected(struct super_block
*sb
)
365 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
366 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
368 return bdi
->dev
== NULL
;
371 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
373 struct super_block
*sb
= journal
->j_private
;
374 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
375 int error
= is_journal_aborted(journal
);
376 struct ext4_journal_cb_entry
*jce
;
378 BUG_ON(txn
->t_state
== T_FINISHED
);
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
)
408 if (sb
->s_flags
& MS_RDONLY
)
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 (ext4_error_ratelimit(sb
)) {
451 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
452 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
455 save_error_info(sb
, function
, line
);
456 ext4_handle_error(sb
);
459 void __ext4_error_inode(struct inode
*inode
, const char *function
,
460 unsigned int line
, ext4_fsblk_t block
,
461 const char *fmt
, ...)
464 struct va_format vaf
;
465 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
467 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
468 es
->s_last_error_block
= cpu_to_le64(block
);
469 if (ext4_error_ratelimit(inode
->i_sb
)) {
474 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: block %llu: comm %s: %pV\n",
476 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
477 block
, current
->comm
, &vaf
);
479 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
480 "inode #%lu: comm %s: %pV\n",
481 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
482 current
->comm
, &vaf
);
485 save_error_info(inode
->i_sb
, function
, line
);
486 ext4_handle_error(inode
->i_sb
);
489 void __ext4_error_file(struct file
*file
, const char *function
,
490 unsigned int line
, ext4_fsblk_t block
,
491 const char *fmt
, ...)
494 struct va_format vaf
;
495 struct ext4_super_block
*es
;
496 struct inode
*inode
= file_inode(file
);
497 char pathname
[80], *path
;
499 es
= EXT4_SB(inode
->i_sb
)->s_es
;
500 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
501 if (ext4_error_ratelimit(inode
->i_sb
)) {
502 path
= file_path(file
, pathname
, sizeof(pathname
));
510 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
511 "block %llu: comm %s: path %s: %pV\n",
512 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
513 block
, current
->comm
, path
, &vaf
);
516 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
517 "comm %s: path %s: %pV\n",
518 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
519 current
->comm
, path
, &vaf
);
522 save_error_info(inode
->i_sb
, function
, line
);
523 ext4_handle_error(inode
->i_sb
);
526 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
533 errstr
= "Corrupt filesystem";
536 errstr
= "Filesystem failed CRC";
539 errstr
= "IO failure";
542 errstr
= "Out of memory";
545 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
546 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
547 errstr
= "Journal has aborted";
549 errstr
= "Readonly filesystem";
552 /* If the caller passed in an extra buffer for unknown
553 * errors, textualise them now. Else we just return
556 /* Check for truncated error codes... */
557 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
566 /* __ext4_std_error decodes expected errors from journaling functions
567 * automatically and invokes the appropriate error response. */
569 void __ext4_std_error(struct super_block
*sb
, const char *function
,
570 unsigned int line
, int errno
)
575 /* Special case: if the error is EROFS, and we're not already
576 * inside a transaction, then there's really no point in logging
578 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
579 (sb
->s_flags
& MS_RDONLY
))
582 if (ext4_error_ratelimit(sb
)) {
583 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
584 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
585 sb
->s_id
, function
, line
, errstr
);
588 save_error_info(sb
, function
, line
);
589 ext4_handle_error(sb
);
593 * ext4_abort is a much stronger failure handler than ext4_error. The
594 * abort function may be used to deal with unrecoverable failures such
595 * as journal IO errors or ENOMEM at a critical moment in log management.
597 * We unconditionally force the filesystem into an ABORT|READONLY state,
598 * unless the error response on the fs has been set to panic in which
599 * case we take the easy way out and panic immediately.
602 void __ext4_abort(struct super_block
*sb
, const char *function
,
603 unsigned int line
, const char *fmt
, ...)
605 struct va_format vaf
;
608 save_error_info(sb
, function
, line
);
612 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
613 sb
->s_id
, function
, line
, &vaf
);
616 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
617 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
618 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
620 * Make sure updated value of ->s_mount_flags will be visible
621 * before ->s_flags update
624 sb
->s_flags
|= MS_RDONLY
;
625 if (EXT4_SB(sb
)->s_journal
)
626 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
627 save_error_info(sb
, function
, line
);
629 if (test_opt(sb
, ERRORS_PANIC
)) {
630 if (EXT4_SB(sb
)->s_journal
&&
631 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
633 panic("EXT4-fs panic from previous error\n");
637 void __ext4_msg(struct super_block
*sb
,
638 const char *prefix
, const char *fmt
, ...)
640 struct va_format vaf
;
643 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
649 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
653 #define ext4_warning_ratelimit(sb) \
654 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
657 void __ext4_warning(struct super_block
*sb
, const char *function
,
658 unsigned int line
, const char *fmt
, ...)
660 struct va_format vaf
;
663 if (!ext4_warning_ratelimit(sb
))
669 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
670 sb
->s_id
, function
, line
, &vaf
);
674 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
675 unsigned int line
, const char *fmt
, ...)
677 struct va_format vaf
;
680 if (!ext4_warning_ratelimit(inode
->i_sb
))
686 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
687 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
688 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
692 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
693 struct super_block
*sb
, ext4_group_t grp
,
694 unsigned long ino
, ext4_fsblk_t block
,
695 const char *fmt
, ...)
699 struct va_format vaf
;
701 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
703 es
->s_last_error_ino
= cpu_to_le32(ino
);
704 es
->s_last_error_block
= cpu_to_le64(block
);
705 __save_error_info(sb
, function
, line
);
707 if (ext4_error_ratelimit(sb
)) {
711 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
712 sb
->s_id
, function
, line
, grp
);
714 printk(KERN_CONT
"inode %lu: ", ino
);
716 printk(KERN_CONT
"block %llu:",
717 (unsigned long long) block
);
718 printk(KERN_CONT
"%pV\n", &vaf
);
722 if (test_opt(sb
, ERRORS_CONT
)) {
723 ext4_commit_super(sb
, 0);
727 ext4_unlock_group(sb
, grp
);
728 ext4_handle_error(sb
);
730 * We only get here in the ERRORS_RO case; relocking the group
731 * may be dangerous, but nothing bad will happen since the
732 * filesystem will have already been marked read/only and the
733 * journal has been aborted. We return 1 as a hint to callers
734 * who might what to use the return value from
735 * ext4_grp_locked_error() to distinguish between the
736 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
737 * aggressively from the ext4 function in question, with a
738 * more appropriate error code.
740 ext4_lock_group(sb
, grp
);
744 void ext4_update_dynamic_rev(struct super_block
*sb
)
746 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
748 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
752 "updating to rev %d because of new feature flag, "
753 "running e2fsck is recommended",
756 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
757 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
758 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
759 /* leave es->s_feature_*compat flags alone */
760 /* es->s_uuid will be set by e2fsck if empty */
763 * The rest of the superblock fields should be zero, and if not it
764 * means they are likely already in use, so leave them alone. We
765 * can leave it up to e2fsck to clean up any inconsistencies there.
770 * Open the external journal device
772 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
774 struct block_device
*bdev
;
775 char b
[BDEVNAME_SIZE
];
777 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
783 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
784 __bdevname(dev
, b
), PTR_ERR(bdev
));
789 * Release the journal device
791 static void ext4_blkdev_put(struct block_device
*bdev
)
793 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
796 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
798 struct block_device
*bdev
;
799 bdev
= sbi
->journal_bdev
;
801 ext4_blkdev_put(bdev
);
802 sbi
->journal_bdev
= NULL
;
806 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
808 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
811 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
815 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
816 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
818 printk(KERN_ERR
"sb_info orphan list:\n");
819 list_for_each(l
, &sbi
->s_orphan
) {
820 struct inode
*inode
= orphan_list_entry(l
);
822 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
823 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
824 inode
->i_mode
, inode
->i_nlink
,
829 static void ext4_put_super(struct super_block
*sb
)
831 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
832 struct ext4_super_block
*es
= sbi
->s_es
;
836 ext4_unregister_li_request(sb
);
837 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
839 flush_workqueue(sbi
->rsv_conversion_wq
);
840 destroy_workqueue(sbi
->rsv_conversion_wq
);
842 if (sbi
->s_journal
) {
843 aborted
= is_journal_aborted(sbi
->s_journal
);
844 err
= jbd2_journal_destroy(sbi
->s_journal
);
845 sbi
->s_journal
= NULL
;
846 if ((err
< 0) && !aborted
)
847 ext4_abort(sb
, "Couldn't clean up the journal");
850 ext4_unregister_sysfs(sb
);
851 ext4_es_unregister_shrinker(sbi
);
852 del_timer_sync(&sbi
->s_err_report
);
853 ext4_release_system_zone(sb
);
855 ext4_ext_release(sb
);
857 if (!(sb
->s_flags
& MS_RDONLY
) && !aborted
) {
858 ext4_clear_feature_journal_needs_recovery(sb
);
859 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
861 if (!(sb
->s_flags
& MS_RDONLY
))
862 ext4_commit_super(sb
, 1);
864 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
865 brelse(sbi
->s_group_desc
[i
]);
866 kvfree(sbi
->s_group_desc
);
867 kvfree(sbi
->s_flex_groups
);
868 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
869 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
870 percpu_counter_destroy(&sbi
->s_dirs_counter
);
871 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
872 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
874 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
875 kfree(sbi
->s_qf_names
[i
]);
878 /* Debugging code just in case the in-memory inode orphan list
879 * isn't empty. The on-disk one can be non-empty if we've
880 * detected an error and taken the fs readonly, but the
881 * in-memory list had better be clean by this point. */
882 if (!list_empty(&sbi
->s_orphan
))
883 dump_orphan_list(sb
, sbi
);
884 J_ASSERT(list_empty(&sbi
->s_orphan
));
886 sync_blockdev(sb
->s_bdev
);
887 invalidate_bdev(sb
->s_bdev
);
888 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
890 * Invalidate the journal device's buffers. We don't want them
891 * floating about in memory - the physical journal device may
892 * hotswapped, and it breaks the `ro-after' testing code.
894 sync_blockdev(sbi
->journal_bdev
);
895 invalidate_bdev(sbi
->journal_bdev
);
896 ext4_blkdev_remove(sbi
);
898 if (sbi
->s_mb_cache
) {
899 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
900 sbi
->s_mb_cache
= NULL
;
903 kthread_stop(sbi
->s_mmp_tsk
);
905 sb
->s_fs_info
= NULL
;
907 * Now that we are completely done shutting down the
908 * superblock, we need to actually destroy the kobject.
910 kobject_put(&sbi
->s_kobj
);
911 wait_for_completion(&sbi
->s_kobj_unregister
);
912 if (sbi
->s_chksum_driver
)
913 crypto_free_shash(sbi
->s_chksum_driver
);
914 kfree(sbi
->s_blockgroup_lock
);
918 static struct kmem_cache
*ext4_inode_cachep
;
921 * Called inside transaction, so use GFP_NOFS
923 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
925 struct ext4_inode_info
*ei
;
927 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
931 ei
->vfs_inode
.i_version
= 1;
932 spin_lock_init(&ei
->i_raw_lock
);
933 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
934 spin_lock_init(&ei
->i_prealloc_lock
);
935 ext4_es_init_tree(&ei
->i_es_tree
);
936 rwlock_init(&ei
->i_es_lock
);
937 INIT_LIST_HEAD(&ei
->i_es_list
);
940 ei
->i_es_shrink_lblk
= 0;
941 ei
->i_reserved_data_blocks
= 0;
942 ei
->i_reserved_meta_blocks
= 0;
943 ei
->i_allocated_meta_blocks
= 0;
944 ei
->i_da_metadata_calc_len
= 0;
945 ei
->i_da_metadata_calc_last_lblock
= 0;
946 spin_lock_init(&(ei
->i_block_reservation_lock
));
948 ei
->i_reserved_quota
= 0;
949 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
952 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
953 spin_lock_init(&ei
->i_completed_io_lock
);
955 ei
->i_datasync_tid
= 0;
956 atomic_set(&ei
->i_unwritten
, 0);
957 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
958 return &ei
->vfs_inode
;
961 static int ext4_drop_inode(struct inode
*inode
)
963 int drop
= generic_drop_inode(inode
);
965 trace_ext4_drop_inode(inode
, drop
);
969 static void ext4_i_callback(struct rcu_head
*head
)
971 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
972 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
975 static void ext4_destroy_inode(struct inode
*inode
)
977 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
978 ext4_msg(inode
->i_sb
, KERN_ERR
,
979 "Inode %lu (%p): orphan list check failed!",
980 inode
->i_ino
, EXT4_I(inode
));
981 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
982 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
986 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
989 static void init_once(void *foo
)
991 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
993 INIT_LIST_HEAD(&ei
->i_orphan
);
994 init_rwsem(&ei
->xattr_sem
);
995 init_rwsem(&ei
->i_data_sem
);
996 init_rwsem(&ei
->i_mmap_sem
);
997 inode_init_once(&ei
->vfs_inode
);
1000 static int __init
init_inodecache(void)
1002 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
1003 sizeof(struct ext4_inode_info
),
1004 0, (SLAB_RECLAIM_ACCOUNT
|
1005 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1007 if (ext4_inode_cachep
== NULL
)
1012 static void destroy_inodecache(void)
1015 * Make sure all delayed rcu free inodes are flushed before we
1019 kmem_cache_destroy(ext4_inode_cachep
);
1022 void ext4_clear_inode(struct inode
*inode
)
1024 invalidate_inode_buffers(inode
);
1027 ext4_discard_preallocations(inode
);
1028 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1029 if (EXT4_I(inode
)->jinode
) {
1030 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1031 EXT4_I(inode
)->jinode
);
1032 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1033 EXT4_I(inode
)->jinode
= NULL
;
1035 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1036 fscrypt_put_encryption_info(inode
, NULL
);
1040 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1041 u64 ino
, u32 generation
)
1043 struct inode
*inode
;
1045 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1046 return ERR_PTR(-ESTALE
);
1047 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1048 return ERR_PTR(-ESTALE
);
1050 /* iget isn't really right if the inode is currently unallocated!!
1052 * ext4_read_inode will return a bad_inode if the inode had been
1053 * deleted, so we should be safe.
1055 * Currently we don't know the generation for parent directory, so
1056 * a generation of 0 means "accept any"
1058 inode
= ext4_iget_normal(sb
, ino
);
1060 return ERR_CAST(inode
);
1061 if (generation
&& inode
->i_generation
!= generation
) {
1063 return ERR_PTR(-ESTALE
);
1069 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1070 int fh_len
, int fh_type
)
1072 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1073 ext4_nfs_get_inode
);
1076 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1077 int fh_len
, int fh_type
)
1079 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1080 ext4_nfs_get_inode
);
1084 * Try to release metadata pages (indirect blocks, directories) which are
1085 * mapped via the block device. Since these pages could have journal heads
1086 * which would prevent try_to_free_buffers() from freeing them, we must use
1087 * jbd2 layer's try_to_free_buffers() function to release them.
