2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
43 #include <linux/user_namespace.h>
45 #include <linux/kthread.h>
46 #include <linux/freezer.h>
49 #include "ext4_extents.h" /* Needed for trace points definition */
50 #include "ext4_jbd2.h"
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/ext4.h>
59 static struct ext4_lazy_init
*ext4_li_info
;
60 static struct mutex ext4_li_mtx
;
61 static struct ratelimit_state ext4_mount_msg_ratelimit
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
66 static int ext4_commit_super(struct super_block
*sb
, int sync
);
67 static void ext4_mark_recovery_complete(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static void ext4_clear_journal_err(struct super_block
*sb
,
70 struct ext4_super_block
*es
);
71 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
72 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
73 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
74 static int ext4_unfreeze(struct super_block
*sb
);
75 static int ext4_freeze(struct super_block
*sb
);
76 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
77 const char *dev_name
, void *data
);
78 static inline int ext2_feature_set_ok(struct super_block
*sb
);
79 static inline int ext3_feature_set_ok(struct super_block
*sb
);
80 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block
*sb
);
83 static void ext4_clear_request_list(void);
84 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
85 unsigned int journal_inum
);
90 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
91 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
95 * page lock -> i_data_sem (rw)
97 * buffered write path:
98 * sb_start_write -> i_mutex -> mmap_sem
99 * sb_start_write -> i_mutex -> transaction start -> page lock ->
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
106 * transaction start -> i_data_sem (rw)
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
110 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
111 * transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 static bool userns_mounts
= false;
118 module_param(userns_mounts
, bool, 0644);
119 MODULE_PARM_DESC(userns_mounts
, "Allow mounts from unprivileged user namespaces");
121 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
122 static struct file_system_type ext2_fs_type
= {
123 .owner
= THIS_MODULE
,
126 .kill_sb
= kill_block_super
,
127 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
129 MODULE_ALIAS_FS("ext2");
130 MODULE_ALIAS("ext2");
131 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
133 #define IS_EXT2_SB(sb) (0)
137 static struct file_system_type ext3_fs_type
= {
138 .owner
= THIS_MODULE
,
141 .kill_sb
= kill_block_super
,
142 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
144 MODULE_ALIAS_FS("ext3");
145 MODULE_ALIAS("ext3");
146 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
148 static int ext4_verify_csum_type(struct super_block
*sb
,
149 struct ext4_super_block
*es
)
151 if (!ext4_has_feature_metadata_csum(sb
))
154 return es
->s_checksum_type
== EXT4_CRC32C_CHKSUM
;
157 static __le32
ext4_superblock_csum(struct super_block
*sb
,
158 struct ext4_super_block
*es
)
160 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
161 int offset
= offsetof(struct ext4_super_block
, s_checksum
);
164 csum
= ext4_chksum(sbi
, ~0, (char *)es
, offset
);
166 return cpu_to_le32(csum
);
169 static int ext4_superblock_csum_verify(struct super_block
*sb
,
170 struct ext4_super_block
*es
)
172 if (!ext4_has_metadata_csum(sb
))
175 return es
->s_checksum
== ext4_superblock_csum(sb
, es
);
178 void ext4_superblock_csum_set(struct super_block
*sb
)
180 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
182 if (!ext4_has_metadata_csum(sb
))
185 es
->s_checksum
= ext4_superblock_csum(sb
, es
);
188 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
192 ret
= kmalloc(size
, flags
| __GFP_NOWARN
);
194 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
198 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
202 ret
= kzalloc(size
, flags
| __GFP_NOWARN
);
204 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
208 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
209 struct ext4_group_desc
*bg
)
211 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
212 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
213 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
216 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
217 struct ext4_group_desc
*bg
)
219 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
220 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
221 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
224 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
225 struct ext4_group_desc
*bg
)
227 return le32_to_cpu(bg
->bg_inode_table_lo
) |
228 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
229 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
232 __u32
ext4_free_group_clusters(struct super_block
*sb
,
233 struct ext4_group_desc
*bg
)
235 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
236 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
237 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
240 __u32
ext4_free_inodes_count(struct super_block
*sb
,
241 struct ext4_group_desc
*bg
)
243 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
244 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
245 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
248 __u32
ext4_used_dirs_count(struct super_block
*sb
,
249 struct ext4_group_desc
*bg
)
251 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
252 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
253 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
256 __u32
ext4_itable_unused_count(struct super_block
*sb
,
257 struct ext4_group_desc
*bg
)
259 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
260 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
261 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
264 void ext4_block_bitmap_set(struct super_block
*sb
,
265 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
267 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
268 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
269 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
272 void ext4_inode_bitmap_set(struct super_block
*sb
,
273 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
275 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
276 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
277 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
280 void ext4_inode_table_set(struct super_block
*sb
,
281 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
283 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
284 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
285 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
288 void ext4_free_group_clusters_set(struct super_block
*sb
,
289 struct ext4_group_desc
*bg
, __u32 count
)
291 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
292 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
293 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
296 void ext4_free_inodes_set(struct super_block
*sb
,
297 struct ext4_group_desc
*bg
, __u32 count
)
299 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
300 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
301 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
304 void ext4_used_dirs_set(struct super_block
*sb
,
305 struct ext4_group_desc
*bg
, __u32 count
)
307 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
308 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
309 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
312 void ext4_itable_unused_set(struct super_block
*sb
,
313 struct ext4_group_desc
*bg
, __u32 count
)
315 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
316 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
317 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
321 static void __save_error_info(struct super_block
*sb
, const char *func
,
324 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
326 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
327 if (bdev_read_only(sb
->s_bdev
))
329 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
330 es
->s_last_error_time
= cpu_to_le32(get_seconds());
331 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
332 es
->s_last_error_line
= cpu_to_le32(line
);
333 if (!es
->s_first_error_time
) {
334 es
->s_first_error_time
= es
->s_last_error_time
;
335 strncpy(es
->s_first_error_func
, func
,
336 sizeof(es
->s_first_error_func
));
337 es
->s_first_error_line
= cpu_to_le32(line
);
338 es
->s_first_error_ino
= es
->s_last_error_ino
;
339 es
->s_first_error_block
= es
->s_last_error_block
;
342 * Start the daily error reporting function if it hasn't been
345 if (!es
->s_error_count
)
346 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
347 le32_add_cpu(&es
->s_error_count
, 1);
350 static void save_error_info(struct super_block
*sb
, const char *func
,
353 __save_error_info(sb
, func
, line
);
354 ext4_commit_super(sb
, 1);
358 * The del_gendisk() function uninitializes the disk-specific data
359 * structures, including the bdi structure, without telling anyone
360 * else. Once this happens, any attempt to call mark_buffer_dirty()
361 * (for example, by ext4_commit_super), will cause a kernel OOPS.
362 * This is a kludge to prevent these oops until we can put in a proper
363 * hook in del_gendisk() to inform the VFS and file system layers.
365 static int block_device_ejected(struct super_block
*sb
)
367 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
368 struct backing_dev_info
*bdi
= inode_to_bdi(bd_inode
);
370 return bdi
->dev
== NULL
;
373 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
375 struct super_block
*sb
= journal
->j_private
;
376 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
377 int error
= is_journal_aborted(journal
);
378 struct ext4_journal_cb_entry
*jce
;
380 BUG_ON(txn
->t_state
== T_FINISHED
);
382 ext4_process_freed_data(sb
, txn
->t_tid
);
384 spin_lock(&sbi
->s_md_lock
);
385 while (!list_empty(&txn
->t_private_list
)) {
386 jce
= list_entry(txn
->t_private_list
.next
,
387 struct ext4_journal_cb_entry
, jce_list
);
388 list_del_init(&jce
->jce_list
);
389 spin_unlock(&sbi
->s_md_lock
);
390 jce
->jce_func(sb
, jce
, error
);
391 spin_lock(&sbi
->s_md_lock
);
393 spin_unlock(&sbi
->s_md_lock
);
396 /* Deal with the reporting of failure conditions on a filesystem such as
397 * inconsistencies detected or read IO failures.
399 * On ext2, we can store the error state of the filesystem in the
400 * superblock. That is not possible on ext4, because we may have other
401 * write ordering constraints on the superblock which prevent us from
402 * writing it out straight away; and given that the journal is about to
403 * be aborted, we can't rely on the current, or future, transactions to
404 * write out the superblock safely.
406 * We'll just use the jbd2_journal_abort() error code to record an error in
407 * the journal instead. On recovery, the journal will complain about
408 * that error until we've noted it down and cleared it.
411 static void ext4_handle_error(struct super_block
*sb
)
413 if (sb
->s_flags
& MS_RDONLY
)
416 if (!test_opt(sb
, ERRORS_CONT
)) {
417 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
419 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
421 jbd2_journal_abort(journal
, -EIO
);
423 if (test_opt(sb
, ERRORS_RO
)) {
424 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
426 * Make sure updated value of ->s_mount_flags will be visible
427 * before ->s_flags update
430 sb
->s_flags
|= MS_RDONLY
;
432 if (test_opt(sb
, ERRORS_PANIC
)) {
433 if (EXT4_SB(sb
)->s_journal
&&
434 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
436 panic("EXT4-fs (device %s): panic forced after error\n",
441 #define ext4_error_ratelimit(sb) \
442 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
445 void __ext4_error(struct super_block
*sb
, const char *function
,
446 unsigned int line
, const char *fmt
, ...)
448 struct va_format vaf
;
451 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
454 if (ext4_error_ratelimit(sb
)) {
459 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
460 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
463 save_error_info(sb
, function
, line
);
464 ext4_handle_error(sb
);
467 void __ext4_error_inode(struct inode
*inode
, const char *function
,
468 unsigned int line
, ext4_fsblk_t block
,
469 const char *fmt
, ...)
472 struct va_format vaf
;
473 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
475 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
478 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
479 es
->s_last_error_block
= cpu_to_le64(block
);
480 if (ext4_error_ratelimit(inode
->i_sb
)) {
485 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: block %llu: comm %s: %pV\n",
487 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
488 block
, current
->comm
, &vaf
);
490 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: "
491 "inode #%lu: comm %s: %pV\n",
492 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
493 current
->comm
, &vaf
);
496 save_error_info(inode
->i_sb
, function
, line
);
497 ext4_handle_error(inode
->i_sb
);
500 void __ext4_error_file(struct file
*file
, const char *function
,
501 unsigned int line
, ext4_fsblk_t block
,
502 const char *fmt
, ...)
505 struct va_format vaf
;
506 struct ext4_super_block
*es
;
507 struct inode
*inode
= file_inode(file
);
508 char pathname
[80], *path
;
510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
513 es
= EXT4_SB(inode
->i_sb
)->s_es
;
514 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
515 if (ext4_error_ratelimit(inode
->i_sb
)) {
516 path
= file_path(file
, pathname
, sizeof(pathname
));
524 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
525 "block %llu: comm %s: path %s: %pV\n",
526 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
527 block
, current
->comm
, path
, &vaf
);
530 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
531 "comm %s: path %s: %pV\n",
532 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
533 current
->comm
, path
, &vaf
);
536 save_error_info(inode
->i_sb
, function
, line
);
537 ext4_handle_error(inode
->i_sb
);
540 const char *ext4_decode_error(struct super_block
*sb
, int errno
,
547 errstr
= "Corrupt filesystem";
550 errstr
= "Filesystem failed CRC";
553 errstr
= "IO failure";
556 errstr
= "Out of memory";
559 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
560 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
561 errstr
= "Journal has aborted";
563 errstr
= "Readonly filesystem";
566 /* If the caller passed in an extra buffer for unknown
567 * errors, textualise them now. Else we just return
570 /* Check for truncated error codes... */
571 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
580 /* __ext4_std_error decodes expected errors from journaling functions
581 * automatically and invokes the appropriate error response. */
583 void __ext4_std_error(struct super_block
*sb
, const char *function
,
584 unsigned int line
, int errno
)
589 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
592 /* Special case: if the error is EROFS, and we're not already
593 * inside a transaction, then there's really no point in logging
595 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
596 (sb
->s_flags
& MS_RDONLY
))
599 if (ext4_error_ratelimit(sb
)) {
600 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
601 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
602 sb
->s_id
, function
, line
, errstr
);
605 save_error_info(sb
, function
, line
);
606 ext4_handle_error(sb
);
610 * ext4_abort is a much stronger failure handler than ext4_error. The
611 * abort function may be used to deal with unrecoverable failures such
612 * as journal IO errors or ENOMEM at a critical moment in log management.
614 * We unconditionally force the filesystem into an ABORT|READONLY state,
615 * unless the error response on the fs has been set to panic in which
616 * case we take the easy way out and panic immediately.
619 void __ext4_abort(struct super_block
*sb
, const char *function
,
620 unsigned int line
, const char *fmt
, ...)
622 struct va_format vaf
;
625 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
628 save_error_info(sb
, function
, line
);
632 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: %pV\n",
633 sb
->s_id
, function
, line
, &vaf
);
636 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
637 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
638 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
640 * Make sure updated value of ->s_mount_flags will be visible
641 * before ->s_flags update
644 sb
->s_flags
|= MS_RDONLY
;
645 if (EXT4_SB(sb
)->s_journal
)
646 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
647 save_error_info(sb
, function
, line
);
649 if (test_opt(sb
, ERRORS_PANIC
)) {
650 if (EXT4_SB(sb
)->s_journal
&&
651 !(EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_REC_ERR
))
653 panic("EXT4-fs panic from previous error\n");
657 void __ext4_msg(struct super_block
*sb
,
658 const char *prefix
, const char *fmt
, ...)
660 struct va_format vaf
;
663 if (!___ratelimit(&(EXT4_SB(sb
)->s_msg_ratelimit_state
), "EXT4-fs"))
669 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
673 #define ext4_warning_ratelimit(sb) \
674 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
677 void __ext4_warning(struct super_block
*sb
, const char *function
,
678 unsigned int line
, const char *fmt
, ...)
680 struct va_format vaf
;
683 if (!ext4_warning_ratelimit(sb
))
689 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
690 sb
->s_id
, function
, line
, &vaf
);
694 void __ext4_warning_inode(const struct inode
*inode
, const char *function
,
695 unsigned int line
, const char *fmt
, ...)
697 struct va_format vaf
;
700 if (!ext4_warning_ratelimit(inode
->i_sb
))
706 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: "
707 "inode #%lu: comm %s: %pV\n", inode
->i_sb
->s_id
,
708 function
, line
, inode
->i_ino
, current
->comm
, &vaf
);
712 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
713 struct super_block
*sb
, ext4_group_t grp
,
714 unsigned long ino
, ext4_fsblk_t block
,
715 const char *fmt
, ...)
719 struct va_format vaf
;
721 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
723 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
726 es
->s_last_error_ino
= cpu_to_le32(ino
);
727 es
->s_last_error_block
= cpu_to_le64(block
);
728 __save_error_info(sb
, function
, line
);
730 if (ext4_error_ratelimit(sb
)) {
734 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
735 sb
->s_id
, function
, line
, grp
);
737 printk(KERN_CONT
"inode %lu: ", ino
);
739 printk(KERN_CONT
"block %llu:",
740 (unsigned long long) block
);
741 printk(KERN_CONT
"%pV\n", &vaf
);
745 if (test_opt(sb
, ERRORS_CONT
)) {
746 ext4_commit_super(sb
, 0);
750 ext4_unlock_group(sb
, grp
);
751 ext4_handle_error(sb
);
753 * We only get here in the ERRORS_RO case; relocking the group
754 * may be dangerous, but nothing bad will happen since the
755 * filesystem will have already been marked read/only and the
756 * journal has been aborted. We return 1 as a hint to callers
757 * who might what to use the return value from
758 * ext4_grp_locked_error() to distinguish between the
759 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
760 * aggressively from the ext4 function in question, with a
761 * more appropriate error code.
763 ext4_lock_group(sb
, grp
);
767 void ext4_update_dynamic_rev(struct super_block
*sb
)
769 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
771 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
775 "updating to rev %d because of new feature flag, "
776 "running e2fsck is recommended",
779 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
780 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
781 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
782 /* leave es->s_feature_*compat flags alone */
783 /* es->s_uuid will be set by e2fsck if empty */
786 * The rest of the superblock fields should be zero, and if not it
787 * means they are likely already in use, so leave them alone. We
788 * can leave it up to e2fsck to clean up any inconsistencies there.
793 * Open the external journal device
795 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
797 struct block_device
*bdev
;
798 char b
[BDEVNAME_SIZE
];
800 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
806 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
807 __bdevname(dev
, b
), PTR_ERR(bdev
));
812 * Release the journal device
814 static void ext4_blkdev_put(struct block_device
*bdev
)
816 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
819 static void ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
821 struct block_device
*bdev
;
822 bdev
= sbi
->journal_bdev
;
824 ext4_blkdev_put(bdev
);
825 sbi
->journal_bdev
= NULL
;
829 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
831 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
834 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
838 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
839 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
841 printk(KERN_ERR
"sb_info orphan list:\n");
842 list_for_each(l
, &sbi
->s_orphan
) {
843 struct inode
*inode
= orphan_list_entry(l
);
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
847 inode
->i_mode
, inode
->i_nlink
,
853 static int ext4_quota_off(struct super_block
*sb
, int type
);
855 static inline void ext4_quota_off_umount(struct super_block
*sb
)
859 /* Use our quota_off function to clear inode flags etc. */
860 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++)
861 ext4_quota_off(sb
, type
);
864 static inline void ext4_quota_off_umount(struct super_block
*sb
)
869 static void ext4_put_super(struct super_block
*sb
)
871 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
872 struct ext4_super_block
*es
= sbi
->s_es
;
876 ext4_unregister_li_request(sb
);
877 ext4_quota_off_umount(sb
);
879 flush_workqueue(sbi
->rsv_conversion_wq
);
880 destroy_workqueue(sbi
->rsv_conversion_wq
);
882 if (sbi
->s_journal
) {
883 aborted
= is_journal_aborted(sbi
->s_journal
);
884 err
= jbd2_journal_destroy(sbi
->s_journal
);
885 sbi
->s_journal
= NULL
;
886 if ((err
< 0) && !aborted
)
887 ext4_abort(sb
, "Couldn't clean up the journal");
890 ext4_unregister_sysfs(sb
);
891 ext4_es_unregister_shrinker(sbi
);
892 del_timer_sync(&sbi
->s_err_report
);
893 ext4_release_system_zone(sb
);
895 ext4_ext_release(sb
);
897 if (!(sb
->s_flags
& MS_RDONLY
) && !aborted
) {
898 ext4_clear_feature_journal_needs_recovery(sb
);
899 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
901 if (!(sb
->s_flags
& MS_RDONLY
))
902 ext4_commit_super(sb
, 1);
904 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
905 brelse(sbi
->s_group_desc
[i
]);
906 kvfree(sbi
->s_group_desc
);
907 kvfree(sbi
->s_flex_groups
);
908 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
909 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
910 percpu_counter_destroy(&sbi
->s_dirs_counter
);
911 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
912 percpu_free_rwsem(&sbi
->s_journal_flag_rwsem
);
914 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
915 kfree(sbi
->s_qf_names
[i
]);
918 /* Debugging code just in case the in-memory inode orphan list
919 * isn't empty. The on-disk one can be non-empty if we've
920 * detected an error and taken the fs readonly, but the
921 * in-memory list had better be clean by this point. */
922 if (!list_empty(&sbi
->s_orphan
))
923 dump_orphan_list(sb
, sbi
);
924 J_ASSERT(list_empty(&sbi
->s_orphan
));
926 sync_blockdev(sb
->s_bdev
);
927 invalidate_bdev(sb
->s_bdev
);
928 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
930 * Invalidate the journal device's buffers. We don't want them
931 * floating about in memory - the physical journal device may
932 * hotswapped, and it breaks the `ro-after' testing code.
