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/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.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/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry
*ext4_proc_root
;
57 static struct kset
*ext4_kset
;
58 static struct ext4_lazy_init
*ext4_li_info
;
59 static struct mutex ext4_li_mtx
;
60 static struct ext4_features
*ext4_feat
;
62 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
63 unsigned long journal_devnum
);
64 static int ext4_commit_super(struct super_block
*sb
, int sync
);
65 static void ext4_mark_recovery_complete(struct super_block
*sb
,
66 struct ext4_super_block
*es
);
67 static void ext4_clear_journal_err(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
70 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
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 void ext4_write_super(struct super_block
*sb
);
76 static int ext4_freeze(struct super_block
*sb
);
77 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
78 const char *dev_name
, void *data
);
79 static inline int ext2_feature_set_ok(struct super_block
*sb
);
80 static inline int ext3_feature_set_ok(struct super_block
*sb
);
81 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block
*sb
);
84 static void ext4_clear_request_list(void);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type
= {
91 .kill_sb
= kill_block_super
,
92 .fs_flags
= FS_REQUIRES_DEV
,
94 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #define IS_EXT2_SB(sb) (0)
100 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101 static struct file_system_type ext3_fs_type
= {
102 .owner
= THIS_MODULE
,
105 .kill_sb
= kill_block_super
,
106 .fs_flags
= FS_REQUIRES_DEV
,
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #define IS_EXT3_SB(sb) (0)
113 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
117 ret
= kmalloc(size
, flags
);
119 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
123 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
127 ret
= kzalloc(size
, flags
);
129 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
133 void ext4_kvfree(void *ptr
)
135 if (is_vmalloc_addr(ptr
))
142 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
143 struct ext4_group_desc
*bg
)
145 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
146 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
147 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
150 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
151 struct ext4_group_desc
*bg
)
153 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
154 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
155 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
158 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
159 struct ext4_group_desc
*bg
)
161 return le32_to_cpu(bg
->bg_inode_table_lo
) |
162 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
163 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
166 __u32
ext4_free_blks_count(struct super_block
*sb
,
167 struct ext4_group_desc
*bg
)
169 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
170 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
171 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
174 __u32
ext4_free_inodes_count(struct super_block
*sb
,
175 struct ext4_group_desc
*bg
)
177 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
178 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
179 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
182 __u32
ext4_used_dirs_count(struct super_block
*sb
,
183 struct ext4_group_desc
*bg
)
185 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
186 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
187 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
190 __u32
ext4_itable_unused_count(struct super_block
*sb
,
191 struct ext4_group_desc
*bg
)
193 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
194 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
195 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
198 void ext4_block_bitmap_set(struct super_block
*sb
,
199 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
201 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
202 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
203 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
206 void ext4_inode_bitmap_set(struct super_block
*sb
,
207 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
209 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
210 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
211 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
214 void ext4_inode_table_set(struct super_block
*sb
,
215 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
217 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
218 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
219 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
222 void ext4_free_blks_set(struct super_block
*sb
,
223 struct ext4_group_desc
*bg
, __u32 count
)
225 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
226 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
227 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
230 void ext4_free_inodes_set(struct super_block
*sb
,
231 struct ext4_group_desc
*bg
, __u32 count
)
233 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
234 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
235 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
238 void ext4_used_dirs_set(struct super_block
*sb
,
239 struct ext4_group_desc
*bg
, __u32 count
)
241 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
242 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
243 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
246 void ext4_itable_unused_set(struct super_block
*sb
,
247 struct ext4_group_desc
*bg
, __u32 count
)
249 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
250 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
251 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
255 /* Just increment the non-pointer handle value */
256 static handle_t
*ext4_get_nojournal(void)
258 handle_t
*handle
= current
->journal_info
;
259 unsigned long ref_cnt
= (unsigned long)handle
;
261 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
264 handle
= (handle_t
*)ref_cnt
;
266 current
->journal_info
= handle
;
271 /* Decrement the non-pointer handle value */
272 static void ext4_put_nojournal(handle_t
*handle
)
274 unsigned long ref_cnt
= (unsigned long)handle
;
276 BUG_ON(ref_cnt
== 0);
279 handle
= (handle_t
*)ref_cnt
;
281 current
->journal_info
= handle
;
285 * Wrappers for jbd2_journal_start/end.
287 * The only special thing we need to do here is to make sure that all
288 * journal_end calls result in the superblock being marked dirty, so
289 * that sync() will call the filesystem's write_super callback if
292 * To avoid j_barrier hold in userspace when a user calls freeze(),
293 * ext4 prevents a new handle from being started by s_frozen, which
294 * is in an upper layer.
296 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
301 trace_ext4_journal_start(sb
, nblocks
, _RET_IP_
);
302 if (sb
->s_flags
& MS_RDONLY
)
303 return ERR_PTR(-EROFS
);
305 journal
= EXT4_SB(sb
)->s_journal
;
306 handle
= ext4_journal_current_handle();
309 * If a handle has been started, it should be allowed to
310 * finish, otherwise deadlock could happen between freeze
311 * and others(e.g. truncate) due to the restart of the
312 * journal handle if the filesystem is forzen and active
313 * handles are not stopped.
316 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
319 return ext4_get_nojournal();
321 * Special case here: if the journal has aborted behind our
322 * backs (eg. EIO in the commit thread), then we still need to
323 * take the FS itself readonly cleanly.
325 if (is_journal_aborted(journal
)) {
326 ext4_abort(sb
, "Detected aborted journal");
327 return ERR_PTR(-EROFS
);
329 return jbd2_journal_start(journal
, nblocks
);
333 * The only special thing we need to do here is to make sure that all
334 * jbd2_journal_stop calls result in the superblock being marked dirty, so
335 * that sync() will call the filesystem's write_super callback if
338 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
340 struct super_block
*sb
;
344 if (!ext4_handle_valid(handle
)) {
345 ext4_put_nojournal(handle
);
348 sb
= handle
->h_transaction
->t_journal
->j_private
;
350 rc
= jbd2_journal_stop(handle
);
355 __ext4_std_error(sb
, where
, line
, err
);
359 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
360 const char *err_fn
, struct buffer_head
*bh
,
361 handle_t
*handle
, int err
)
364 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
366 BUG_ON(!ext4_handle_valid(handle
));
369 BUFFER_TRACE(bh
, "abort");
374 if (is_handle_aborted(handle
))
377 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
378 caller
, line
, errstr
, err_fn
);
380 jbd2_journal_abort_handle(handle
);
383 static void __save_error_info(struct super_block
*sb
, const char *func
,
386 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
388 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
389 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
390 es
->s_last_error_time
= cpu_to_le32(get_seconds());
391 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
392 es
->s_last_error_line
= cpu_to_le32(line
);
393 if (!es
->s_first_error_time
) {
394 es
->s_first_error_time
= es
->s_last_error_time
;
395 strncpy(es
->s_first_error_func
, func
,
396 sizeof(es
->s_first_error_func
));
397 es
->s_first_error_line
= cpu_to_le32(line
);
398 es
->s_first_error_ino
= es
->s_last_error_ino
;
399 es
->s_first_error_block
= es
->s_last_error_block
;
402 * Start the daily error reporting function if it hasn't been
405 if (!es
->s_error_count
)
406 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
407 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
410 static void save_error_info(struct super_block
*sb
, const char *func
,
413 __save_error_info(sb
, func
, line
);
414 ext4_commit_super(sb
, 1);
418 * The del_gendisk() function uninitializes the disk-specific data
419 * structures, including the bdi structure, without telling anyone
420 * else. Once this happens, any attempt to call mark_buffer_dirty()
421 * (for example, by ext4_commit_super), will cause a kernel OOPS.
422 * This is a kludge to prevent these oops until we can put in a proper
423 * hook in del_gendisk() to inform the VFS and file system layers.
425 static int block_device_ejected(struct super_block
*sb
)
427 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
428 struct backing_dev_info
*bdi
= bd_inode
->i_mapping
->backing_dev_info
;
430 return bdi
->dev
== NULL
;
434 /* Deal with the reporting of failure conditions on a filesystem such as
435 * inconsistencies detected or read IO failures.
437 * On ext2, we can store the error state of the filesystem in the
438 * superblock. That is not possible on ext4, because we may have other
439 * write ordering constraints on the superblock which prevent us from
440 * writing it out straight away; and given that the journal is about to
441 * be aborted, we can't rely on the current, or future, transactions to
442 * write out the superblock safely.
444 * We'll just use the jbd2_journal_abort() error code to record an error in
445 * the journal instead. On recovery, the journal will complain about
446 * that error until we've noted it down and cleared it.
449 static void ext4_handle_error(struct super_block
*sb
)
451 if (sb
->s_flags
& MS_RDONLY
)
454 if (!test_opt(sb
, ERRORS_CONT
)) {
455 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
457 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
459 jbd2_journal_abort(journal
, -EIO
);
461 if (test_opt(sb
, ERRORS_RO
)) {
462 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
463 sb
->s_flags
|= MS_RDONLY
;
465 if (test_opt(sb
, ERRORS_PANIC
))
466 panic("EXT4-fs (device %s): panic forced after error\n",
470 void __ext4_error(struct super_block
*sb
, const char *function
,
471 unsigned int line
, const char *fmt
, ...)