1089 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1092 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1094 WARN_ON(PageChecked(page
));
1095 if (!page_has_buffers(page
))
1098 return jbd2_journal_try_to_free_buffers(journal
, page
,
1099 wait
& ~__GFP_DIRECT_RECLAIM
);
1100 return try_to_free_buffers(page
);
1103 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1104 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1106 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1107 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1110 static int ext4_key_prefix(struct inode
*inode
, u8
**key
)
1112 *key
= EXT4_SB(inode
->i_sb
)->key_prefix
;
1113 return EXT4_SB(inode
->i_sb
)->key_prefix_size
;
1116 static int ext4_prepare_context(struct inode
*inode
)
1118 return ext4_convert_inline_data(inode
);
1121 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1124 handle_t
*handle
= fs_data
;
1125 int res
, res2
, retries
= 0;
1128 * If a journal handle was specified, then the encryption context is
1129 * being set on a new inode via inheritance and is part of a larger
1130 * transaction to create the inode. Otherwise the encryption context is
1131 * being set on an existing inode in its own transaction. Only in the
1132 * latter case should the "retry on ENOSPC" logic be used.
1136 res
= ext4_xattr_set_handle(handle
, inode
,
1137 EXT4_XATTR_INDEX_ENCRYPTION
,
1138 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1141 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1142 ext4_clear_inode_state(inode
,
1143 EXT4_STATE_MAY_INLINE_DATA
);
1145 * Update inode->i_flags - e.g. S_DAX may get disabled
1147 ext4_set_inode_flags(inode
);
1153 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
,
1154 ext4_jbd2_credits_xattr(inode
));
1156 return PTR_ERR(handle
);
1158 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1159 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1162 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1163 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1164 ext4_set_inode_flags(inode
);
1165 res
= ext4_mark_inode_dirty(handle
, inode
);
1167 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1169 res2
= ext4_journal_stop(handle
);
1171 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1178 static int ext4_dummy_context(struct inode
*inode
)
1180 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1183 static unsigned ext4_max_namelen(struct inode
*inode
)
1185 return S_ISLNK(inode
->i_mode
) ? inode
->i_sb
->s_blocksize
:
1189 static struct fscrypt_operations ext4_cryptops
= {
1190 .get_context
= ext4_get_context
,
1191 .key_prefix
= ext4_key_prefix
,
1192 .prepare_context
= ext4_prepare_context
,
1193 .set_context
= ext4_set_context
,
1194 .dummy_context
= ext4_dummy_context
,
1195 .is_encrypted
= ext4_encrypted_inode
,
1196 .empty_dir
= ext4_empty_dir
,
1197 .max_namelen
= ext4_max_namelen
,
1200 static struct fscrypt_operations ext4_cryptops
= {
1201 .is_encrypted
= ext4_encrypted_inode
,
1206 static char *quotatypes
[] = INITQFNAMES
;
1207 #define QTYPE2NAME(t) (quotatypes[t])
1209 static int ext4_write_dquot(struct dquot
*dquot
);
1210 static int ext4_acquire_dquot(struct dquot
*dquot
);
1211 static int ext4_release_dquot(struct dquot
*dquot
);
1212 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1213 static int ext4_write_info(struct super_block
*sb
, int type
);
1214 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1215 const struct path
*path
);
1216 static int ext4_quota_off(struct super_block
*sb
, int type
);
1217 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1218 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1219 size_t len
, loff_t off
);
1220 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1221 const char *data
, size_t len
, loff_t off
);
1222 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1223 unsigned int flags
);
1224 static int ext4_enable_quotas(struct super_block
*sb
);
1225 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1227 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1229 return EXT4_I(inode
)->i_dquot
;
1232 static const struct dquot_operations ext4_quota_operations
= {
1233 .get_reserved_space
= ext4_get_reserved_space
,
1234 .write_dquot
= ext4_write_dquot
,
1235 .acquire_dquot
= ext4_acquire_dquot
,
1236 .release_dquot
= ext4_release_dquot
,
1237 .mark_dirty
= ext4_mark_dquot_dirty
,
1238 .write_info
= ext4_write_info
,
1239 .alloc_dquot
= dquot_alloc
,
1240 .destroy_dquot
= dquot_destroy
,
1241 .get_projid
= ext4_get_projid
,
1242 .get_next_id
= ext4_get_next_id
,
1245 static const struct quotactl_ops ext4_qctl_operations
= {
1246 .quota_on
= ext4_quota_on
,
1247 .quota_off
= ext4_quota_off
,
1248 .quota_sync
= dquot_quota_sync
,
1249 .get_state
= dquot_get_state
,
1250 .set_info
= dquot_set_dqinfo
,
1251 .get_dqblk
= dquot_get_dqblk
,
1252 .set_dqblk
= dquot_set_dqblk
,
1253 .get_nextdqblk
= dquot_get_next_dqblk
,
1257 static const struct super_operations ext4_sops
= {
1258 .alloc_inode
= ext4_alloc_inode
,
1259 .destroy_inode
= ext4_destroy_inode
,
1260 .write_inode
= ext4_write_inode
,
1261 .dirty_inode
= ext4_dirty_inode
,
1262 .drop_inode
= ext4_drop_inode
,
1263 .evict_inode
= ext4_evict_inode
,
1264 .put_super
= ext4_put_super
,
1265 .sync_fs
= ext4_sync_fs
,
1266 .freeze_fs
= ext4_freeze
,
1267 .unfreeze_fs
= ext4_unfreeze
,
1268 .statfs
= ext4_statfs
,
1269 .remount_fs
= ext4_remount
,
1270 .show_options
= ext4_show_options
,
1272 .quota_read
= ext4_quota_read
,
1273 .quota_write
= ext4_quota_write
,
1274 .get_dquots
= ext4_get_dquots
,
1276 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1279 static const struct export_operations ext4_export_ops
= {
1280 .fh_to_dentry
= ext4_fh_to_dentry
,
1281 .fh_to_parent
= ext4_fh_to_parent
,
1282 .get_parent
= ext4_get_parent
,
1286 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1287 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1288 Opt_nouid32
, Opt_debug
, Opt_removed
,
1289 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1290 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1291 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1292 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1293 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1294 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1295 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1296 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1297 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1298 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1299 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1300 Opt_lazytime
, Opt_nolazytime
,
1301 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1302 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1303 Opt_dioread_nolock
, Opt_dioread_lock
,
1304 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1305 Opt_max_dir_size_kb
, Opt_nojournal_checksum
,
1308 static const match_table_t tokens
= {
1309 {Opt_bsd_df
, "bsddf"},
1310 {Opt_minix_df
, "minixdf"},
1311 {Opt_grpid
, "grpid"},
1312 {Opt_grpid
, "bsdgroups"},
1313 {Opt_nogrpid
, "nogrpid"},
1314 {Opt_nogrpid
, "sysvgroups"},
1315 {Opt_resgid
, "resgid=%u"},
1316 {Opt_resuid
, "resuid=%u"},
1318 {Opt_err_cont
, "errors=continue"},
1319 {Opt_err_panic
, "errors=panic"},
1320 {Opt_err_ro
, "errors=remount-ro"},
1321 {Opt_nouid32
, "nouid32"},
1322 {Opt_debug
, "debug"},
1323 {Opt_removed
, "oldalloc"},
1324 {Opt_removed
, "orlov"},
1325 {Opt_user_xattr
, "user_xattr"},
1326 {Opt_nouser_xattr
, "nouser_xattr"},
1328 {Opt_noacl
, "noacl"},
1329 {Opt_noload
, "norecovery"},
1330 {Opt_noload
, "noload"},
1331 {Opt_removed
, "nobh"},
1332 {Opt_removed
, "bh"},
1333 {Opt_commit
, "commit=%u"},
1334 {Opt_min_batch_time
, "min_batch_time=%u"},
1335 {Opt_max_batch_time
, "max_batch_time=%u"},
1336 {Opt_journal_dev
, "journal_dev=%u"},
1337 {Opt_journal_path
, "journal_path=%s"},
1338 {Opt_journal_checksum
, "journal_checksum"},
1339 {Opt_nojournal_checksum
, "nojournal_checksum"},
1340 {Opt_journal_async_commit
, "journal_async_commit"},
1341 {Opt_abort
, "abort"},
1342 {Opt_data_journal
, "data=journal"},
1343 {Opt_data_ordered
, "data=ordered"},
1344 {Opt_data_writeback
, "data=writeback"},
1345 {Opt_data_err_abort
, "data_err=abort"},
1346 {Opt_data_err_ignore
, "data_err=ignore"},
1347 {Opt_offusrjquota
, "usrjquota="},
1348 {Opt_usrjquota
, "usrjquota=%s"},
1349 {Opt_offgrpjquota
, "grpjquota="},
1350 {Opt_grpjquota
, "grpjquota=%s"},
1351 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1352 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1353 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1354 {Opt_grpquota
, "grpquota"},
1355 {Opt_noquota
, "noquota"},
1356 {Opt_quota
, "quota"},
1357 {Opt_usrquota
, "usrquota"},
1358 {Opt_prjquota
, "prjquota"},
1359 {Opt_barrier
, "barrier=%u"},
1360 {Opt_barrier
, "barrier"},
1361 {Opt_nobarrier
, "nobarrier"},
1362 {Opt_i_version
, "i_version"},
1364 {Opt_stripe
, "stripe=%u"},
1365 {Opt_delalloc
, "delalloc"},
1366 {Opt_lazytime
, "lazytime"},
1367 {Opt_nolazytime
, "nolazytime"},
1368 {Opt_nodelalloc
, "nodelalloc"},
1369 {Opt_removed
, "mblk_io_submit"},
1370 {Opt_removed
, "nomblk_io_submit"},
1371 {Opt_block_validity
, "block_validity"},
1372 {Opt_noblock_validity
, "noblock_validity"},
1373 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1374 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1375 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1376 {Opt_auto_da_alloc
, "auto_da_alloc"},
1377 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1378 {Opt_dioread_nolock
, "dioread_nolock"},
1379 {Opt_dioread_lock
, "dioread_lock"},
1380 {Opt_discard
, "discard"},
1381 {Opt_nodiscard
, "nodiscard"},
1382 {Opt_init_itable
, "init_itable=%u"},
1383 {Opt_init_itable
, "init_itable"},
1384 {Opt_noinit_itable
, "noinit_itable"},
1385 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1386 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1387 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1388 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1389 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1390 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1391 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1395 static ext4_fsblk_t
get_sb_block(void **data
)
1397 ext4_fsblk_t sb_block
;
1398 char *options
= (char *) *data
;
1400 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1401 return 1; /* Default location */
1404 /* TODO: use simple_strtoll with >32bit ext4 */
1405 sb_block
= simple_strtoul(options
, &options
, 0);
1406 if (*options
&& *options
!= ',') {
1407 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1411 if (*options
== ',')
1413 *data
= (void *) options
;
1418 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1419 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1420 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1423 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1425 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1429 if (sb_any_quota_loaded(sb
) &&
1430 !sbi
->s_qf_names
[qtype
]) {
1431 ext4_msg(sb
, KERN_ERR
,
1432 "Cannot change journaled "
1433 "quota options when quota turned on");
1436 if (ext4_has_feature_quota(sb
)) {
1437 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1438 "ignored when QUOTA feature is enabled");
1441 qname
= match_strdup(args
);
1443 ext4_msg(sb
, KERN_ERR
,
1444 "Not enough memory for storing quotafile name");
1447 if (sbi
->s_qf_names
[qtype
]) {
1448 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1451 ext4_msg(sb
, KERN_ERR
,
1452 "%s quota file already specified",
1456 if (strchr(qname
, '/')) {
1457 ext4_msg(sb
, KERN_ERR
,
1458 "quotafile must be on filesystem root");
1461 sbi
->s_qf_names
[qtype
] = qname
;
1469 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1472 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1474 if (sb_any_quota_loaded(sb
) &&
1475 sbi
->s_qf_names
[qtype
]) {
1476 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1477 " when quota turned on");
1480 kfree(sbi
->s_qf_names
[qtype
]);
1481 sbi
->s_qf_names
[qtype
] = NULL
;
1486 #define MOPT_SET 0x0001
1487 #define MOPT_CLEAR 0x0002
1488 #define MOPT_NOSUPPORT 0x0004
1489 #define MOPT_EXPLICIT 0x0008
1490 #define MOPT_CLEAR_ERR 0x0010
1491 #define MOPT_GTE0 0x0020
1494 #define MOPT_QFMT 0x0040
1496 #define MOPT_Q MOPT_NOSUPPORT
1497 #define MOPT_QFMT MOPT_NOSUPPORT
1499 #define MOPT_DATAJ 0x0080
1500 #define MOPT_NO_EXT2 0x0100
1501 #define MOPT_NO_EXT3 0x0200
1502 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1503 #define MOPT_STRING 0x0400
1505 static const struct mount_opts
{
1509 } ext4_mount_opts
[] = {
1510 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1511 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1512 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1513 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1514 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1515 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1516 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1517 MOPT_EXT4_ONLY
| MOPT_SET
},
1518 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1519 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1520 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1521 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1522 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1523 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1524 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1525 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1526 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1527 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1528 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1529 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1530 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1531 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1532 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1533 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1534 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1535 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1536 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1537 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1539 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1541 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1542 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1543 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1544 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1545 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1546 {Opt_commit
, 0, MOPT_GTE0
},
1547 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1548 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1549 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1550 {Opt_init_itable
, 0, MOPT_GTE0
},
1551 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1552 {Opt_stripe
, 0, MOPT_GTE0
},
1553 {Opt_resuid
, 0, MOPT_GTE0
},
1554 {Opt_resgid
, 0, MOPT_GTE0
},
1555 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1556 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1557 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1558 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1559 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1560 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1561 MOPT_NO_EXT2
| MOPT_DATAJ
},
1562 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1563 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1564 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1565 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1566 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1568 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1569 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1571 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1572 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1573 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1574 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1576 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1578 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1580 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1581 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1582 MOPT_CLEAR
| MOPT_Q
},
1583 {Opt_usrjquota
, 0, MOPT_Q
},
1584 {Opt_grpjquota
, 0, MOPT_Q
},
1585 {Opt_offusrjquota
, 0, MOPT_Q
},
1586 {Opt_offgrpjquota
, 0, MOPT_Q
},
1587 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1588 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1589 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1590 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1591 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1595 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1596 substring_t
*args
, unsigned long *journal_devnum
,
1597 unsigned int *journal_ioprio
, int is_remount
)
1599 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1600 const struct mount_opts
*m
;
1606 if (token
== Opt_usrjquota
)
1607 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1608 else if (token
== Opt_grpjquota
)
1609 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1610 else if (token
== Opt_offusrjquota
)
1611 return clear_qf_name(sb
, USRQUOTA
);
1612 else if (token
== Opt_offgrpjquota
)
1613 return clear_qf_name(sb
, GRPQUOTA
);
1617 case Opt_nouser_xattr
:
1618 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1621 return 1; /* handled by get_sb_block() */
1623 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1626 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1629 sb
->s_flags
|= MS_I_VERSION
;
1632 sb
->s_flags
|= MS_LAZYTIME
;
1634 case Opt_nolazytime
:
1635 sb
->s_flags
&= ~MS_LAZYTIME
;
1639 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1640 if (token
== m
->token
)
1643 if (m
->token
== Opt_err
) {
1644 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1645 "or missing value", opt
);
1649 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1650 ext4_msg(sb
, KERN_ERR
,
1651 "Mount option \"%s\" incompatible with ext2", opt
);
1654 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1655 ext4_msg(sb
, KERN_ERR
,
1656 "Mount option \"%s\" incompatible with ext3", opt
);
1660 if (token
== Opt_err_panic
&& !capable(CAP_SYS_ADMIN
)) {
1661 ext4_msg(sb
, KERN_ERR
,
1662 "Mount option \"%s\" not allowed for unprivileged mounts",
1667 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1669 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1671 if (m
->flags
& MOPT_EXPLICIT
) {
1672 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1673 set_opt2(sb
, EXPLICIT_DELALLOC
);
1674 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1675 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1679 if (m
->flags
& MOPT_CLEAR_ERR
)
1680 clear_opt(sb
, ERRORS_MASK
);
1681 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1682 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1683 "options when quota turned on");
1687 if (m
->flags
& MOPT_NOSUPPORT
) {