934 sync_blockdev(sbi
->journal_bdev
);
935 invalidate_bdev(sbi
->journal_bdev
);
936 ext4_blkdev_remove(sbi
);
938 if (sbi
->s_ea_inode_cache
) {
939 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
940 sbi
->s_ea_inode_cache
= NULL
;
942 if (sbi
->s_ea_block_cache
) {
943 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
944 sbi
->s_ea_block_cache
= NULL
;
947 kthread_stop(sbi
->s_mmp_tsk
);
949 sb
->s_fs_info
= NULL
;
951 * Now that we are completely done shutting down the
952 * superblock, we need to actually destroy the kobject.
954 kobject_put(&sbi
->s_kobj
);
955 wait_for_completion(&sbi
->s_kobj_unregister
);
956 if (sbi
->s_chksum_driver
)
957 crypto_free_shash(sbi
->s_chksum_driver
);
958 kfree(sbi
->s_blockgroup_lock
);
962 static struct kmem_cache
*ext4_inode_cachep
;
965 * Called inside transaction, so use GFP_NOFS
967 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
969 struct ext4_inode_info
*ei
;
971 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
975 ei
->vfs_inode
.i_version
= 1;
976 spin_lock_init(&ei
->i_raw_lock
);
977 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
978 spin_lock_init(&ei
->i_prealloc_lock
);
979 ext4_es_init_tree(&ei
->i_es_tree
);
980 rwlock_init(&ei
->i_es_lock
);
981 INIT_LIST_HEAD(&ei
->i_es_list
);
984 ei
->i_es_shrink_lblk
= 0;
985 ei
->i_reserved_data_blocks
= 0;
986 ei
->i_da_metadata_calc_len
= 0;
987 ei
->i_da_metadata_calc_last_lblock
= 0;
988 spin_lock_init(&(ei
->i_block_reservation_lock
));
990 ei
->i_reserved_quota
= 0;
991 memset(&ei
->i_dquot
, 0, sizeof(ei
->i_dquot
));
994 INIT_LIST_HEAD(&ei
->i_rsv_conversion_list
);
995 spin_lock_init(&ei
->i_completed_io_lock
);
997 ei
->i_datasync_tid
= 0;
998 atomic_set(&ei
->i_unwritten
, 0);
999 INIT_WORK(&ei
->i_rsv_conversion_work
, ext4_end_io_rsv_work
);
1000 return &ei
->vfs_inode
;
1003 static int ext4_drop_inode(struct inode
*inode
)
1005 int drop
= generic_drop_inode(inode
);
1007 trace_ext4_drop_inode(inode
, drop
);
1011 static void ext4_i_callback(struct rcu_head
*head
)
1013 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1014 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
1017 static void ext4_destroy_inode(struct inode
*inode
)
1019 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
1020 ext4_msg(inode
->i_sb
, KERN_ERR
,
1021 "Inode %lu (%p): orphan list check failed!",
1022 inode
->i_ino
, EXT4_I(inode
));
1023 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
1024 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
1028 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
1031 static void init_once(void *foo
)
1033 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
1035 INIT_LIST_HEAD(&ei
->i_orphan
);
1036 init_rwsem(&ei
->xattr_sem
);
1037 init_rwsem(&ei
->i_data_sem
);
1038 init_rwsem(&ei
->i_mmap_sem
);
1039 inode_init_once(&ei
->vfs_inode
);
1042 static int __init
init_inodecache(void)
1044 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
1045 sizeof(struct ext4_inode_info
),
1046 0, (SLAB_RECLAIM_ACCOUNT
|
1047 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1049 if (ext4_inode_cachep
== NULL
)
1054 static void destroy_inodecache(void)
1057 * Make sure all delayed rcu free inodes are flushed before we
1061 kmem_cache_destroy(ext4_inode_cachep
);
1064 void ext4_clear_inode(struct inode
*inode
)
1066 invalidate_inode_buffers(inode
);
1069 ext4_discard_preallocations(inode
);
1070 ext4_es_remove_extent(inode
, 0, EXT_MAX_BLOCKS
);
1071 if (EXT4_I(inode
)->jinode
) {
1072 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1073 EXT4_I(inode
)->jinode
);
1074 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1075 EXT4_I(inode
)->jinode
= NULL
;
1077 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1078 fscrypt_put_encryption_info(inode
, NULL
);
1082 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1083 u64 ino
, u32 generation
)
1085 struct inode
*inode
;
1087 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1088 return ERR_PTR(-ESTALE
);
1089 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1090 return ERR_PTR(-ESTALE
);
1092 /* iget isn't really right if the inode is currently unallocated!!
1094 * ext4_read_inode will return a bad_inode if the inode had been
1095 * deleted, so we should be safe.
1097 * Currently we don't know the generation for parent directory, so
1098 * a generation of 0 means "accept any"
1100 inode
= ext4_iget_normal(sb
, ino
);
1102 return ERR_CAST(inode
);
1103 if (generation
&& inode
->i_generation
!= generation
) {
1105 return ERR_PTR(-ESTALE
);
1111 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1112 int fh_len
, int fh_type
)
1114 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1115 ext4_nfs_get_inode
);
1118 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1119 int fh_len
, int fh_type
)
1121 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1122 ext4_nfs_get_inode
);
1126 * Try to release metadata pages (indirect blocks, directories) which are
1127 * mapped via the block device. Since these pages could have journal heads
1128 * which would prevent try_to_free_buffers() from freeing them, we must use
1129 * jbd2 layer's try_to_free_buffers() function to release them.
1131 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1134 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1136 WARN_ON(PageChecked(page
));
1137 if (!page_has_buffers(page
))
1140 return jbd2_journal_try_to_free_buffers(journal
, page
,
1141 wait
& ~__GFP_DIRECT_RECLAIM
);
1142 return try_to_free_buffers(page
);
1145 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1146 static int ext4_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1148 return ext4_xattr_get(inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1149 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
, ctx
, len
);
1152 static int ext4_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1155 handle_t
*handle
= fs_data
;
1156 int res
, res2
, credits
, retries
= 0;
1159 * Encrypting the root directory is not allowed because e2fsck expects
1160 * lost+found to exist and be unencrypted, and encrypting the root
1161 * directory would imply encrypting the lost+found directory as well as
1162 * the filename "lost+found" itself.
1164 if (inode
->i_ino
== EXT4_ROOT_INO
)
1167 res
= ext4_convert_inline_data(inode
);
1172 * If a journal handle was specified, then the encryption context is
1173 * being set on a new inode via inheritance and is part of a larger
1174 * transaction to create the inode. Otherwise the encryption context is
1175 * being set on an existing inode in its own transaction. Only in the
1176 * latter case should the "retry on ENOSPC" logic be used.
1180 res
= ext4_xattr_set_handle(handle
, inode
,
1181 EXT4_XATTR_INDEX_ENCRYPTION
,
1182 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1185 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1186 ext4_clear_inode_state(inode
,
1187 EXT4_STATE_MAY_INLINE_DATA
);
1189 * Update inode->i_flags - e.g. S_DAX may get disabled
1191 ext4_set_inode_flags(inode
);
1196 res
= dquot_initialize(inode
);
1200 res
= ext4_xattr_set_credits(inode
, len
, false /* is_create */,
1205 handle
= ext4_journal_start(inode
, EXT4_HT_MISC
, credits
);
1207 return PTR_ERR(handle
);
1209 res
= ext4_xattr_set_handle(handle
, inode
, EXT4_XATTR_INDEX_ENCRYPTION
,
1210 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT
,
1213 ext4_set_inode_flag(inode
, EXT4_INODE_ENCRYPT
);
1214 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1215 ext4_set_inode_flags(inode
);
1216 res
= ext4_mark_inode_dirty(handle
, inode
);
1218 EXT4_ERROR_INODE(inode
, "Failed to mark inode dirty");
1220 res2
= ext4_journal_stop(handle
);
1222 if (res
== -ENOSPC
&& ext4_should_retry_alloc(inode
->i_sb
, &retries
))
1229 static bool ext4_dummy_context(struct inode
*inode
)
1231 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode
->i_sb
));
1234 static unsigned ext4_max_namelen(struct inode
*inode
)
1236 return S_ISLNK(inode
->i_mode
) ? inode
->i_sb
->s_blocksize
:
1240 static const struct fscrypt_operations ext4_cryptops
= {
1241 .key_prefix
= "ext4:",
1242 .get_context
= ext4_get_context
,
1243 .set_context
= ext4_set_context
,
1244 .dummy_context
= ext4_dummy_context
,
1245 .is_encrypted
= ext4_encrypted_inode
,
1246 .empty_dir
= ext4_empty_dir
,
1247 .max_namelen
= ext4_max_namelen
,
1250 static const struct fscrypt_operations ext4_cryptops
= {
1251 .is_encrypted
= ext4_encrypted_inode
,
1256 static const char * const quotatypes
[] = INITQFNAMES
;
1257 #define QTYPE2NAME(t) (quotatypes[t])
1259 static int ext4_write_dquot(struct dquot
*dquot
);
1260 static int ext4_acquire_dquot(struct dquot
*dquot
);
1261 static int ext4_release_dquot(struct dquot
*dquot
);
1262 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1263 static int ext4_write_info(struct super_block
*sb
, int type
);
1264 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1265 const struct path
*path
);
1266 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1267 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1268 size_t len
, loff_t off
);
1269 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1270 const char *data
, size_t len
, loff_t off
);
1271 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
1272 unsigned int flags
);
1273 static int ext4_enable_quotas(struct super_block
*sb
);
1274 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
);
1276 static struct dquot
**ext4_get_dquots(struct inode
*inode
)
1278 return EXT4_I(inode
)->i_dquot
;
1281 static const struct dquot_operations ext4_quota_operations
= {
1282 .get_reserved_space
= ext4_get_reserved_space
,
1283 .write_dquot
= ext4_write_dquot
,
1284 .acquire_dquot
= ext4_acquire_dquot
,
1285 .release_dquot
= ext4_release_dquot
,
1286 .mark_dirty
= ext4_mark_dquot_dirty
,
1287 .write_info
= ext4_write_info
,
1288 .alloc_dquot
= dquot_alloc
,
1289 .destroy_dquot
= dquot_destroy
,
1290 .get_projid
= ext4_get_projid
,
1291 .get_inode_usage
= ext4_get_inode_usage
,
1292 .get_next_id
= ext4_get_next_id
,
1295 static const struct quotactl_ops ext4_qctl_operations
= {
1296 .quota_on
= ext4_quota_on
,
1297 .quota_off
= ext4_quota_off
,
1298 .quota_sync
= dquot_quota_sync
,
1299 .get_state
= dquot_get_state
,
1300 .set_info
= dquot_set_dqinfo
,
1301 .get_dqblk
= dquot_get_dqblk
,
1302 .set_dqblk
= dquot_set_dqblk
,
1303 .get_nextdqblk
= dquot_get_next_dqblk
,
1307 static const struct super_operations ext4_sops
= {
1308 .alloc_inode
= ext4_alloc_inode
,
1309 .destroy_inode
= ext4_destroy_inode
,
1310 .write_inode
= ext4_write_inode
,
1311 .dirty_inode
= ext4_dirty_inode
,
1312 .drop_inode
= ext4_drop_inode
,
1313 .evict_inode
= ext4_evict_inode
,
1314 .put_super
= ext4_put_super
,
1315 .sync_fs
= ext4_sync_fs
,
1316 .freeze_fs
= ext4_freeze
,
1317 .unfreeze_fs
= ext4_unfreeze
,
1318 .statfs
= ext4_statfs
,
1319 .remount_fs
= ext4_remount
,
1320 .show_options
= ext4_show_options
,
1322 .quota_read
= ext4_quota_read
,
1323 .quota_write
= ext4_quota_write
,
1324 .get_dquots
= ext4_get_dquots
,
1326 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1329 static const struct export_operations ext4_export_ops
= {
1330 .fh_to_dentry
= ext4_fh_to_dentry
,
1331 .fh_to_parent
= ext4_fh_to_parent
,
1332 .get_parent
= ext4_get_parent
,
1336 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1337 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1338 Opt_nouid32
, Opt_debug
, Opt_removed
,
1339 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1340 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1341 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
, Opt_journal_dev
,
1342 Opt_journal_path
, Opt_journal_checksum
, Opt_journal_async_commit
,
1343 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1344 Opt_data_err_abort
, Opt_data_err_ignore
, Opt_test_dummy_encryption
,
1345 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1346 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1347 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1348 Opt_usrquota
, Opt_grpquota
, Opt_prjquota
, Opt_i_version
, Opt_dax
,
1349 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1350 Opt_lazytime
, Opt_nolazytime
, Opt_debug_want_extra_isize
,
1351 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1352 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1353 Opt_dioread_nolock
, Opt_dioread_lock
,
1354 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1355 Opt_max_dir_size_kb
, Opt_nojournal_checksum
, Opt_nombcache
,
1358 static const match_table_t tokens
= {
1359 {Opt_bsd_df
, "bsddf"},
1360 {Opt_minix_df
, "minixdf"},
1361 {Opt_grpid
, "grpid"},
1362 {Opt_grpid
, "bsdgroups"},
1363 {Opt_nogrpid
, "nogrpid"},
1364 {Opt_nogrpid
, "sysvgroups"},
1365 {Opt_resgid
, "resgid=%u"},
1366 {Opt_resuid
, "resuid=%u"},
1368 {Opt_err_cont
, "errors=continue"},
1369 {Opt_err_panic
, "errors=panic"},
1370 {Opt_err_ro
, "errors=remount-ro"},
1371 {Opt_nouid32
, "nouid32"},
1372 {Opt_debug
, "debug"},
1373 {Opt_removed
, "oldalloc"},
1374 {Opt_removed
, "orlov"},
1375 {Opt_user_xattr
, "user_xattr"},
1376 {Opt_nouser_xattr
, "nouser_xattr"},
1378 {Opt_noacl
, "noacl"},
1379 {Opt_noload
, "norecovery"},
1380 {Opt_noload
, "noload"},
1381 {Opt_removed
, "nobh"},
1382 {Opt_removed
, "bh"},
1383 {Opt_commit
, "commit=%u"},
1384 {Opt_min_batch_time
, "min_batch_time=%u"},
1385 {Opt_max_batch_time
, "max_batch_time=%u"},
1386 {Opt_journal_dev
, "journal_dev=%u"},
1387 {Opt_journal_path
, "journal_path=%s"},
1388 {Opt_journal_checksum
, "journal_checksum"},
1389 {Opt_nojournal_checksum
, "nojournal_checksum"},
1390 {Opt_journal_async_commit
, "journal_async_commit"},
1391 {Opt_abort
, "abort"},
1392 {Opt_data_journal
, "data=journal"},
1393 {Opt_data_ordered
, "data=ordered"},
1394 {Opt_data_writeback
, "data=writeback"},
1395 {Opt_data_err_abort
, "data_err=abort"},
1396 {Opt_data_err_ignore
, "data_err=ignore"},
1397 {Opt_offusrjquota
, "usrjquota="},
1398 {Opt_usrjquota
, "usrjquota=%s"},
1399 {Opt_offgrpjquota
, "grpjquota="},
1400 {Opt_grpjquota
, "grpjquota=%s"},
1401 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1402 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1403 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1404 {Opt_grpquota
, "grpquota"},
1405 {Opt_noquota
, "noquota"},
1406 {Opt_quota
, "quota"},
1407 {Opt_usrquota
, "usrquota"},
1408 {Opt_prjquota
, "prjquota"},
1409 {Opt_barrier
, "barrier=%u"},
1410 {Opt_barrier
, "barrier"},
1411 {Opt_nobarrier
, "nobarrier"},
1412 {Opt_i_version
, "i_version"},
1414 {Opt_stripe
, "stripe=%u"},
1415 {Opt_delalloc
, "delalloc"},
1416 {Opt_lazytime
, "lazytime"},
1417 {Opt_nolazytime
, "nolazytime"},
1418 {Opt_debug_want_extra_isize
, "debug_want_extra_isize=%u"},
1419 {Opt_nodelalloc
, "nodelalloc"},
1420 {Opt_removed
, "mblk_io_submit"},
1421 {Opt_removed
, "nomblk_io_submit"},
1422 {Opt_block_validity
, "block_validity"},
1423 {Opt_noblock_validity
, "noblock_validity"},
1424 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1425 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1426 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1427 {Opt_auto_da_alloc
, "auto_da_alloc"},
1428 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1429 {Opt_dioread_nolock
, "dioread_nolock"},
1430 {Opt_dioread_lock
, "dioread_lock"},
1431 {Opt_discard
, "discard"},
1432 {Opt_nodiscard
, "nodiscard"},
1433 {Opt_init_itable
, "init_itable=%u"},
1434 {Opt_init_itable
, "init_itable"},
1435 {Opt_noinit_itable
, "noinit_itable"},
1436 {Opt_max_dir_size_kb
, "max_dir_size_kb=%u"},
1437 {Opt_test_dummy_encryption
, "test_dummy_encryption"},
1438 {Opt_nombcache
, "nombcache"},
1439 {Opt_nombcache
, "no_mbcache"}, /* for backward compatibility */
1440 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1441 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1442 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1443 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1444 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1448 static ext4_fsblk_t
get_sb_block(void **data
)
1450 ext4_fsblk_t sb_block
;
1451 char *options
= (char *) *data
;
1453 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1454 return 1; /* Default location */
1457 /* TODO: use simple_strtoll with >32bit ext4 */
1458 sb_block
= simple_strtoul(options
, &options
, 0);
1459 if (*options
&& *options
!= ',') {
1460 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1464 if (*options
== ',')
1466 *data
= (void *) options
;
1471 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1472 static const char deprecated_msg
[] =
1473 "Mount option \"%s\" will be removed by %s\n"
1474 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1477 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1479 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1483 if (sb_any_quota_loaded(sb
) &&
1484 !sbi
->s_qf_names
[qtype
]) {
1485 ext4_msg(sb
, KERN_ERR
,
1486 "Cannot change journaled "
1487 "quota options when quota turned on");
1490 if (ext4_has_feature_quota(sb
)) {
1491 ext4_msg(sb
, KERN_INFO
, "Journaled quota options "
1492 "ignored when QUOTA feature is enabled");
1495 qname
= match_strdup(args
);
1497 ext4_msg(sb
, KERN_ERR
,
1498 "Not enough memory for storing quotafile name");
1501 if (sbi
->s_qf_names
[qtype
]) {
1502 if (strcmp(sbi
->s_qf_names
[qtype
], qname
) == 0)
1505 ext4_msg(sb
, KERN_ERR
,
1506 "%s quota file already specified",
1510 if (strchr(qname
, '/')) {
1511 ext4_msg(sb
, KERN_ERR
,
1512 "quotafile must be on filesystem root");
1515 sbi
->s_qf_names
[qtype
] = qname
;
1523 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1526 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1528 if (sb_any_quota_loaded(sb
) &&
1529 sbi
->s_qf_names
[qtype
]) {
1530 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1531 " when quota turned on");
1534 kfree(sbi
->s_qf_names
[qtype
]);
1535 sbi
->s_qf_names
[qtype
] = NULL
;
1540 #define MOPT_SET 0x0001
1541 #define MOPT_CLEAR 0x0002
1542 #define MOPT_NOSUPPORT 0x0004
1543 #define MOPT_EXPLICIT 0x0008
1544 #define MOPT_CLEAR_ERR 0x0010
1545 #define MOPT_GTE0 0x0020
1548 #define MOPT_QFMT 0x0040
1550 #define MOPT_Q MOPT_NOSUPPORT
1551 #define MOPT_QFMT MOPT_NOSUPPORT
1553 #define MOPT_DATAJ 0x0080
1554 #define MOPT_NO_EXT2 0x0100
1555 #define MOPT_NO_EXT3 0x0200
1556 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1557 #define MOPT_STRING 0x0400
1559 static const struct mount_opts
{
1563 } ext4_mount_opts
[] = {
1564 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1565 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1566 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1567 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1568 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1569 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1570 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1571 MOPT_EXT4_ONLY
| MOPT_SET
},
1572 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
,
1573 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1574 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1575 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1576 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
,
1577 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1578 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
,
1579 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1580 {Opt_nojournal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1581 MOPT_EXT4_ONLY
| MOPT_CLEAR
},
1582 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
,
1583 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1584 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1585 EXT4_MOUNT_JOURNAL_CHECKSUM
),
1586 MOPT_EXT4_ONLY
| MOPT_SET
| MOPT_EXPLICIT
},
1587 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_NO_EXT2
| MOPT_SET
},
1588 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1589 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1590 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1591 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
,
1593 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
,
1595 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1596 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1597 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1598 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1599 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1600 {Opt_commit
, 0, MOPT_GTE0
},
1601 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1602 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1603 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1604 {Opt_init_itable
, 0, MOPT_GTE0
},
1605 {Opt_dax
, EXT4_MOUNT_DAX
, MOPT_SET
},
1606 {Opt_stripe
, 0, MOPT_GTE0
},
1607 {Opt_resuid
, 0, MOPT_GTE0
},
1608 {Opt_resgid
, 0, MOPT_GTE0
},
1609 {Opt_journal_dev
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1610 {Opt_journal_path
, 0, MOPT_NO_EXT2
| MOPT_STRING
},
1611 {Opt_journal_ioprio
, 0, MOPT_NO_EXT2
| MOPT_GTE0
},
1612 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1613 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_NO_EXT2
| MOPT_DATAJ
},
1614 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
,
1615 MOPT_NO_EXT2
| MOPT_DATAJ
},
1616 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1617 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1618 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1619 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1620 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1622 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1623 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1625 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1626 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1627 {Opt_debug_want_extra_isize
, 0, MOPT_GTE0
},
1628 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1629 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1631 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1633 {Opt_prjquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_PRJQUOTA
,
1635 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1636 EXT4_MOUNT_GRPQUOTA
| EXT4_MOUNT_PRJQUOTA
),
1637 MOPT_CLEAR
| MOPT_Q
},
1638 {Opt_usrjquota
, 0, MOPT_Q
},
1639 {Opt_grpjquota
, 0, MOPT_Q
},
1640 {Opt_offusrjquota