473 struct va_format vaf
;
479 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
480 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
483 ext4_handle_error(sb
);
486 void ext4_error_inode(struct inode
*inode
, const char *function
,
487 unsigned int line
, ext4_fsblk_t block
,
488 const char *fmt
, ...)
491 struct va_format vaf
;
492 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
494 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
495 es
->s_last_error_block
= cpu_to_le64(block
);
496 save_error_info(inode
->i_sb
, function
, line
);
500 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
501 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
503 printk(KERN_CONT
"block %llu: ", block
);
504 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
507 ext4_handle_error(inode
->i_sb
);
510 void ext4_error_file(struct file
*file
, const char *function
,
511 unsigned int line
, ext4_fsblk_t block
,
512 const char *fmt
, ...)
515 struct va_format vaf
;
516 struct ext4_super_block
*es
;
517 struct inode
*inode
= file
->f_dentry
->d_inode
;
518 char pathname
[80], *path
;
520 es
= EXT4_SB(inode
->i_sb
)->s_es
;
521 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
522 save_error_info(inode
->i_sb
, function
, line
);
523 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
527 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
528 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
530 printk(KERN_CONT
"block %llu: ", block
);
534 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
537 ext4_handle_error(inode
->i_sb
);
540 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
547 errstr
= "IO failure";
550 errstr
= "Out of memory";
553 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
554 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
555 errstr
= "Journal has aborted";
557 errstr
= "Readonly filesystem";
560 /* If the caller passed in an extra buffer for unknown
561 * errors, textualise them now. Else we just return
564 /* Check for truncated error codes... */
565 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
574 /* __ext4_std_error decodes expected errors from journaling functions
575 * automatically and invokes the appropriate error response. */
577 void __ext4_std_error(struct super_block
*sb
, const char *function
,
578 unsigned int line
, int errno
)
583 /* Special case: if the error is EROFS, and we're not already
584 * inside a transaction, then there's really no point in logging
586 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
587 (sb
->s_flags
& MS_RDONLY
))
590 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
591 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb
->s_id
, function
, line
, errstr
);
593 save_error_info(sb
, function
, line
);
595 ext4_handle_error(sb
);
599 * ext4_abort is a much stronger failure handler than ext4_error. The
600 * abort function may be used to deal with unrecoverable failures such
601 * as journal IO errors or ENOMEM at a critical moment in log management.
603 * We unconditionally force the filesystem into an ABORT|READONLY state,
604 * unless the error response on the fs has been set to panic in which
605 * case we take the easy way out and panic immediately.
608 void __ext4_abort(struct super_block
*sb
, const char *function
,
609 unsigned int line
, const char *fmt
, ...)
613 save_error_info(sb
, function
, line
);
615 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
621 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
622 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
623 sb
->s_flags
|= MS_RDONLY
;
624 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
625 if (EXT4_SB(sb
)->s_journal
)
626 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
627 save_error_info(sb
, function
, line
);
629 if (test_opt(sb
, ERRORS_PANIC
))
630 panic("EXT4-fs panic from previous error\n");
633 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
635 struct va_format vaf
;
641 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
645 void __ext4_warning(struct super_block
*sb
, const char *function
,
646 unsigned int line
, const char *fmt
, ...)
648 struct va_format vaf
;
654 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
655 sb
->s_id
, function
, line
, &vaf
);
659 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
660 struct super_block
*sb
, ext4_group_t grp
,
661 unsigned long ino
, ext4_fsblk_t block
,
662 const char *fmt
, ...)
666 struct va_format vaf
;
668 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
670 es
->s_last_error_ino
= cpu_to_le32(ino
);
671 es
->s_last_error_block
= cpu_to_le64(block
);
672 __save_error_info(sb
, function
, line
);
678 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
679 sb
->s_id
, function
, line
, grp
);
681 printk(KERN_CONT
"inode %lu: ", ino
);
683 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
684 printk(KERN_CONT
"%pV\n", &vaf
);
687 if (test_opt(sb
, ERRORS_CONT
)) {
688 ext4_commit_super(sb
, 0);
692 ext4_unlock_group(sb
, grp
);
693 ext4_handle_error(sb
);
695 * We only get here in the ERRORS_RO case; relocking the group
696 * may be dangerous, but nothing bad will happen since the
697 * filesystem will have already been marked read/only and the
698 * journal has been aborted. We return 1 as a hint to callers
699 * who might what to use the return value from
700 * ext4_grp_locked_error() to distinguish between the
701 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
702 * aggressively from the ext4 function in question, with a
703 * more appropriate error code.
705 ext4_lock_group(sb
, grp
);
709 void ext4_update_dynamic_rev(struct super_block
*sb
)
711 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
713 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
717 "updating to rev %d because of new feature flag, "
718 "running e2fsck is recommended",
721 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
722 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
723 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
724 /* leave es->s_feature_*compat flags alone */
725 /* es->s_uuid will be set by e2fsck if empty */
728 * The rest of the superblock fields should be zero, and if not it
729 * means they are likely already in use, so leave them alone. We
730 * can leave it up to e2fsck to clean up any inconsistencies there.
735 * Open the external journal device
737 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
739 struct block_device
*bdev
;
740 char b
[BDEVNAME_SIZE
];
742 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
748 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
749 __bdevname(dev
, b
), PTR_ERR(bdev
));
754 * Release the journal device
756 static int ext4_blkdev_put(struct block_device
*bdev
)
758 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
761 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
763 struct block_device
*bdev
;
766 bdev
= sbi
->journal_bdev
;
768 ret
= ext4_blkdev_put(bdev
);
769 sbi
->journal_bdev
= NULL
;
774 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
776 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
779 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
783 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
784 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
786 printk(KERN_ERR
"sb_info orphan list:\n");
787 list_for_each(l
, &sbi
->s_orphan
) {
788 struct inode
*inode
= orphan_list_entry(l
);
790 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
791 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
792 inode
->i_mode
, inode
->i_nlink
,
797 static void ext4_put_super(struct super_block
*sb
)
799 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
800 struct ext4_super_block
*es
= sbi
->s_es
;
803 ext4_unregister_li_request(sb
);
804 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
806 flush_workqueue(sbi
->dio_unwritten_wq
);
807 destroy_workqueue(sbi
->dio_unwritten_wq
);
811 ext4_commit_super(sb
, 1);
813 if (sbi
->s_journal
) {
814 err
= jbd2_journal_destroy(sbi
->s_journal
);
815 sbi
->s_journal
= NULL
;
817 ext4_abort(sb
, "Couldn't clean up the journal");
820 del_timer(&sbi
->s_err_report
);
821 ext4_release_system_zone(sb
);
823 ext4_ext_release(sb
);
824 ext4_xattr_put_super(sb
);
826 if (!(sb
->s_flags
& MS_RDONLY
)) {
827 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
828 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
829 ext4_commit_super(sb
, 1);
832 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
834 kobject_del(&sbi
->s_kobj
);
836 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
837 brelse(sbi
->s_group_desc
[i
]);
838 ext4_kvfree(sbi
->s_group_desc
);
839 ext4_kvfree(sbi
->s_flex_groups
);
840 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
841 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
842 percpu_counter_destroy(&sbi
->s_dirs_counter
);
843 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
846 for (i
= 0; i
< MAXQUOTAS
; i
++)
847 kfree(sbi
->s_qf_names
[i
]);
850 /* Debugging code just in case the in-memory inode orphan list
851 * isn't empty. The on-disk one can be non-empty if we've
852 * detected an error and taken the fs readonly, but the
853 * in-memory list had better be clean by this point. */
854 if (!list_empty(&sbi
->s_orphan
))
855 dump_orphan_list(sb
, sbi
);
856 J_ASSERT(list_empty(&sbi
->s_orphan
));
858 invalidate_bdev(sb
->s_bdev
);
859 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
861 * Invalidate the journal device's buffers. We don't want them
862 * floating about in memory - the physical journal device may
863 * hotswapped, and it breaks the `ro-after' testing code.
865 sync_blockdev(sbi
->journal_bdev
);
866 invalidate_bdev(sbi
->journal_bdev
);
867 ext4_blkdev_remove(sbi
);
870 kthread_stop(sbi
->s_mmp_tsk
);
871 sb
->s_fs_info
= NULL
;
873 * Now that we are completely done shutting down the
874 * superblock, we need to actually destroy the kobject.