1688 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1689 } else if (token
== Opt_commit
) {
1691 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1692 sbi
->s_commit_interval
= HZ
* arg
;
1693 } else if (token
== Opt_max_batch_time
) {
1694 sbi
->s_max_batch_time
= arg
;
1695 } else if (token
== Opt_min_batch_time
) {
1696 sbi
->s_min_batch_time
= arg
;
1697 } else if (token
== Opt_inode_readahead_blks
) {
1698 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1699 ext4_msg(sb
, KERN_ERR
,
1700 "EXT4-fs: inode_readahead_blks must be "
1701 "0 or a power of 2 smaller than 2^31");
1704 sbi
->s_inode_readahead_blks
= arg
;
1705 } else if (token
== Opt_init_itable
) {
1706 set_opt(sb
, INIT_INODE_TABLE
);
1708 arg
= EXT4_DEF_LI_WAIT_MULT
;
1709 sbi
->s_li_wait_mult
= arg
;
1710 } else if (token
== Opt_max_dir_size_kb
) {
1711 sbi
->s_max_dir_size_kb
= arg
;
1712 } else if (token
== Opt_stripe
) {
1713 sbi
->s_stripe
= arg
;
1714 } else if (token
== Opt_resuid
) {
1715 uid
= make_kuid(sb
->s_user_ns
, arg
);
1716 if (!uid_valid(uid
)) {
1717 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1720 sbi
->s_resuid
= uid
;
1721 } else if (token
== Opt_resgid
) {
1722 gid
= make_kgid(sb
->s_user_ns
, arg
);
1723 if (!gid_valid(gid
)) {
1724 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1727 sbi
->s_resgid
= gid
;
1728 } else if (token
== Opt_journal_dev
) {
1730 ext4_msg(sb
, KERN_ERR
,
1731 "Cannot specify journal on remount");
1734 *journal_devnum
= arg
;
1735 } else if (token
== Opt_journal_path
) {
1737 struct inode
*journal_inode
;
1742 ext4_msg(sb
, KERN_ERR
,
1743 "Cannot specify journal on remount");
1746 journal_path
= match_strdup(&args
[0]);
1747 if (!journal_path
) {
1748 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1749 "journal device string");
1753 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1755 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1756 "journal device path: error %d", error
);
1757 kfree(journal_path
);
1762 * Refuse access for unprivileged mounts if the user does
1763 * not have rw access to the journal device via the supplied
1766 if (!capable(CAP_SYS_ADMIN
) &&
1767 inode_permission(d_inode(path
.dentry
), MAY_READ
|MAY_WRITE
)) {
1768 ext4_msg(sb
, KERN_ERR
,
1769 "error: Insufficient access to journal path %s",
1774 journal_inode
= d_inode(path
.dentry
);
1775 if (!S_ISBLK(journal_inode
->i_mode
)) {
1776 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1777 "is not a block device", journal_path
);
1779 kfree(journal_path
);
1783 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1785 kfree(journal_path
);
1786 } else if (token
== Opt_journal_ioprio
) {
1788 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1793 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1794 } else if (token
== Opt_test_dummy_encryption
) {
1795 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1796 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1797 ext4_msg(sb
, KERN_WARNING
,
1798 "Test dummy encryption mode enabled");
1800 ext4_msg(sb
, KERN_WARNING
,
1801 "Test dummy encryption mount option ignored");
1803 } else if (m
->flags
& MOPT_DATAJ
) {
1805 if (!sbi
->s_journal
)
1806 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1807 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1808 ext4_msg(sb
, KERN_ERR
,
1809 "Cannot change data mode on remount");
1813 clear_opt(sb
, DATA_FLAGS
);
1814 sbi
->s_mount_opt
|= m
->mount_opt
;
1817 } else if (m
->flags
& MOPT_QFMT
) {
1818 if (sb_any_quota_loaded(sb
) &&
1819 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1820 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1821 "quota options when quota turned on");
1824 if (ext4_has_feature_quota(sb
)) {
1825 ext4_msg(sb
, KERN_INFO
,
1826 "Quota format mount options ignored "
1827 "when QUOTA feature is enabled");
1830 sbi
->s_jquota_fmt
= m
->mount_opt
;
1832 } else if (token
== Opt_dax
) {
1833 #ifdef CONFIG_FS_DAX
1834 ext4_msg(sb
, KERN_WARNING
,
1835 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1836 sbi
->s_mount_opt
|= m
->mount_opt
;
1838 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1841 } else if (token
== Opt_data_err_abort
) {
1842 sbi
->s_mount_opt
|= m
->mount_opt
;
1843 } else if (token
== Opt_data_err_ignore
) {
1844 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1848 if (m
->flags
& MOPT_CLEAR
)
1850 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1851 ext4_msg(sb
, KERN_WARNING
,
1852 "buggy handling of option %s", opt
);
1857 sbi
->s_mount_opt
|= m
->mount_opt
;
1859 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1864 static int parse_options(char *options
, struct super_block
*sb
,
1865 unsigned long *journal_devnum
,
1866 unsigned int *journal_ioprio
,
1869 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1871 substring_t args
[MAX_OPT_ARGS
];
1877 while ((p
= strsep(&options
, ",")) != NULL
) {
1881 * Initialize args struct so we know whether arg was
1882 * found; some options take optional arguments.
1884 args
[0].to
= args
[0].from
= NULL
;
1885 token
= match_token(p
, tokens
, args
);
1886 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1887 journal_ioprio
, is_remount
) < 0)
1892 * We do the test below only for project quotas. 'usrquota' and
1893 * 'grpquota' mount options are allowed even without quota feature
1894 * to support legacy quotas in quota files.
1896 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
1897 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
1898 "Cannot enable project quota enforcement.");
1901 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1902 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1903 clear_opt(sb
, USRQUOTA
);
1905 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1906 clear_opt(sb
, GRPQUOTA
);
1908 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1909 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1914 if (!sbi
->s_jquota_fmt
) {
1915 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1921 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1923 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1925 if (blocksize
< PAGE_SIZE
) {
1926 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1927 "dioread_nolock if block size != PAGE_SIZE");
1934 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1935 struct super_block
*sb
)
1937 #if defined(CONFIG_QUOTA)
1938 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1940 if (sbi
->s_jquota_fmt
) {
1943 switch (sbi
->s_jquota_fmt
) {
1954 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1957 if (sbi
->s_qf_names
[USRQUOTA
])
1958 seq_show_option(seq
, "usrjquota", sbi
->s_qf_names
[USRQUOTA
]);
1960 if (sbi
->s_qf_names
[GRPQUOTA
])
1961 seq_show_option(seq
, "grpjquota", sbi
->s_qf_names
[GRPQUOTA
]);
1965 static const char *token2str(int token
)
1967 const struct match_token
*t
;
1969 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1970 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1977 * - it's set to a non-default value OR
1978 * - if the per-sb default is different from the global default
1980 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1983 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1984 struct ext4_super_block
*es
= sbi
->s_es
;
1985 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1986 const struct mount_opts
*m
;
1987 char sep
= nodefs
? '\n' : ',';
1989 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1990 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1992 if (sbi
->s_sb_block
!= 1)
1993 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1995 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1996 int want_set
= m
->flags
& MOPT_SET
;
1997 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1998 (m
->flags
& MOPT_CLEAR_ERR
))
2000 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2001 continue; /* skip if same as the default */
2003 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2004 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2005 continue; /* select Opt_noFoo vs Opt_Foo */
2006 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2009 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
)) ||
2010 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2011 SEQ_OPTS_PRINT("resuid=%u",
2012 from_kuid_munged(sb
->s_user_ns
, sbi
->s_resuid
));
2013 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
)) ||
2014 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2015 SEQ_OPTS_PRINT("resgid=%u",
2016 from_kgid_munged(sb
->s_user_ns
, sbi
->s_resgid
));
2017 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2018 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2019 SEQ_OPTS_PUTS("errors=remount-ro");
2020 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2021 SEQ_OPTS_PUTS("errors=continue");
2022 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2023 SEQ_OPTS_PUTS("errors=panic");
2024 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2025 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2026 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2027 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2028 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2029 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2030 if (sb
->s_flags
& MS_I_VERSION
)
2031 SEQ_OPTS_PUTS("i_version");
2032 if (nodefs
|| sbi
->s_stripe
)
2033 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2034 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
2035 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2036 SEQ_OPTS_PUTS("data=journal");
2037 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2038 SEQ_OPTS_PUTS("data=ordered");
2039 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2040 SEQ_OPTS_PUTS("data=writeback");
2043 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2044 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2045 sbi
->s_inode_readahead_blks
);
2047 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
2048 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2049 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2050 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2051 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2052 if (test_opt(sb
, DATA_ERR_ABORT
))
2053 SEQ_OPTS_PUTS("data_err=abort");
2055 ext4_show_quota_options(seq
, sb
);
2059 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2061 return _ext4_show_options(seq
, root
->d_sb
, 0);
2064 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2066 struct super_block
*sb
= seq
->private;
2069 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
2070 rc
= _ext4_show_options(seq
, sb
, 1);
2071 seq_puts(seq
, "\n");
2075 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2078 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2081 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2082 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2083 "forcing read-only mode");
2088 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2089 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2090 "running e2fsck is recommended");
2091 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2092 ext4_msg(sb
, KERN_WARNING
,
2093 "warning: mounting fs with errors, "
2094 "running e2fsck is recommended");
2095 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2096 le16_to_cpu(es
->s_mnt_count
) >=
2097 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2098 ext4_msg(sb
, KERN_WARNING
,
2099 "warning: maximal mount count reached, "
2100 "running e2fsck is recommended");
2101 else if (le32_to_cpu(es
->s_checkinterval
) &&
2102 (le32_to_cpu(es
->s_lastcheck
) +
2103 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
2104 ext4_msg(sb
, KERN_WARNING
,
2105 "warning: checktime reached, "
2106 "running e2fsck is recommended");
2107 if (!sbi
->s_journal
)
2108 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2109 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2110 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2111 le16_add_cpu(&es
->s_mnt_count
, 1);
2112 es
->s_mtime
= cpu_to_le32(get_seconds());
2113 ext4_update_dynamic_rev(sb
);
2115 ext4_set_feature_journal_needs_recovery(sb
);
2117 ext4_commit_super(sb
, 1);
2119 if (test_opt(sb
, DEBUG
))
2120 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2121 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2123 sbi
->s_groups_count
,
2124 EXT4_BLOCKS_PER_GROUP(sb
),
2125 EXT4_INODES_PER_GROUP(sb
),
2126 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2128 cleancache_init_fs(sb
);
2132 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2134 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2135 struct flex_groups
*new_groups
;
2138 if (!sbi
->s_log_groups_per_flex
)
2141 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2142 if (size
<= sbi
->s_flex_groups_allocated
)
2145 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2146 new_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
2148 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2149 size
/ (int) sizeof(struct flex_groups
));
2153 if (sbi
->s_flex_groups
) {
2154 memcpy(new_groups
, sbi
->s_flex_groups
,
2155 (sbi
->s_flex_groups_allocated
*
2156 sizeof(struct flex_groups
)));
2157 kvfree(sbi
->s_flex_groups
);
2159 sbi
->s_flex_groups
= new_groups
;
2160 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2164 static int ext4_fill_flex_info(struct super_block
*sb
)
2166 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2167 struct ext4_group_desc
*gdp
= NULL
;
2168 ext4_group_t flex_group
;
2171 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2172 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2173 sbi
->s_log_groups_per_flex
= 0;
2177 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2181 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2182 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2184 flex_group
= ext4_flex_group(sbi
, i
);
2185 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2186 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2187 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2188 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2189 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2190 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2198 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2199 struct ext4_group_desc
*gdp
)
2201 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2203 __le32 le_group
= cpu_to_le32(block_group
);
2204 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2206 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2207 /* Use new metadata_csum algorithm */
2209 __u16 dummy_csum
= 0;
2211 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2213 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2214 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2215 sizeof(dummy_csum
));
2216 offset
+= sizeof(dummy_csum
);
2217 if (offset
< sbi
->s_desc_size
)
2218 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2219 sbi
->s_desc_size
- offset
);
2221 crc
= csum32
& 0xFFFF;
2225 /* old crc16 code */
2226 if (!ext4_has_feature_gdt_csum(sb
))
2229 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2230 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2231 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2232 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2233 /* for checksum of struct ext4_group_desc do the rest...*/
2234 if (ext4_has_feature_64bit(sb
) &&
2235 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2236 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2237 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2241 return cpu_to_le16(crc
);
2244 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2245 struct ext4_group_desc
*gdp
)
2247 if (ext4_has_group_desc_csum(sb
) &&
2248 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2254 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2255 struct ext4_group_desc
*gdp
)
2257 if (!ext4_has_group_desc_csum(sb
))
2259 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2262 /* Called at mount-time, super-block is locked */
2263 static int ext4_check_descriptors(struct super_block
*sb
,
2264 ext4_fsblk_t sb_block
,
2265 ext4_group_t
*first_not_zeroed
)
2267 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2268 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2269 ext4_fsblk_t last_block
;
2270 ext4_fsblk_t block_bitmap
;
2271 ext4_fsblk_t inode_bitmap
;
2272 ext4_fsblk_t inode_table
;
2273 int flexbg_flag
= 0;
2274 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2276 if (ext4_has_feature_flex_bg(sb
))
2279 ext4_debug("Checking group descriptors");
2281 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2282 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2284 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2285 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2287 last_block
= first_block
+
2288 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2290 if ((grp
== sbi
->s_groups_count
) &&
2291 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2294 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2295 if (block_bitmap
== sb_block
) {
2296 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2297 "Block bitmap for group %u overlaps "
2300 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2301 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2302 "Block bitmap for group %u not in group "
2303 "(block %llu)!", i
, block_bitmap
);
2306 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2307 if (inode_bitmap
== sb_block
) {
2308 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2309 "Inode bitmap for group %u overlaps "
2312 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2313 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2314 "Inode bitmap for group %u not in group "
2315 "(block %llu)!", i
, inode_bitmap
);
2318 inode_table
= ext4_inode_table(sb
, gdp
);
2319 if (inode_table
== sb_block
) {
2320 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2321 "Inode table for group %u overlaps "
2324 if (inode_table
< first_block
||
2325 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2326 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2327 "Inode table for group %u not in group "
2328 "(block %llu)!", i
, inode_table
);
2331 ext4_lock_group(sb
, i
);
2332 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2333 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2334 "Checksum for group %u failed (%u!=%u)",
2335 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2336 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2337 if (!(sb
->s_flags
& MS_RDONLY
)) {
2338 ext4_unlock_group(sb
, i
);
2342 ext4_unlock_group(sb
, i
);
2344 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2346 if (NULL
!= first_not_zeroed
)
2347 *first_not_zeroed
= grp
;
2351 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2352 * the superblock) which were deleted from all directories, but held open by
2353 * a process at the time of a crash. We walk the list and try to delete these
2354 * inodes at recovery time (only with a read-write filesystem).