, 0, MOPT_Q
},
1641 {Opt_offgrpjquota
, 0, MOPT_Q
},
1642 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1643 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1644 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1645 {Opt_max_dir_size_kb
, 0, MOPT_GTE0
},
1646 {Opt_test_dummy_encryption
, 0, MOPT_GTE0
},
1647 {Opt_nombcache
, EXT4_MOUNT_NO_MBCACHE
, MOPT_SET
},
1651 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1652 substring_t
*args
, unsigned long *journal_devnum
,
1653 unsigned int *journal_ioprio
, int is_remount
)
1655 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1656 const struct mount_opts
*m
;
1662 if (token
== Opt_usrjquota
)
1663 return set_qf_name(sb
, USRQUOTA
, &args
[0]);
1664 else if (token
== Opt_grpjquota
)
1665 return set_qf_name(sb
, GRPQUOTA
, &args
[0]);
1666 else if (token
== Opt_offusrjquota
)
1667 return clear_qf_name(sb
, USRQUOTA
);
1668 else if (token
== Opt_offgrpjquota
)
1669 return clear_qf_name(sb
, GRPQUOTA
);
1673 case Opt_nouser_xattr
:
1674 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, opt
, "3.5");
1677 return 1; /* handled by get_sb_block() */
1679 ext4_msg(sb
, KERN_WARNING
, "Ignoring removed %s option", opt
);
1682 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1685 sb
->s_flags
|= MS_I_VERSION
;
1688 sb
->s_flags
|= MS_LAZYTIME
;
1690 case Opt_nolazytime
:
1691 sb
->s_flags
&= ~MS_LAZYTIME
;
1695 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++)
1696 if (token
== m
->token
)
1699 if (m
->token
== Opt_err
) {
1700 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1701 "or missing value", opt
);
1705 if ((m
->flags
& MOPT_NO_EXT2
) && IS_EXT2_SB(sb
)) {
1706 ext4_msg(sb
, KERN_ERR
,
1707 "Mount option \"%s\" incompatible with ext2", opt
);
1710 if ((m
->flags
& MOPT_NO_EXT3
) && IS_EXT3_SB(sb
)) {
1711 ext4_msg(sb
, KERN_ERR
,
1712 "Mount option \"%s\" incompatible with ext3", opt
);
1716 if (token
== Opt_err_panic
&& !capable(CAP_SYS_ADMIN
)) {
1717 ext4_msg(sb
, KERN_ERR
,
1718 "Mount option \"%s\" not allowed for unprivileged mounts",
1723 if (args
->from
&& !(m
->flags
& MOPT_STRING
) && match_int(args
, &arg
))
1725 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1727 if (m
->flags
& MOPT_EXPLICIT
) {
1728 if (m
->mount_opt
& EXT4_MOUNT_DELALLOC
) {
1729 set_opt2(sb
, EXPLICIT_DELALLOC
);
1730 } else if (m
->mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) {
1731 set_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
);
1735 if (m
->flags
& MOPT_CLEAR_ERR
)
1736 clear_opt(sb
, ERRORS_MASK
);
1737 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1738 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1739 "options when quota turned on");
1743 if (m
->flags
& MOPT_NOSUPPORT
) {
1744 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1745 } else if (token
== Opt_commit
) {
1747 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1748 sbi
->s_commit_interval
= HZ
* arg
;
1749 } else if (token
== Opt_debug_want_extra_isize
) {
1750 sbi
->s_want_extra_isize
= arg
;
1751 } else if (token
== Opt_max_batch_time
) {
1752 sbi
->s_max_batch_time
= arg
;
1753 } else if (token
== Opt_min_batch_time
) {
1754 sbi
->s_min_batch_time
= arg
;
1755 } else if (token
== Opt_inode_readahead_blks
) {
1756 if (arg
&& (arg
> (1 << 30) || !is_power_of_2(arg
))) {
1757 ext4_msg(sb
, KERN_ERR
,
1758 "EXT4-fs: inode_readahead_blks must be "
1759 "0 or a power of 2 smaller than 2^31");
1762 sbi
->s_inode_readahead_blks
= arg
;
1763 } else if (token
== Opt_init_itable
) {
1764 set_opt(sb
, INIT_INODE_TABLE
);
1766 arg
= EXT4_DEF_LI_WAIT_MULT
;
1767 sbi
->s_li_wait_mult
= arg
;
1768 } else if (token
== Opt_max_dir_size_kb
) {
1769 sbi
->s_max_dir_size_kb
= arg
;
1770 } else if (token
== Opt_stripe
) {
1771 sbi
->s_stripe
= arg
;
1772 } else if (token
== Opt_resuid
) {
1773 uid
= make_kuid(sb
->s_user_ns
, arg
);
1774 if (!uid_valid(uid
)) {
1775 ext4_msg(sb
, KERN_ERR
, "Invalid uid value %d", arg
);
1778 sbi
->s_resuid
= uid
;
1779 } else if (token
== Opt_resgid
) {
1780 gid
= make_kgid(sb
->s_user_ns
, arg
);
1781 if (!gid_valid(gid
)) {
1782 ext4_msg(sb
, KERN_ERR
, "Invalid gid value %d", arg
);
1785 sbi
->s_resgid
= gid
;
1786 } else if (token
== Opt_journal_dev
) {
1788 ext4_msg(sb
, KERN_ERR
,
1789 "Cannot specify journal on remount");
1792 *journal_devnum
= arg
;
1793 } else if (token
== Opt_journal_path
) {
1795 struct inode
*journal_inode
;
1800 ext4_msg(sb
, KERN_ERR
,
1801 "Cannot specify journal on remount");
1804 journal_path
= match_strdup(&args
[0]);
1805 if (!journal_path
) {
1806 ext4_msg(sb
, KERN_ERR
, "error: could not dup "
1807 "journal device string");
1811 error
= kern_path(journal_path
, LOOKUP_FOLLOW
, &path
);
1813 ext4_msg(sb
, KERN_ERR
, "error: could not find "
1814 "journal device path: error %d", error
);
1815 kfree(journal_path
);
1820 * Refuse access for unprivileged mounts if the user does
1821 * not have rw access to the journal device via the supplied
1824 if (!capable(CAP_SYS_ADMIN
) &&
1825 inode_permission(d_inode(path
.dentry
), MAY_READ
|MAY_WRITE
)) {
1826 ext4_msg(sb
, KERN_ERR
,
1827 "error: Insufficient access to journal path %s",
1832 journal_inode
= d_inode(path
.dentry
);
1833 if (!S_ISBLK(journal_inode
->i_mode
)) {
1834 ext4_msg(sb
, KERN_ERR
, "error: journal path %s "
1835 "is not a block device", journal_path
);
1837 kfree(journal_path
);
1841 *journal_devnum
= new_encode_dev(journal_inode
->i_rdev
);
1843 kfree(journal_path
);
1844 } else if (token
== Opt_journal_ioprio
) {
1846 ext4_msg(sb
, KERN_ERR
, "Invalid journal IO priority"
1851 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1852 } else if (token
== Opt_test_dummy_encryption
) {
1853 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1854 sbi
->s_mount_flags
|= EXT4_MF_TEST_DUMMY_ENCRYPTION
;
1855 ext4_msg(sb
, KERN_WARNING
,
1856 "Test dummy encryption mode enabled");
1858 ext4_msg(sb
, KERN_WARNING
,
1859 "Test dummy encryption mount option ignored");
1861 } else if (m
->flags
& MOPT_DATAJ
) {
1863 if (!sbi
->s_journal
)
1864 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1865 else if (test_opt(sb
, DATA_FLAGS
) != m
->mount_opt
) {
1866 ext4_msg(sb
, KERN_ERR
,
1867 "Cannot change data mode on remount");
1871 clear_opt(sb
, DATA_FLAGS
);
1872 sbi
->s_mount_opt
|= m
->mount_opt
;
1875 } else if (m
->flags
& MOPT_QFMT
) {
1876 if (sb_any_quota_loaded(sb
) &&
1877 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1878 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled "
1879 "quota options when quota turned on");
1882 if (ext4_has_feature_quota(sb
)) {
1883 ext4_msg(sb
, KERN_INFO
,
1884 "Quota format mount options ignored "
1885 "when QUOTA feature is enabled");
1888 sbi
->s_jquota_fmt
= m
->mount_opt
;
1890 } else if (token
== Opt_dax
) {
1891 #ifdef CONFIG_FS_DAX
1892 ext4_msg(sb
, KERN_WARNING
,
1893 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1894 sbi
->s_mount_opt
|= m
->mount_opt
;
1896 ext4_msg(sb
, KERN_INFO
, "dax option not supported");
1899 } else if (token
== Opt_data_err_abort
) {
1900 sbi
->s_mount_opt
|= m
->mount_opt
;
1901 } else if (token
== Opt_data_err_ignore
) {
1902 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1906 if (m
->flags
& MOPT_CLEAR
)
1908 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1909 ext4_msg(sb
, KERN_WARNING
,
1910 "buggy handling of option %s", opt
);
1915 sbi
->s_mount_opt
|= m
->mount_opt
;
1917 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1922 static int parse_options(char *options
, struct super_block
*sb
,
1923 unsigned long *journal_devnum
,
1924 unsigned int *journal_ioprio
,
1927 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1929 substring_t args
[MAX_OPT_ARGS
];
1935 while ((p
= strsep(&options
, ",")) != NULL
) {
1939 * Initialize args struct so we know whether arg was
1940 * found; some options take optional arguments.
1942 args
[0].to
= args
[0].from
= NULL
;
1943 token
= match_token(p
, tokens
, args
);
1944 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1945 journal_ioprio
, is_remount
) < 0)
1950 * We do the test below only for project quotas. 'usrquota' and
1951 * 'grpquota' mount options are allowed even without quota feature
1952 * to support legacy quotas in quota files.
1954 if (test_opt(sb
, PRJQUOTA
) && !ext4_has_feature_project(sb
)) {
1955 ext4_msg(sb
, KERN_ERR
, "Project quota feature not enabled. "
1956 "Cannot enable project quota enforcement.");
1959 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1960 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1961 clear_opt(sb
, USRQUOTA
);
1963 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1964 clear_opt(sb
, GRPQUOTA
);
1966 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1967 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1972 if (!sbi
->s_jquota_fmt
) {
1973 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1979 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
1981 BLOCK_SIZE
<< le32_to_cpu(sbi
->s_es
->s_log_block_size
);
1983 if (blocksize
< PAGE_SIZE
) {
1984 ext4_msg(sb
, KERN_ERR
, "can't mount with "
1985 "dioread_nolock if block size != PAGE_SIZE");
1992 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1993 struct super_block
*sb
)
1995 #if defined(CONFIG_QUOTA)
1996 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1998 if (sbi
->s_jquota_fmt
) {
2001 switch (sbi
->s_jquota_fmt
) {
2012 seq_printf(seq
, ",jqfmt=%s", fmtname
);
2015 if (sbi
->s_qf_names
[USRQUOTA
])
2016 seq_show_option(seq
, "usrjquota", sbi
->s_qf_names
[USRQUOTA
]);
2018 if (sbi
->s_qf_names
[GRPQUOTA
])
2019 seq_show_option(seq
, "grpjquota", sbi
->s_qf_names
[GRPQUOTA
]);
2023 static const char *token2str(int token
)
2025 const struct match_token
*t
;
2027 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
2028 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
2035 * - it's set to a non-default value OR
2036 * - if the per-sb default is different from the global default
2038 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
2041 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2042 struct ext4_super_block
*es
= sbi
->s_es
;
2043 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
2044 const struct mount_opts
*m
;
2045 char sep
= nodefs
? '\n' : ',';
2047 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2048 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2050 if (sbi
->s_sb_block
!= 1)
2051 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
2053 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
2054 int want_set
= m
->flags
& MOPT_SET
;
2055 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
2056 (m
->flags
& MOPT_CLEAR_ERR
))
2058 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
2059 continue; /* skip if same as the default */
2061 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
2062 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
2063 continue; /* select Opt_noFoo vs Opt_Foo */
2064 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
2067 if (nodefs
|| !uid_eq(sbi
->s_resuid
, make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
)) ||
2068 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
2069 SEQ_OPTS_PRINT("resuid=%u",
2070 from_kuid_munged(sb
->s_user_ns
, sbi
->s_resuid
));
2071 if (nodefs
|| !gid_eq(sbi
->s_resgid
, make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
)) ||
2072 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
2073 SEQ_OPTS_PRINT("resgid=%u",
2074 from_kgid_munged(sb
->s_user_ns
, sbi
->s_resgid
));
2075 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
2076 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
2077 SEQ_OPTS_PUTS("errors=remount-ro");
2078 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
2079 SEQ_OPTS_PUTS("errors=continue");
2080 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
2081 SEQ_OPTS_PUTS("errors=panic");
2082 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
2083 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
2084 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
2085 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
2086 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
2087 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
2088 if (sb
->s_flags
& MS_I_VERSION
)
2089 SEQ_OPTS_PUTS("i_version");
2090 if (nodefs
|| sbi
->s_stripe
)
2091 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
2092 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
2093 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
2094 SEQ_OPTS_PUTS("data=journal");
2095 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
2096 SEQ_OPTS_PUTS("data=ordered");
2097 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
2098 SEQ_OPTS_PUTS("data=writeback");
2101 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
2102 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2103 sbi
->s_inode_readahead_blks
);
2105 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
2106 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
2107 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
2108 if (nodefs
|| sbi
->s_max_dir_size_kb
)
2109 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi
->s_max_dir_size_kb
);
2110 if (test_opt(sb
, DATA_ERR_ABORT
))
2111 SEQ_OPTS_PUTS("data_err=abort");
2113 ext4_show_quota_options(seq
, sb
);
2117 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
2119 return _ext4_show_options(seq
, root
->d_sb
, 0);
2122 int ext4_seq_options_show(struct seq_file
*seq
, void *offset
)
2124 struct super_block
*sb
= seq
->private;
2127 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
2128 rc
= _ext4_show_options(seq
, sb
, 1);
2129 seq_puts(seq
, "\n");
2133 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
2136 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2139 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
2140 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
2141 "forcing read-only mode");
2146 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
2147 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
2148 "running e2fsck is recommended");
2149 else if (sbi
->s_mount_state
& EXT4_ERROR_FS
)
2150 ext4_msg(sb
, KERN_WARNING
,
2151 "warning: mounting fs with errors, "
2152 "running e2fsck is recommended");
2153 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
2154 le16_to_cpu(es
->s_mnt_count
) >=
2155 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2156 ext4_msg(sb
, KERN_WARNING
,
2157 "warning: maximal mount count reached, "
2158 "running e2fsck is recommended");
2159 else if (le32_to_cpu(es
->s_checkinterval
) &&
2160 (le32_to_cpu(es
->s_lastcheck
) +
2161 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
2162 ext4_msg(sb
, KERN_WARNING
,
2163 "warning: checktime reached, "
2164 "running e2fsck is recommended");
2165 if (!sbi
->s_journal
)
2166 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
2167 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
2168 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
2169 le16_add_cpu(&es
->s_mnt_count
, 1);
2170 es
->s_mtime
= cpu_to_le32(get_seconds());
2171 ext4_update_dynamic_rev(sb
);
2173 ext4_set_feature_journal_needs_recovery(sb
);
2175 ext4_commit_super(sb
, 1);
2177 if (test_opt(sb
, DEBUG
))
2178 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
2179 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2181 sbi
->s_groups_count
,
2182 EXT4_BLOCKS_PER_GROUP(sb
),
2183 EXT4_INODES_PER_GROUP(sb
),
2184 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
2186 cleancache_init_fs(sb
);
2190 int ext4_alloc_flex_bg_array(struct super_block
*sb
, ext4_group_t ngroup
)
2192 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2193 struct flex_groups
*new_groups
;
2196 if (!sbi
->s_log_groups_per_flex
)
2199 size
= ext4_flex_group(sbi
, ngroup
- 1) + 1;
2200 if (size
<= sbi
->s_flex_groups_allocated
)
2203 size
= roundup_pow_of_two(size
* sizeof(struct flex_groups
));
2204 new_groups
= kvzalloc(size
, GFP_KERNEL
);
2206 ext4_msg(sb
, KERN_ERR
, "not enough memory for %d flex groups",
2207 size
/ (int) sizeof(struct flex_groups
));
2211 if (sbi
->s_flex_groups
) {
2212 memcpy(new_groups
, sbi
->s_flex_groups
,
2213 (sbi
->s_flex_groups_allocated
*
2214 sizeof(struct flex_groups
)));
2215 kvfree(sbi
->s_flex_groups
);
2217 sbi
->s_flex_groups
= new_groups
;
2218 sbi
->s_flex_groups_allocated
= size
/ sizeof(struct flex_groups
);
2222 static int ext4_fill_flex_info(struct super_block
*sb
)
2224 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2225 struct ext4_group_desc
*gdp
= NULL
;
2226 ext4_group_t flex_group
;
2229 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2230 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
2231 sbi
->s_log_groups_per_flex
= 0;
2235 err
= ext4_alloc_flex_bg_array(sb
, sbi
->s_groups_count
);
2239 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2240 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2242 flex_group
= ext4_flex_group(sbi
, i
);
2243 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2244 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2245 atomic64_add(ext4_free_group_clusters(sb
, gdp
),
2246 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2247 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2248 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2256 static __le16
ext4_group_desc_csum(struct super_block
*sb
, __u32 block_group
,
2257 struct ext4_group_desc
*gdp
)
2259 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2261 __le32 le_group
= cpu_to_le32(block_group
);
2262 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2264 if (ext4_has_metadata_csum(sbi
->s_sb
)) {
2265 /* Use new metadata_csum algorithm */
2267 __u16 dummy_csum
= 0;
2269 csum32
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&le_group
,
2271 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
, offset
);
2272 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)&dummy_csum
,
2273 sizeof(dummy_csum
));
2274 offset
+= sizeof(dummy_csum
);
2275 if (offset
< sbi
->s_desc_size
)
2276 csum32
= ext4_chksum(sbi
, csum32
, (__u8
*)gdp
+ offset
,
2277 sbi
->s_desc_size
- offset
);
2279 crc
= csum32
& 0xFFFF;
2283 /* old crc16 code */
2284 if (!ext4_has_feature_gdt_csum(sb
))
2287 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2288 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2289 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2290 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2291 /* for checksum of struct ext4_group_desc do the rest...*/
2292 if (ext4_has_feature_64bit(sb
) &&
2293 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2294 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2295 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2299 return cpu_to_le16(crc
);
2302 int ext4_group_desc_csum_verify(struct super_block
*sb
, __u32 block_group
,
2303 struct ext4_group_desc
*gdp
)
2305 if (ext4_has_group_desc_csum(sb
) &&
2306 (gdp
->bg_checksum
!= ext4_group_desc_csum(sb
, block_group
, gdp
)))
2312 void ext4_group_desc_csum_set(struct super_block
*sb
, __u32 block_group
,
2313 struct ext4_group_desc
*gdp
)
2315 if (!ext4_has_group_desc_csum(sb
))
2317 gdp
->bg_checksum
= ext4_group_desc_csum(sb
, block_group
, gdp
);
2320 /* Called at mount-time, super-block is locked */
2321 static int ext4_check_descriptors(struct super_block
*sb
,
2322 ext4_fsblk_t sb_block
,
2323 ext4_group_t
*first_not_zeroed
)
2325 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2326 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2327 ext4_fsblk_t last_block
;
2328 ext4_fsblk_t block_bitmap
;
2329 ext4_fsblk_t inode_bitmap
;
2330 ext4_fsblk_t inode_table
;
2331 int flexbg_flag
= 0;
2332 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2334 if (ext4_has_feature_flex_bg(sb
))
2337 ext4_debug("Checking group descriptors");
2339 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2340 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2342 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2343 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2345 last_block
= first_block
+
2346 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2348 if ((grp
== sbi
->s_groups_count
) &&
2349 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2352 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2353 if (block_bitmap
== sb_block
) {
2354 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2355 "Block bitmap for group %u overlaps "
2358 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2359 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2360 "Block bitmap for group %u not in group "
2361 "(block %llu)!", i
, block_bitmap
);
2364 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2365 if (inode_bitmap
== sb_block
) {
2366 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2367 "Inode bitmap for group %u overlaps "
2370 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2371 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2372 "Inode bitmap for group %u not in group "
2373 "(block %llu)!", i
, inode_bitmap
);
2376 inode_table
= ext4_inode_table(sb
, gdp
);
2377 if (inode_table
== sb_block
) {
2378 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2379 "Inode table for group %u overlaps "
2382 if (inode_table
< first_block
||
2383 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2384 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2385 "Inode table for group %u not in group "
2386 "(block %llu)!", i
, inode_table
);
2389 ext4_lock_group(sb
, i
);
2390 if (!ext4_group_desc_csum_verify(sb
, i
, gdp
)) {
2391 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2392 "Checksum for group %u failed (%u!=%u)",
2393 i
, le16_to_cpu(ext4_group_desc_csum(sb
, i
,
2394 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2395 if (!(sb
->s_flags
& MS_RDONLY
)) {
2396 ext4_unlock_group(sb
, i
);
2400 ext4_unlock_group(sb
, i
);
2402 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2404 if (NULL
!= first_not_zeroed
)
2405 *first_not_zeroed
= grp
;
2409 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2410 * the superblock) which were deleted from all directories, but held open by
2411 * a process at the time of a crash. We walk the list and try to delete these
2412 * inodes at recovery time (only with a read-write filesystem).