877 kobject_put(&sbi
->s_kobj
);
878 wait_for_completion(&sbi
->s_kobj_unregister
);
879 kfree(sbi
->s_blockgroup_lock
);
883 static struct kmem_cache
*ext4_inode_cachep
;
886 * Called inside transaction, so use GFP_NOFS
888 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
890 struct ext4_inode_info
*ei
;
892 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
896 ei
->vfs_inode
.i_version
= 1;
897 ei
->vfs_inode
.i_data
.writeback_index
= 0;
898 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
899 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
900 spin_lock_init(&ei
->i_prealloc_lock
);
901 ei
->i_reserved_data_blocks
= 0;
902 ei
->i_reserved_meta_blocks
= 0;
903 ei
->i_allocated_meta_blocks
= 0;
904 ei
->i_da_metadata_calc_len
= 0;
905 spin_lock_init(&(ei
->i_block_reservation_lock
));
907 ei
->i_reserved_quota
= 0;
910 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
911 spin_lock_init(&ei
->i_completed_io_lock
);
912 ei
->cur_aio_dio
= NULL
;
914 ei
->i_datasync_tid
= 0;
915 atomic_set(&ei
->i_ioend_count
, 0);
916 atomic_set(&ei
->i_aiodio_unwritten
, 0);
918 return &ei
->vfs_inode
;
921 static int ext4_drop_inode(struct inode
*inode
)
923 int drop
= generic_drop_inode(inode
);
925 trace_ext4_drop_inode(inode
, drop
);
929 static void ext4_i_callback(struct rcu_head
*head
)
931 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
932 INIT_LIST_HEAD(&inode
->i_dentry
);
933 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
936 static void ext4_destroy_inode(struct inode
*inode
)
938 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
939 ext4_msg(inode
->i_sb
, KERN_ERR
,
940 "Inode %lu (%p): orphan list check failed!",
941 inode
->i_ino
, EXT4_I(inode
));
942 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
943 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
947 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
950 static void init_once(void *foo
)
952 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
954 INIT_LIST_HEAD(&ei
->i_orphan
);
955 #ifdef CONFIG_EXT4_FS_XATTR
956 init_rwsem(&ei
->xattr_sem
);
958 init_rwsem(&ei
->i_data_sem
);
959 inode_init_once(&ei
->vfs_inode
);
962 static int init_inodecache(void)
964 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
965 sizeof(struct ext4_inode_info
),
966 0, (SLAB_RECLAIM_ACCOUNT
|
969 if (ext4_inode_cachep
== NULL
)
974 static void destroy_inodecache(void)
976 kmem_cache_destroy(ext4_inode_cachep
);
979 void ext4_clear_inode(struct inode
*inode
)
981 invalidate_inode_buffers(inode
);
982 end_writeback(inode
);
984 ext4_discard_preallocations(inode
);
985 if (EXT4_I(inode
)->jinode
) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
987 EXT4_I(inode
)->jinode
);
988 jbd2_free_inode(EXT4_I(inode
)->jinode
);
989 EXT4_I(inode
)->jinode
= NULL
;
993 static inline void ext4_show_quota_options(struct seq_file
*seq
,
994 struct super_block
*sb
)
996 #if defined(CONFIG_QUOTA)
997 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
999 if (sbi
->s_jquota_fmt
) {
1002 switch (sbi
->s_jquota_fmt
) {
1013 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1016 if (sbi
->s_qf_names
[USRQUOTA
])
1017 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1019 if (sbi
->s_qf_names
[GRPQUOTA
])
1020 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1022 if (test_opt(sb
, USRQUOTA
))
1023 seq_puts(seq
, ",usrquota");
1025 if (test_opt(sb
, GRPQUOTA
))
1026 seq_puts(seq
, ",grpquota");
1032 * - it's set to a non-default value OR
1033 * - if the per-sb default is different from the global default
1035 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
1038 unsigned long def_mount_opts
;
1039 struct super_block
*sb
= vfs
->mnt_sb
;
1040 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1041 struct ext4_super_block
*es
= sbi
->s_es
;
1043 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1044 def_errors
= le16_to_cpu(es
->s_errors
);
1046 if (sbi
->s_sb_block
!= 1)
1047 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
1048 if (test_opt(sb
, MINIX_DF
))
1049 seq_puts(seq
, ",minixdf");
1050 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1051 seq_puts(seq
, ",grpid");
1052 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1053 seq_puts(seq
, ",nogrpid");
1054 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
1055 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
1056 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
1058 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
1059 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
1060 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
1062 if (test_opt(sb
, ERRORS_RO
)) {
1063 if (def_errors
== EXT4_ERRORS_PANIC
||
1064 def_errors
== EXT4_ERRORS_CONTINUE
) {
1065 seq_puts(seq
, ",errors=remount-ro");
1068 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1069 seq_puts(seq
, ",errors=continue");
1070 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1071 seq_puts(seq
, ",errors=panic");
1072 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
1073 seq_puts(seq
, ",nouid32");
1074 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
1075 seq_puts(seq
, ",debug");
1076 if (test_opt(sb
, OLDALLOC
))
1077 seq_puts(seq
, ",oldalloc");
1078 #ifdef CONFIG_EXT4_FS_XATTR
1079 if (test_opt(sb
, XATTR_USER
))
1080 seq_puts(seq
, ",user_xattr");
1081 if (!test_opt(sb
, XATTR_USER
))
1082 seq_puts(seq
, ",nouser_xattr");
1084 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1085 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
1086 seq_puts(seq
, ",acl");
1087 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
1088 seq_puts(seq
, ",noacl");
1090 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
1091 seq_printf(seq
, ",commit=%u",
1092 (unsigned) (sbi
->s_commit_interval
/ HZ
));
1094 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
1095 seq_printf(seq
, ",min_batch_time=%u",
1096 (unsigned) sbi
->s_min_batch_time
);
1098 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1099 seq_printf(seq
, ",max_batch_time=%u",
1100 (unsigned) sbi
->s_min_batch_time
);
1104 * We're changing the default of barrier mount option, so
1105 * let's always display its mount state so it's clear what its
1108 seq_puts(seq
, ",barrier=");
1109 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1110 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1111 seq_puts(seq
, ",journal_async_commit");
1112 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1113 seq_puts(seq
, ",journal_checksum");
1114 if (test_opt(sb
, I_VERSION
))
1115 seq_puts(seq
, ",i_version");
1116 if (!test_opt(sb
, DELALLOC
) &&
1117 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1118 seq_puts(seq
, ",nodelalloc");
1120 if (!test_opt(sb
, MBLK_IO_SUBMIT
))
1121 seq_puts(seq
, ",nomblk_io_submit");
1123 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1125 * journal mode get enabled in different ways
1126 * So just print the value even if we didn't specify it
1128 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1129 seq_puts(seq
, ",data=journal");
1130 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1131 seq_puts(seq
, ",data=ordered");
1132 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1133 seq_puts(seq
, ",data=writeback");
1135 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1136 seq_printf(seq
, ",inode_readahead_blks=%u",
1137 sbi
->s_inode_readahead_blks
);
1139 if (test_opt(sb
, DATA_ERR_ABORT
))
1140 seq_puts(seq
, ",data_err=abort");
1142 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1143 seq_puts(seq
, ",noauto_da_alloc");
1145 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1146 seq_puts(seq
, ",discard");
1148 if (test_opt(sb
, NOLOAD
))
1149 seq_puts(seq
, ",norecovery");
1151 if (test_opt(sb
, DIOREAD_NOLOCK
))
1152 seq_puts(seq
, ",dioread_nolock");
1154 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1155 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1156 seq_puts(seq
, ",block_validity");
1158 if (!test_opt(sb
, INIT_INODE_TABLE
))
1159 seq_puts(seq
, ",noinit_inode_table");
1160 else if (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)
1161 seq_printf(seq
, ",init_inode_table=%u",
1162 (unsigned) sbi
->s_li_wait_mult
);
1164 ext4_show_quota_options(seq
, sb
);
1169 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1170 u64 ino
, u32 generation
)
1172 struct inode
*inode
;
1174 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1175 return ERR_PTR(-ESTALE
);
1176 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1177 return ERR_PTR(-ESTALE
);
1179 /* iget isn't really right if the inode is currently unallocated!!
1181 * ext4_read_inode will return a bad_inode if the inode had been
1182 * deleted, so we should be safe.
1184 * Currently we don't know the generation for parent directory, so
1185 * a generation of 0 means "accept any"
1187 inode
= ext4_iget(sb
, ino
);
1189 return ERR_CAST(inode
);
1190 if (generation
&& inode
->i_generation
!= generation
) {
1192 return ERR_PTR(-ESTALE
);
1198 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1199 int fh_len
, int fh_type
)
1201 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1202 ext4_nfs_get_inode
);
1205 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1206 int fh_len
, int fh_type
)
1208 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1209 ext4_nfs_get_inode
);
1213 * Try to release metadata pages (indirect blocks, directories) which are
1214 * mapped via the block device. Since these pages could have journal heads
1215 * which would prevent try_to_free_buffers() from freeing them, we must use
1216 * jbd2 layer's try_to_free_buffers() function to release them.