2356 * In order to keep the orphan inode chain consistent during traversal (in
2357 * case of crash during recovery), we link each inode into the superblock
2358 * orphan list_head and handle it the same way as an inode deletion during
2359 * normal operation (which journals the operations for us).
2361 * We only do an iget() and an iput() on each inode, which is very safe if we
2362 * accidentally point at an in-use or already deleted inode. The worst that
2363 * can happen in this case is that we get a "bit already cleared" message from
2364 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2365 * e2fsck was run on this filesystem, and it must have already done the orphan
2366 * inode cleanup for us, so we can safely abort without any further action.
2368 static void ext4_orphan_cleanup(struct super_block
*sb
,
2369 struct ext4_super_block
*es
)
2371 unsigned int s_flags
= sb
->s_flags
;
2372 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2376 if (!es
->s_last_orphan
) {
2377 jbd_debug(4, "no orphan inodes to clean up\n");
2381 if (bdev_read_only(sb
->s_bdev
)) {
2382 ext4_msg(sb
, KERN_ERR
, "write access "
2383 "unavailable, skipping orphan cleanup");
2387 /* Check if feature set would not allow a r/w mount */
2388 if (!ext4_feature_set_ok(sb
, 0)) {
2389 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2390 "unknown ROCOMPAT features");
2394 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2395 /* don't clear list on RO mount w/ errors */
2396 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2397 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2398 "clearing orphan list.\n");
2399 es
->s_last_orphan
= 0;
2401 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2405 if (s_flags
& MS_RDONLY
) {
2406 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2407 sb
->s_flags
&= ~MS_RDONLY
;
2410 /* Needed for iput() to work correctly and not trash data */
2411 sb
->s_flags
|= MS_ACTIVE
;
2412 /* Turn on quotas so that they are updated correctly */
2413 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2414 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2415 int ret
= ext4_quota_on_mount(sb
, i
);
2417 ext4_msg(sb
, KERN_ERR
,
2418 "Cannot turn on journaled "
2419 "quota: error %d", ret
);
2424 while (es
->s_last_orphan
) {
2425 struct inode
*inode
;
2428 * We may have encountered an error during cleanup; if
2429 * so, skip the rest.
2431 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2432 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2433 es
->s_last_orphan
= 0;
2437 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2438 if (IS_ERR(inode
)) {
2439 es
->s_last_orphan
= 0;
2443 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2444 dquot_initialize(inode
);
2445 if (inode
->i_nlink
) {
2446 if (test_opt(sb
, DEBUG
))
2447 ext4_msg(sb
, KERN_DEBUG
,
2448 "%s: truncating inode %lu to %lld bytes",
2449 __func__
, inode
->i_ino
, inode
->i_size
);
2450 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2451 inode
->i_ino
, inode
->i_size
);
2453 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2454 ret
= ext4_truncate(inode
);
2456 ext4_std_error(inode
->i_sb
, ret
);
2457 inode_unlock(inode
);
2460 if (test_opt(sb
, DEBUG
))
2461 ext4_msg(sb
, KERN_DEBUG
,
2462 "%s: deleting unreferenced inode %lu",
2463 __func__
, inode
->i_ino
);
2464 jbd_debug(2, "deleting unreferenced inode %lu\n",
2468 iput(inode
); /* The delete magic happens here! */
2471 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2474 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2475 PLURAL(nr_orphans
));
2477 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2478 PLURAL(nr_truncates
));
2480 /* Turn quotas off */
2481 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2482 if (sb_dqopt(sb
)->files
[i
])
2483 dquot_quota_off(sb
, i
);
2486 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2490 * Maximal extent format file size.
2491 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2492 * extent format containers, within a sector_t, and within i_blocks
2493 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2494 * so that won't be a limiting factor.
2496 * However there is other limiting factor. We do store extents in the form
2497 * of starting block and length, hence the resulting length of the extent
2498 * covering maximum file size must fit into on-disk format containers as
2499 * well. Given that length is always by 1 unit bigger than max unit (because
2500 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2502 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2504 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2507 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2509 /* small i_blocks in vfs inode? */
2510 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2512 * CONFIG_LBDAF is not enabled implies the inode
2513 * i_block represent total blocks in 512 bytes
2514 * 32 == size of vfs inode i_blocks * 8
2516 upper_limit
= (1LL << 32) - 1;
2518 /* total blocks in file system block size */
2519 upper_limit
>>= (blkbits
- 9);
2520 upper_limit
<<= blkbits
;
2524 * 32-bit extent-start container, ee_block. We lower the maxbytes
2525 * by one fs block, so ee_len can cover the extent of maximum file
2528 res
= (1LL << 32) - 1;
2531 /* Sanity check against vm- & vfs- imposed limits */
2532 if (res
> upper_limit
)
2539 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2540 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2541 * We need to be 1 filesystem block less than the 2^48 sector limit.
2543 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2545 loff_t res
= EXT4_NDIR_BLOCKS
;
2548 /* This is calculated to be the largest file size for a dense, block
2549 * mapped file such that the file's total number of 512-byte sectors,
2550 * including data and all indirect blocks, does not exceed (2^48 - 1).
2552 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2553 * number of 512-byte sectors of the file.
2556 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2558 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2559 * the inode i_block field represents total file blocks in
2560 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2562 upper_limit
= (1LL << 32) - 1;
2564 /* total blocks in file system block size */
2565 upper_limit
>>= (bits
- 9);
2569 * We use 48 bit ext4_inode i_blocks
2570 * With EXT4_HUGE_FILE_FL set the i_blocks
2571 * represent total number of blocks in
2572 * file system block size
2574 upper_limit
= (1LL << 48) - 1;
2578 /* indirect blocks */
2580 /* double indirect blocks */
2581 meta_blocks
+= 1 + (1LL << (bits
-2));
2582 /* tripple indirect blocks */
2583 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2585 upper_limit
-= meta_blocks
;
2586 upper_limit
<<= bits
;
2588 res
+= 1LL << (bits
-2);
2589 res
+= 1LL << (2*(bits
-2));
2590 res
+= 1LL << (3*(bits
-2));
2592 if (res
> upper_limit
)
2595 if (res
> MAX_LFS_FILESIZE
)
2596 res
= MAX_LFS_FILESIZE
;
2601 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2602 ext4_fsblk_t logical_sb_block
, int nr
)
2604 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2605 ext4_group_t bg
, first_meta_bg
;
2608 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2610 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2611 return logical_sb_block
+ nr
+ 1;
2612 bg
= sbi
->s_desc_per_block
* nr
;
2613 if (ext4_bg_has_super(sb
, bg
))
2617 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2618 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2619 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2622 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2623 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2626 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2630 * ext4_get_stripe_size: Get the stripe size.
2631 * @sbi: In memory super block info
2633 * If we have specified it via mount option, then
2634 * use the mount option value. If the value specified at mount time is
2635 * greater than the blocks per group use the super block value.
2636 * If the super block value is greater than blocks per group return 0.
2637 * Allocator needs it be less than blocks per group.
2640 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2642 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2643 unsigned long stripe_width
=
2644 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2647 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2648 ret
= sbi
->s_stripe
;
2649 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2651 else if (stride
<= sbi
->s_blocks_per_group
)
2657 * If the stripe width is 1, this makes no sense and
2658 * we set it to 0 to turn off stripe handling code.
2667 * Check whether this filesystem can be mounted based on
2668 * the features present and the RDONLY/RDWR mount requested.
2669 * Returns 1 if this filesystem can be mounted as requested,
2670 * 0 if it cannot be.
2672 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2674 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2675 ext4_msg(sb
, KERN_ERR
,
2676 "Couldn't mount because of "
2677 "unsupported optional features (%x)",
2678 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2679 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2686 if (ext4_has_feature_readonly(sb
)) {
2687 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2688 sb
->s_flags
|= MS_RDONLY
;
2692 /* Check that feature set is OK for a read-write mount */
2693 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2694 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2695 "unsupported optional features (%x)",
2696 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2697 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2701 * Large file size enabled file system can only be mounted
2702 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2704 if (ext4_has_feature_huge_file(sb
)) {
2705 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2706 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2707 "cannot be mounted RDWR without "
2712 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2713 ext4_msg(sb
, KERN_ERR
,
2714 "Can't support bigalloc feature without "
2715 "extents feature\n");
2719 #ifndef CONFIG_QUOTA
2720 if (ext4_has_feature_quota(sb
) && !readonly
) {
2721 ext4_msg(sb
, KERN_ERR
,
2722 "Filesystem with quota feature cannot be mounted RDWR "
2723 "without CONFIG_QUOTA");
2726 if (ext4_has_feature_project(sb
) && !readonly
) {
2727 ext4_msg(sb
, KERN_ERR
,
2728 "Filesystem with project quota feature cannot be mounted RDWR "
2729 "without CONFIG_QUOTA");
2732 #endif /* CONFIG_QUOTA */
2737 * This function is called once a day if we have errors logged
2738 * on the file system
2740 static void print_daily_error_info(unsigned long arg
)
2742 struct super_block
*sb
= (struct super_block
*) arg
;
2743 struct ext4_sb_info
*sbi
;
2744 struct ext4_super_block
*es
;
2749 if (es
->s_error_count
)
2750 /* fsck newer than v1.41.13 is needed to clean this condition. */
2751 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2752 le32_to_cpu(es
->s_error_count
));
2753 if (es
->s_first_error_time
) {
2754 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2755 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2756 (int) sizeof(es
->s_first_error_func
),
2757 es
->s_first_error_func
,
2758 le32_to_cpu(es
->s_first_error_line
));
2759 if (es
->s_first_error_ino
)
2760 printk(KERN_CONT
": inode %u",
2761 le32_to_cpu(es
->s_first_error_ino
));
2762 if (es
->s_first_error_block
)
2763 printk(KERN_CONT
": block %llu", (unsigned long long)
2764 le64_to_cpu(es
->s_first_error_block
));
2765 printk(KERN_CONT
"\n");
2767 if (es
->s_last_error_time
) {
2768 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2769 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2770 (int) sizeof(es
->s_last_error_func
),
2771 es
->s_last_error_func
,
2772 le32_to_cpu(es
->s_last_error_line
));
2773 if (es
->s_last_error_ino
)
2774 printk(KERN_CONT
": inode %u",
2775 le32_to_cpu(es
->s_last_error_ino
));
2776 if (es
->s_last_error_block
)
2777 printk(KERN_CONT
": block %llu", (unsigned long long)
2778 le64_to_cpu(es
->s_last_error_block
));
2779 printk(KERN_CONT
"\n");
2781 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2784 /* Find next suitable group and run ext4_init_inode_table */
2785 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2787 struct ext4_group_desc
*gdp
= NULL
;
2788 ext4_group_t group
, ngroups
;
2789 struct super_block
*sb
;
2790 unsigned long timeout
= 0;
2794 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2796 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2797 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2803 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2807 if (group
>= ngroups
)
2812 ret
= ext4_init_inode_table(sb
, group
,
2813 elr
->lr_timeout
? 0 : 1);
2814 if (elr
->lr_timeout
== 0) {
2815 timeout
= (jiffies
- timeout
) *
2816 elr
->lr_sbi
->s_li_wait_mult
;
2817 elr
->lr_timeout
= timeout
;
2819 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2820 elr
->lr_next_group
= group
+ 1;
2826 * Remove lr_request from the list_request and free the
2827 * request structure. Should be called with li_list_mtx held
2829 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2831 struct ext4_sb_info
*sbi
;
2838 list_del(&elr
->lr_request
);
2839 sbi
->s_li_request
= NULL
;
2843 static void ext4_unregister_li_request(struct super_block
*sb
)
2845 mutex_lock(&ext4_li_mtx
);
2846 if (!ext4_li_info
) {
2847 mutex_unlock(&ext4_li_mtx
);
2851 mutex_lock(&ext4_li_info
->li_list_mtx
);
2852 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2853 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2854 mutex_unlock(&ext4_li_mtx
);
2857 static struct task_struct
*ext4_lazyinit_task
;
2860 * This is the function where ext4lazyinit thread lives. It walks
2861 * through the request list searching for next scheduled filesystem.
2862 * When such a fs is found, run the lazy initialization request
2863 * (ext4_rn_li_request) and keep track of the time spend in this
2864 * function. Based on that time we compute next schedule time of
2865 * the request. When walking through the list is complete, compute
2866 * next waking time and put itself into sleep.