2414 * In order to keep the orphan inode chain consistent during traversal (in
2415 * case of crash during recovery), we link each inode into the superblock
2416 * orphan list_head and handle it the same way as an inode deletion during
2417 * normal operation (which journals the operations for us).
2419 * We only do an iget() and an iput() on each inode, which is very safe if we
2420 * accidentally point at an in-use or already deleted inode. The worst that
2421 * can happen in this case is that we get a "bit already cleared" message from
2422 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2423 * e2fsck was run on this filesystem, and it must have already done the orphan
2424 * inode cleanup for us, so we can safely abort without any further action.
2426 static void ext4_orphan_cleanup(struct super_block
*sb
,
2427 struct ext4_super_block
*es
)
2429 unsigned int s_flags
= sb
->s_flags
;
2430 int ret
, nr_orphans
= 0, nr_truncates
= 0;
2434 if (!es
->s_last_orphan
) {
2435 jbd_debug(4, "no orphan inodes to clean up\n");
2439 if (bdev_read_only(sb
->s_bdev
)) {
2440 ext4_msg(sb
, KERN_ERR
, "write access "
2441 "unavailable, skipping orphan cleanup");
2445 /* Check if feature set would not allow a r/w mount */
2446 if (!ext4_feature_set_ok(sb
, 0)) {
2447 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2448 "unknown ROCOMPAT features");
2452 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2453 /* don't clear list on RO mount w/ errors */
2454 if (es
->s_last_orphan
&& !(s_flags
& MS_RDONLY
)) {
2455 ext4_msg(sb
, KERN_INFO
, "Errors on filesystem, "
2456 "clearing orphan list.\n");
2457 es
->s_last_orphan
= 0;
2459 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2463 if (s_flags
& MS_RDONLY
) {
2464 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2465 sb
->s_flags
&= ~MS_RDONLY
;
2468 /* Needed for iput() to work correctly and not trash data */
2469 sb
->s_flags
|= MS_ACTIVE
;
2470 /* Turn on quotas so that they are updated correctly */
2471 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2472 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2473 int ret
= ext4_quota_on_mount(sb
, i
);
2475 ext4_msg(sb
, KERN_ERR
,
2476 "Cannot turn on journaled "
2477 "quota: error %d", ret
);
2482 while (es
->s_last_orphan
) {
2483 struct inode
*inode
;
2486 * We may have encountered an error during cleanup; if
2487 * so, skip the rest.
2489 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2490 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2491 es
->s_last_orphan
= 0;
2495 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2496 if (IS_ERR(inode
)) {
2497 es
->s_last_orphan
= 0;
2501 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2502 dquot_initialize(inode
);
2503 if (inode
->i_nlink
) {
2504 if (test_opt(sb
, DEBUG
))
2505 ext4_msg(sb
, KERN_DEBUG
,
2506 "%s: truncating inode %lu to %lld bytes",
2507 __func__
, inode
->i_ino
, inode
->i_size
);
2508 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2509 inode
->i_ino
, inode
->i_size
);
2511 truncate_inode_pages(inode
->i_mapping
, inode
->i_size
);
2512 ret
= ext4_truncate(inode
);
2514 ext4_std_error(inode
->i_sb
, ret
);
2515 inode_unlock(inode
);
2518 if (test_opt(sb
, DEBUG
))
2519 ext4_msg(sb
, KERN_DEBUG
,
2520 "%s: deleting unreferenced inode %lu",
2521 __func__
, inode
->i_ino
);
2522 jbd_debug(2, "deleting unreferenced inode %lu\n",
2526 iput(inode
); /* The delete magic happens here! */
2529 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2532 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2533 PLURAL(nr_orphans
));
2535 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2536 PLURAL(nr_truncates
));
2538 /* Turn quotas off */
2539 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
2540 if (sb_dqopt(sb
)->files
[i
])
2541 dquot_quota_off(sb
, i
);
2544 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2548 * Maximal extent format file size.
2549 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2550 * extent format containers, within a sector_t, and within i_blocks
2551 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2552 * so that won't be a limiting factor.
2554 * However there is other limiting factor. We do store extents in the form
2555 * of starting block and length, hence the resulting length of the extent
2556 * covering maximum file size must fit into on-disk format containers as
2557 * well. Given that length is always by 1 unit bigger than max unit (because
2558 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2560 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2562 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2565 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2567 /* small i_blocks in vfs inode? */
2568 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2570 * CONFIG_LBDAF is not enabled implies the inode
2571 * i_block represent total blocks in 512 bytes
2572 * 32 == size of vfs inode i_blocks * 8
2574 upper_limit
= (1LL << 32) - 1;
2576 /* total blocks in file system block size */
2577 upper_limit
>>= (blkbits
- 9);
2578 upper_limit
<<= blkbits
;
2582 * 32-bit extent-start container, ee_block. We lower the maxbytes
2583 * by one fs block, so ee_len can cover the extent of maximum file
2586 res
= (1LL << 32) - 1;
2589 /* Sanity check against vm- & vfs- imposed limits */
2590 if (res
> upper_limit
)
2597 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2598 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2599 * We need to be 1 filesystem block less than the 2^48 sector limit.
2601 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2603 loff_t res
= EXT4_NDIR_BLOCKS
;
2606 /* This is calculated to be the largest file size for a dense, block
2607 * mapped file such that the file's total number of 512-byte sectors,
2608 * including data and all indirect blocks, does not exceed (2^48 - 1).
2610 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2611 * number of 512-byte sectors of the file.
2614 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2616 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2617 * the inode i_block field represents total file blocks in
2618 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2620 upper_limit
= (1LL << 32) - 1;
2622 /* total blocks in file system block size */
2623 upper_limit
>>= (bits
- 9);
2627 * We use 48 bit ext4_inode i_blocks
2628 * With EXT4_HUGE_FILE_FL set the i_blocks
2629 * represent total number of blocks in
2630 * file system block size
2632 upper_limit
= (1LL << 48) - 1;
2636 /* indirect blocks */
2638 /* double indirect blocks */
2639 meta_blocks
+= 1 + (1LL << (bits
-2));
2640 /* tripple indirect blocks */
2641 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2643 upper_limit
-= meta_blocks
;
2644 upper_limit
<<= bits
;
2646 res
+= 1LL << (bits
-2);
2647 res
+= 1LL << (2*(bits
-2));
2648 res
+= 1LL << (3*(bits
-2));
2650 if (res
> upper_limit
)
2653 if (res
> MAX_LFS_FILESIZE
)
2654 res
= MAX_LFS_FILESIZE
;
2659 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2660 ext4_fsblk_t logical_sb_block
, int nr
)
2662 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2663 ext4_group_t bg
, first_meta_bg
;
2666 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2668 if (!ext4_has_feature_meta_bg(sb
) || nr
< first_meta_bg
)
2669 return logical_sb_block
+ nr
+ 1;
2670 bg
= sbi
->s_desc_per_block
* nr
;
2671 if (ext4_bg_has_super(sb
, bg
))
2675 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2676 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2677 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2680 if (sb
->s_blocksize
== 1024 && nr
== 0 &&
2681 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_first_data_block
) == 0)
2684 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2688 * ext4_get_stripe_size: Get the stripe size.
2689 * @sbi: In memory super block info
2691 * If we have specified it via mount option, then
2692 * use the mount option value. If the value specified at mount time is
2693 * greater than the blocks per group use the super block value.
2694 * If the super block value is greater than blocks per group return 0.
2695 * Allocator needs it be less than blocks per group.
2698 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2700 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2701 unsigned long stripe_width
=
2702 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2705 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2706 ret
= sbi
->s_stripe
;
2707 else if (stripe_width
&& stripe_width
<= sbi
->s_blocks_per_group
)
2709 else if (stride
&& stride
<= sbi
->s_blocks_per_group
)
2715 * If the stripe width is 1, this makes no sense and
2716 * we set it to 0 to turn off stripe handling code.
2725 * Check whether this filesystem can be mounted based on
2726 * the features present and the RDONLY/RDWR mount requested.
2727 * Returns 1 if this filesystem can be mounted as requested,
2728 * 0 if it cannot be.
2730 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2732 if (ext4_has_unknown_ext4_incompat_features(sb
)) {
2733 ext4_msg(sb
, KERN_ERR
,
2734 "Couldn't mount because of "
2735 "unsupported optional features (%x)",
2736 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2737 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2744 if (ext4_has_feature_readonly(sb
)) {
2745 ext4_msg(sb
, KERN_INFO
, "filesystem is read-only");
2746 sb
->s_flags
|= MS_RDONLY
;
2750 /* Check that feature set is OK for a read-write mount */
2751 if (ext4_has_unknown_ext4_ro_compat_features(sb
)) {
2752 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2753 "unsupported optional features (%x)",
2754 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2755 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2759 * Large file size enabled file system can only be mounted
2760 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2762 if (ext4_has_feature_huge_file(sb
)) {
2763 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2764 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2765 "cannot be mounted RDWR without "
2770 if (ext4_has_feature_bigalloc(sb
) && !ext4_has_feature_extents(sb
)) {
2771 ext4_msg(sb
, KERN_ERR
,
2772 "Can't support bigalloc feature without "
2773 "extents feature\n");
2777 #ifndef CONFIG_QUOTA
2778 if (ext4_has_feature_quota(sb
) && !readonly
) {
2779 ext4_msg(sb
, KERN_ERR
,
2780 "Filesystem with quota feature cannot be mounted RDWR "
2781 "without CONFIG_QUOTA");
2784 if (ext4_has_feature_project(sb
) && !readonly
) {
2785 ext4_msg(sb
, KERN_ERR
,
2786 "Filesystem with project quota feature cannot be mounted RDWR "
2787 "without CONFIG_QUOTA");
2790 #endif /* CONFIG_QUOTA */
2795 * This function is called once a day if we have errors logged
2796 * on the file system
2798 static void print_daily_error_info(unsigned long arg
)
2800 struct super_block
*sb
= (struct super_block
*) arg
;
2801 struct ext4_sb_info
*sbi
;
2802 struct ext4_super_block
*es
;
2807 if (es
->s_error_count
)
2808 /* fsck newer than v1.41.13 is needed to clean this condition. */
2809 ext4_msg(sb
, KERN_NOTICE
, "error count since last fsck: %u",
2810 le32_to_cpu(es
->s_error_count
));
2811 if (es
->s_first_error_time
) {
2812 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at time %u: %.*s:%d",
2813 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2814 (int) sizeof(es
->s_first_error_func
),
2815 es
->s_first_error_func
,
2816 le32_to_cpu(es
->s_first_error_line
));
2817 if (es
->s_first_error_ino
)
2818 printk(KERN_CONT
": inode %u",
2819 le32_to_cpu(es
->s_first_error_ino
));
2820 if (es
->s_first_error_block
)
2821 printk(KERN_CONT
": block %llu", (unsigned long long)
2822 le64_to_cpu(es
->s_first_error_block
));
2823 printk(KERN_CONT
"\n");
2825 if (es
->s_last_error_time
) {
2826 printk(KERN_NOTICE
"EXT4-fs (%s): last error at time %u: %.*s:%d",
2827 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2828 (int) sizeof(es
->s_last_error_func
),
2829 es
->s_last_error_func
,
2830 le32_to_cpu(es
->s_last_error_line
));
2831 if (es
->s_last_error_ino
)
2832 printk(KERN_CONT
": inode %u",
2833 le32_to_cpu(es
->s_last_error_ino
));
2834 if (es
->s_last_error_block
)
2835 printk(KERN_CONT
": block %llu", (unsigned long long)
2836 le64_to_cpu(es
->s_last_error_block
));
2837 printk(KERN_CONT
"\n");
2839 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2842 /* Find next suitable group and run ext4_init_inode_table */
2843 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2845 struct ext4_group_desc
*gdp
= NULL
;
2846 ext4_group_t group
, ngroups
;
2847 struct super_block
*sb
;
2848 unsigned long timeout
= 0;
2852 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2854 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2855 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2861 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2865 if (group
>= ngroups
)
2870 ret
= ext4_init_inode_table(sb
, group
,
2871 elr
->lr_timeout
? 0 : 1);
2872 if (elr
->lr_timeout
== 0) {
2873 timeout
= (jiffies
- timeout
) *
2874 elr
->lr_sbi
->s_li_wait_mult
;
2875 elr
->lr_timeout
= timeout
;
2877 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2878 elr
->lr_next_group
= group
+ 1;
2884 * Remove lr_request from the list_request and free the
2885 * request structure. Should be called with li_list_mtx held
2887 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2889 struct ext4_sb_info
*sbi
;
2896 list_del(&elr
->lr_request
);
2897 sbi
->s_li_request
= NULL
;
2901 static void ext4_unregister_li_request(struct super_block
*sb
)
2903 mutex_lock(&ext4_li_mtx
);
2904 if (!ext4_li_info
) {
2905 mutex_unlock(&ext4_li_mtx
);
2909 mutex_lock(&ext4_li_info
->li_list_mtx
);
2910 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2911 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2912 mutex_unlock(&ext4_li_mtx
);
2915 static struct task_struct
*ext4_lazyinit_task
;
2918 * This is the function where ext4lazyinit thread lives. It walks
2919 * through the request list searching for next scheduled filesystem.
2920 * When such a fs is found, run the lazy initialization request
2921 * (ext4_rn_li_request) and keep track of the time spend in this
2922 * function. Based on that time we compute next schedule time of
2923 * the request. When walking through the list is complete, compute
2924 * next waking time and put itself into sleep.