1218 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1221 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1223 WARN_ON(PageChecked(page
));
1224 if (!page_has_buffers(page
))
1227 return jbd2_journal_try_to_free_buffers(journal
, page
,
1228 wait
& ~__GFP_WAIT
);
1229 return try_to_free_buffers(page
);
1233 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1234 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1236 static int ext4_write_dquot(struct dquot
*dquot
);
1237 static int ext4_acquire_dquot(struct dquot
*dquot
);
1238 static int ext4_release_dquot(struct dquot
*dquot
);
1239 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1240 static int ext4_write_info(struct super_block
*sb
, int type
);
1241 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1243 static int ext4_quota_off(struct super_block
*sb
, int type
);
1244 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1245 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1246 size_t len
, loff_t off
);
1247 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1248 const char *data
, size_t len
, loff_t off
);
1250 static const struct dquot_operations ext4_quota_operations
= {
1251 .get_reserved_space
= ext4_get_reserved_space
,
1252 .write_dquot
= ext4_write_dquot
,
1253 .acquire_dquot
= ext4_acquire_dquot
,
1254 .release_dquot
= ext4_release_dquot
,
1255 .mark_dirty
= ext4_mark_dquot_dirty
,
1256 .write_info
= ext4_write_info
,
1257 .alloc_dquot
= dquot_alloc
,
1258 .destroy_dquot
= dquot_destroy
,
1261 static const struct quotactl_ops ext4_qctl_operations
= {
1262 .quota_on
= ext4_quota_on
,
1263 .quota_off
= ext4_quota_off
,
1264 .quota_sync
= dquot_quota_sync
,
1265 .get_info
= dquot_get_dqinfo
,
1266 .set_info
= dquot_set_dqinfo
,
1267 .get_dqblk
= dquot_get_dqblk
,
1268 .set_dqblk
= dquot_set_dqblk
1272 static const struct super_operations ext4_sops
= {
1273 .alloc_inode
= ext4_alloc_inode
,
1274 .destroy_inode
= ext4_destroy_inode
,
1275 .write_inode
= ext4_write_inode
,
1276 .dirty_inode
= ext4_dirty_inode
,
1277 .drop_inode
= ext4_drop_inode
,
1278 .evict_inode
= ext4_evict_inode
,
1279 .put_super
= ext4_put_super
,
1280 .sync_fs
= ext4_sync_fs
,
1281 .freeze_fs
= ext4_freeze
,
1282 .unfreeze_fs
= ext4_unfreeze
,
1283 .statfs
= ext4_statfs
,
1284 .remount_fs
= ext4_remount
,
1285 .show_options
= ext4_show_options
,
1287 .quota_read
= ext4_quota_read
,
1288 .quota_write
= ext4_quota_write
,
1290 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1293 static const struct super_operations ext4_nojournal_sops
= {
1294 .alloc_inode
= ext4_alloc_inode
,
1295 .destroy_inode
= ext4_destroy_inode
,
1296 .write_inode
= ext4_write_inode
,
1297 .dirty_inode
= ext4_dirty_inode
,
1298 .drop_inode
= ext4_drop_inode
,
1299 .evict_inode
= ext4_evict_inode
,
1300 .write_super
= ext4_write_super
,
1301 .put_super
= ext4_put_super
,
1302 .statfs
= ext4_statfs
,
1303 .remount_fs
= ext4_remount
,
1304 .show_options
= ext4_show_options
,
1306 .quota_read
= ext4_quota_read
,
1307 .quota_write
= ext4_quota_write
,
1309 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1312 static const struct export_operations ext4_export_ops
= {
1313 .fh_to_dentry
= ext4_fh_to_dentry
,
1314 .fh_to_parent
= ext4_fh_to_parent
,
1315 .get_parent
= ext4_get_parent
,
1319 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1320 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1321 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1322 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1323 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1324 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1325 Opt_journal_update
, Opt_journal_dev
,
1326 Opt_journal_checksum
, Opt_journal_async_commit
,
1327 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1328 Opt_data_err_abort
, Opt_data_err_ignore
,
1329 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1330 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1331 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1332 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1333 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1334 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1335 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1336 Opt_dioread_nolock
, Opt_dioread_lock
,
1337 Opt_discard
, Opt_nodiscard
,
1338 Opt_init_inode_table
, Opt_noinit_inode_table
,
1341 static const match_table_t tokens
= {
1342 {Opt_bsd_df
, "bsddf"},
1343 {Opt_minix_df
, "minixdf"},
1344 {Opt_grpid
, "grpid"},
1345 {Opt_grpid
, "bsdgroups"},
1346 {Opt_nogrpid
, "nogrpid"},
1347 {Opt_nogrpid
, "sysvgroups"},
1348 {Opt_resgid
, "resgid=%u"},
1349 {Opt_resuid
, "resuid=%u"},
1351 {Opt_err_cont
, "errors=continue"},
1352 {Opt_err_panic
, "errors=panic"},
1353 {Opt_err_ro
, "errors=remount-ro"},
1354 {Opt_nouid32
, "nouid32"},
1355 {Opt_debug
, "debug"},
1356 {Opt_oldalloc
, "oldalloc"},
1357 {Opt_orlov
, "orlov"},
1358 {Opt_user_xattr
, "user_xattr"},
1359 {Opt_nouser_xattr
, "nouser_xattr"},
1361 {Opt_noacl
, "noacl"},
1362 {Opt_noload
, "noload"},
1363 {Opt_noload
, "norecovery"},
1366 {Opt_commit
, "commit=%u"},
1367 {Opt_min_batch_time
, "min_batch_time=%u"},
1368 {Opt_max_batch_time
, "max_batch_time=%u"},
1369 {Opt_journal_update
, "journal=update"},
1370 {Opt_journal_dev
, "journal_dev=%u"},
1371 {Opt_journal_checksum
, "journal_checksum"},
1372 {Opt_journal_async_commit
, "journal_async_commit"},
1373 {Opt_abort
, "abort"},
1374 {Opt_data_journal
, "data=journal"},
1375 {Opt_data_ordered
, "data=ordered"},
1376 {Opt_data_writeback
, "data=writeback"},
1377 {Opt_data_err_abort
, "data_err=abort"},
1378 {Opt_data_err_ignore
, "data_err=ignore"},
1379 {Opt_offusrjquota
, "usrjquota="},
1380 {Opt_usrjquota
, "usrjquota=%s"},
1381 {Opt_offgrpjquota
, "grpjquota="},
1382 {Opt_grpjquota
, "grpjquota=%s"},
1383 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1384 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1385 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1386 {Opt_grpquota
, "grpquota"},
1387 {Opt_noquota
, "noquota"},
1388 {Opt_quota
, "quota"},
1389 {Opt_usrquota
, "usrquota"},
1390 {Opt_barrier
, "barrier=%u"},
1391 {Opt_barrier
, "barrier"},
1392 {Opt_nobarrier
, "nobarrier"},
1393 {Opt_i_version
, "i_version"},
1394 {Opt_stripe
, "stripe=%u"},
1395 {Opt_resize
, "resize"},
1396 {Opt_delalloc
, "delalloc"},
1397 {Opt_nodelalloc
, "nodelalloc"},
1398 {Opt_mblk_io_submit
, "mblk_io_submit"},
1399 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1400 {Opt_block_validity
, "block_validity"},
1401 {Opt_noblock_validity
, "noblock_validity"},
1402 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1403 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1404 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1405 {Opt_auto_da_alloc
, "auto_da_alloc"},
1406 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1407 {Opt_dioread_nolock
, "dioread_nolock"},
1408 {Opt_dioread_lock
, "dioread_lock"},
1409 {Opt_discard
, "discard"},
1410 {Opt_nodiscard
, "nodiscard"},
1411 {Opt_init_inode_table
, "init_itable=%u"},
1412 {Opt_init_inode_table
, "init_itable"},
1413 {Opt_noinit_inode_table
, "noinit_itable"},
1417 static ext4_fsblk_t
get_sb_block(void **data
)
1419 ext4_fsblk_t sb_block
;
1420 char *options
= (char *) *data
;
1422 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1423 return 1; /* Default location */
1426 /* TODO: use simple_strtoll with >32bit ext4 */
1427 sb_block
= simple_strtoul(options
, &options
, 0);
1428 if (*options
&& *options
!= ',') {
1429 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1433 if (*options
== ',')
1435 *data
= (void *) options
;
1440 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1441 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1442 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1445 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1447 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1450 if (sb_any_quota_loaded(sb
) &&
1451 !sbi
->s_qf_names
[qtype
]) {
1452 ext4_msg(sb
, KERN_ERR
,
1453 "Cannot change journaled "
1454 "quota options when quota turned on");
1457 qname
= match_strdup(args
);
1459 ext4_msg(sb
, KERN_ERR
,
1460 "Not enough memory for storing quotafile name");
1463 if (sbi
->s_qf_names
[qtype
] &&
1464 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1465 ext4_msg(sb
, KERN_ERR
,
1466 "%s quota file already specified", QTYPE2NAME(qtype
));
1470 sbi
->s_qf_names
[qtype
] = qname
;
1471 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1472 ext4_msg(sb
, KERN_ERR
,
1473 "quotafile must be on filesystem root");
1474 kfree(sbi
->s_qf_names
[qtype
]);
1475 sbi
->s_qf_names
[qtype
] = NULL
;
1482 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1485 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1487 if (sb_any_quota_loaded(sb
) &&
1488 sbi
->s_qf_names
[qtype
]) {
1489 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1490 " when quota turned on");
1494 * The space will be released later when all options are confirmed
1497 sbi
->s_qf_names
[qtype
] = NULL
;
1502 static int parse_options(char *options
, struct super_block
*sb
,
1503 unsigned long *journal_devnum
,
1504 unsigned int *journal_ioprio
,
1505 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1507 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1509 substring_t args
[MAX_OPT_ARGS
];
1519 while ((p
= strsep(&options
, ",")) != NULL
) {
1525 * Initialize args struct so we know whether arg was
1526 * found; some options take optional arguments.