2868 static int ext4_lazyinit_thread(void *arg
)
2870 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2871 struct list_head
*pos
, *n
;
2872 struct ext4_li_request
*elr
;
2873 unsigned long next_wakeup
, cur
;
2875 BUG_ON(NULL
== eli
);
2879 next_wakeup
= MAX_JIFFY_OFFSET
;
2881 mutex_lock(&eli
->li_list_mtx
);
2882 if (list_empty(&eli
->li_request_list
)) {
2883 mutex_unlock(&eli
->li_list_mtx
);
2886 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2889 elr
= list_entry(pos
, struct ext4_li_request
,
2892 if (time_before(jiffies
, elr
->lr_next_sched
)) {
2893 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2894 next_wakeup
= elr
->lr_next_sched
;
2897 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
2898 if (sb_start_write_trylock(elr
->lr_super
)) {
2901 * We hold sb->s_umount, sb can not
2902 * be removed from the list, it is
2903 * now safe to drop li_list_mtx
2905 mutex_unlock(&eli
->li_list_mtx
);
2906 err
= ext4_run_li_request(elr
);
2907 sb_end_write(elr
->lr_super
);
2908 mutex_lock(&eli
->li_list_mtx
);
2911 up_read((&elr
->lr_super
->s_umount
));
2913 /* error, remove the lazy_init job */
2915 ext4_remove_li_request(elr
);
2919 elr
->lr_next_sched
= jiffies
+
2921 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
2923 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2924 next_wakeup
= elr
->lr_next_sched
;
2926 mutex_unlock(&eli
->li_list_mtx
);
2931 if ((time_after_eq(cur
, next_wakeup
)) ||
2932 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2937 schedule_timeout_interruptible(next_wakeup
- cur
);
2939 if (kthread_should_stop()) {
2940 ext4_clear_request_list();
2947 * It looks like the request list is empty, but we need
2948 * to check it under the li_list_mtx lock, to prevent any
2949 * additions into it, and of course we should lock ext4_li_mtx
2950 * to atomically free the list and ext4_li_info, because at
2951 * this point another ext4 filesystem could be registering
2954 mutex_lock(&ext4_li_mtx
);
2955 mutex_lock(&eli
->li_list_mtx
);
2956 if (!list_empty(&eli
->li_request_list
)) {
2957 mutex_unlock(&eli
->li_list_mtx
);
2958 mutex_unlock(&ext4_li_mtx
);
2961 mutex_unlock(&eli
->li_list_mtx
);
2962 kfree(ext4_li_info
);
2963 ext4_li_info
= NULL
;
2964 mutex_unlock(&ext4_li_mtx
);
2969 static void ext4_clear_request_list(void)
2971 struct list_head
*pos
, *n
;
2972 struct ext4_li_request
*elr
;
2974 mutex_lock(&ext4_li_info
->li_list_mtx
);
2975 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2976 elr
= list_entry(pos
, struct ext4_li_request
,
2978 ext4_remove_li_request(elr
);
2980 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2983 static int ext4_run_lazyinit_thread(void)
2985 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2986 ext4_li_info
, "ext4lazyinit");
2987 if (IS_ERR(ext4_lazyinit_task
)) {
2988 int err
= PTR_ERR(ext4_lazyinit_task
);
2989 ext4_clear_request_list();
2990 kfree(ext4_li_info
);
2991 ext4_li_info
= NULL
;
2992 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
2993 "initialization thread\n",
2997 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3002 * Check whether it make sense to run itable init. thread or not.
3003 * If there is at least one uninitialized inode table, return
3004 * corresponding group number, else the loop goes through all
3005 * groups and return total number of groups.
3007 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3009 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3010 struct ext4_group_desc
*gdp
= NULL
;
3012 for (group
= 0; group
< ngroups
; group
++) {
3013 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3017 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3024 static int ext4_li_info_new(void)
3026 struct ext4_lazy_init
*eli
= NULL
;
3028 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3032 INIT_LIST_HEAD(&eli
->li_request_list
);
3033 mutex_init(&eli
->li_list_mtx
);
3035 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3042 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3045 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3046 struct ext4_li_request
*elr
;
3048 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3054 elr
->lr_next_group
= start
;
3057 * Randomize first schedule time of the request to
3058 * spread the inode table initialization requests
3061 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3062 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3066 int ext4_register_li_request(struct super_block
*sb
,
3067 ext4_group_t first_not_zeroed
)
3069 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3070 struct ext4_li_request
*elr
= NULL
;
3071 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3074 mutex_lock(&ext4_li_mtx
);
3075 if (sbi
->s_li_request
!= NULL
) {
3077 * Reset timeout so it can be computed again, because
3078 * s_li_wait_mult might have changed.
3080 sbi
->s_li_request
->lr_timeout
= 0;
3084 if (first_not_zeroed
== ngroups
||
3085 (sb
->s_flags
& MS_RDONLY
) ||
3086 !test_opt(sb
, INIT_INODE_TABLE
))
3089 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3095 if (NULL
== ext4_li_info
) {
3096 ret
= ext4_li_info_new();
3101 mutex_lock(&ext4_li_info
->li_list_mtx
);
3102 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3103 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3105 sbi
->s_li_request
= elr
;
3107 * set elr to NULL here since it has been inserted to
3108 * the request_list and the removal and free of it is
3109 * handled by ext4_clear_request_list from now on.
3113 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3114 ret
= ext4_run_lazyinit_thread();
3119 mutex_unlock(&ext4_li_mtx
);
3126 * We do not need to lock anything since this is called on
3129 static void ext4_destroy_lazyinit_thread(void)
3132 * If thread exited earlier
3133 * there's nothing to be done.
3135 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3138 kthread_stop(ext4_lazyinit_task
);
3141 static int set_journal_csum_feature_set(struct super_block
*sb
)
3144 int compat
, incompat
;
3145 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3147 if (ext4_has_metadata_csum(sb
)) {
3148 /* journal checksum v3 */
3150 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3152 /* journal checksum v1 */
3153 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3157 jbd2_journal_clear_features(sbi
->s_journal
,
3158 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3159 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3160 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3161 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3162 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3164 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3166 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3167 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3170 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3171 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3173 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3174 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3181 * Note: calculating the overhead so we can be compatible with
3182 * historical BSD practice is quite difficult in the face of
3183 * clusters/bigalloc. This is because multiple metadata blocks from
3184 * different block group can end up in the same allocation cluster.
3185 * Calculating the exact overhead in the face of clustered allocation
3186 * requires either O(all block bitmaps) in memory or O(number of block
3187 * groups**2) in time. We will still calculate the superblock for
3188 * older file systems --- and if we come across with a bigalloc file
3189 * system with zero in s_overhead_clusters the estimate will be close to
3190 * correct especially for very large cluster sizes --- but for newer
3191 * file systems, it's better to calculate this figure once at mkfs
3192 * time, and store it in the superblock. If the superblock value is
3193 * present (even for non-bigalloc file systems), we will use it.
3195 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3198 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3199 struct ext4_group_desc
*gdp
;
3200 ext4_fsblk_t first_block
, last_block
, b
;
3201 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3202 int s
, j
, count
= 0;
3204 if (!ext4_has_feature_bigalloc(sb
))
3205 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3206 sbi
->s_itb_per_group
+ 2);
3208 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3209 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3210 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3211 for (i
= 0; i
< ngroups
; i
++) {
3212 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3213 b
= ext4_block_bitmap(sb
, gdp
);
3214 if (b
>= first_block
&& b
<= last_block
) {
3215 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3218 b
= ext4_inode_bitmap(sb
, gdp
);
3219 if (b
>= first_block
&& b
<= last_block
) {
3220 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3223 b
= ext4_inode_table(sb
, gdp
);
3224 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3225 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3226 int c
= EXT4_B2C(sbi
, b
- first_block
);
3227 ext4_set_bit(c
, buf
);
3233 if (ext4_bg_has_super(sb
, grp
)) {
3234 ext4_set_bit(s
++, buf
);
3237 j
= ext4_bg_num_gdb(sb
, grp
);
3238 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3239 ext4_error(sb
, "Invalid number of block group "
3240 "descriptor blocks: %d", j
);
3241 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3245 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3249 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3250 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3254 * Compute the overhead and stash it in sbi->s_overhead
3256 int ext4_calculate_overhead(struct super_block
*sb
)
3258 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3259 struct ext4_super_block
*es
= sbi
->s_es
;
3260 struct inode
*j_inode
;
3261 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3262 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3263 ext4_fsblk_t overhead
= 0;
3264 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3270 * Compute the overhead (FS structures). This is constant
3271 * for a given filesystem unless the number of block groups
3272 * changes so we cache the previous value until it does.
3276 * All of the blocks before first_data_block are overhead
3278 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3281 * Add the overhead found in each block group
3283 for (i
= 0; i
< ngroups
; i
++) {
3286 blks
= count_overhead(sb
, i
, buf
);
3289 memset(buf
, 0, PAGE_SIZE
);
3294 * Add the internal journal blocks whether the journal has been
3297 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3298 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3299 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3300 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3302 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3303 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3306 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3309 sbi
->s_overhead
= overhead
;
3311 free_page((unsigned long) buf
);
3315 static void ext4_set_resv_clusters(struct super_block
*sb
)
3317 ext4_fsblk_t resv_clusters
;
3318 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3321 * There's no need to reserve anything when we aren't using extents.
3322 * The space estimates are exact, there are no unwritten extents,
3323 * hole punching doesn't need new metadata... This is needed especially
3324 * to keep ext2/3 backward compatibility.
3326 if (!ext4_has_feature_extents(sb
))
3329 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3330 * This should cover the situations where we can not afford to run
3331 * out of space like for example punch hole, or converting
3332 * unwritten extents in delalloc path. In most cases such
3333 * allocation would require 1, or 2 blocks, higher numbers are
3336 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3337 sbi
->s_cluster_bits
);
3339 do_div(resv_clusters
, 50);
3340 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3342 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3345 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3347 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3348 struct buffer_head
*bh
;
3349 struct ext4_super_block
*es
= NULL
;
3350 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3352 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3353 ext4_fsblk_t logical_sb_block
;
3354 unsigned long offset
= 0;
3355 unsigned long journal_devnum
= 0;
3356 unsigned long def_mount_opts
;
3360 int blocksize
, clustersize
;
3361 unsigned int db_count
;
3363 int needs_recovery
, has_huge_files
, has_bigalloc
;
3366 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3367 ext4_group_t first_not_zeroed
;
3369 if ((data
&& !orig_data
) || !sbi
)
3372 if (!userns_mounts
&& !capable(CAP_SYS_ADMIN
)) {
3377 sbi
->s_blockgroup_lock
=
3378 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3379 if (!sbi
->s_blockgroup_lock
)
3382 sb
->s_fs_info
= sbi
;
3384 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3385 sbi
->s_sb_block
= sb_block
;
3386 if (sb
->s_bdev
->bd_part
)
3387 sbi
->s_sectors_written_start
=
3388 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3390 /* Cleanup superblock name */
3391 strreplace(sb
->s_id
, '/', '!');
3393 /* -EINVAL is default */
3395 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3397 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3402 * The ext4 superblock will not be buffer aligned for other than 1kB
3403 * block sizes. We need to calculate the offset from buffer start.
3405 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3406 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3407 offset
= do_div(logical_sb_block
, blocksize
);
3409 logical_sb_block
= sb_block
;
3412 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3413 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3417 * Note: s_es must be initialized as soon as possible because
3418 * some ext4 macro-instructions depend on its value
3420 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3422 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3423 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3425 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3427 /* Warn if metadata_csum and gdt_csum are both set. */
3428 if (ext4_has_feature_metadata_csum(sb
) &&
3429 ext4_has_feature_gdt_csum(sb
))
3430 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3431 "redundant flags; please run fsck.");
3433 /* Check for a known checksum algorithm */
3434 if (!ext4_verify_csum_type(sb
, es
)) {
3435 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3436 "unknown checksum algorithm.");
3441 /* Load the checksum driver */
3442 if (ext4_has_feature_metadata_csum(sb
)) {
3443 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3444 if (IS_ERR(sbi
->s_chksum_driver
)) {
3445 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3446 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3447 sbi
->s_chksum_driver
= NULL
;
3452 /* Check superblock checksum */
3453 if (!ext4_superblock_csum_verify(sb
, es
)) {
3454 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3455 "invalid superblock checksum. Run e2fsck?");
3461 /* Precompute checksum seed for all metadata */
3462 if (ext4_has_feature_csum_seed(sb
))
3463 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3464 else if (ext4_has_metadata_csum(sb
))
3465 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3466 sizeof(es
->s_uuid
));
3468 /* Set defaults before we parse the mount options */
3469 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3470 set_opt(sb
, INIT_INODE_TABLE
);
3471 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3473 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3475 if (def_mount_opts
& EXT4_DEFM_UID16
)
3476 set_opt(sb
, NO_UID32
);
3477 /* xattr user namespace & acls are now defaulted on */
3478 set_opt(sb
, XATTR_USER
);
3479 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3480 set_opt(sb
, POSIX_ACL
);
3482 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3483 if (ext4_has_metadata_csum(sb
))
3484 set_opt(sb
, JOURNAL_CHECKSUM
);
3486 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3487 set_opt(sb
, JOURNAL_DATA
);
3488 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3489 set_opt(sb
, ORDERED_DATA
);
3490 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3491 set_opt(sb
, WRITEBACK_DATA
);
3493 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
) {
3494 if (!capable(CAP_SYS_ADMIN
))
3496 set_opt(sb
, ERRORS_PANIC
);
3497 } else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
) {
3498 set_opt(sb
, ERRORS_CONT
);
3500 set_opt(sb
, ERRORS_RO
);
3502 /* block_validity enabled by default; disable with noblock_validity */
3503 set_opt(sb
, BLOCK_VALIDITY
);
3504 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3505 set_opt(sb
, DISCARD
);
3507 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3508 if (!uid_valid(sbi
->s_resuid
))
3509 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
);
3510 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3511 if (!gid_valid(sbi
->s_resgid
))
3512 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
);
3513 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3514 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3515 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3517 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3518 set_opt(sb
, BARRIER
);
3521 * enable delayed allocation by default
3522 * Use -o nodelalloc to turn it off
3524 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3525 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3526 set_opt(sb
, DELALLOC
);
3529 * set default s_li_wait_mult for lazyinit, for the case there is
3530 * no mount option specified.
3532 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3534 if (sbi
->s_es
->s_mount_opts
[0]) {
3535 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3536 sizeof(sbi
->s_es
->s_mount_opts
),
3540 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3541 &journal_ioprio
, 0)) {
3542 ext4_msg(sb
, KERN_WARNING
,
3543 "failed to parse options in superblock: %s",
3546 kfree(s_mount_opts
);
3548 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3549 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3550 &journal_ioprio
, 0))
3553 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3554 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3555 "with data=journal disables delayed "
3556 "allocation and O_DIRECT support!\n");
3557 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3558 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3559 "both data=journal and delalloc");
3562 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3563 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3564 "both data=journal and dioread_nolock");
3567 if (test_opt(sb
, DAX
)) {
3568 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3569 "both data=journal and dax");
3572 if (ext4_has_feature_encrypt(sb
)) {
3573 ext4_msg(sb
, KERN_WARNING
,
3574 "encrypted files will use data=ordered "
3575 "instead of data journaling mode");
3577 if (test_opt(sb
, DELALLOC
))
3578 clear_opt(sb
, DELALLOC
);
3580 sb
->s_iflags
|= SB_I_CGROUPWB
;
3583 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3584 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3586 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3587 (ext4_has_compat_features(sb
) ||
3588 ext4_has_ro_compat_features(sb
) ||
3589 ext4_has_incompat_features(sb
)))
3590 ext4_msg(sb
, KERN_WARNING
,
3591 "feature flags set on rev 0 fs, "
3592 "running e2fsck is recommended");
3594 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3595 set_opt2(sb
, HURD_COMPAT
);
3596 if (ext4_has_feature_64bit(sb
)) {
3597 ext4_msg(sb
, KERN_ERR
,
3598 "The Hurd can't support 64-bit file systems");
3603 if (IS_EXT2_SB(sb
)) {
3604 if (ext2_feature_set_ok(sb
))
3605 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3606 "using the ext4 subsystem");
3608 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3609 "to feature incompatibilities");
3614 if (IS_EXT3_SB(sb
)) {
3615 if (ext3_feature_set_ok(sb
))
3616 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3617 "using the ext4 subsystem");
3619 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3620 "to feature incompatibilities");
3626 * Check feature flags regardless of the revision level, since we
3627 * previously didn't change the revision level when setting the flags,
3628 * so there is a chance incompat flags are set on a rev 0 filesystem.