2926 static int ext4_lazyinit_thread(void *arg
)
2928 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2929 struct list_head
*pos
, *n
;
2930 struct ext4_li_request
*elr
;
2931 unsigned long next_wakeup
, cur
;
2933 BUG_ON(NULL
== eli
);
2937 next_wakeup
= MAX_JIFFY_OFFSET
;
2939 mutex_lock(&eli
->li_list_mtx
);
2940 if (list_empty(&eli
->li_request_list
)) {
2941 mutex_unlock(&eli
->li_list_mtx
);
2944 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2947 elr
= list_entry(pos
, struct ext4_li_request
,
2950 if (time_before(jiffies
, elr
->lr_next_sched
)) {
2951 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2952 next_wakeup
= elr
->lr_next_sched
;
2955 if (down_read_trylock(&elr
->lr_super
->s_umount
)) {
2956 if (sb_start_write_trylock(elr
->lr_super
)) {
2959 * We hold sb->s_umount, sb can not
2960 * be removed from the list, it is
2961 * now safe to drop li_list_mtx
2963 mutex_unlock(&eli
->li_list_mtx
);
2964 err
= ext4_run_li_request(elr
);
2965 sb_end_write(elr
->lr_super
);
2966 mutex_lock(&eli
->li_list_mtx
);
2969 up_read((&elr
->lr_super
->s_umount
));
2971 /* error, remove the lazy_init job */
2973 ext4_remove_li_request(elr
);
2977 elr
->lr_next_sched
= jiffies
+
2979 % (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
2981 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2982 next_wakeup
= elr
->lr_next_sched
;
2984 mutex_unlock(&eli
->li_list_mtx
);
2989 if ((time_after_eq(cur
, next_wakeup
)) ||
2990 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2995 schedule_timeout_interruptible(next_wakeup
- cur
);
2997 if (kthread_should_stop()) {
2998 ext4_clear_request_list();
3005 * It looks like the request list is empty, but we need
3006 * to check it under the li_list_mtx lock, to prevent any
3007 * additions into it, and of course we should lock ext4_li_mtx
3008 * to atomically free the list and ext4_li_info, because at
3009 * this point another ext4 filesystem could be registering
3012 mutex_lock(&ext4_li_mtx
);
3013 mutex_lock(&eli
->li_list_mtx
);
3014 if (!list_empty(&eli
->li_request_list
)) {
3015 mutex_unlock(&eli
->li_list_mtx
);
3016 mutex_unlock(&ext4_li_mtx
);
3019 mutex_unlock(&eli
->li_list_mtx
);
3020 kfree(ext4_li_info
);
3021 ext4_li_info
= NULL
;
3022 mutex_unlock(&ext4_li_mtx
);
3027 static void ext4_clear_request_list(void)
3029 struct list_head
*pos
, *n
;
3030 struct ext4_li_request
*elr
;
3032 mutex_lock(&ext4_li_info
->li_list_mtx
);
3033 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
3034 elr
= list_entry(pos
, struct ext4_li_request
,
3036 ext4_remove_li_request(elr
);
3038 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3041 static int ext4_run_lazyinit_thread(void)
3043 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
3044 ext4_li_info
, "ext4lazyinit");
3045 if (IS_ERR(ext4_lazyinit_task
)) {
3046 int err
= PTR_ERR(ext4_lazyinit_task
);
3047 ext4_clear_request_list();
3048 kfree(ext4_li_info
);
3049 ext4_li_info
= NULL
;
3050 printk(KERN_CRIT
"EXT4-fs: error %d creating inode table "
3051 "initialization thread\n",
3055 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
3060 * Check whether it make sense to run itable init. thread or not.
3061 * If there is at least one uninitialized inode table, return
3062 * corresponding group number, else the loop goes through all
3063 * groups and return total number of groups.
3065 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
3067 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
3068 struct ext4_group_desc
*gdp
= NULL
;
3070 for (group
= 0; group
< ngroups
; group
++) {
3071 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
3075 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
3082 static int ext4_li_info_new(void)
3084 struct ext4_lazy_init
*eli
= NULL
;
3086 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
3090 INIT_LIST_HEAD(&eli
->li_request_list
);
3091 mutex_init(&eli
->li_list_mtx
);
3093 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
3100 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3103 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3104 struct ext4_li_request
*elr
;
3106 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3112 elr
->lr_next_group
= start
;
3115 * Randomize first schedule time of the request to
3116 * spread the inode table initialization requests
3119 elr
->lr_next_sched
= jiffies
+ (prandom_u32() %
3120 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
));
3124 int ext4_register_li_request(struct super_block
*sb
,
3125 ext4_group_t first_not_zeroed
)
3127 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3128 struct ext4_li_request
*elr
= NULL
;
3129 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3132 mutex_lock(&ext4_li_mtx
);
3133 if (sbi
->s_li_request
!= NULL
) {
3135 * Reset timeout so it can be computed again, because
3136 * s_li_wait_mult might have changed.
3138 sbi
->s_li_request
->lr_timeout
= 0;
3142 if (first_not_zeroed
== ngroups
||
3143 (sb
->s_flags
& MS_RDONLY
) ||
3144 !test_opt(sb
, INIT_INODE_TABLE
))
3147 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3153 if (NULL
== ext4_li_info
) {
3154 ret
= ext4_li_info_new();
3159 mutex_lock(&ext4_li_info
->li_list_mtx
);
3160 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3161 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3163 sbi
->s_li_request
= elr
;
3165 * set elr to NULL here since it has been inserted to
3166 * the request_list and the removal and free of it is
3167 * handled by ext4_clear_request_list from now on.
3171 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3172 ret
= ext4_run_lazyinit_thread();
3177 mutex_unlock(&ext4_li_mtx
);
3184 * We do not need to lock anything since this is called on
3187 static void ext4_destroy_lazyinit_thread(void)
3190 * If thread exited earlier
3191 * there's nothing to be done.
3193 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3196 kthread_stop(ext4_lazyinit_task
);
3199 static int set_journal_csum_feature_set(struct super_block
*sb
)
3202 int compat
, incompat
;
3203 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3205 if (ext4_has_metadata_csum(sb
)) {
3206 /* journal checksum v3 */
3208 incompat
= JBD2_FEATURE_INCOMPAT_CSUM_V3
;
3210 /* journal checksum v1 */
3211 compat
= JBD2_FEATURE_COMPAT_CHECKSUM
;
3215 jbd2_journal_clear_features(sbi
->s_journal
,
3216 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3217 JBD2_FEATURE_INCOMPAT_CSUM_V3
|
3218 JBD2_FEATURE_INCOMPAT_CSUM_V2
);
3219 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3220 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3222 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
|
3224 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3225 ret
= jbd2_journal_set_features(sbi
->s_journal
,
3228 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3229 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3231 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3232 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3239 * Note: calculating the overhead so we can be compatible with
3240 * historical BSD practice is quite difficult in the face of
3241 * clusters/bigalloc. This is because multiple metadata blocks from
3242 * different block group can end up in the same allocation cluster.
3243 * Calculating the exact overhead in the face of clustered allocation
3244 * requires either O(all block bitmaps) in memory or O(number of block
3245 * groups**2) in time. We will still calculate the superblock for
3246 * older file systems --- and if we come across with a bigalloc file
3247 * system with zero in s_overhead_clusters the estimate will be close to
3248 * correct especially for very large cluster sizes --- but for newer
3249 * file systems, it's better to calculate this figure once at mkfs
3250 * time, and store it in the superblock. If the superblock value is
3251 * present (even for non-bigalloc file systems), we will use it.
3253 static int count_overhead(struct super_block
*sb
, ext4_group_t grp
,
3256 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3257 struct ext4_group_desc
*gdp
;
3258 ext4_fsblk_t first_block
, last_block
, b
;
3259 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3260 int s
, j
, count
= 0;
3262 if (!ext4_has_feature_bigalloc(sb
))
3263 return (ext4_bg_has_super(sb
, grp
) + ext4_bg_num_gdb(sb
, grp
) +
3264 sbi
->s_itb_per_group
+ 2);
3266 first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
) +
3267 (grp
* EXT4_BLOCKS_PER_GROUP(sb
));
3268 last_block
= first_block
+ EXT4_BLOCKS_PER_GROUP(sb
) - 1;
3269 for (i
= 0; i
< ngroups
; i
++) {
3270 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
3271 b
= ext4_block_bitmap(sb
, gdp
);
3272 if (b
>= first_block
&& b
<= last_block
) {
3273 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3276 b
= ext4_inode_bitmap(sb
, gdp
);
3277 if (b
>= first_block
&& b
<= last_block
) {
3278 ext4_set_bit(EXT4_B2C(sbi
, b
- first_block
), buf
);
3281 b
= ext4_inode_table(sb
, gdp
);
3282 if (b
>= first_block
&& b
+ sbi
->s_itb_per_group
<= last_block
)
3283 for (j
= 0; j
< sbi
->s_itb_per_group
; j
++, b
++) {
3284 int c
= EXT4_B2C(sbi
, b
- first_block
);
3285 ext4_set_bit(c
, buf
);
3291 if (ext4_bg_has_super(sb
, grp
)) {
3292 ext4_set_bit(s
++, buf
);
3295 j
= ext4_bg_num_gdb(sb
, grp
);
3296 if (s
+ j
> EXT4_BLOCKS_PER_GROUP(sb
)) {
3297 ext4_error(sb
, "Invalid number of block group "
3298 "descriptor blocks: %d", j
);
3299 j
= EXT4_BLOCKS_PER_GROUP(sb
) - s
;
3303 ext4_set_bit(EXT4_B2C(sbi
, s
++), buf
);
3307 return EXT4_CLUSTERS_PER_GROUP(sb
) -
3308 ext4_count_free(buf
, EXT4_CLUSTERS_PER_GROUP(sb
) / 8);
3312 * Compute the overhead and stash it in sbi->s_overhead
3314 int ext4_calculate_overhead(struct super_block
*sb
)
3316 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3317 struct ext4_super_block
*es
= sbi
->s_es
;
3318 struct inode
*j_inode
;
3319 unsigned int j_blocks
, j_inum
= le32_to_cpu(es
->s_journal_inum
);
3320 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3321 ext4_fsblk_t overhead
= 0;
3322 char *buf
= (char *) get_zeroed_page(GFP_NOFS
);
3328 * Compute the overhead (FS structures). This is constant
3329 * for a given filesystem unless the number of block groups
3330 * changes so we cache the previous value until it does.
3334 * All of the blocks before first_data_block are overhead
3336 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
3339 * Add the overhead found in each block group
3341 for (i
= 0; i
< ngroups
; i
++) {
3344 blks
= count_overhead(sb
, i
, buf
);
3347 memset(buf
, 0, PAGE_SIZE
);
3352 * Add the internal journal blocks whether the journal has been
3355 if (sbi
->s_journal
&& !sbi
->journal_bdev
)
3356 overhead
+= EXT4_NUM_B2C(sbi
, sbi
->s_journal
->j_maxlen
);
3357 else if (ext4_has_feature_journal(sb
) && !sbi
->s_journal
) {
3358 j_inode
= ext4_get_journal_inode(sb
, j_inum
);
3360 j_blocks
= j_inode
->i_size
>> sb
->s_blocksize_bits
;
3361 overhead
+= EXT4_NUM_B2C(sbi
, j_blocks
);
3364 ext4_msg(sb
, KERN_ERR
, "can't get journal size");
3367 sbi
->s_overhead
= overhead
;
3369 free_page((unsigned long) buf
);
3373 static void ext4_set_resv_clusters(struct super_block
*sb
)
3375 ext4_fsblk_t resv_clusters
;
3376 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3379 * There's no need to reserve anything when we aren't using extents.
3380 * The space estimates are exact, there are no unwritten extents,
3381 * hole punching doesn't need new metadata... This is needed especially
3382 * to keep ext2/3 backward compatibility.
3384 if (!ext4_has_feature_extents(sb
))
3387 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3388 * This should cover the situations where we can not afford to run
3389 * out of space like for example punch hole, or converting
3390 * unwritten extents in delalloc path. In most cases such
3391 * allocation would require 1, or 2 blocks, higher numbers are
3394 resv_clusters
= (ext4_blocks_count(sbi
->s_es
) >>
3395 sbi
->s_cluster_bits
);
3397 do_div(resv_clusters
, 50);
3398 resv_clusters
= min_t(ext4_fsblk_t
, resv_clusters
, 4096);
3400 atomic64_set(&sbi
->s_resv_clusters
, resv_clusters
);
3403 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3405 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3406 struct buffer_head
*bh
;
3407 struct ext4_super_block
*es
= NULL
;
3408 struct ext4_sb_info
*sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3410 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3411 ext4_fsblk_t logical_sb_block
;
3412 unsigned long offset
= 0;
3413 unsigned long journal_devnum
= 0;
3414 unsigned long def_mount_opts
;
3418 int blocksize
, clustersize
;
3419 unsigned int db_count
;
3421 int needs_recovery
, has_huge_files
, has_bigalloc
;
3424 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3425 ext4_group_t first_not_zeroed
;
3427 if ((data
&& !orig_data
) || !sbi
)
3430 if (!userns_mounts
&& !capable(CAP_SYS_ADMIN
)) {
3435 sbi
->s_blockgroup_lock
=
3436 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3437 if (!sbi
->s_blockgroup_lock
)
3440 sb
->s_fs_info
= sbi
;
3442 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3443 sbi
->s_sb_block
= sb_block
;
3444 if (sb
->s_bdev
->bd_part
)
3445 sbi
->s_sectors_written_start
=
3446 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3448 /* Cleanup superblock name */
3449 strreplace(sb
->s_id
, '/', '!');
3451 /* -EINVAL is default */
3453 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3455 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3460 * The ext4 superblock will not be buffer aligned for other than 1kB
3461 * block sizes. We need to calculate the offset from buffer start.
3463 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3464 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3465 offset
= do_div(logical_sb_block
, blocksize
);
3467 logical_sb_block
= sb_block
;
3470 if (!(bh
= sb_bread_unmovable(sb
, logical_sb_block
))) {
3471 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3475 * Note: s_es must be initialized as soon as possible because
3476 * some ext4 macro-instructions depend on its value
3478 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
3480 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3481 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3483 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3485 /* Warn if metadata_csum and gdt_csum are both set. */
3486 if (ext4_has_feature_metadata_csum(sb
) &&
3487 ext4_has_feature_gdt_csum(sb
))
3488 ext4_warning(sb
, "metadata_csum and uninit_bg are "
3489 "redundant flags; please run fsck.");
3491 /* Check for a known checksum algorithm */
3492 if (!ext4_verify_csum_type(sb
, es
)) {
3493 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3494 "unknown checksum algorithm.");
3499 /* Load the checksum driver */
3500 if (ext4_has_feature_metadata_csum(sb
) ||
3501 ext4_has_feature_ea_inode(sb
)) {
3502 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32c", 0, 0);
3503 if (IS_ERR(sbi
->s_chksum_driver
)) {
3504 ext4_msg(sb
, KERN_ERR
, "Cannot load crc32c driver.");
3505 ret
= PTR_ERR(sbi
->s_chksum_driver
);
3506 sbi
->s_chksum_driver
= NULL
;
3511 /* Check superblock checksum */
3512 if (!ext4_superblock_csum_verify(sb
, es
)) {
3513 ext4_msg(sb
, KERN_ERR
, "VFS: Found ext4 filesystem with "
3514 "invalid superblock checksum. Run e2fsck?");
3520 /* Precompute checksum seed for all metadata */
3521 if (ext4_has_feature_csum_seed(sb
))
3522 sbi
->s_csum_seed
= le32_to_cpu(es
->s_checksum_seed
);
3523 else if (ext4_has_metadata_csum(sb
) || ext4_has_feature_ea_inode(sb
))
3524 sbi
->s_csum_seed
= ext4_chksum(sbi
, ~0, es
->s_uuid
,
3525 sizeof(es
->s_uuid
));
3527 /* Set defaults before we parse the mount options */
3528 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3529 set_opt(sb
, INIT_INODE_TABLE
);
3530 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3532 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3534 if (def_mount_opts
& EXT4_DEFM_UID16
)
3535 set_opt(sb
, NO_UID32
);
3536 /* xattr user namespace & acls are now defaulted on */
3537 set_opt(sb
, XATTR_USER
);
3538 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3539 set_opt(sb
, POSIX_ACL
);
3541 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3542 if (ext4_has_metadata_csum(sb
))
3543 set_opt(sb
, JOURNAL_CHECKSUM
);
3545 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3546 set_opt(sb
, JOURNAL_DATA
);
3547 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3548 set_opt(sb
, ORDERED_DATA
);
3549 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3550 set_opt(sb
, WRITEBACK_DATA
);
3552 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
) {
3553 if (!capable(CAP_SYS_ADMIN
))
3555 set_opt(sb
, ERRORS_PANIC
);
3556 } else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
) {
3557 set_opt(sb
, ERRORS_CONT
);
3559 set_opt(sb
, ERRORS_RO
);
3561 /* block_validity enabled by default; disable with noblock_validity */
3562 set_opt(sb
, BLOCK_VALIDITY
);
3563 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3564 set_opt(sb
, DISCARD
);
3566 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resuid
));
3567 if (!uid_valid(sbi
->s_resuid
))
3568 sbi
->s_resuid
= make_kuid(sb
->s_user_ns
, EXT4_DEF_RESUID
);
3569 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, le16_to_cpu(es
->s_def_resgid
));
3570 if (!gid_valid(sbi
->s_resgid
))
3571 sbi
->s_resgid
= make_kgid(sb
->s_user_ns
, EXT4_DEF_RESGID
);
3572 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3573 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3574 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3576 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3577 set_opt(sb
, BARRIER
);
3580 * enable delayed allocation by default
3581 * Use -o nodelalloc to turn it off
3583 if (!IS_EXT3_SB(sb
) && !IS_EXT2_SB(sb
) &&
3584 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3585 set_opt(sb
, DELALLOC
);
3588 * set default s_li_wait_mult for lazyinit, for the case there is
3589 * no mount option specified.
3591 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3593 if (sbi
->s_es
->s_mount_opts
[0]) {
3594 char *s_mount_opts
= kstrndup(sbi
->s_es
->s_mount_opts
,
3595 sizeof(sbi
->s_es
->s_mount_opts
),
3599 if (!parse_options(s_mount_opts
, sb
, &journal_devnum
,
3600 &journal_ioprio
, 0)) {
3601 ext4_msg(sb
, KERN_WARNING
,
3602 "failed to parse options in superblock: %s",
3605 kfree(s_mount_opts
);
3607 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3608 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3609 &journal_ioprio
, 0))
3612 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3613 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3614 "with data=journal disables delayed "
3615 "allocation and O_DIRECT support!\n");
3616 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3617 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3618 "both data=journal and delalloc");
3621 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3622 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3623 "both data=journal and dioread_nolock");
3626 if (test_opt(sb
, DAX
)) {
3627 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3628 "both data=journal and dax");
3631 if (ext4_has_feature_encrypt(sb
)) {
3632 ext4_msg(sb
, KERN_WARNING
,
3633 "encrypted files will use data=ordered "
3634 "instead of data journaling mode");
3636 if (test_opt(sb
, DELALLOC
))
3637 clear_opt(sb
, DELALLOC
);
3639 sb
->s_iflags
|= SB_I_CGROUPWB
;
3642 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3643 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3645 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3646 (ext4_has_compat_features(sb
) ||
3647 ext4_has_ro_compat_features(sb
) ||
3648 ext4_has_incompat_features(sb
)))
3649 ext4_msg(sb
, KERN_WARNING
,
3650 "feature flags set on rev 0 fs, "
3651 "running e2fsck is recommended");
3653 if (es
->s_creator_os
== cpu_to_le32(EXT4_OS_HURD
)) {
3654 set_opt2(sb
, HURD_COMPAT
);
3655 if (ext4_has_feature_64bit(sb
)) {
3656 ext4_msg(sb
, KERN_ERR
,
3657 "The Hurd can't support 64-bit file systems");
3662 * ea_inode feature uses l_i_version field which is not
3663 * available in HURD_COMPAT mode.