1528 args
[0].to
= args
[0].from
= NULL
;
1529 token
= match_token(p
, tokens
, args
);
1532 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1533 clear_opt(sb
, MINIX_DF
);
1536 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1537 set_opt(sb
, MINIX_DF
);
1541 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1546 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1547 clear_opt(sb
, GRPID
);
1551 if (match_int(&args
[0], &option
))
1553 sbi
->s_resuid
= option
;
1556 if (match_int(&args
[0], &option
))
1558 sbi
->s_resgid
= option
;
1561 /* handled by get_sb_block() instead of here */
1562 /* *sb_block = match_int(&args[0]); */
1565 clear_opt(sb
, ERRORS_CONT
);
1566 clear_opt(sb
, ERRORS_RO
);
1567 set_opt(sb
, ERRORS_PANIC
);
1570 clear_opt(sb
, ERRORS_CONT
);
1571 clear_opt(sb
, ERRORS_PANIC
);
1572 set_opt(sb
, ERRORS_RO
);
1575 clear_opt(sb
, ERRORS_RO
);
1576 clear_opt(sb
, ERRORS_PANIC
);
1577 set_opt(sb
, ERRORS_CONT
);
1580 set_opt(sb
, NO_UID32
);
1586 set_opt(sb
, OLDALLOC
);
1589 clear_opt(sb
, OLDALLOC
);
1591 #ifdef CONFIG_EXT4_FS_XATTR
1592 case Opt_user_xattr
:
1593 set_opt(sb
, XATTR_USER
);
1595 case Opt_nouser_xattr
:
1596 clear_opt(sb
, XATTR_USER
);
1599 case Opt_user_xattr
:
1600 case Opt_nouser_xattr
:
1601 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1604 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1606 set_opt(sb
, POSIX_ACL
);
1609 clear_opt(sb
, POSIX_ACL
);
1614 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1617 case Opt_journal_update
:
1619 /* Eventually we will want to be able to create
1620 a journal file here. For now, only allow the
1621 user to specify an existing inode to be the
1624 ext4_msg(sb
, KERN_ERR
,
1625 "Cannot specify journal on remount");
1628 set_opt(sb
, UPDATE_JOURNAL
);
1630 case Opt_journal_dev
:
1632 ext4_msg(sb
, KERN_ERR
,
1633 "Cannot specify journal on remount");
1636 if (match_int(&args
[0], &option
))
1638 *journal_devnum
= option
;
1640 case Opt_journal_checksum
:
1641 set_opt(sb
, JOURNAL_CHECKSUM
);
1643 case Opt_journal_async_commit
:
1644 set_opt(sb
, JOURNAL_ASYNC_COMMIT
);
1645 set_opt(sb
, JOURNAL_CHECKSUM
);
1648 set_opt(sb
, NOLOAD
);
1651 if (match_int(&args
[0], &option
))
1656 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1657 sbi
->s_commit_interval
= HZ
* option
;
1659 case Opt_max_batch_time
:
1660 if (match_int(&args
[0], &option
))
1665 option
= EXT4_DEF_MAX_BATCH_TIME
;
1666 sbi
->s_max_batch_time
= option
;
1668 case Opt_min_batch_time
:
1669 if (match_int(&args
[0], &option
))
1673 sbi
->s_min_batch_time
= option
;
1675 case Opt_data_journal
:
1676 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1678 case Opt_data_ordered
:
1679 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1681 case Opt_data_writeback
:
1682 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1685 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1686 ext4_msg(sb
, KERN_ERR
,
1687 "Cannot change data mode on remount");
1691 clear_opt(sb
, DATA_FLAGS
);
1692 sbi
->s_mount_opt
|= data_opt
;
1695 case Opt_data_err_abort
:
1696 set_opt(sb
, DATA_ERR_ABORT
);
1698 case Opt_data_err_ignore
:
1699 clear_opt(sb
, DATA_ERR_ABORT
);
1703 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1707 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1710 case Opt_offusrjquota
:
1711 if (!clear_qf_name(sb
, USRQUOTA
))
1714 case Opt_offgrpjquota
:
1715 if (!clear_qf_name(sb
, GRPQUOTA
))
1719 case Opt_jqfmt_vfsold
:
1720 qfmt
= QFMT_VFS_OLD
;
1722 case Opt_jqfmt_vfsv0
:
1725 case Opt_jqfmt_vfsv1
:
1728 if (sb_any_quota_loaded(sb
) &&
1729 sbi
->s_jquota_fmt
!= qfmt
) {
1730 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1731 "journaled quota options when "
1735 sbi
->s_jquota_fmt
= qfmt
;
1740 set_opt(sb
, USRQUOTA
);
1744 set_opt(sb
, GRPQUOTA
);
1747 if (sb_any_quota_loaded(sb
)) {
1748 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1749 "options when quota turned on");
1752 clear_opt(sb
, QUOTA
);
1753 clear_opt(sb
, USRQUOTA
);
1754 clear_opt(sb
, GRPQUOTA
);
1760 ext4_msg(sb
, KERN_ERR
,
1761 "quota options not supported");
1765 case Opt_offusrjquota
:
1766 case Opt_offgrpjquota
:
1767 case Opt_jqfmt_vfsold
:
1768 case Opt_jqfmt_vfsv0
:
1769 case Opt_jqfmt_vfsv1
:
1770 ext4_msg(sb
, KERN_ERR
,
1771 "journaled quota options not supported");
1777 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1780 clear_opt(sb
, BARRIER
);
1784 if (match_int(&args
[0], &option
))
1787 option
= 1; /* No argument, default to 1 */
1789 set_opt(sb
, BARRIER
);
1791 clear_opt(sb
, BARRIER
);
1797 ext4_msg(sb
, KERN_ERR
,
1798 "resize option only available "
1802 if (match_int(&args
[0], &option
) != 0)
1804 *n_blocks_count
= option
;
1807 ext4_msg(sb
, KERN_WARNING
,
1808 "Ignoring deprecated nobh option");
1811 ext4_msg(sb
, KERN_WARNING
,
1812 "Ignoring deprecated bh option");
1815 set_opt(sb
, I_VERSION
);
1816 sb
->s_flags
|= MS_I_VERSION
;
1818 case Opt_nodelalloc
:
1819 clear_opt(sb
, DELALLOC
);
1820 clear_opt2(sb
, EXPLICIT_DELALLOC
);
1822 case Opt_mblk_io_submit
:
1823 set_opt(sb
, MBLK_IO_SUBMIT
);
1825 case Opt_nomblk_io_submit
:
1826 clear_opt(sb
, MBLK_IO_SUBMIT
);
1829 if (match_int(&args
[0], &option
))
1833 sbi
->s_stripe
= option
;
1836 set_opt(sb
, DELALLOC
);
1837 set_opt2(sb
, EXPLICIT_DELALLOC
);
1839 case Opt_block_validity
:
1840 set_opt(sb
, BLOCK_VALIDITY
);
1842 case Opt_noblock_validity
:
1843 clear_opt(sb
, BLOCK_VALIDITY
);
1845 case Opt_inode_readahead_blks
:
1846 if (match_int(&args
[0], &option
))
1848 if (option
< 0 || option
> (1 << 30))
1850 if (option
&& !is_power_of_2(option
)) {
1851 ext4_msg(sb
, KERN_ERR
,
1852 "EXT4-fs: inode_readahead_blks"
1853 " must be a power of 2");
1856 sbi
->s_inode_readahead_blks
= option
;
1858 case Opt_journal_ioprio
:
1859 if (match_int(&args
[0], &option
))
1861 if (option
< 0 || option
> 7)
1863 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1866 case Opt_noauto_da_alloc
:
1867 set_opt(sb
, NO_AUTO_DA_ALLOC
);
1869 case Opt_auto_da_alloc
:
1871 if (match_int(&args
[0], &option
))
1874 option
= 1; /* No argument, default to 1 */
1876 clear_opt(sb
, NO_AUTO_DA_ALLOC
);
1878 set_opt(sb
,NO_AUTO_DA_ALLOC
);
1881 set_opt(sb
, DISCARD
);
1884 clear_opt(sb
, DISCARD
);
1886 case Opt_dioread_nolock
:
1887 set_opt(sb
, DIOREAD_NOLOCK
);
1889 case Opt_dioread_lock
:
1890 clear_opt(sb
, DIOREAD_NOLOCK
);
1892 case Opt_init_inode_table
:
1893 set_opt(sb
, INIT_INODE_TABLE
);
1895 if (match_int(&args
[0], &option
))
1898 option
= EXT4_DEF_LI_WAIT_MULT
;
1901 sbi
->s_li_wait_mult
= option
;
1903 case Opt_noinit_inode_table
:
1904 clear_opt(sb
, INIT_INODE_TABLE
);
1907 ext4_msg(sb
, KERN_ERR
,
1908 "Unrecognized mount option \"%s\" "
1909 "or missing value", p
);
1914 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1915 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1916 clear_opt(sb
, USRQUOTA
);
1918 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1919 clear_opt(sb
, GRPQUOTA
);
1921 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1922 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1927 if (!sbi
->s_jquota_fmt
) {
1928 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1933 if (sbi
->s_jquota_fmt
) {
1934 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1935 "specified with no journaling "
1944 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1947 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1950 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1951 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1952 "forcing read-only mode");
1957 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1958 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1959 "running e2fsck is recommended");
1960 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1961 ext4_msg(sb
, KERN_WARNING
,
1962 "warning: mounting fs with errors, "
1963 "running e2fsck is recommended");
1964 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1965 le16_to_cpu(es
->s_mnt_count
) >=
1966 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1967 ext4_msg(sb
, KERN_WARNING
,
1968 "warning: maximal mount count reached, "
1969 "running e2fsck is recommended");
1970 else if (le32_to_cpu(es
->s_checkinterval
) &&
1971 (le32_to_cpu(es
->s_lastcheck
) +
1972 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1973 ext4_msg(sb
, KERN_WARNING
,
1974 "warning: checktime reached, "
1975 "running e2fsck is recommended");
1976 if (!sbi
->s_journal
)
1977 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1978 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1979 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1980 le16_add_cpu(&es
->s_mnt_count
, 1);
1981 es
->s_mtime
= cpu_to_le32(get_seconds());
1982 ext4_update_dynamic_rev(sb
);
1984 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1986 ext4_commit_super(sb
, 1);
1988 if (test_opt(sb
, DEBUG
))
1989 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1990 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1992 sbi
->s_groups_count
,
1993 EXT4_BLOCKS_PER_GROUP(sb