3630 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3633 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3634 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3635 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3636 ext4_msg(sb
, KERN_ERR
,
3637 "Unsupported filesystem blocksize %d (%d log_block_size)",
3638 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3641 if (le32_to_cpu(es
->s_log_block_size
) >
3642 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3643 ext4_msg(sb
, KERN_ERR
,
3644 "Invalid log block size: %u",
3645 le32_to_cpu(es
->s_log_block_size
));
3649 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3650 ext4_msg(sb
, KERN_ERR
,
3651 "Number of reserved GDT blocks insanely large: %d",
3652 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3656 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3657 err
= bdev_dax_supported(sb
, blocksize
);
3662 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3663 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3664 es
->s_encryption_level
);
3668 if (sb
->s_blocksize
!= blocksize
) {
3669 /* Validate the filesystem blocksize */
3670 if (!sb_set_blocksize(sb
, blocksize
)) {
3671 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3677 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3678 offset
= do_div(logical_sb_block
, blocksize
);
3679 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3681 ext4_msg(sb
, KERN_ERR
,
3682 "Can't read superblock on 2nd try");
3685 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3687 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3688 ext4_msg(sb
, KERN_ERR
,
3689 "Magic mismatch, very weird!");
3694 has_huge_files
= ext4_has_feature_huge_file(sb
);
3695 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3697 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3699 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3700 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3701 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3703 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3704 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3705 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3706 (!is_power_of_2(sbi
->s_inode_size
)) ||
3707 (sbi
->s_inode_size
> blocksize
)) {
3708 ext4_msg(sb
, KERN_ERR
,
3709 "unsupported inode size: %d",
3713 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3714 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3717 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3718 if (ext4_has_feature_64bit(sb
)) {
3719 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3720 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3721 !is_power_of_2(sbi
->s_desc_size
)) {
3722 ext4_msg(sb
, KERN_ERR
,
3723 "unsupported descriptor size %lu",
3728 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3730 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3731 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3733 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3734 if (sbi
->s_inodes_per_block
== 0)
3736 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3737 sbi
->s_inodes_per_group
> blocksize
* 8) {
3738 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3739 sbi
->s_blocks_per_group
);
3742 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3743 sbi
->s_inodes_per_block
;
3744 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3746 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3747 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3748 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3750 for (i
= 0; i
< 4; i
++)
3751 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3752 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3753 if (ext4_has_feature_dir_index(sb
)) {
3754 i
= le32_to_cpu(es
->s_flags
);
3755 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3756 sbi
->s_hash_unsigned
= 3;
3757 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3758 #ifdef __CHAR_UNSIGNED__
3759 if (!(sb
->s_flags
& MS_RDONLY
))
3761 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3762 sbi
->s_hash_unsigned
= 3;
3764 if (!(sb
->s_flags
& MS_RDONLY
))
3766 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3771 /* Handle clustersize */
3772 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3773 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
3775 if (clustersize
< blocksize
) {
3776 ext4_msg(sb
, KERN_ERR
,
3777 "cluster size (%d) smaller than "
3778 "block size (%d)", clustersize
, blocksize
);
3781 if (le32_to_cpu(es
->s_log_cluster_size
) >
3782 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3783 ext4_msg(sb
, KERN_ERR
,
3784 "Invalid log cluster size: %u",
3785 le32_to_cpu(es
->s_log_cluster_size
));
3788 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3789 le32_to_cpu(es
->s_log_block_size
);
3790 sbi
->s_clusters_per_group
=
3791 le32_to_cpu(es
->s_clusters_per_group
);
3792 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3793 ext4_msg(sb
, KERN_ERR
,
3794 "#clusters per group too big: %lu",
3795 sbi
->s_clusters_per_group
);
3798 if (sbi
->s_blocks_per_group
!=
3799 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3800 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3801 "clusters per group (%lu) inconsistent",
3802 sbi
->s_blocks_per_group
,
3803 sbi
->s_clusters_per_group
);
3807 if (clustersize
!= blocksize
) {
3808 ext4_warning(sb
, "fragment/cluster size (%d) != "
3809 "block size (%d)", clustersize
,
3811 clustersize
= blocksize
;
3813 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3814 ext4_msg(sb
, KERN_ERR
,
3815 "#blocks per group too big: %lu",
3816 sbi
->s_blocks_per_group
);
3819 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3820 sbi
->s_cluster_bits
= 0;
3822 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3824 /* Do we have standard group size of clustersize * 8 blocks ? */
3825 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3826 set_opt2(sb
, STD_GROUP_SIZE
);
3829 * Test whether we have more sectors than will fit in sector_t,
3830 * and whether the max offset is addressable by the page cache.
3832 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3833 ext4_blocks_count(es
));
3835 ext4_msg(sb
, KERN_ERR
, "filesystem"
3836 " too large to mount safely on this system");
3837 if (sizeof(sector_t
) < 8)
3838 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3842 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3845 /* check blocks count against device size */
3846 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3847 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3848 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3849 "exceeds size of device (%llu blocks)",
3850 ext4_blocks_count(es
), blocks_count
);
3855 * It makes no sense for the first data block to be beyond the end
3856 * of the filesystem.
3858 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3859 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3860 "block %u is beyond end of filesystem (%llu)",
3861 le32_to_cpu(es
->s_first_data_block
),
3862 ext4_blocks_count(es
));
3865 blocks_count
= (ext4_blocks_count(es
) -
3866 le32_to_cpu(es
->s_first_data_block
) +
3867 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3868 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3869 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3870 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3871 "(block count %llu, first data block %u, "
3872 "blocks per group %lu)", sbi
->s_groups_count
,
3873 ext4_blocks_count(es
),
3874 le32_to_cpu(es
->s_first_data_block
),
3875 EXT4_BLOCKS_PER_GROUP(sb
));
3878 sbi
->s_groups_count
= blocks_count
;
3879 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3880 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3881 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3882 EXT4_DESC_PER_BLOCK(sb
);
3883 if (ext4_has_feature_meta_bg(sb
)) {
3884 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
3885 ext4_msg(sb
, KERN_WARNING
,
3886 "first meta block group too large: %u "
3887 "(group descriptor block count %u)",
3888 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
3892 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3893 sizeof(struct buffer_head
*),
3895 if (sbi
->s_group_desc
== NULL
) {
3896 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3901 bgl_lock_init(sbi
->s_blockgroup_lock
);
3903 for (i
= 0; i
< db_count
; i
++) {
3904 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3905 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3906 if (!sbi
->s_group_desc
[i
]) {
3907 ext4_msg(sb
, KERN_ERR
,
3908 "can't read group descriptor %d", i
);
3913 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
3914 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3915 ret
= -EFSCORRUPTED
;
3919 sbi
->s_gdb_count
= db_count
;
3920 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3921 spin_lock_init(&sbi
->s_next_gen_lock
);
3923 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3924 (unsigned long) sb
);
3926 /* Register extent status tree shrinker */
3927 if (ext4_es_register_shrinker(sbi
))
3930 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3931 sbi
->s_extent_max_zeroout_kb
= 32;
3934 * set up enough so that it can read an inode
3936 sb
->s_op
= &ext4_sops
;
3937 sb
->s_export_op
= &ext4_export_ops
;
3938 sb
->s_xattr
= ext4_xattr_handlers
;
3939 sb
->s_cop
= &ext4_cryptops
;
3941 sb
->dq_op
= &ext4_quota_operations
;
3942 if (ext4_has_feature_quota(sb
))
3943 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
3945 sb
->s_qcop
= &ext4_qctl_operations
;
3946 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
3948 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3950 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3951 mutex_init(&sbi
->s_orphan_lock
);
3955 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3956 ext4_has_feature_journal_needs_recovery(sb
));
3958 if (ext4_has_feature_mmp(sb
) && !(sb
->s_flags
& MS_RDONLY
))
3959 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3960 goto failed_mount3a
;
3963 * The first inode we look at is the journal inode. Don't try
3964 * root first: it may be modified in the journal!
3966 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
3967 err
= ext4_load_journal(sb
, es
, journal_devnum
);
3969 goto failed_mount3a
;
3970 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3971 ext4_has_feature_journal_needs_recovery(sb
)) {
3972 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3973 "suppressed and not mounted read-only");
3974 goto failed_mount_wq
;
3976 /* Nojournal mode, all journal mount options are illegal */
3977 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
3978 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3979 "journal_checksum, fs mounted w/o journal");
3980 goto failed_mount_wq
;
3982 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3983 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3984 "journal_async_commit, fs mounted w/o journal");
3985 goto failed_mount_wq
;
3987 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
3988 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3989 "commit=%lu, fs mounted w/o journal",
3990 sbi
->s_commit_interval
/ HZ
);
3991 goto failed_mount_wq
;
3993 if (EXT4_MOUNT_DATA_FLAGS
&
3994 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
3995 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3996 "data=, fs mounted w/o journal");
3997 goto failed_mount_wq
;
3999 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4000 clear_opt(sb
, JOURNAL_CHECKSUM
);
4001 clear_opt(sb
, DATA_FLAGS
);
4002 sbi
->s_journal
= NULL
;
4007 if (ext4_has_feature_64bit(sb
) &&
4008 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4009 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4010 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4011 goto failed_mount_wq
;
4014 if (!set_journal_csum_feature_set(sb
)) {
4015 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4017 goto failed_mount_wq
;
4020 /* We have now updated the journal if required, so we can
4021 * validate the data journaling mode. */
4022 switch (test_opt(sb
, DATA_FLAGS
)) {
4024 /* No mode set, assume a default based on the journal
4025 * capabilities: ORDERED_DATA if the journal can
4026 * cope, else JOURNAL_DATA
4028 if (jbd2_journal_check_available_features
4029 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4030 set_opt(sb
, ORDERED_DATA
);
4032 set_opt(sb
, JOURNAL_DATA
);
4035 case EXT4_MOUNT_ORDERED_DATA
:
4036 case EXT4_MOUNT_WRITEBACK_DATA
:
4037 if (!jbd2_journal_check_available_features
4038 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4039 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4040 "requested data journaling mode");
4041 goto failed_mount_wq
;
4047 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4048 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4049 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4050 "journal_async_commit in data=ordered mode");
4051 goto failed_mount_wq
;
4054 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4056 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4059 sbi
->s_mb_cache
= ext4_xattr_create_cache();
4060 if (!sbi
->s_mb_cache
) {
4061 ext4_msg(sb
, KERN_ERR
, "Failed to create an mb_cache");
4062 goto failed_mount_wq
;
4065 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4066 (blocksize
!= PAGE_SIZE
)) {
4067 ext4_msg(sb
, KERN_ERR
,
4068 "Unsupported blocksize for fs encryption");
4069 goto failed_mount_wq
;
4072 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !(sb
->s_flags
& MS_RDONLY
) &&
4073 !ext4_has_feature_encrypt(sb
)) {
4074 ext4_set_feature_encrypt(sb
);
4075 ext4_commit_super(sb
, 1);
4079 * Get the # of file system overhead blocks from the
4080 * superblock if present.
4082 if (es
->s_overhead_clusters
)
4083 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4085 err
= ext4_calculate_overhead(sb
);
4087 goto failed_mount_wq
;
4091 * The maximum number of concurrent works can be high and
4092 * concurrency isn't really necessary. Limit it to 1.
4094 EXT4_SB(sb
)->rsv_conversion_wq
=
4095 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4096 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4097 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4103 * The jbd2_journal_load will have done any necessary log recovery,
4104 * so we can safely mount the rest of the filesystem now.
4107 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4109 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4110 ret
= PTR_ERR(root
);
4114 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4115 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4119 sb
->s_root
= d_make_root(root
);
4121 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4126 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4127 sb
->s_flags
|= MS_RDONLY
;
4129 /* determine the minimum size of new large inodes, if present */
4130 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
4131 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4132 EXT4_GOOD_OLD_INODE_SIZE
;
4133 if (ext4_has_feature_extra_isize(sb
)) {
4134 if (sbi
->s_want_extra_isize
<
4135 le16_to_cpu(es
->s_want_extra_isize
))
4136 sbi
->s_want_extra_isize
=
4137 le16_to_cpu(es
->s_want_extra_isize
);
4138 if (sbi
->s_want_extra_isize
<
4139 le16_to_cpu(es
->s_min_extra_isize
))
4140 sbi
->s_want_extra_isize
=
4141 le16_to_cpu(es
->s_min_extra_isize
);
4144 /* Check if enough inode space is available */
4145 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4146 sbi
->s_inode_size
) {
4147 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4148 EXT4_GOOD_OLD_INODE_SIZE
;
4149 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4153 ext4_set_resv_clusters(sb
);
4155 err
= ext4_setup_system_zone(sb
);
4157 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4159 goto failed_mount4a
;
4163 err
= ext4_mb_init(sb
);
4165 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4170 block
= ext4_count_free_clusters(sb
);
4171 ext4_free_blocks_count_set(sbi
->s_es
,
4172 EXT4_C2B(sbi
, block
));
4173 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4176 unsigned long freei
= ext4_count_free_inodes(sb
);
4177 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4178 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4182 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4183 ext4_count_dirs(sb
), GFP_KERNEL
);
4185 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4188 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4191 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4195 if (ext4_has_feature_flex_bg(sb
))
4196 if (!ext4_fill_flex_info(sb
)) {
4197 ext4_msg(sb
, KERN_ERR
,
4198 "unable to initialize "
4199 "flex_bg meta info!");
4203 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4207 err
= ext4_register_sysfs(sb
);
4212 /* Enable quota usage during mount. */
4213 if (ext4_has_feature_quota(sb
) && !(sb
->s_flags
& MS_RDONLY
)) {
4214 err
= ext4_enable_quotas(sb
);
4218 #endif /* CONFIG_QUOTA */
4220 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4221 ext4_orphan_cleanup(sb
, es
);
4222 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4223 if (needs_recovery
) {
4224 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4225 ext4_mark_recovery_complete(sb
, es
);
4227 if (EXT4_SB(sb
)->s_journal
) {
4228 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4229 descr
= " journalled data mode";
4230 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4231 descr
= " ordered data mode";
4233 descr
= " writeback data mode";
4235 descr
= "out journal";
4237 if (test_opt(sb
, DISCARD
)) {
4238 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4239 if (!blk_queue_discard(q
))
4240 ext4_msg(sb
, KERN_WARNING
,
4241 "mounting with \"discard\" option, but "
4242 "the device does not support discard");
4245 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4246 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4247 "Opts: %.*s%s%s", descr
,
4248 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4249 sbi
->s_es
->s_mount_opts
,
4250 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4252 if (es
->s_error_count
)
4253 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4255 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4256 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4257 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4258 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4261 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4262 memcpy(sbi
->key_prefix
, EXT4_KEY_DESC_PREFIX
,
4263 EXT4_KEY_DESC_PREFIX_SIZE
);
4264 sbi
->key_prefix_size
= EXT4_KEY_DESC_PREFIX_SIZE
;
4270 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4275 ext4_unregister_sysfs(sb
);
4278 ext4_unregister_li_request(sb
);
4280 ext4_mb_release(sb
);
4281 if (sbi
->s_flex_groups
)
4282 kvfree(sbi
->s_flex_groups
);
4283 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4284 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4285 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4286 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4288 ext4_ext_release(sb
);
4289 ext4_release_system_zone(sb
);
4294 ext4_msg(sb
, KERN_ERR
, "mount failed");
4295 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4296 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4298 if (sbi
->s_mb_cache
) {
4299 ext4_xattr_destroy_cache(sbi
->s_mb_cache
);
4300 sbi
->s_mb_cache
= NULL
;
4302 if (sbi
->s_journal
) {
4303 jbd2_journal_destroy(sbi
->s_journal
);
4304 sbi
->s_journal
= NULL
;
4307 ext4_es_unregister_shrinker(sbi
);
4309 del_timer_sync(&sbi
->s_err_report
);
4311 kthread_stop(sbi
->s_mmp_tsk
);
4313 for (i
= 0; i
< db_count
; i
++)
4314 brelse(sbi
->s_group_desc
[i
]);
4315 kvfree(sbi
->s_group_desc
);
4317 if (sbi
->s_chksum_driver
)
4318 crypto_free_shash(sbi
->s_chksum_driver
);
4320 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4321 kfree(sbi
->s_qf_names
[i
]);
4323 ext4_blkdev_remove(sbi
);
4326 /* sb->s_user_ns will be put when sb is destroyed */
4327 sb
->s_fs_info
= NULL
;
4328 kfree(sbi
->s_blockgroup_lock
);
4332 return err
? err
: ret
;
4336 * Setup any per-fs journal parameters now. We'll do this both on
4337 * initial mount, once the journal has been initialised but before we've
4338 * done any recovery; and again on any subsequent remount.