3665 if (ext4_has_feature_ea_inode(sb
)) {
3666 ext4_msg(sb
, KERN_ERR
,
3667 "ea_inode feature is not supported for Hurd");
3672 if (IS_EXT2_SB(sb
)) {
3673 if (ext2_feature_set_ok(sb
))
3674 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3675 "using the ext4 subsystem");
3677 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3678 "to feature incompatibilities");
3683 if (IS_EXT3_SB(sb
)) {
3684 if (ext3_feature_set_ok(sb
))
3685 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3686 "using the ext4 subsystem");
3688 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3689 "to feature incompatibilities");
3695 * Check feature flags regardless of the revision level, since we
3696 * previously didn't change the revision level when setting the flags,
3697 * so there is a chance incompat flags are set on a rev 0 filesystem.
3699 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3702 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3703 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3704 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3705 ext4_msg(sb
, KERN_ERR
,
3706 "Unsupported filesystem blocksize %d (%d log_block_size)",
3707 blocksize
, le32_to_cpu(es
->s_log_block_size
));
3710 if (le32_to_cpu(es
->s_log_block_size
) >
3711 (EXT4_MAX_BLOCK_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3712 ext4_msg(sb
, KERN_ERR
,
3713 "Invalid log block size: %u",
3714 le32_to_cpu(es
->s_log_block_size
));
3718 if (le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) > (blocksize
/ 4)) {
3719 ext4_msg(sb
, KERN_ERR
,
3720 "Number of reserved GDT blocks insanely large: %d",
3721 le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
));
3725 if (sbi
->s_mount_opt
& EXT4_MOUNT_DAX
) {
3726 err
= bdev_dax_supported(sb
, blocksize
);
3731 if (ext4_has_feature_encrypt(sb
) && es
->s_encryption_level
) {
3732 ext4_msg(sb
, KERN_ERR
, "Unsupported encryption level %d",
3733 es
->s_encryption_level
);
3737 if (sb
->s_blocksize
!= blocksize
) {
3738 /* Validate the filesystem blocksize */
3739 if (!sb_set_blocksize(sb
, blocksize
)) {
3740 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3746 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3747 offset
= do_div(logical_sb_block
, blocksize
);
3748 bh
= sb_bread_unmovable(sb
, logical_sb_block
);
3750 ext4_msg(sb
, KERN_ERR
,
3751 "Can't read superblock on 2nd try");
3754 es
= (struct ext4_super_block
*)(bh
->b_data
+ offset
);
3756 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3757 ext4_msg(sb
, KERN_ERR
,
3758 "Magic mismatch, very weird!");
3763 has_huge_files
= ext4_has_feature_huge_file(sb
);
3764 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3766 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3768 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3769 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3770 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3772 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3773 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3774 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3775 (!is_power_of_2(sbi
->s_inode_size
)) ||
3776 (sbi
->s_inode_size
> blocksize
)) {
3777 ext4_msg(sb
, KERN_ERR
,
3778 "unsupported inode size: %d",
3782 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3783 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3786 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3787 if (ext4_has_feature_64bit(sb
)) {
3788 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3789 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3790 !is_power_of_2(sbi
->s_desc_size
)) {
3791 ext4_msg(sb
, KERN_ERR
,
3792 "unsupported descriptor size %lu",
3797 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3799 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3800 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3802 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3803 if (sbi
->s_inodes_per_block
== 0)
3805 if (sbi
->s_inodes_per_group
< sbi
->s_inodes_per_block
||
3806 sbi
->s_inodes_per_group
> blocksize
* 8) {
3807 ext4_msg(sb
, KERN_ERR
, "invalid inodes per group: %lu\n",
3808 sbi
->s_blocks_per_group
);
3811 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3812 sbi
->s_inodes_per_block
;
3813 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3815 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3816 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3817 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3819 for (i
= 0; i
< 4; i
++)
3820 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3821 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3822 if (ext4_has_feature_dir_index(sb
)) {
3823 i
= le32_to_cpu(es
->s_flags
);
3824 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3825 sbi
->s_hash_unsigned
= 3;
3826 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3827 #ifdef __CHAR_UNSIGNED__
3828 if (!(sb
->s_flags
& MS_RDONLY
))
3830 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3831 sbi
->s_hash_unsigned
= 3;
3833 if (!(sb
->s_flags
& MS_RDONLY
))
3835 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3840 /* Handle clustersize */
3841 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3842 has_bigalloc
= ext4_has_feature_bigalloc(sb
);
3844 if (clustersize
< blocksize
) {
3845 ext4_msg(sb
, KERN_ERR
,
3846 "cluster size (%d) smaller than "
3847 "block size (%d)", clustersize
, blocksize
);
3850 if (le32_to_cpu(es
->s_log_cluster_size
) >
3851 (EXT4_MAX_CLUSTER_LOG_SIZE
- EXT4_MIN_BLOCK_LOG_SIZE
)) {
3852 ext4_msg(sb
, KERN_ERR
,
3853 "Invalid log cluster size: %u",
3854 le32_to_cpu(es
->s_log_cluster_size
));
3857 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3858 le32_to_cpu(es
->s_log_block_size
);
3859 sbi
->s_clusters_per_group
=
3860 le32_to_cpu(es
->s_clusters_per_group
);
3861 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3862 ext4_msg(sb
, KERN_ERR
,
3863 "#clusters per group too big: %lu",
3864 sbi
->s_clusters_per_group
);
3867 if (sbi
->s_blocks_per_group
!=
3868 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3869 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3870 "clusters per group (%lu) inconsistent",
3871 sbi
->s_blocks_per_group
,
3872 sbi
->s_clusters_per_group
);
3876 if (clustersize
!= blocksize
) {
3877 ext4_warning(sb
, "fragment/cluster size (%d) != "
3878 "block size (%d)", clustersize
,
3880 clustersize
= blocksize
;
3882 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3883 ext4_msg(sb
, KERN_ERR
,
3884 "#blocks per group too big: %lu",
3885 sbi
->s_blocks_per_group
);
3888 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3889 sbi
->s_cluster_bits
= 0;
3891 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3893 /* Do we have standard group size of clustersize * 8 blocks ? */
3894 if (sbi
->s_blocks_per_group
== clustersize
<< 3)
3895 set_opt2(sb
, STD_GROUP_SIZE
);
3898 * Test whether we have more sectors than will fit in sector_t,
3899 * and whether the max offset is addressable by the page cache.
3901 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3902 ext4_blocks_count(es
));
3904 ext4_msg(sb
, KERN_ERR
, "filesystem"
3905 " too large to mount safely on this system");
3906 if (sizeof(sector_t
) < 8)
3907 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3911 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3914 /* check blocks count against device size */
3915 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3916 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3917 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3918 "exceeds size of device (%llu blocks)",
3919 ext4_blocks_count(es
), blocks_count
);
3924 * It makes no sense for the first data block to be beyond the end
3925 * of the filesystem.
3927 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3928 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3929 "block %u is beyond end of filesystem (%llu)",
3930 le32_to_cpu(es
->s_first_data_block
),
3931 ext4_blocks_count(es
));
3934 blocks_count
= (ext4_blocks_count(es
) -
3935 le32_to_cpu(es
->s_first_data_block
) +
3936 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3937 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3938 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3939 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3940 "(block count %llu, first data block %u, "
3941 "blocks per group %lu)", sbi
->s_groups_count
,
3942 ext4_blocks_count(es
),
3943 le32_to_cpu(es
->s_first_data_block
),
3944 EXT4_BLOCKS_PER_GROUP(sb
));
3947 sbi
->s_groups_count
= blocks_count
;
3948 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3949 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3950 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3951 EXT4_DESC_PER_BLOCK(sb
);
3952 if (ext4_has_feature_meta_bg(sb
)) {
3953 if (le32_to_cpu(es
->s_first_meta_bg
) > db_count
) {
3954 ext4_msg(sb
, KERN_WARNING
,
3955 "first meta block group too large: %u "
3956 "(group descriptor block count %u)",
3957 le32_to_cpu(es
->s_first_meta_bg
), db_count
);
3961 sbi
->s_group_desc
= kvmalloc(db_count
*
3962 sizeof(struct buffer_head
*),
3964 if (sbi
->s_group_desc
== NULL
) {
3965 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3970 bgl_lock_init(sbi
->s_blockgroup_lock
);
3972 /* Pre-read the descriptors into the buffer cache */
3973 for (i
= 0; i
< db_count
; i
++) {
3974 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3975 sb_breadahead(sb
, block
);
3978 for (i
= 0; i
< db_count
; i
++) {
3979 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3980 sbi
->s_group_desc
[i
] = sb_bread_unmovable(sb
, block
);
3981 if (!sbi
->s_group_desc
[i
]) {
3982 ext4_msg(sb
, KERN_ERR
,
3983 "can't read group descriptor %d", i
);
3988 if (!ext4_check_descriptors(sb
, logical_sb_block
, &first_not_zeroed
)) {
3989 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3990 ret
= -EFSCORRUPTED
;
3994 sbi
->s_gdb_count
= db_count
;
3995 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3996 spin_lock_init(&sbi
->s_next_gen_lock
);
3998 setup_timer(&sbi
->s_err_report
, print_daily_error_info
,
3999 (unsigned long) sb
);
4001 /* Register extent status tree shrinker */
4002 if (ext4_es_register_shrinker(sbi
))
4005 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
4006 sbi
->s_extent_max_zeroout_kb
= 32;
4009 * set up enough so that it can read an inode
4011 sb
->s_op
= &ext4_sops
;
4012 sb
->s_export_op
= &ext4_export_ops
;
4013 sb
->s_xattr
= ext4_xattr_handlers
;
4014 sb
->s_cop
= &ext4_cryptops
;
4016 sb
->dq_op
= &ext4_quota_operations
;
4017 if (ext4_has_feature_quota(sb
))
4018 sb
->s_qcop
= &dquot_quotactl_sysfile_ops
;
4020 sb
->s_qcop
= &ext4_qctl_operations
;
4021 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
4023 memcpy(&sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
4025 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
4026 mutex_init(&sbi
->s_orphan_lock
);
4030 needs_recovery
= (es
->s_last_orphan
!= 0 ||
4031 ext4_has_feature_journal_needs_recovery(sb
));
4033 if (ext4_has_feature_mmp(sb
) && !(sb
->s_flags
& MS_RDONLY
))
4034 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
4035 goto failed_mount3a
;
4038 * The first inode we look at is the journal inode. Don't try
4039 * root first: it may be modified in the journal!
4041 if (!test_opt(sb
, NOLOAD
) && ext4_has_feature_journal(sb
)) {
4042 err
= ext4_load_journal(sb
, es
, journal_devnum
);
4044 goto failed_mount3a
;
4045 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
4046 ext4_has_feature_journal_needs_recovery(sb
)) {
4047 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
4048 "suppressed and not mounted read-only");
4049 goto failed_mount_wq
;
4051 /* Nojournal mode, all journal mount options are illegal */
4052 if (test_opt2(sb
, EXPLICIT_JOURNAL_CHECKSUM
)) {
4053 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4054 "journal_checksum, fs mounted w/o journal");
4055 goto failed_mount_wq
;
4057 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4058 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4059 "journal_async_commit, fs mounted w/o journal");
4060 goto failed_mount_wq
;
4062 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
4063 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4064 "commit=%lu, fs mounted w/o journal",
4065 sbi
->s_commit_interval
/ HZ
);
4066 goto failed_mount_wq
;
4068 if (EXT4_MOUNT_DATA_FLAGS
&
4069 (sbi
->s_mount_opt
^ sbi
->s_def_mount_opt
)) {
4070 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4071 "data=, fs mounted w/o journal");
4072 goto failed_mount_wq
;
4074 sbi
->s_def_mount_opt
&= EXT4_MOUNT_JOURNAL_CHECKSUM
;
4075 clear_opt(sb
, JOURNAL_CHECKSUM
);
4076 clear_opt(sb
, DATA_FLAGS
);
4077 sbi
->s_journal
= NULL
;
4082 if (ext4_has_feature_64bit(sb
) &&
4083 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
4084 JBD2_FEATURE_INCOMPAT_64BIT
)) {
4085 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
4086 goto failed_mount_wq
;
4089 if (!set_journal_csum_feature_set(sb
)) {
4090 ext4_msg(sb
, KERN_ERR
, "Failed to set journal checksum "
4092 goto failed_mount_wq
;
4095 /* We have now updated the journal if required, so we can
4096 * validate the data journaling mode. */
4097 switch (test_opt(sb
, DATA_FLAGS
)) {
4099 /* No mode set, assume a default based on the journal
4100 * capabilities: ORDERED_DATA if the journal can
4101 * cope, else JOURNAL_DATA
4103 if (jbd2_journal_check_available_features
4104 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
4105 set_opt(sb
, ORDERED_DATA
);
4107 set_opt(sb
, JOURNAL_DATA
);
4110 case EXT4_MOUNT_ORDERED_DATA
:
4111 case EXT4_MOUNT_WRITEBACK_DATA
:
4112 if (!jbd2_journal_check_available_features
4113 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
4114 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
4115 "requested data journaling mode");
4116 goto failed_mount_wq
;
4122 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
&&
4123 test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
4124 ext4_msg(sb
, KERN_ERR
, "can't mount with "
4125 "journal_async_commit in data=ordered mode");
4126 goto failed_mount_wq
;
4129 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4131 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
4134 if (!test_opt(sb
, NO_MBCACHE
)) {
4135 sbi
->s_ea_block_cache
= ext4_xattr_create_cache();
4136 if (!sbi
->s_ea_block_cache
) {
4137 ext4_msg(sb
, KERN_ERR
,
4138 "Failed to create ea_block_cache");
4139 goto failed_mount_wq
;
4142 if (ext4_has_feature_ea_inode(sb
)) {
4143 sbi
->s_ea_inode_cache
= ext4_xattr_create_cache();
4144 if (!sbi
->s_ea_inode_cache
) {
4145 ext4_msg(sb
, KERN_ERR
,
4146 "Failed to create ea_inode_cache");
4147 goto failed_mount_wq
;
4152 if ((DUMMY_ENCRYPTION_ENABLED(sbi
) || ext4_has_feature_encrypt(sb
)) &&
4153 (blocksize
!= PAGE_SIZE
)) {
4154 ext4_msg(sb
, KERN_ERR
,
4155 "Unsupported blocksize for fs encryption");
4156 goto failed_mount_wq
;
4159 if (DUMMY_ENCRYPTION_ENABLED(sbi
) && !(sb
->s_flags
& MS_RDONLY
) &&
4160 !ext4_has_feature_encrypt(sb
)) {
4161 ext4_set_feature_encrypt(sb
);
4162 ext4_commit_super(sb
, 1);
4166 * Get the # of file system overhead blocks from the
4167 * superblock if present.
4169 if (es
->s_overhead_clusters
)
4170 sbi
->s_overhead
= le32_to_cpu(es
->s_overhead_clusters
);
4172 err
= ext4_calculate_overhead(sb
);
4174 goto failed_mount_wq
;
4178 * The maximum number of concurrent works can be high and
4179 * concurrency isn't really necessary. Limit it to 1.
4181 EXT4_SB(sb
)->rsv_conversion_wq
=
4182 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
4183 if (!EXT4_SB(sb
)->rsv_conversion_wq
) {
4184 printk(KERN_ERR
"EXT4-fs: failed to create workqueue\n");
4190 * The jbd2_journal_load will have done any necessary log recovery,
4191 * so we can safely mount the rest of the filesystem now.
4194 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
4196 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
4197 ret
= PTR_ERR(root
);
4201 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
4202 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
4206 sb
->s_root
= d_make_root(root
);
4208 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
4213 if (ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
))
4214 sb
->s_flags
|= MS_RDONLY
;
4216 /* determine the minimum size of new large inodes, if present */
4217 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
&&
4218 sbi
->s_want_extra_isize
== 0) {
4219 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4220 EXT4_GOOD_OLD_INODE_SIZE
;
4221 if (ext4_has_feature_extra_isize(sb
)) {
4222 if (sbi
->s_want_extra_isize
<
4223 le16_to_cpu(es
->s_want_extra_isize
))
4224 sbi
->s_want_extra_isize
=
4225 le16_to_cpu(es
->s_want_extra_isize
);
4226 if (sbi
->s_want_extra_isize
<
4227 le16_to_cpu(es
->s_min_extra_isize
))
4228 sbi
->s_want_extra_isize
=
4229 le16_to_cpu(es
->s_min_extra_isize
);
4232 /* Check if enough inode space is available */
4233 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
4234 sbi
->s_inode_size
) {
4235 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
4236 EXT4_GOOD_OLD_INODE_SIZE
;
4237 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
4241 ext4_set_resv_clusters(sb
);
4243 err
= ext4_setup_system_zone(sb
);
4245 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
4247 goto failed_mount4a
;
4251 err
= ext4_mb_init(sb
);
4253 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
4258 block
= ext4_count_free_clusters(sb
);
4259 ext4_free_blocks_count_set(sbi
->s_es
,
4260 EXT4_C2B(sbi
, block
));
4261 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
, block
,
4264 unsigned long freei
= ext4_count_free_inodes(sb
);
4265 sbi
->s_es
->s_free_inodes_count
= cpu_to_le32(freei
);
4266 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
, freei
,
4270 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
4271 ext4_count_dirs(sb
), GFP_KERNEL
);
4273 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0,
4276 err
= percpu_init_rwsem(&sbi
->s_journal_flag_rwsem
);
4279 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
4283 if (ext4_has_feature_flex_bg(sb
))
4284 if (!ext4_fill_flex_info(sb
)) {
4285 ext4_msg(sb
, KERN_ERR
,
4286 "unable to initialize "
4287 "flex_bg meta info!");
4291 err
= ext4_register_li_request(sb
, first_not_zeroed
);
4295 err
= ext4_register_sysfs(sb
);
4300 /* Enable quota usage during mount. */
4301 if (ext4_has_feature_quota(sb
) && !(sb
->s_flags
& MS_RDONLY
)) {
4302 err
= ext4_enable_quotas(sb
);
4306 #endif /* CONFIG_QUOTA */
4308 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
4309 ext4_orphan_cleanup(sb
, es
);
4310 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
4311 if (needs_recovery
) {
4312 ext4_msg(sb
, KERN_INFO
, "recovery complete");
4313 ext4_mark_recovery_complete(sb
, es
);
4315 if (EXT4_SB(sb
)->s_journal
) {
4316 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
4317 descr
= " journalled data mode";
4318 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
4319 descr
= " ordered data mode";
4321 descr
= " writeback data mode";
4323 descr
= "out journal";
4325 if (test_opt(sb
, DISCARD
)) {
4326 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
4327 if (!blk_queue_discard(q
))
4328 ext4_msg(sb
, KERN_WARNING
,
4329 "mounting with \"discard\" option, but "
4330 "the device does not support discard");
4333 if (___ratelimit(&ext4_mount_msg_ratelimit
, "EXT4-fs mount"))
4334 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
4335 "Opts: %.*s%s%s", descr
,
4336 (int) sizeof(sbi
->s_es
->s_mount_opts
),
4337 sbi
->s_es
->s_mount_opts
,
4338 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
4340 if (es
->s_error_count
)
4341 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
4343 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4344 ratelimit_state_init(&sbi
->s_err_ratelimit_state
, 5 * HZ
, 10);
4345 ratelimit_state_init(&sbi
->s_warning_ratelimit_state
, 5 * HZ
, 10);
4346 ratelimit_state_init(&sbi
->s_msg_ratelimit_state
, 5 * HZ
, 10);
4353 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
4358 ext4_unregister_sysfs(sb
);
4361 ext4_unregister_li_request(sb
);
4363 ext4_mb_release(sb
);
4364 if (sbi
->s_flex_groups
)
4365 kvfree(sbi
->s_flex_groups
);
4366 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
4367 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
4368 percpu_counter_destroy(&sbi
->s_dirs_counter
);
4369 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
4371 ext4_ext_release(sb
);
4372 ext4_release_system_zone(sb
);
4377 ext4_msg(sb
, KERN_ERR
, "mount failed");
4378 if (EXT4_SB(sb
)->rsv_conversion_wq
)
4379 destroy_workqueue(EXT4_SB(sb
)->rsv_conversion_wq
);
4381 if (sbi
->s_ea_inode_cache
) {
4382 ext4_xattr_destroy_cache(sbi
->s_ea_inode_cache
);
4383 sbi
->s_ea_inode_cache
= NULL
;
4385 if (sbi
->s_ea_block_cache
) {
4386 ext4_xattr_destroy_cache(sbi
->s_ea_block_cache
);
4387 sbi
->s_ea_block_cache
= NULL
;
4389 if (sbi
->s_journal
) {
4390 jbd2_journal_destroy(sbi
->s_journal
);
4391 sbi
->s_journal
= NULL
;
4394 ext4_es_unregister_shrinker(sbi
);
4396 del_timer_sync(&sbi
->s_err_report
);
4398 kthread_stop(sbi
->s_mmp_tsk
);
4400 for (i
= 0; i
< db_count
; i
++)
4401 brelse(sbi
->s_group_desc
[i
]);
4402 kvfree(sbi
->s_group_desc
);
4404 if (sbi
->s_chksum_driver
)
4405 crypto_free_shash(sbi
->s_chksum_driver
);
4407 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4408 kfree(sbi
->s_qf_names
[i
]);
4410 ext4_blkdev_remove(sbi
);
4413 /* sb->s_user_ns will be put when sb is destroyed */
4414 sb
->s_fs_info
= NULL
;
4415 kfree(sbi
->s_blockgroup_lock
);
4419 return err
? err
: ret
;
4423 * Setup any per-fs journal parameters now. We'll do this both on
4424 * initial mount, once the journal has been initialised but before we've
4425 * done any recovery; and again on any subsequent remount.