),
1994 EXT4_INODES_PER_GROUP(sb
),
1995 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1997 cleancache_init_fs(sb
);
2001 static int ext4_fill_flex_info(struct super_block
*sb
)
2003 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2004 struct ext4_group_desc
*gdp
= NULL
;
2005 ext4_group_t flex_group_count
;
2006 ext4_group_t flex_group
;
2007 int groups_per_flex
= 0;
2011 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
2012 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
2014 if (groups_per_flex
< 2) {
2015 sbi
->s_log_groups_per_flex
= 0;
2019 /* We allocate both existing and potentially added groups */
2020 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
2021 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
2022 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
2023 size
= flex_group_count
* sizeof(struct flex_groups
);
2024 sbi
->s_flex_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
2025 if (sbi
->s_flex_groups
== NULL
) {
2026 ext4_msg(sb
, KERN_ERR
, "not enough memory for %u flex groups",
2031 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2032 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2034 flex_group
= ext4_flex_group(sbi
, i
);
2035 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2036 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2037 atomic_add(ext4_free_blks_count(sb
, gdp
),
2038 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
2039 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2040 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2048 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
2049 struct ext4_group_desc
*gdp
)
2053 if (sbi
->s_es
->s_feature_ro_compat
&
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
2055 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2056 __le32 le_group
= cpu_to_le32(block_group
);
2058 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2059 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2060 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2061 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi
->s_es
->s_feature_incompat
&
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2065 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2066 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2067 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2071 return cpu_to_le16(crc
);
2074 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
2075 struct ext4_group_desc
*gdp
)
2077 if ((sbi
->s_es
->s_feature_ro_compat
&
2078 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
2079 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
2085 /* Called at mount-time, super-block is locked */
2086 static int ext4_check_descriptors(struct super_block
*sb
,
2087 ext4_group_t
*first_not_zeroed
)
2089 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2090 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2091 ext4_fsblk_t last_block
;
2092 ext4_fsblk_t block_bitmap
;
2093 ext4_fsblk_t inode_bitmap
;
2094 ext4_fsblk_t inode_table
;
2095 int flexbg_flag
= 0;
2096 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2098 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2101 ext4_debug("Checking group descriptors");
2103 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2104 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2106 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2107 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2109 last_block
= first_block
+
2110 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2112 if ((grp
== sbi
->s_groups_count
) &&
2113 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2116 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2117 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2118 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2119 "Block bitmap for group %u not in group "
2120 "(block %llu)!", i
, block_bitmap
);
2123 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2124 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2125 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2126 "Inode bitmap for group %u not in group "
2127 "(block %llu)!", i
, inode_bitmap
);
2130 inode_table
= ext4_inode_table(sb
, gdp
);
2131 if (inode_table
< first_block
||
2132 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2133 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2134 "Inode table for group %u not in group "
2135 "(block %llu)!", i
, inode_table
);
2138 ext4_lock_group(sb
, i
);
2139 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2140 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2141 "Checksum for group %u failed (%u!=%u)",
2142 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2143 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2144 if (!(sb
->s_flags
& MS_RDONLY
)) {
2145 ext4_unlock_group(sb
, i
);
2149 ext4_unlock_group(sb
, i
);
2151 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2153 if (NULL
!= first_not_zeroed
)
2154 *first_not_zeroed
= grp
;
2156 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2157 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2161 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2162 * the superblock) which were deleted from all directories, but held open by
2163 * a process at the time of a crash. We walk the list and try to delete these
2164 * inodes at recovery time (only with a read-write filesystem).
2166 * In order to keep the orphan inode chain consistent during traversal (in
2167 * case of crash during recovery), we link each inode into the superblock
2168 * orphan list_head and handle it the same way as an inode deletion during
2169 * normal operation (which journals the operations for us).
2171 * We only do an iget() and an iput() on each inode, which is very safe if we
2172 * accidentally point at an in-use or already deleted inode. The worst that
2173 * can happen in this case is that we get a "bit already cleared" message from
2174 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2175 * e2fsck was run on this filesystem, and it must have already done the orphan
2176 * inode cleanup for us, so we can safely abort without any further action.
2178 static void ext4_orphan_cleanup(struct super_block
*sb
,
2179 struct ext4_super_block
*es
)
2181 unsigned int s_flags
= sb
->s_flags
;
2182 int nr_orphans
= 0, nr_truncates
= 0;
2186 if (!es
->s_last_orphan
) {
2187 jbd_debug(4, "no orphan inodes to clean up\n");
2191 if (bdev_read_only(sb
->s_bdev
)) {
2192 ext4_msg(sb
, KERN_ERR
, "write access "
2193 "unavailable, skipping orphan cleanup");
2197 /* Check if feature set would not allow a r/w mount */
2198 if (!ext4_feature_set_ok(sb
, 0)) {
2199 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2200 "unknown ROCOMPAT features");
2204 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2205 if (es
->s_last_orphan
)
2206 jbd_debug(1, "Errors on filesystem, "
2207 "clearing orphan list.\n");
2208 es
->s_last_orphan
= 0;
2209 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2213 if (s_flags
& MS_RDONLY
) {
2214 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2215 sb
->s_flags
&= ~MS_RDONLY
;
2218 /* Needed for iput() to work correctly and not trash data */
2219 sb
->s_flags
|= MS_ACTIVE
;
2220 /* Turn on quotas so that they are updated correctly */
2221 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2222 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2223 int ret
= ext4_quota_on_mount(sb
, i
);
2225 ext4_msg(sb
, KERN_ERR
,
2226 "Cannot turn on journaled "
2227 "quota: error %d", ret
);
2232 while (es
->s_last_orphan
) {
2233 struct inode
*inode
;
2235 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2236 if (IS_ERR(inode
)) {
2237 es
->s_last_orphan
= 0;
2241 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2242 dquot_initialize(inode
);
2243 if (inode
->i_nlink
) {
2244 ext4_msg(sb
, KERN_DEBUG
,
2245 "%s: truncating inode %lu to %lld bytes",
2246 __func__
, inode
->i_ino
, inode
->i_size
);
2247 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2248 inode
->i_ino
, inode
->i_size
);
2249 ext4_truncate(inode
);
2252 ext4_msg(sb
, KERN_DEBUG
,
2253 "%s: deleting unreferenced inode %lu",
2254 __func__
, inode
->i_ino
);
2255 jbd_debug(2, "deleting unreferenced inode %lu\n",
2259 iput(inode
); /* The delete magic happens here! */
2262 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2265 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2266 PLURAL(nr_orphans
));
2268 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2269 PLURAL(nr_truncates
));
2271 /* Turn quotas off */
2272 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2273 if (sb_dqopt(sb
)->files
[i
])
2274 dquot_quota_off(sb
, i
);
2277 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2281 * Maximal extent format file size.
2282 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2283 * extent format containers, within a sector_t, and within i_blocks
2284 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2285 * so that won't be a limiting factor.