4340 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4342 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4344 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4345 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4346 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4348 write_lock(&journal
->j_state_lock
);
4349 if (test_opt(sb
, BARRIER
))
4350 journal
->j_flags
|= JBD2_BARRIER
;
4352 journal
->j_flags
&= ~JBD2_BARRIER
;
4353 if (test_opt(sb
, DATA_ERR_ABORT
))
4354 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4356 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4357 write_unlock(&journal
->j_state_lock
);
4360 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4361 unsigned int journal_inum
)
4363 struct inode
*journal_inode
;
4366 * Test for the existence of a valid inode on disk. Bad things
4367 * happen if we iget() an unused inode, as the subsequent iput()
4368 * will try to delete it.
4370 journal_inode
= ext4_iget(sb
, journal_inum
);
4371 if (IS_ERR(journal_inode
)) {
4372 ext4_msg(sb
, KERN_ERR
, "no journal found");
4375 if (!journal_inode
->i_nlink
) {
4376 make_bad_inode(journal_inode
);
4377 iput(journal_inode
);
4378 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4382 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4383 journal_inode
, journal_inode
->i_size
);
4384 if (!S_ISREG(journal_inode
->i_mode
)) {
4385 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4386 iput(journal_inode
);
4389 return journal_inode
;
4392 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4393 unsigned int journal_inum
)
4395 struct inode
*journal_inode
;
4398 BUG_ON(!ext4_has_feature_journal(sb
));
4400 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4404 journal
= jbd2_journal_init_inode(journal_inode
);
4406 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4407 iput(journal_inode
);
4410 journal
->j_private
= sb
;
4411 ext4_init_journal_params(sb
, journal
);
4415 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4418 struct buffer_head
*bh
;
4422 int hblock
, blocksize
;
4423 ext4_fsblk_t sb_block
;
4424 unsigned long offset
;
4425 struct ext4_super_block
*es
;
4426 struct block_device
*bdev
;
4428 BUG_ON(!ext4_has_feature_journal(sb
));
4430 bdev
= ext4_blkdev_get(j_dev
, sb
);
4434 blocksize
= sb
->s_blocksize
;
4435 hblock
= bdev_logical_block_size(bdev
);
4436 if (blocksize
< hblock
) {
4437 ext4_msg(sb
, KERN_ERR
,
4438 "blocksize too small for journal device");
4442 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4443 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4444 set_blocksize(bdev
, blocksize
);
4445 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4446 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4447 "external journal");
4451 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4452 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4453 !(le32_to_cpu(es
->s_feature_incompat
) &
4454 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4455 ext4_msg(sb
, KERN_ERR
, "external journal has "
4461 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4462 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4463 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4464 ext4_msg(sb
, KERN_ERR
, "external journal has "
4465 "corrupt superblock");
4470 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4471 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4476 len
= ext4_blocks_count(es
);
4477 start
= sb_block
+ 1;
4478 brelse(bh
); /* we're done with the superblock */
4480 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4481 start
, len
, blocksize
);
4483 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4486 journal
->j_private
= sb
;
4487 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4488 wait_on_buffer(journal
->j_sb_buffer
);
4489 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4490 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4493 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4494 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4495 "user (unsupported) - %d",
4496 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4499 EXT4_SB(sb
)->journal_bdev
= bdev
;
4500 ext4_init_journal_params(sb
, journal
);
4504 jbd2_journal_destroy(journal
);
4506 ext4_blkdev_put(bdev
);
4510 static int ext4_load_journal(struct super_block
*sb
,
4511 struct ext4_super_block
*es
,
4512 unsigned long journal_devnum
)
4515 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4518 int really_read_only
;
4520 BUG_ON(!ext4_has_feature_journal(sb
));
4522 if (journal_devnum
&&
4523 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4524 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4525 "numbers have changed");
4526 journal_dev
= new_decode_dev(journal_devnum
);
4528 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4530 really_read_only
= bdev_read_only(sb
->s_bdev
);
4533 * Are we loading a blank journal or performing recovery after a
4534 * crash? For recovery, we need to check in advance whether we
4535 * can get read-write access to the device.
4537 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4538 if (sb
->s_flags
& MS_RDONLY
) {
4539 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4540 "required on readonly filesystem");
4541 if (really_read_only
) {
4542 ext4_msg(sb
, KERN_ERR
, "write access "
4543 "unavailable, cannot proceed");
4546 ext4_msg(sb
, KERN_INFO
, "write access will "
4547 "be enabled during recovery");
4551 if (journal_inum
&& journal_dev
) {
4552 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4553 "and inode journals!");
4558 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4561 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4565 if (!(journal
->j_flags
& JBD2_BARRIER
))
4566 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4568 if (!ext4_has_feature_journal_needs_recovery(sb
))
4569 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4571 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4573 memcpy(save
, ((char *) es
) +
4574 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4575 err
= jbd2_journal_load(journal
);
4577 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4578 save
, EXT4_S_ERR_LEN
);
4583 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4584 jbd2_journal_destroy(journal
);
4588 EXT4_SB(sb
)->s_journal
= journal
;
4589 ext4_clear_journal_err(sb
, es
);
4591 if (!really_read_only
&& journal_devnum
&&
4592 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4593 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4595 /* Make sure we flush the recovery flag to disk. */
4596 ext4_commit_super(sb
, 1);
4602 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4604 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4605 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4608 if (!sbh
|| block_device_ejected(sb
))
4611 * If the file system is mounted read-only, don't update the
4612 * superblock write time. This avoids updating the superblock
4613 * write time when we are mounting the root file system
4614 * read/only but we need to replay the journal; at that point,
4615 * for people who are east of GMT and who make their clock
4616 * tick in localtime for Windows bug-for-bug compatibility,
4617 * the clock is set in the future, and this will cause e2fsck
4618 * to complain and force a full file system check.
4620 if (!(sb
->s_flags
& MS_RDONLY
))
4621 es
->s_wtime
= cpu_to_le32(get_seconds());
4622 if (sb
->s_bdev
->bd_part
)
4623 es
->s_kbytes_written
=
4624 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4625 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4626 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4628 es
->s_kbytes_written
=
4629 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4630 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4631 ext4_free_blocks_count_set(es
,
4632 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4633 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4634 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4635 es
->s_free_inodes_count
=
4636 cpu_to_le32(percpu_counter_sum_positive(
4637 &EXT4_SB(sb
)->s_freeinodes_counter
));
4638 BUFFER_TRACE(sbh
, "marking dirty");
4639 ext4_superblock_csum_set(sb
);
4642 if (buffer_write_io_error(sbh
)) {
4644 * Oh, dear. A previous attempt to write the
4645 * superblock failed. This could happen because the
4646 * USB device was yanked out. Or it could happen to
4647 * be a transient write error and maybe the block will
4648 * be remapped. Nothing we can do but to retry the
4649 * write and hope for the best.
4651 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4652 "superblock detected");
4653 clear_buffer_write_io_error(sbh
);
4654 set_buffer_uptodate(sbh
);
4656 mark_buffer_dirty(sbh
);
4659 error
= __sync_dirty_buffer(sbh
,
4660 test_opt(sb
, BARRIER
) ? REQ_FUA
: REQ_SYNC
);
4664 error
= buffer_write_io_error(sbh
);
4666 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4668 clear_buffer_write_io_error(sbh
);
4669 set_buffer_uptodate(sbh
);
4676 * Have we just finished recovery? If so, and if we are mounting (or
4677 * remounting) the filesystem readonly, then we will end up with a
4678 * consistent fs on disk. Record that fact.
4680 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4681 struct ext4_super_block
*es
)
4683 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4685 if (!ext4_has_feature_journal(sb
)) {
4686 BUG_ON(journal
!= NULL
);
4689 jbd2_journal_lock_updates(journal
);
4690 if (jbd2_journal_flush(journal
) < 0)
4693 if (ext4_has_feature_journal_needs_recovery(sb
) &&
4694 sb
->s_flags
& MS_RDONLY
) {
4695 ext4_clear_feature_journal_needs_recovery(sb
);
4696 ext4_commit_super(sb
, 1);
4700 jbd2_journal_unlock_updates(journal
);
4704 * If we are mounting (or read-write remounting) a filesystem whose journal
4705 * has recorded an error from a previous lifetime, move that error to the
4706 * main filesystem now.
4708 static void ext4_clear_journal_err(struct super_block
*sb
,
4709 struct ext4_super_block
*es
)
4715 BUG_ON(!ext4_has_feature_journal(sb
));
4717 journal
= EXT4_SB(sb
)->s_journal
;
4720 * Now check for any error status which may have been recorded in the
4721 * journal by a prior ext4_error() or ext4_abort()
4724 j_errno
= jbd2_journal_errno(journal
);
4728 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4729 ext4_warning(sb
, "Filesystem error recorded "
4730 "from previous mount: %s", errstr
);
4731 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4733 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4734 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4735 ext4_commit_super(sb
, 1);
4737 jbd2_journal_clear_err(journal
);
4738 jbd2_journal_update_sb_errno(journal
);
4743 * Force the running and committing transactions to commit,
4744 * and wait on the commit.
4746 int ext4_force_commit(struct super_block
*sb
)
4750 if (sb
->s_flags
& MS_RDONLY
)
4753 journal
= EXT4_SB(sb
)->s_journal
;
4754 return ext4_journal_force_commit(journal
);
4757 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4761 bool needs_barrier
= false;
4762 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4764 trace_ext4_sync_fs(sb
, wait
);
4765 flush_workqueue(sbi
->rsv_conversion_wq
);
4767 * Writeback quota in non-journalled quota case - journalled quota has
4770 dquot_writeback_dquots(sb
, -1);
4772 * Data writeback is possible w/o journal transaction, so barrier must
4773 * being sent at the end of the function. But we can skip it if
4774 * transaction_commit will do it for us.
4776 if (sbi
->s_journal
) {
4777 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4778 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4779 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4780 needs_barrier
= true;
4782 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4784 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4787 } else if (wait
&& test_opt(sb
, BARRIER
))
4788 needs_barrier
= true;
4789 if (needs_barrier
) {
4791 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4800 * LVM calls this function before a (read-only) snapshot is created. This
4801 * gives us a chance to flush the journal completely and mark the fs clean.
4803 * Note that only this function cannot bring a filesystem to be in a clean
4804 * state independently. It relies on upper layer to stop all data & metadata
4807 static int ext4_freeze(struct super_block
*sb
)
4812 if (sb
->s_flags
& MS_RDONLY
)
4815 journal
= EXT4_SB(sb
)->s_journal
;
4818 /* Now we set up the journal barrier. */
4819 jbd2_journal_lock_updates(journal
);
4822 * Don't clear the needs_recovery flag if we failed to
4823 * flush the journal.
4825 error
= jbd2_journal_flush(journal
);
4829 /* Journal blocked and flushed, clear needs_recovery flag. */
4830 ext4_clear_feature_journal_needs_recovery(sb
);
4833 error
= ext4_commit_super(sb
, 1);
4836 /* we rely on upper layer to stop further updates */
4837 jbd2_journal_unlock_updates(journal
);
4842 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4843 * flag here, even though the filesystem is not technically dirty yet.
4845 static int ext4_unfreeze(struct super_block
*sb
)
4847 if (sb
->s_flags
& MS_RDONLY
)
4850 if (EXT4_SB(sb
)->s_journal
) {
4851 /* Reset the needs_recovery flag before the fs is unlocked. */
4852 ext4_set_feature_journal_needs_recovery(sb
);
4855 ext4_commit_super(sb
, 1);
4860 * Structure to save mount options for ext4_remount's benefit
4862 struct ext4_mount_options
{
4863 unsigned long s_mount_opt
;
4864 unsigned long s_mount_opt2
;
4867 unsigned long s_commit_interval
;
4868 u32 s_min_batch_time
, s_max_batch_time
;
4871 char *s_qf_names
[EXT4_MAXQUOTAS
];
4875 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4877 struct ext4_super_block
*es
;
4878 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4879 unsigned long old_sb_flags
;
4880 struct ext4_mount_options old_opts
;
4881 int enable_quota
= 0;
4883 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4888 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4890 /* Store the original options */
4891 old_sb_flags
= sb
->s_flags
;
4892 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4893 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4894 old_opts
.s_resuid
= sbi
->s_resuid
;
4895 old_opts
.s_resgid
= sbi
->s_resgid
;
4896 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4897 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4898 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4900 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4901 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4902 if (sbi
->s_qf_names
[i
]) {
4903 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4905 if (!old_opts
.s_qf_names
[i
]) {
4906 for (j
= 0; j
< i
; j
++)
4907 kfree(old_opts
.s_qf_names
[j
]);
4912 old_opts
.s_qf_names
[i
] = NULL
;
4914 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4915 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4917 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4922 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
4923 test_opt(sb
, JOURNAL_CHECKSUM
)) {
4924 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
4925 "during remount not supported; ignoring");
4926 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4929 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
4930 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
4931 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4932 "both data=journal and delalloc");
4936 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
4937 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4938 "both data=journal and dioread_nolock");
4942 if (test_opt(sb
, DAX
)) {
4943 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4944 "both data=journal and dax");
4948 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
4949 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4950 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4951 "journal_async_commit in data=ordered mode");
4957 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
4958 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
4959 "dax flag with busy inodes while remounting");
4960 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
4963 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4964 ext4_abort(sb
, "Abort forced by user");
4966 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4967 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4971 if (sbi
->s_journal
) {
4972 ext4_init_journal_params(sb
, sbi
->s_journal
);
4973 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4976 if (*flags
& MS_LAZYTIME
)
4977 sb
->s_flags
|= MS_LAZYTIME
;
4979 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4980 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4985 if (*flags
& MS_RDONLY
) {
4986 err
= sync_filesystem(sb
);
4989 err
= dquot_suspend(sb
, -1);
4994 * First of all, the unconditional stuff we have to do
4995 * to disable replay of the journal when we next remount
4997 sb
->s_flags
|= MS_RDONLY
;
5000 * OK, test if we are remounting a valid rw partition
5001 * readonly, and if so set the rdonly flag and then
5002 * mark the partition as valid again.