4427 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
4429 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4431 journal
->j_commit_interval
= sbi
->s_commit_interval
;
4432 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
4433 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
4435 write_lock(&journal
->j_state_lock
);
4436 if (test_opt(sb
, BARRIER
))
4437 journal
->j_flags
|= JBD2_BARRIER
;
4439 journal
->j_flags
&= ~JBD2_BARRIER
;
4440 if (test_opt(sb
, DATA_ERR_ABORT
))
4441 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
4443 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
4444 write_unlock(&journal
->j_state_lock
);
4447 static struct inode
*ext4_get_journal_inode(struct super_block
*sb
,
4448 unsigned int journal_inum
)
4450 struct inode
*journal_inode
;
4453 * Test for the existence of a valid inode on disk. Bad things
4454 * happen if we iget() an unused inode, as the subsequent iput()
4455 * will try to delete it.
4457 journal_inode
= ext4_iget(sb
, journal_inum
);
4458 if (IS_ERR(journal_inode
)) {
4459 ext4_msg(sb
, KERN_ERR
, "no journal found");
4462 if (!journal_inode
->i_nlink
) {
4463 make_bad_inode(journal_inode
);
4464 iput(journal_inode
);
4465 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
4469 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4470 journal_inode
, journal_inode
->i_size
);
4471 if (!S_ISREG(journal_inode
->i_mode
)) {
4472 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
4473 iput(journal_inode
);
4476 return journal_inode
;
4479 static journal_t
*ext4_get_journal(struct super_block
*sb
,
4480 unsigned int journal_inum
)
4482 struct inode
*journal_inode
;
4485 BUG_ON(!ext4_has_feature_journal(sb
));
4487 journal_inode
= ext4_get_journal_inode(sb
, journal_inum
);
4491 journal
= jbd2_journal_init_inode(journal_inode
);
4493 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
4494 iput(journal_inode
);
4497 journal
->j_private
= sb
;
4498 ext4_init_journal_params(sb
, journal
);
4502 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
4505 struct buffer_head
*bh
;
4509 int hblock
, blocksize
;
4510 ext4_fsblk_t sb_block
;
4511 unsigned long offset
;
4512 struct ext4_super_block
*es
;
4513 struct block_device
*bdev
;
4515 BUG_ON(!ext4_has_feature_journal(sb
));
4517 bdev
= ext4_blkdev_get(j_dev
, sb
);
4521 blocksize
= sb
->s_blocksize
;
4522 hblock
= bdev_logical_block_size(bdev
);
4523 if (blocksize
< hblock
) {
4524 ext4_msg(sb
, KERN_ERR
,
4525 "blocksize too small for journal device");
4529 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
4530 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
4531 set_blocksize(bdev
, blocksize
);
4532 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
4533 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
4534 "external journal");
4538 es
= (struct ext4_super_block
*) (bh
->b_data
+ offset
);
4539 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
4540 !(le32_to_cpu(es
->s_feature_incompat
) &
4541 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
4542 ext4_msg(sb
, KERN_ERR
, "external journal has "
4548 if ((le32_to_cpu(es
->s_feature_ro_compat
) &
4549 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM
) &&
4550 es
->s_checksum
!= ext4_superblock_csum(sb
, es
)) {
4551 ext4_msg(sb
, KERN_ERR
, "external journal has "
4552 "corrupt superblock");
4557 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
4558 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
4563 len
= ext4_blocks_count(es
);
4564 start
= sb_block
+ 1;
4565 brelse(bh
); /* we're done with the superblock */
4567 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4568 start
, len
, blocksize
);
4570 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4573 journal
->j_private
= sb
;
4574 ll_rw_block(REQ_OP_READ
, REQ_META
| REQ_PRIO
, 1, &journal
->j_sb_buffer
);
4575 wait_on_buffer(journal
->j_sb_buffer
);
4576 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4577 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4580 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4581 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4582 "user (unsupported) - %d",
4583 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4586 EXT4_SB(sb
)->journal_bdev
= bdev
;
4587 ext4_init_journal_params(sb
, journal
);
4591 jbd2_journal_destroy(journal
);
4593 ext4_blkdev_put(bdev
);
4597 static int ext4_load_journal(struct super_block
*sb
,
4598 struct ext4_super_block
*es
,
4599 unsigned long journal_devnum
)
4602 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4605 int really_read_only
;
4607 BUG_ON(!ext4_has_feature_journal(sb
));
4609 if (journal_devnum
&&
4610 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4611 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4612 "numbers have changed");
4613 journal_dev
= new_decode_dev(journal_devnum
);
4615 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4617 really_read_only
= bdev_read_only(sb
->s_bdev
);
4620 * Are we loading a blank journal or performing recovery after a
4621 * crash? For recovery, we need to check in advance whether we
4622 * can get read-write access to the device.
4624 if (ext4_has_feature_journal_needs_recovery(sb
)) {
4625 if (sb
->s_flags
& MS_RDONLY
) {
4626 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4627 "required on readonly filesystem");
4628 if (really_read_only
) {
4629 ext4_msg(sb
, KERN_ERR
, "write access "
4630 "unavailable, cannot proceed");
4633 ext4_msg(sb
, KERN_INFO
, "write access will "
4634 "be enabled during recovery");
4638 if (journal_inum
&& journal_dev
) {
4639 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4640 "and inode journals!");
4645 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4648 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4652 if (!(journal
->j_flags
& JBD2_BARRIER
))
4653 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4655 if (!ext4_has_feature_journal_needs_recovery(sb
))
4656 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4658 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4660 memcpy(save
, ((char *) es
) +
4661 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4662 err
= jbd2_journal_load(journal
);
4664 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4665 save
, EXT4_S_ERR_LEN
);
4670 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4671 jbd2_journal_destroy(journal
);
4675 EXT4_SB(sb
)->s_journal
= journal
;
4676 ext4_clear_journal_err(sb
, es
);
4678 if (!really_read_only
&& journal_devnum
&&
4679 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4680 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4682 /* Make sure we flush the recovery flag to disk. */
4683 ext4_commit_super(sb
, 1);
4689 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4691 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4692 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4695 if (!sbh
|| block_device_ejected(sb
))
4698 * If the file system is mounted read-only, don't update the
4699 * superblock write time. This avoids updating the superblock
4700 * write time when we are mounting the root file system
4701 * read/only but we need to replay the journal; at that point,
4702 * for people who are east of GMT and who make their clock
4703 * tick in localtime for Windows bug-for-bug compatibility,
4704 * the clock is set in the future, and this will cause e2fsck
4705 * to complain and force a full file system check.
4707 if (!(sb
->s_flags
& MS_RDONLY
))
4708 es
->s_wtime
= cpu_to_le32(get_seconds());
4709 if (sb
->s_bdev
->bd_part
)
4710 es
->s_kbytes_written
=
4711 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4712 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4713 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4715 es
->s_kbytes_written
=
4716 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4717 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeclusters_counter
))
4718 ext4_free_blocks_count_set(es
,
4719 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4720 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4721 if (percpu_counter_initialized(&EXT4_SB(sb
)->s_freeinodes_counter
))
4722 es
->s_free_inodes_count
=
4723 cpu_to_le32(percpu_counter_sum_positive(
4724 &EXT4_SB(sb
)->s_freeinodes_counter
));
4725 BUFFER_TRACE(sbh
, "marking dirty");
4726 ext4_superblock_csum_set(sb
);
4729 if (buffer_write_io_error(sbh
)) {
4731 * Oh, dear. A previous attempt to write the
4732 * superblock failed. This could happen because the
4733 * USB device was yanked out. Or it could happen to
4734 * be a transient write error and maybe the block will
4735 * be remapped. Nothing we can do but to retry the
4736 * write and hope for the best.
4738 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4739 "superblock detected");
4740 clear_buffer_write_io_error(sbh
);
4741 set_buffer_uptodate(sbh
);
4743 mark_buffer_dirty(sbh
);
4746 error
= __sync_dirty_buffer(sbh
,
4747 REQ_SYNC
| (test_opt(sb
, BARRIER
) ? REQ_FUA
: 0));
4751 error
= buffer_write_io_error(sbh
);
4753 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4755 clear_buffer_write_io_error(sbh
);
4756 set_buffer_uptodate(sbh
);
4763 * Have we just finished recovery? If so, and if we are mounting (or
4764 * remounting) the filesystem readonly, then we will end up with a
4765 * consistent fs on disk. Record that fact.
4767 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4768 struct ext4_super_block
*es
)
4770 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4772 if (!ext4_has_feature_journal(sb
)) {
4773 BUG_ON(journal
!= NULL
);
4776 jbd2_journal_lock_updates(journal
);
4777 if (jbd2_journal_flush(journal
) < 0)
4780 if (ext4_has_feature_journal_needs_recovery(sb
) &&
4781 sb
->s_flags
& MS_RDONLY
) {
4782 ext4_clear_feature_journal_needs_recovery(sb
);
4783 ext4_commit_super(sb
, 1);
4787 jbd2_journal_unlock_updates(journal
);
4791 * If we are mounting (or read-write remounting) a filesystem whose journal
4792 * has recorded an error from a previous lifetime, move that error to the
4793 * main filesystem now.
4795 static void ext4_clear_journal_err(struct super_block
*sb
,
4796 struct ext4_super_block
*es
)
4802 BUG_ON(!ext4_has_feature_journal(sb
));
4804 journal
= EXT4_SB(sb
)->s_journal
;
4807 * Now check for any error status which may have been recorded in the
4808 * journal by a prior ext4_error() or ext4_abort()
4811 j_errno
= jbd2_journal_errno(journal
);
4815 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4816 ext4_warning(sb
, "Filesystem error recorded "
4817 "from previous mount: %s", errstr
);
4818 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4820 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4821 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4822 ext4_commit_super(sb
, 1);
4824 jbd2_journal_clear_err(journal
);
4825 jbd2_journal_update_sb_errno(journal
);
4830 * Force the running and committing transactions to commit,
4831 * and wait on the commit.
4833 int ext4_force_commit(struct super_block
*sb
)
4837 if (sb
->s_flags
& MS_RDONLY
)
4840 journal
= EXT4_SB(sb
)->s_journal
;
4841 return ext4_journal_force_commit(journal
);
4844 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4848 bool needs_barrier
= false;
4849 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4851 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb
))))
4854 trace_ext4_sync_fs(sb
, wait
);
4855 flush_workqueue(sbi
->rsv_conversion_wq
);
4857 * Writeback quota in non-journalled quota case - journalled quota has
4860 dquot_writeback_dquots(sb
, -1);
4862 * Data writeback is possible w/o journal transaction, so barrier must
4863 * being sent at the end of the function. But we can skip it if
4864 * transaction_commit will do it for us.
4866 if (sbi
->s_journal
) {
4867 target
= jbd2_get_latest_transaction(sbi
->s_journal
);
4868 if (wait
&& sbi
->s_journal
->j_flags
& JBD2_BARRIER
&&
4869 !jbd2_trans_will_send_data_barrier(sbi
->s_journal
, target
))
4870 needs_barrier
= true;
4872 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4874 ret
= jbd2_log_wait_commit(sbi
->s_journal
,
4877 } else if (wait
&& test_opt(sb
, BARRIER
))
4878 needs_barrier
= true;
4879 if (needs_barrier
) {
4881 err
= blkdev_issue_flush(sb
->s_bdev
, GFP_KERNEL
, NULL
);
4890 * LVM calls this function before a (read-only) snapshot is created. This
4891 * gives us a chance to flush the journal completely and mark the fs clean.
4893 * Note that only this function cannot bring a filesystem to be in a clean
4894 * state independently. It relies on upper layer to stop all data & metadata
4897 static int ext4_freeze(struct super_block
*sb
)
4902 if (sb
->s_flags
& MS_RDONLY
)
4905 journal
= EXT4_SB(sb
)->s_journal
;
4908 /* Now we set up the journal barrier. */
4909 jbd2_journal_lock_updates(journal
);
4912 * Don't clear the needs_recovery flag if we failed to
4913 * flush the journal.
4915 error
= jbd2_journal_flush(journal
);
4919 /* Journal blocked and flushed, clear needs_recovery flag. */
4920 ext4_clear_feature_journal_needs_recovery(sb
);
4923 error
= ext4_commit_super(sb
, 1);
4926 /* we rely on upper layer to stop further updates */
4927 jbd2_journal_unlock_updates(journal
);
4932 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4933 * flag here, even though the filesystem is not technically dirty yet.
4935 static int ext4_unfreeze(struct super_block
*sb
)
4937 if ((sb
->s_flags
& MS_RDONLY
) || ext4_forced_shutdown(EXT4_SB(sb
)))
4940 if (EXT4_SB(sb
)->s_journal
) {
4941 /* Reset the needs_recovery flag before the fs is unlocked. */
4942 ext4_set_feature_journal_needs_recovery(sb
);
4945 ext4_commit_super(sb
, 1);
4950 * Structure to save mount options for ext4_remount's benefit
4952 struct ext4_mount_options
{
4953 unsigned long s_mount_opt
;
4954 unsigned long s_mount_opt2
;
4957 unsigned long s_commit_interval
;
4958 u32 s_min_batch_time
, s_max_batch_time
;
4961 char *s_qf_names
[EXT4_MAXQUOTAS
];
4965 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4967 struct ext4_super_block
*es
;
4968 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4969 unsigned long old_sb_flags
;
4970 struct ext4_mount_options old_opts
;
4971 int enable_quota
= 0;
4973 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4978 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4980 /* Store the original options */
4981 old_sb_flags
= sb
->s_flags
;
4982 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4983 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4984 old_opts
.s_resuid
= sbi
->s_resuid
;
4985 old_opts
.s_resgid
= sbi
->s_resgid
;
4986 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4987 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4988 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4990 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4991 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
4992 if (sbi
->s_qf_names
[i
]) {
4993 old_opts
.s_qf_names
[i
] = kstrdup(sbi
->s_qf_names
[i
],
4995 if (!old_opts
.s_qf_names
[i
]) {
4996 for (j
= 0; j
< i
; j
++)
4997 kfree(old_opts
.s_qf_names
[j
]);
5002 old_opts
.s_qf_names
[i
] = NULL
;
5004 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
5005 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
5007 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
5012 if ((old_opts
.s_mount_opt
& EXT4_MOUNT_JOURNAL_CHECKSUM
) ^
5013 test_opt(sb
, JOURNAL_CHECKSUM
)) {
5014 ext4_msg(sb
, KERN_ERR
, "changing journal_checksum "
5015 "during remount not supported; ignoring");
5016 sbi
->s_mount_opt
^= EXT4_MOUNT_JOURNAL_CHECKSUM
;
5019 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
5020 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
5021 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5022 "both data=journal and delalloc");
5026 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
5027 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5028 "both data=journal and dioread_nolock");
5032 if (test_opt(sb
, DAX
)) {
5033 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5034 "both data=journal and dax");
5038 } else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
) {
5039 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
5040 ext4_msg(sb
, KERN_ERR
, "can't mount with "
5041 "journal_async_commit in data=ordered mode");
5047 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_NO_MBCACHE
) {
5048 ext4_msg(sb
, KERN_ERR
, "can't enable nombcache during remount");
5053 if ((sbi
->s_mount_opt
^ old_opts
.s_mount_opt
) & EXT4_MOUNT_DAX
) {
5054 ext4_msg(sb
, KERN_WARNING
, "warning: refusing change of "
5055 "dax flag with busy inodes while remounting");
5056 sbi
->s_mount_opt
^= EXT4_MOUNT_DAX
;
5059 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
5060 ext4_abort(sb
, "Abort forced by user");
5062 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
5063 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
5067 if (sbi
->s_journal
) {
5068 ext4_init_journal_params(sb
, sbi
->s_journal
);
5069 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
5072 if (*flags
& MS_LAZYTIME
)
5073 sb
->s_flags
|= MS_LAZYTIME
;
5075 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
5076 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
5081 if (*flags
& MS_RDONLY
) {
5082 err
= sync_filesystem(sb
);
5085 err
= dquot_suspend(sb
, -1);
5090 * First of all, the unconditional stuff we have to do
5091 * to disable replay of the journal when we next remount
5093 sb
->s_flags
|= MS_RDONLY
;
5096 * OK, test if we are remounting a valid rw partition
5097 * readonly, and if so set the rdonly flag and then
5098 * mark the partition as valid again.
5100 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
5101 (sbi
->s_mount_state
& EXT4_VALID_FS
))
5102 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
5105 ext4_mark_recovery_complete(sb
, es
);
5107 /* Make sure we can mount this feature set readwrite */
5108 if (ext4_has_feature_readonly(sb
) ||
5109 !ext4_feature_set_ok(sb
, 0)) {
5114 * Make sure the group descriptor checksums
5115 * are sane. If they aren't, refuse to remount r/w.
5117 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
5118 struct ext4_group_desc
*gdp
=
5119 ext4_get_group_desc(sb
, g
, NULL
);
5121 if (!ext4_group_desc_csum_verify(sb
, g
, gdp
)) {
5122 ext4_msg(sb
, KERN_ERR
,
5123 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5124 g
, le16_to_cpu(ext4_group_desc_csum(sb
, g
, gdp
)),
5125 le16_to_cpu(gdp
->bg_checksum
));
5132 * If we have an unprocessed orphan list hanging
5133 * around from a previously readonly bdev mount,
5134 * require a full umount/remount for now.