2287 * However there is other limiting factor. We do store extents in the form
2288 * of starting block and length, hence the resulting length of the extent
2289 * covering maximum file size must fit into on-disk format containers as
2290 * well. Given that length is always by 1 unit bigger than max unit (because
2291 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2293 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2295 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2298 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2300 /* small i_blocks in vfs inode? */
2301 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2303 * CONFIG_LBDAF is not enabled implies the inode
2304 * i_block represent total blocks in 512 bytes
2305 * 32 == size of vfs inode i_blocks * 8
2307 upper_limit
= (1LL << 32) - 1;
2309 /* total blocks in file system block size */
2310 upper_limit
>>= (blkbits
- 9);
2311 upper_limit
<<= blkbits
;
2315 * 32-bit extent-start container, ee_block. We lower the maxbytes
2316 * by one fs block, so ee_len can cover the extent of maximum file
2319 res
= (1LL << 32) - 1;
2322 /* Sanity check against vm- & vfs- imposed limits */
2323 if (res
> upper_limit
)
2330 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2331 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2332 * We need to be 1 filesystem block less than the 2^48 sector limit.
2334 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2336 loff_t res
= EXT4_NDIR_BLOCKS
;
2339 /* This is calculated to be the largest file size for a dense, block
2340 * mapped file such that the file's total number of 512-byte sectors,
2341 * including data and all indirect blocks, does not exceed (2^48 - 1).
2343 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2344 * number of 512-byte sectors of the file.
2347 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2349 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2350 * the inode i_block field represents total file blocks in
2351 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2353 upper_limit
= (1LL << 32) - 1;
2355 /* total blocks in file system block size */
2356 upper_limit
>>= (bits
- 9);
2360 * We use 48 bit ext4_inode i_blocks
2361 * With EXT4_HUGE_FILE_FL set the i_blocks
2362 * represent total number of blocks in
2363 * file system block size
2365 upper_limit
= (1LL << 48) - 1;
2369 /* indirect blocks */
2371 /* double indirect blocks */
2372 meta_blocks
+= 1 + (1LL << (bits
-2));
2373 /* tripple indirect blocks */
2374 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2376 upper_limit
-= meta_blocks
;
2377 upper_limit
<<= bits
;
2379 res
+= 1LL << (bits
-2);
2380 res
+= 1LL << (2*(bits
-2));
2381 res
+= 1LL << (3*(bits
-2));
2383 if (res
> upper_limit
)
2386 if (res
> MAX_LFS_FILESIZE
)
2387 res
= MAX_LFS_FILESIZE
;
2392 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2393 ext4_fsblk_t logical_sb_block
, int nr
)
2395 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2396 ext4_group_t bg
, first_meta_bg
;
2399 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2401 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2403 return logical_sb_block
+ nr
+ 1;
2404 bg
= sbi
->s_desc_per_block
* nr
;
2405 if (ext4_bg_has_super(sb
, bg
))
2408 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2412 * ext4_get_stripe_size: Get the stripe size.
2413 * @sbi: In memory super block info
2415 * If we have specified it via mount option, then
2416 * use the mount option value. If the value specified at mount time is
2417 * greater than the blocks per group use the super block value.
2418 * If the super block value is greater than blocks per group return 0.
2419 * Allocator needs it be less than blocks per group.
2422 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2424 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2425 unsigned long stripe_width
=
2426 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2429 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2430 ret
= sbi
->s_stripe
;
2431 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2433 else if (stride
<= sbi
->s_blocks_per_group
)
2439 * If the stripe width is 1, this makes no sense and
2440 * we set it to 0 to turn off stripe handling code.
2451 struct attribute attr
;
2452 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2453 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2454 const char *, size_t);
2458 static int parse_strtoul(const char *buf
,
2459 unsigned long max
, unsigned long *value
)
2463 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2464 endp
= skip_spaces(endp
);
2465 if (*endp
|| *value
> max
)
2471 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2472 struct ext4_sb_info
*sbi
,
2475 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2476 (s64
) percpu_counter_sum(&sbi
->s_dirtyclusters_counter
));
2479 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2480 struct ext4_sb_info
*sbi
, char *buf
)
2482 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2484 if (!sb
->s_bdev
->bd_part
)
2485 return snprintf(buf
, PAGE_SIZE
, "0\n");
2486 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2487 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2488 sbi
->s_sectors_written_start
) >> 1);
2491 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2492 struct ext4_sb_info
*sbi
, char *buf
)
2494 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2496 if (!sb
->s_bdev
->bd_part
)
2497 return snprintf(buf
, PAGE_SIZE
, "0\n");
2498 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2499 (unsigned long long)(sbi
->s_kbytes_written
+
2500 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2501 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2504 static ssize_t
extent_cache_hits_show(struct ext4_attr
*a
,
2505 struct ext4_sb_info
*sbi
, char *buf
)
2507 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_hits
);
2510 static ssize_t
extent_cache_misses_show(struct ext4_attr
*a
,
2511 struct ext4_sb_info
*sbi
, char *buf
)
2513 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_misses
);
2516 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2517 struct ext4_sb_info
*sbi
,
2518 const char *buf
, size_t count
)
2522 if (parse_strtoul(buf
, 0x40000000, &t
))
2525 if (t
&& !is_power_of_2(t
))
2528 sbi
->s_inode_readahead_blks
= t
;
2532 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2533 struct ext4_sb_info
*sbi
, char *buf
)
2535 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2537 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2540 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2541 struct ext4_sb_info
*sbi
,
2542 const char *buf
, size_t count
)
2544 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2547 if (parse_strtoul(buf
, 0xffffffff, &t
))
2553 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2554 static struct ext4_attr ext4_attr_##_name = { \
2555 .attr = {.name = __stringify(_name), .mode = _mode }, \
2558 .offset = offsetof(struct ext4_sb_info, _elname), \
2560 #define EXT4_ATTR(name, mode, show, store) \
2561 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2563 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2564 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2565 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2566 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2567 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2568 #define ATTR_LIST(name) &ext4_attr_##name.attr
2570 EXT4_RO_ATTR(delayed_allocation_blocks
);
2571 EXT4_RO_ATTR(session_write_kbytes
);
2572 EXT4_RO_ATTR(lifetime_write_kbytes
);
2573 EXT4_RO_ATTR(extent_cache_hits
);
2574 EXT4_RO_ATTR(extent_cache_misses
);
2575 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2576 inode_readahead_blks_store
, s_inode_readahead_blks
);
2577 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2578 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2579 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2580 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2581 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2582 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2583 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2584 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2586 static struct attribute
*ext4_attrs
[] = {
2587 ATTR_LIST(delayed_allocation_blocks
),
2588 ATTR_LIST(session_write_kbytes
),
2589 ATTR_LIST(lifetime_write_kbytes
),
2590 ATTR_LIST(extent_cache_hits
),
2591 ATTR_LIST(extent_cache_misses
),
2592 ATTR_LIST(inode_readahead_blks
),
2593 ATTR_LIST(inode_goal
),
2594 ATTR_LIST(mb_stats
),
2595 ATTR_LIST(mb_max_to_scan
),
2596 ATTR_LIST(mb_min_to_scan
),
2597 ATTR_LIST(mb_order2_req
),
2598 ATTR_LIST(mb_stream_req
),
2599 ATTR_LIST(mb_group_prealloc
),
2600 ATTR_LIST(max_writeback_mb_bump
),
2604 /* Features this copy of ext4 supports */
2605 EXT4_INFO_ATTR(lazy_itable_init
);
2606 EXT4_INFO_ATTR(batched_discard
);
2608 static struct attribute
*ext4_feat_attrs
[] = {
2609 ATTR_LIST(lazy_itable_init
),
2610 ATTR_LIST(batched_discard
),
2614 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2615 struct attribute
*attr
, char *buf
)
2617 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2619 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2621 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2624 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2625 struct attribute
*attr
,
2626 const char *buf
, size_t len
)
2628 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2630 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2632 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2635 static void ext4_sb_release(struct kobject
*kobj
)
2637 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2639 complete(&sbi
->s_kobj_unregister
);
2642 static const struct sysfs_ops ext4_attr_ops
= {
2643 .show
= ext4_attr_show
,
2644 .store
= ext4_attr_store
,
2647 static struct kobj_type ext4_ktype
= {
2648 .default_attrs
= ext4_attrs
,
2649 .sysfs_ops
= &ext4_attr_ops
,
2650 .release
= ext4_sb_release
,
2653 static void ext4_feat_release(struct kobject
*kobj
)
2655 complete(&ext4_feat
->f_kobj_unregister
);
2658 static struct kobj_type ext4_feat_ktype
= {
2659 .default_attrs
= ext4_feat_attrs
,
2660 .sysfs_ops
= &ext4_attr_ops
,
2661 .release
= ext4_feat_release
,
2665 * Check whether this filesystem can be mounted based on
2666 * the features present and the RDONLY/RDWR mount requested.
2667 * Returns 1 if this filesystem can be mounted as requested,
2668 * 0 if it cannot be.