5004 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5005 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5006 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5009 ext4_mark_recovery_complete(sb
, es
);
5011 /* Make sure we can mount this feature set readwrite */
5012 if (ext4_has_feature_readonly(sb
) ||
5013 !ext4_feature_set_ok(sb
, 0)) {
5018 * Make sure the group descriptor checksums
5019 * are sane. If they aren't, refuse to remount r/w.
5021 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5022 struct ext4_group_desc
*gdp
=
5023 ext4_get_group_desc(sb
, g
, NULL
);
5025 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5026 ext4_msg(sb
, KERN_ERR
,
5027 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5028 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5029 le16_to_cpu(gdp
->bg_checksum
));
5036 * If we have an unprocessed orphan list hanging
5037 * around from a previously readonly bdev mount,
5038 * require a full umount/remount for now.
5040 if (es
->s_last_orphan
) {
5041 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5042 "remount RDWR because of unprocessed "
5043 "orphan inode list. Please "
5044 "umount/remount instead");
5050 * Mounting a RDONLY partition read-write, so reread
5051 * and store the current valid flag. (It may have
5052 * been changed by e2fsck since we originally mounted
5056 ext4_clear_journal_err(sb
, es
);
5057 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5058 if (!ext4_setup_super(sb
, es
, 0))
5059 sb
->s_flags
&= ~MS_RDONLY
;
5060 if (ext4_has_feature_mmp(sb
))
5061 if (ext4_multi_mount_protect(sb
,
5062 le64_to_cpu(es
->s_mmp_block
))) {
5071 * Reinitialize lazy itable initialization thread based on
5074 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
5075 ext4_unregister_li_request(sb
);
5077 ext4_group_t first_not_zeroed
;
5078 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5079 ext4_register_li_request(sb
, first_not_zeroed
);
5082 ext4_setup_system_zone(sb
);
5083 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5084 ext4_commit_super(sb
, 1);
5087 /* Release old quota file names */
5088 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5089 kfree(old_opts
.s_qf_names
[i
]);
5091 if (sb_any_quota_suspended(sb
))
5092 dquot_resume(sb
, -1);
5093 else if (ext4_has_feature_quota(sb
)) {
5094 err
= ext4_enable_quotas(sb
);
5101 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5102 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5107 sb
->s_flags
= old_sb_flags
;
5108 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5109 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5110 sbi
->s_resuid
= old_opts
.s_resuid
;
5111 sbi
->s_resgid
= old_opts
.s_resgid
;
5112 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5113 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5114 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5116 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5117 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5118 kfree(sbi
->s_qf_names
[i
]);
5119 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5127 static int ext4_statfs_project(struct super_block
*sb
,
5128 kprojid_t projid
, struct kstatfs
*buf
)
5131 struct dquot
*dquot
;
5135 qid
= make_kqid_projid(projid
);
5136 dquot
= dqget(sb
, qid
);
5138 return PTR_ERR(dquot
);
5139 spin_lock(&dq_data_lock
);
5141 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5142 dquot
->dq_dqb
.dqb_bsoftlimit
:
5143 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5144 if (limit
&& buf
->f_blocks
> limit
) {
5145 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
5146 buf
->f_blocks
= limit
;
5147 buf
->f_bfree
= buf
->f_bavail
=
5148 (buf
->f_blocks
> curblock
) ?
5149 (buf
->f_blocks
- curblock
) : 0;
5152 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5153 dquot
->dq_dqb
.dqb_isoftlimit
:
5154 dquot
->dq_dqb
.dqb_ihardlimit
;
5155 if (limit
&& buf
->f_files
> limit
) {
5156 buf
->f_files
= limit
;
5158 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5159 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5162 spin_unlock(&dq_data_lock
);
5168 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5170 struct super_block
*sb
= dentry
->d_sb
;
5171 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5172 struct ext4_super_block
*es
= sbi
->s_es
;
5173 ext4_fsblk_t overhead
= 0, resv_blocks
;
5176 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5178 if (!test_opt(sb
, MINIX_DF
))
5179 overhead
= sbi
->s_overhead
;
5181 buf
->f_type
= EXT4_SUPER_MAGIC
;
5182 buf
->f_bsize
= sb
->s_blocksize
;
5183 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5184 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5185 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5186 /* prevent underflow in case that few free space is available */
5187 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5188 buf
->f_bavail
= buf
->f_bfree
-
5189 (ext4_r_blocks_count(es
) + resv_blocks
);
5190 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5192 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5193 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5194 buf
->f_namelen
= EXT4_NAME_LEN
;
5195 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5196 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5197 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5198 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5201 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5202 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5203 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5208 /* Helper function for writing quotas on sync - we need to start transaction
5209 * before quota file is locked for write. Otherwise the are possible deadlocks:
5210 * Process 1 Process 2
5211 * ext4_create() quota_sync()
5212 * jbd2_journal_start() write_dquot()
5213 * dquot_initialize() down(dqio_mutex)
5214 * down(dqio_mutex) jbd2_journal_start()
5220 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5222 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5225 static int ext4_write_dquot(struct dquot
*dquot
)
5229 struct inode
*inode
;
5231 inode
= dquot_to_inode(dquot
);
5232 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5233 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5235 return PTR_ERR(handle
);
5236 ret
= dquot_commit(dquot
);
5237 err
= ext4_journal_stop(handle
);
5243 static int ext4_acquire_dquot(struct dquot
*dquot
)
5248 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5249 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5251 return PTR_ERR(handle
);
5252 ret
= dquot_acquire(dquot
);
5253 err
= ext4_journal_stop(handle
);
5259 static int ext4_release_dquot(struct dquot
*dquot
)
5264 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5265 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5266 if (IS_ERR(handle
)) {
5267 /* Release dquot anyway to avoid endless cycle in dqput() */
5268 dquot_release(dquot
);
5269 return PTR_ERR(handle
);
5271 ret
= dquot_release(dquot
);
5272 err
= ext4_journal_stop(handle
);
5278 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5280 struct super_block
*sb
= dquot
->dq_sb
;
5281 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5283 /* Are we journaling quotas? */
5284 if (ext4_has_feature_quota(sb
) ||
5285 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5286 dquot_mark_dquot_dirty(dquot
);
5287 return ext4_write_dquot(dquot
);
5289 return dquot_mark_dquot_dirty(dquot
);
5293 static int ext4_write_info(struct super_block
*sb
, int type
)
5298 /* Data block + inode block */
5299 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5301 return PTR_ERR(handle
);
5302 ret
= dquot_commit_info(sb
, type
);
5303 err
= ext4_journal_stop(handle
);
5310 * Turn on quotas during mount time - we need to find
5311 * the quota file and such...
5313 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5315 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5316 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5319 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5321 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5323 /* The first argument of lockdep_set_subclass has to be
5324 * *exactly* the same as the argument to init_rwsem() --- in
5325 * this case, in init_once() --- or lockdep gets unhappy
5326 * because the name of the lock is set using the
5327 * stringification of the argument to init_rwsem().
5329 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5330 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5334 * Standard function to be called on quota_on
5336 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5337 const struct path
*path
)
5341 if (!test_opt(sb
, QUOTA
))
5344 /* Quotafile not on the same filesystem? */
5345 if (path
->dentry
->d_sb
!= sb
)
5347 /* Journaling quota? */
5348 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5349 /* Quotafile not in fs root? */
5350 if (path
->dentry
->d_parent
!= sb
->s_root
)
5351 ext4_msg(sb
, KERN_WARNING
,
5352 "Quota file not on filesystem root. "
5353 "Journaled quota will not work");
5357 * When we journal data on quota file, we have to flush journal to see
5358 * all updates to the file when we bypass pagecache...
5360 if (EXT4_SB(sb
)->s_journal
&&
5361 ext4_should_journal_data(d_inode(path
->dentry
))) {
5363 * We don't need to lock updates but journal_flush() could
5364 * otherwise be livelocked...
5366 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5367 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5368 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5372 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5373 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5375 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5380 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5384 struct inode
*qf_inode
;
5385 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5386 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5387 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5388 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5391 BUG_ON(!ext4_has_feature_quota(sb
));
5393 if (!qf_inums
[type
])
5396 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5397 if (IS_ERR(qf_inode
)) {
5398 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5399 return PTR_ERR(qf_inode
);
5402 /* Don't account quota for quota files to avoid recursion */
5403 qf_inode
->i_flags
|= S_NOQUOTA
;
5404 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5405 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5408 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5413 /* Enable usage tracking for all quota types. */
5414 static int ext4_enable_quotas(struct super_block
*sb
)
5417 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5418 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5419 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5420 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5422 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5423 test_opt(sb
, USRQUOTA
),
5424 test_opt(sb
, GRPQUOTA
),
5425 test_opt(sb
, PRJQUOTA
),
5428 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5429 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5430 if (qf_inums
[type
]) {
5431 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5432 DQUOT_USAGE_ENABLED
|
5433 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5436 "Failed to enable quota tracking "
5437 "(type=%d, err=%d). Please run "
5438 "e2fsck to fix.", type
, err
);
5446 static int ext4_quota_off(struct super_block
*sb
, int type
)
5448 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5451 /* Force all delayed allocation blocks to be allocated.
5452 * Caller already holds s_umount sem */
5453 if (test_opt(sb
, DELALLOC
))
5454 sync_filesystem(sb
);
5459 /* Update modification times of quota files when userspace can
5460 * start looking at them */
5461 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5464 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5465 ext4_mark_inode_dirty(handle
, inode
);
5466 ext4_journal_stop(handle
);
5469 return dquot_quota_off(sb
, type
);
5472 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5473 * acquiring the locks... As quota files are never truncated and quota code
5474 * itself serializes the operations (and no one else should touch the files)
5475 * we don't have to be afraid of races */
5476 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5477 size_t len
, loff_t off
)
5479 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5480 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5481 int offset
= off
& (sb
->s_blocksize
- 1);
5484 struct buffer_head
*bh
;
5485 loff_t i_size
= i_size_read(inode
);
5489 if (off
+len
> i_size
)
5492 while (toread
> 0) {
5493 tocopy
= sb
->s_blocksize
- offset
< toread
?
5494 sb
->s_blocksize
- offset
: toread
;
5495 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5498 if (!bh
) /* A hole? */
5499 memset(data
, 0, tocopy
);
5501 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5511 /* Write to quotafile (we know the transaction is already started and has
5512 * enough credits) */
5513 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5514 const char *data
, size_t len
, loff_t off
)
5516 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5517 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5518 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5520 struct buffer_head
*bh
;
5521 handle_t
*handle
= journal_current_handle();
5523 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5524 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5525 " cancelled because transaction is not started",
5526 (unsigned long long)off
, (unsigned long long)len
);
5530 * Since we account only one data block in transaction credits,
5531 * then it is impossible to cross a block boundary.
5533 if (sb
->s_blocksize
- offset
< len
) {
5534 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5535 " cancelled because not block aligned",
5536 (unsigned long long)off
, (unsigned long long)len
);
5541 bh
= ext4_bread(handle
, inode
, blk
,
5542 EXT4_GET_BLOCKS_CREATE
|
5543 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5544 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5545 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5550 BUFFER_TRACE(bh
, "get write access");
5551 err
= ext4_journal_get_write_access(handle
, bh
);
5557 memcpy(bh
->b_data
+offset
, data
, len
);
5558 flush_dcache_page(bh
->b_page
);
5560 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5563 if (inode
->i_size
< off
+ len
) {
5564 i_size_write(inode
, off
+ len
);
5565 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5566 ext4_mark_inode_dirty(handle
, inode
);
5571 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5573 const struct quota_format_ops
*ops
;
5575 if (!sb_has_quota_loaded(sb
, qid
->type
))
5577 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5578 if (!ops
|| !ops
->get_next_id
)
5580 return dquot_get_next_id(sb
, qid
);
5584 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5585 const char *dev_name
, void *data
)
5587 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5590 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5591 static inline void register_as_ext2(void)
5593 int err
= register_filesystem(&ext2_fs_type
);
5596 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5599 static inline void unregister_as_ext2(void)
5601 unregister_filesystem(&ext2_fs_type
);
5604 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5606 if (ext4_has_unknown_ext2_incompat_features(sb
))
5608 if (sb
->s_flags
& MS_RDONLY
)
5610 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5615 static inline void register_as_ext2(void) { }
5616 static inline void unregister_as_ext2(void) { }
5617 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5620 static inline void register_as_ext3(void)
5622 int err
= register_filesystem(&ext3_fs_type
);
5625 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5628 static inline void unregister_as_ext3(void)
5630 unregister_filesystem(&ext3_fs_type
);
5633 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5635 if (ext4_has_unknown_ext3_incompat_features(sb
))
5637 if (!ext4_has_feature_journal(sb
))
5639 if (sb
->s_flags
& MS_RDONLY
)
5641 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5646 static struct file_system_type ext4_fs_type
= {
5647 .owner
= THIS_MODULE
,
5649 .mount
= ext4_mount
,
5650 .kill_sb
= kill_block_super
,
5651 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
5653 MODULE_ALIAS_FS("ext4");
5655 /* Shared across all ext4 file systems */
5656 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5658 static int __init
ext4_init_fs(void)
5662 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5663 ext4_li_info
= NULL
;
5664 mutex_init(&ext4_li_mtx
);
5666 /* Build-time check for flags consistency */
5667 ext4_check_flag_values();
5669 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5670 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5672 err
= ext4_init_es();
5676 err
= ext4_init_pageio();
5680 err
= ext4_init_system_zone();
5684 err
= ext4_init_sysfs();
5688 err
= ext4_init_mballoc();
5691 err
= init_inodecache();
5696 err
= register_filesystem(&ext4_fs_type
);
5702 unregister_as_ext2();
5703 unregister_as_ext3();
5704 destroy_inodecache();
5706 ext4_exit_mballoc();
5710 ext4_exit_system_zone();
5719 static void __exit
ext4_exit_fs(void)
5721 ext4_destroy_lazyinit_thread();
5722 unregister_as_ext2();
5723 unregister_as_ext3();
5724 unregister_filesystem(&ext4_fs_type
);
5725 destroy_inodecache();
5726 ext4_exit_mballoc();
5728 ext4_exit_system_zone();
5733 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5734 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5735 MODULE_LICENSE("GPL");
5736 module_init(ext4_init_fs
)
5737 module_exit(ext4_exit_fs
)