5136 if (es
->s_last_orphan
) {
5137 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
5138 "remount RDWR because of unprocessed "
5139 "orphan inode list. Please "
5140 "umount/remount instead");
5146 * Mounting a RDONLY partition read-write, so reread
5147 * and store the current valid flag. (It may have
5148 * been changed by e2fsck since we originally mounted
5152 ext4_clear_journal_err(sb
, es
);
5153 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
5154 if (!ext4_setup_super(sb
, es
, 0))
5155 sb
->s_flags
&= ~MS_RDONLY
;
5156 if (ext4_has_feature_mmp(sb
))
5157 if (ext4_multi_mount_protect(sb
,
5158 le64_to_cpu(es
->s_mmp_block
))) {
5167 * Reinitialize lazy itable initialization thread based on
5170 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
5171 ext4_unregister_li_request(sb
);
5173 ext4_group_t first_not_zeroed
;
5174 first_not_zeroed
= ext4_has_uninit_itable(sb
);
5175 ext4_register_li_request(sb
, first_not_zeroed
);
5178 ext4_setup_system_zone(sb
);
5179 if (sbi
->s_journal
== NULL
&& !(old_sb_flags
& MS_RDONLY
))
5180 ext4_commit_super(sb
, 1);
5183 /* Release old quota file names */
5184 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++)
5185 kfree(old_opts
.s_qf_names
[i
]);
5187 if (sb_any_quota_suspended(sb
))
5188 dquot_resume(sb
, -1);
5189 else if (ext4_has_feature_quota(sb
)) {
5190 err
= ext4_enable_quotas(sb
);
5197 *flags
= (*flags
& ~MS_LAZYTIME
) | (sb
->s_flags
& MS_LAZYTIME
);
5198 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
5203 sb
->s_flags
= old_sb_flags
;
5204 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
5205 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
5206 sbi
->s_resuid
= old_opts
.s_resuid
;
5207 sbi
->s_resgid
= old_opts
.s_resgid
;
5208 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
5209 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
5210 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
5212 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
5213 for (i
= 0; i
< EXT4_MAXQUOTAS
; i
++) {
5214 kfree(sbi
->s_qf_names
[i
]);
5215 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
5223 static int ext4_statfs_project(struct super_block
*sb
,
5224 kprojid_t projid
, struct kstatfs
*buf
)
5227 struct dquot
*dquot
;
5231 qid
= make_kqid_projid(projid
);
5232 dquot
= dqget(sb
, qid
);
5234 return PTR_ERR(dquot
);
5235 spin_lock(&dq_data_lock
);
5237 limit
= (dquot
->dq_dqb
.dqb_bsoftlimit
?
5238 dquot
->dq_dqb
.dqb_bsoftlimit
:
5239 dquot
->dq_dqb
.dqb_bhardlimit
) >> sb
->s_blocksize_bits
;
5240 if (limit
&& buf
->f_blocks
> limit
) {
5241 curblock
= dquot
->dq_dqb
.dqb_curspace
>> sb
->s_blocksize_bits
;
5242 buf
->f_blocks
= limit
;
5243 buf
->f_bfree
= buf
->f_bavail
=
5244 (buf
->f_blocks
> curblock
) ?
5245 (buf
->f_blocks
- curblock
) : 0;
5248 limit
= dquot
->dq_dqb
.dqb_isoftlimit
?
5249 dquot
->dq_dqb
.dqb_isoftlimit
:
5250 dquot
->dq_dqb
.dqb_ihardlimit
;
5251 if (limit
&& buf
->f_files
> limit
) {
5252 buf
->f_files
= limit
;
5254 (buf
->f_files
> dquot
->dq_dqb
.dqb_curinodes
) ?
5255 (buf
->f_files
- dquot
->dq_dqb
.dqb_curinodes
) : 0;
5258 spin_unlock(&dq_data_lock
);
5264 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
5266 struct super_block
*sb
= dentry
->d_sb
;
5267 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5268 struct ext4_super_block
*es
= sbi
->s_es
;
5269 ext4_fsblk_t overhead
= 0, resv_blocks
;
5272 resv_blocks
= EXT4_C2B(sbi
, atomic64_read(&sbi
->s_resv_clusters
));
5274 if (!test_opt(sb
, MINIX_DF
))
5275 overhead
= sbi
->s_overhead
;
5277 buf
->f_type
= EXT4_SUPER_MAGIC
;
5278 buf
->f_bsize
= sb
->s_blocksize
;
5279 buf
->f_blocks
= ext4_blocks_count(es
) - EXT4_C2B(sbi
, overhead
);
5280 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
5281 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
5282 /* prevent underflow in case that few free space is available */
5283 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
5284 buf
->f_bavail
= buf
->f_bfree
-
5285 (ext4_r_blocks_count(es
) + resv_blocks
);
5286 if (buf
->f_bfree
< (ext4_r_blocks_count(es
) + resv_blocks
))
5288 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
5289 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
5290 buf
->f_namelen
= EXT4_NAME_LEN
;
5291 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
5292 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
5293 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
5294 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
5297 if (ext4_test_inode_flag(dentry
->d_inode
, EXT4_INODE_PROJINHERIT
) &&
5298 sb_has_quota_limits_enabled(sb
, PRJQUOTA
))
5299 ext4_statfs_project(sb
, EXT4_I(dentry
->d_inode
)->i_projid
, buf
);
5304 /* Helper function for writing quotas on sync - we need to start transaction
5305 * before quota file is locked for write. Otherwise the are possible deadlocks:
5306 * Process 1 Process 2
5307 * ext4_create() quota_sync()
5308 * jbd2_journal_start() write_dquot()
5309 * dquot_initialize() down(dqio_mutex)
5310 * down(dqio_mutex) jbd2_journal_start()
5316 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
5318 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_id
.type
];
5321 static int ext4_write_dquot(struct dquot
*dquot
)
5325 struct inode
*inode
;
5327 inode
= dquot_to_inode(dquot
);
5328 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
,
5329 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
5331 return PTR_ERR(handle
);
5332 ret
= dquot_commit(dquot
);
5333 err
= ext4_journal_stop(handle
);
5339 static int ext4_acquire_dquot(struct dquot
*dquot
)
5344 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5345 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
5347 return PTR_ERR(handle
);
5348 ret
= dquot_acquire(dquot
);
5349 err
= ext4_journal_stop(handle
);
5355 static int ext4_release_dquot(struct dquot
*dquot
)
5360 handle
= ext4_journal_start(dquot_to_inode(dquot
), EXT4_HT_QUOTA
,
5361 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
5362 if (IS_ERR(handle
)) {
5363 /* Release dquot anyway to avoid endless cycle in dqput() */
5364 dquot_release(dquot
);
5365 return PTR_ERR(handle
);
5367 ret
= dquot_release(dquot
);
5368 err
= ext4_journal_stop(handle
);
5374 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
5376 struct super_block
*sb
= dquot
->dq_sb
;
5377 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
5379 /* Are we journaling quotas? */
5380 if (ext4_has_feature_quota(sb
) ||
5381 sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
5382 dquot_mark_dquot_dirty(dquot
);
5383 return ext4_write_dquot(dquot
);
5385 return dquot_mark_dquot_dirty(dquot
);
5389 static int ext4_write_info(struct super_block
*sb
, int type
)
5394 /* Data block + inode block */
5395 handle
= ext4_journal_start(d_inode(sb
->s_root
), EXT4_HT_QUOTA
, 2);
5397 return PTR_ERR(handle
);
5398 ret
= dquot_commit_info(sb
, type
);
5399 err
= ext4_journal_stop(handle
);
5406 * Turn on quotas during mount time - we need to find
5407 * the quota file and such...
5409 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
5411 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
5412 EXT4_SB(sb
)->s_jquota_fmt
, type
);
5415 static void lockdep_set_quota_inode(struct inode
*inode
, int subclass
)
5417 struct ext4_inode_info
*ei
= EXT4_I(inode
);
5419 /* The first argument of lockdep_set_subclass has to be
5420 * *exactly* the same as the argument to init_rwsem() --- in
5421 * this case, in init_once() --- or lockdep gets unhappy
5422 * because the name of the lock is set using the
5423 * stringification of the argument to init_rwsem().
5425 (void) ei
; /* shut up clang warning if !CONFIG_LOCKDEP */
5426 lockdep_set_subclass(&ei
->i_data_sem
, subclass
);
5430 * Standard function to be called on quota_on
5432 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
5433 const struct path
*path
)
5437 if (!test_opt(sb
, QUOTA
))
5440 /* Quotafile not on the same filesystem? */
5441 if (path
->dentry
->d_sb
!= sb
)
5443 /* Journaling quota? */
5444 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
5445 /* Quotafile not in fs root? */
5446 if (path
->dentry
->d_parent
!= sb
->s_root
)
5447 ext4_msg(sb
, KERN_WARNING
,
5448 "Quota file not on filesystem root. "
5449 "Journaled quota will not work");
5453 * When we journal data on quota file, we have to flush journal to see
5454 * all updates to the file when we bypass pagecache...
5456 if (EXT4_SB(sb
)->s_journal
&&
5457 ext4_should_journal_data(d_inode(path
->dentry
))) {
5459 * We don't need to lock updates but journal_flush() could
5460 * otherwise be livelocked...
5462 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
5463 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
5464 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
5469 lockdep_set_quota_inode(path
->dentry
->d_inode
, I_DATA_SEM_QUOTA
);
5470 err
= dquot_quota_on(sb
, type
, format_id
, path
);
5472 lockdep_set_quota_inode(path
->dentry
->d_inode
,
5475 struct inode
*inode
= d_inode(path
->dentry
);
5479 * Set inode flags to prevent userspace from messing with quota
5480 * files. If this fails, we return success anyway since quotas
5481 * are already enabled and this is not a hard failure.
5484 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5487 EXT4_I(inode
)->i_flags
|= EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
;
5488 inode_set_flags(inode
, S_NOATIME
| S_IMMUTABLE
,
5489 S_NOATIME
| S_IMMUTABLE
);
5490 ext4_mark_inode_dirty(handle
, inode
);
5491 ext4_journal_stop(handle
);
5493 inode_unlock(inode
);
5498 static int ext4_quota_enable(struct super_block
*sb
, int type
, int format_id
,
5502 struct inode
*qf_inode
;
5503 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5504 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5505 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5506 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5509 BUG_ON(!ext4_has_feature_quota(sb
));
5511 if (!qf_inums
[type
])
5514 qf_inode
= ext4_iget(sb
, qf_inums
[type
]);
5515 if (IS_ERR(qf_inode
)) {
5516 ext4_error(sb
, "Bad quota inode # %lu", qf_inums
[type
]);
5517 return PTR_ERR(qf_inode
);
5520 /* Don't account quota for quota files to avoid recursion */
5521 qf_inode
->i_flags
|= S_NOQUOTA
;
5522 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_QUOTA
);
5523 err
= dquot_enable(qf_inode
, type
, format_id
, flags
);
5526 lockdep_set_quota_inode(qf_inode
, I_DATA_SEM_NORMAL
);
5531 /* Enable usage tracking for all quota types. */
5532 static int ext4_enable_quotas(struct super_block
*sb
)
5535 unsigned long qf_inums
[EXT4_MAXQUOTAS
] = {
5536 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_usr_quota_inum
),
5537 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_grp_quota_inum
),
5538 le32_to_cpu(EXT4_SB(sb
)->s_es
->s_prj_quota_inum
)
5540 bool quota_mopt
[EXT4_MAXQUOTAS
] = {
5541 test_opt(sb
, USRQUOTA
),
5542 test_opt(sb
, GRPQUOTA
),
5543 test_opt(sb
, PRJQUOTA
),
5546 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
5547 for (type
= 0; type
< EXT4_MAXQUOTAS
; type
++) {
5548 if (qf_inums
[type
]) {
5549 err
= ext4_quota_enable(sb
, type
, QFMT_VFS_V1
,
5550 DQUOT_USAGE_ENABLED
|
5551 (quota_mopt
[type
] ? DQUOT_LIMITS_ENABLED
: 0));
5554 "Failed to enable quota tracking "
5555 "(type=%d, err=%d). Please run "
5556 "e2fsck to fix.", type
, err
);
5564 static int ext4_quota_off(struct super_block
*sb
, int type
)
5566 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5570 /* Force all delayed allocation blocks to be allocated.
5571 * Caller already holds s_umount sem */
5572 if (test_opt(sb
, DELALLOC
))
5573 sync_filesystem(sb
);
5575 if (!inode
|| !igrab(inode
))
5578 err
= dquot_quota_off(sb
, type
);
5579 if (err
|| ext4_has_feature_quota(sb
))
5584 * Update modification times of quota files when userspace can
5585 * start looking at them. If we fail, we return success anyway since
5586 * this is not a hard failure and quotas are already disabled.
5588 handle
= ext4_journal_start(inode
, EXT4_HT_QUOTA
, 1);
5591 EXT4_I(inode
)->i_flags
&= ~(EXT4_NOATIME_FL
| EXT4_IMMUTABLE_FL
);
5592 inode_set_flags(inode
, 0, S_NOATIME
| S_IMMUTABLE
);
5593 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
5594 ext4_mark_inode_dirty(handle
, inode
);
5595 ext4_journal_stop(handle
);
5597 inode_unlock(inode
);
5599 lockdep_set_quota_inode(inode
, I_DATA_SEM_NORMAL
);
5603 return dquot_quota_off(sb
, type
);
5606 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5607 * acquiring the locks... As quota files are never truncated and quota code
5608 * itself serializes the operations (and no one else should touch the files)
5609 * we don't have to be afraid of races */
5610 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
5611 size_t len
, loff_t off
)
5613 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5614 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5615 int offset
= off
& (sb
->s_blocksize
- 1);
5618 struct buffer_head
*bh
;
5619 loff_t i_size
= i_size_read(inode
);
5623 if (off
+len
> i_size
)
5626 while (toread
> 0) {
5627 tocopy
= sb
->s_blocksize
- offset
< toread
?
5628 sb
->s_blocksize
- offset
: toread
;
5629 bh
= ext4_bread(NULL
, inode
, blk
, 0);
5632 if (!bh
) /* A hole? */
5633 memset(data
, 0, tocopy
);
5635 memcpy(data
, bh
->b_data
+offset
, tocopy
);
5645 /* Write to quotafile (we know the transaction is already started and has
5646 * enough credits) */
5647 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
5648 const char *data
, size_t len
, loff_t off
)
5650 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
5651 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
5652 int err
, offset
= off
& (sb
->s_blocksize
- 1);
5654 struct buffer_head
*bh
;
5655 handle_t
*handle
= journal_current_handle();
5657 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
5658 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5659 " cancelled because transaction is not started",
5660 (unsigned long long)off
, (unsigned long long)len
);
5664 * Since we account only one data block in transaction credits,
5665 * then it is impossible to cross a block boundary.
5667 if (sb
->s_blocksize
- offset
< len
) {
5668 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
5669 " cancelled because not block aligned",
5670 (unsigned long long)off
, (unsigned long long)len
);
5675 bh
= ext4_bread(handle
, inode
, blk
,
5676 EXT4_GET_BLOCKS_CREATE
|
5677 EXT4_GET_BLOCKS_METADATA_NOFAIL
);
5678 } while (IS_ERR(bh
) && (PTR_ERR(bh
) == -ENOSPC
) &&
5679 ext4_should_retry_alloc(inode
->i_sb
, &retries
));
5684 BUFFER_TRACE(bh
, "get write access");
5685 err
= ext4_journal_get_write_access(handle
, bh
);
5691 memcpy(bh
->b_data
+offset
, data
, len
);
5692 flush_dcache_page(bh
->b_page
);
5694 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
5697 if (inode
->i_size
< off
+ len
) {
5698 i_size_write(inode
, off
+ len
);
5699 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
5700 ext4_mark_inode_dirty(handle
, inode
);
5705 static int ext4_get_next_id(struct super_block
*sb
, struct kqid
*qid
)
5707 const struct quota_format_ops
*ops
;
5709 if (!sb_has_quota_loaded(sb
, qid
->type
))
5711 ops
= sb_dqopt(sb
)->ops
[qid
->type
];
5712 if (!ops
|| !ops
->get_next_id
)
5714 return dquot_get_next_id(sb
, qid
);
5718 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
5719 const char *dev_name
, void *data
)
5721 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
5724 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5725 static inline void register_as_ext2(void)
5727 int err
= register_filesystem(&ext2_fs_type
);
5730 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
5733 static inline void unregister_as_ext2(void)
5735 unregister_filesystem(&ext2_fs_type
);
5738 static inline int ext2_feature_set_ok(struct super_block
*sb
)
5740 if (ext4_has_unknown_ext2_incompat_features(sb
))
5742 if (sb
->s_flags
& MS_RDONLY
)
5744 if (ext4_has_unknown_ext2_ro_compat_features(sb
))
5749 static inline void register_as_ext2(void) { }
5750 static inline void unregister_as_ext2(void) { }
5751 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
5754 static inline void register_as_ext3(void)
5756 int err
= register_filesystem(&ext3_fs_type
);
5759 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
5762 static inline void unregister_as_ext3(void)
5764 unregister_filesystem(&ext3_fs_type
);
5767 static inline int ext3_feature_set_ok(struct super_block
*sb
)
5769 if (ext4_has_unknown_ext3_incompat_features(sb
))
5771 if (!ext4_has_feature_journal(sb
))
5773 if (sb
->s_flags
& MS_RDONLY
)
5775 if (ext4_has_unknown_ext3_ro_compat_features(sb
))
5780 static struct file_system_type ext4_fs_type
= {
5781 .owner
= THIS_MODULE
,
5783 .mount
= ext4_mount
,
5784 .kill_sb
= kill_block_super
,
5785 .fs_flags
= FS_REQUIRES_DEV
| FS_USERNS_MOUNT
,
5787 MODULE_ALIAS_FS("ext4");
5789 /* Shared across all ext4 file systems */
5790 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5792 static int __init
ext4_init_fs(void)
5796 ratelimit_state_init(&ext4_mount_msg_ratelimit
, 30 * HZ
, 64);
5797 ext4_li_info
= NULL
;
5798 mutex_init(&ext4_li_mtx
);
5800 /* Build-time check for flags consistency */
5801 ext4_check_flag_values();
5803 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++)
5804 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5806 err
= ext4_init_es();
5810 err
= ext4_init_pageio();
5814 err
= ext4_init_system_zone();
5818 err
= ext4_init_sysfs();
5822 err
= ext4_init_mballoc();
5825 err
= init_inodecache();
5830 err
= register_filesystem(&ext4_fs_type
);
5836 unregister_as_ext2();
5837 unregister_as_ext3();
5838 destroy_inodecache();
5840 ext4_exit_mballoc();
5844 ext4_exit_system_zone();
5853 static void __exit
ext4_exit_fs(void)
5855 ext4_destroy_lazyinit_thread();
5856 unregister_as_ext2();
5857 unregister_as_ext3();
5858 unregister_filesystem(&ext4_fs_type
);
5859 destroy_inodecache();
5860 ext4_exit_mballoc();
5862 ext4_exit_system_zone();
5867 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5868 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5869 MODULE_LICENSE("GPL");
5870 module_init(ext4_init_fs
)
5871 module_exit(ext4_exit_fs
)