2670 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2672 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2673 ext4_msg(sb
, KERN_ERR
,
2674 "Couldn't mount because of "
2675 "unsupported optional features (%x)",
2676 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2677 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2684 /* Check that feature set is OK for a read-write mount */
2685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2686 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2687 "unsupported optional features (%x)",
2688 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2689 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2693 * Large file size enabled file system can only be mounted
2694 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2696 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2697 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2698 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2699 "cannot be mounted RDWR without "
2704 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2705 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2706 ext4_msg(sb
, KERN_ERR
,
2707 "Can't support bigalloc feature without "
2708 "extents feature\n");
2715 * This function is called once a day if we have errors logged
2716 * on the file system
2718 static void print_daily_error_info(unsigned long arg
)
2720 struct super_block
*sb
= (struct super_block
*) arg
;
2721 struct ext4_sb_info
*sbi
;
2722 struct ext4_super_block
*es
;
2727 if (es
->s_error_count
)
2728 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2729 le32_to_cpu(es
->s_error_count
));
2730 if (es
->s_first_error_time
) {
2731 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2732 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2733 (int) sizeof(es
->s_first_error_func
),
2734 es
->s_first_error_func
,
2735 le32_to_cpu(es
->s_first_error_line
));
2736 if (es
->s_first_error_ino
)
2737 printk(": inode %u",
2738 le32_to_cpu(es
->s_first_error_ino
));
2739 if (es
->s_first_error_block
)
2740 printk(": block %llu", (unsigned long long)
2741 le64_to_cpu(es
->s_first_error_block
));
2744 if (es
->s_last_error_time
) {
2745 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2746 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2747 (int) sizeof(es
->s_last_error_func
),
2748 es
->s_last_error_func
,
2749 le32_to_cpu(es
->s_last_error_line
));
2750 if (es
->s_last_error_ino
)
2751 printk(": inode %u",
2752 le32_to_cpu(es
->s_last_error_ino
));
2753 if (es
->s_last_error_block
)
2754 printk(": block %llu", (unsigned long long)
2755 le64_to_cpu(es
->s_last_error_block
));
2758 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2761 /* Find next suitable group and run ext4_init_inode_table */
2762 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2764 struct ext4_group_desc
*gdp
= NULL
;
2765 ext4_group_t group
, ngroups
;
2766 struct super_block
*sb
;
2767 unsigned long timeout
= 0;
2771 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2773 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2774 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2780 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2784 if (group
== ngroups
)
2789 ret
= ext4_init_inode_table(sb
, group
,
2790 elr
->lr_timeout
? 0 : 1);
2791 if (elr
->lr_timeout
== 0) {
2792 timeout
= (jiffies
- timeout
) *
2793 elr
->lr_sbi
->s_li_wait_mult
;
2794 elr
->lr_timeout
= timeout
;
2796 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2797 elr
->lr_next_group
= group
+ 1;
2804 * Remove lr_request from the list_request and free the
2805 * request structure. Should be called with li_list_mtx held
2807 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2809 struct ext4_sb_info
*sbi
;
2816 list_del(&elr
->lr_request
);
2817 sbi
->s_li_request
= NULL
;
2821 static void ext4_unregister_li_request(struct super_block
*sb
)
2823 mutex_lock(&ext4_li_mtx
);
2824 if (!ext4_li_info
) {
2825 mutex_unlock(&ext4_li_mtx
);
2829 mutex_lock(&ext4_li_info
->li_list_mtx
);
2830 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2831 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2832 mutex_unlock(&ext4_li_mtx
);
2835 static struct task_struct
*ext4_lazyinit_task
;
2838 * This is the function where ext4lazyinit thread lives. It walks
2839 * through the request list searching for next scheduled filesystem.
2840 * When such a fs is found, run the lazy initialization request
2841 * (ext4_rn_li_request) and keep track of the time spend in this
2842 * function. Based on that time we compute next schedule time of
2843 * the request. When walking through the list is complete, compute
2844 * next waking time and put itself into sleep.
2846 static int ext4_lazyinit_thread(void *arg
)
2848 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2849 struct list_head
*pos
, *n
;
2850 struct ext4_li_request
*elr
;
2851 unsigned long next_wakeup
, cur
;
2853 BUG_ON(NULL
== eli
);
2857 next_wakeup
= MAX_JIFFY_OFFSET
;
2859 mutex_lock(&eli
->li_list_mtx
);
2860 if (list_empty(&eli
->li_request_list
)) {
2861 mutex_unlock(&eli
->li_list_mtx
);
2865 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2866 elr
= list_entry(pos
, struct ext4_li_request
,
2869 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2870 if (ext4_run_li_request(elr
) != 0) {
2871 /* error, remove the lazy_init job */
2872 ext4_remove_li_request(elr
);
2877 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2878 next_wakeup
= elr
->lr_next_sched
;
2880 mutex_unlock(&eli
->li_list_mtx
);
2882 if (freezing(current
))
2886 if ((time_after_eq(cur
, next_wakeup
)) ||
2887 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2892 schedule_timeout_interruptible(next_wakeup
- cur
);
2894 if (kthread_should_stop()) {
2895 ext4_clear_request_list();
2902 * It looks like the request list is empty, but we need
2903 * to check it under the li_list_mtx lock, to prevent any
2904 * additions into it, and of course we should lock ext4_li_mtx
2905 * to atomically free the list and ext4_li_info, because at
2906 * this point another ext4 filesystem could be registering
2909 mutex_lock(&ext4_li_mtx
);
2910 mutex_lock(&eli
->li_list_mtx
);
2911 if (!list_empty(&eli
->li_request_list
)) {
2912 mutex_unlock(&eli
->li_list_mtx
);
2913 mutex_unlock(&ext4_li_mtx
);
2916 mutex_unlock(&eli
->li_list_mtx
);
2917 kfree(ext4_li_info
);
2918 ext4_li_info
= NULL
;
2919 mutex_unlock(&ext4_li_mtx
);
2924 static void ext4_clear_request_list(void)
2926 struct list_head
*pos
, *n
;
2927 struct ext4_li_request
*elr
;
2929 mutex_lock(&ext4_li_info
->li_list_mtx
);
2930 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2931 elr
= list_entry(pos
, struct ext4_li_request
,
2933 ext4_remove_li_request(elr
);
2935 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2938 static int ext4_run_lazyinit_thread(void)
2940 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2941 ext4_li_info
, "ext4lazyinit");
2942 if (IS_ERR(ext4_lazyinit_task
)) {
2943 int err
= PTR_ERR(ext4_lazyinit_task
);
2944 ext4_clear_request_list();
2945 kfree(ext4_li_info
);
2946 ext4_li_info
= NULL
;
2947 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2948 "initialization thread\n",
2952 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2957 * Check whether it make sense to run itable init. thread or not.
2958 * If there is at least one uninitialized inode table, return
2959 * corresponding group number, else the loop goes through all
2960 * groups and return total number of groups.
2962 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2964 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2965 struct ext4_group_desc
*gdp
= NULL
;
2967 for (group
= 0; group
< ngroups
; group
++) {
2968 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2972 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2979 static int ext4_li_info_new(void)
2981 struct ext4_lazy_init
*eli
= NULL
;
2983 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2987 INIT_LIST_HEAD(&eli
->li_request_list
);
2988 mutex_init(&eli
->li_list_mtx
);
2990 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2997 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
3000 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3001 struct ext4_li_request
*elr
;
3004 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
3010 elr
->lr_next_group
= start
;
3013 * Randomize first schedule time of the request to
3014 * spread the inode table initialization requests
3017 get_random_bytes(&rnd
, sizeof(rnd
));
3018 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
3019 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
3024 static int ext4_register_li_request(struct super_block
*sb
,
3025 ext4_group_t first_not_zeroed
)
3027 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3028 struct ext4_li_request
*elr
;
3029 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3032 if (sbi
->s_li_request
!= NULL
) {
3034 * Reset timeout so it can be computed again, because
3035 * s_li_wait_mult might have changed.
3037 sbi
->s_li_request
->lr_timeout
= 0;
3041 if (first_not_zeroed
== ngroups
||
3042 (sb
->s_flags
& MS_RDONLY
) ||
3043 !test_opt(sb
, INIT_INODE_TABLE
))
3046 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3050 mutex_lock(&ext4_li_mtx
);
3052 if (NULL
== ext4_li_info
) {
3053 ret
= ext4_li_info_new();
3058 mutex_lock(&ext4_li_info
->li_list_mtx
);
3059 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3060 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3062 sbi
->s_li_request
= elr
;
3064 * set elr to NULL here since it has been inserted to
3065 * the request_list and the removal and free of it is
3066 * handled by ext4_clear_request_list from now on.
3070 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3071 ret
= ext4_run_lazyinit_thread();
3076 mutex_unlock(&ext4_li_mtx
);
3083 * We do not need to lock anything since this is called on
3086 static void ext4_destroy_lazyinit_thread(void)
3089 * If thread exited earlier
3090 * there's nothing to be done.
3092 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3095 kthread_stop(ext4_lazyinit_task
);
3098 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3099 __releases(kernel_lock
)
3100 __acquires(kernel_lock
)
3102 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3103 struct buffer_head
*bh
;
3104 struct ext4_super_block
*es
= NULL
;
3105 struct ext4_sb_info
*sbi
;
3107 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3108 ext4_fsblk_t logical_sb_block
;
3109 unsigned long offset
= 0;
3110 unsigned long journal_devnum
= 0;
3111 unsigned long def_mount_opts
;
3116 int blocksize
, clustersize
;
3117 unsigned int db_count
;
3119 int needs_recovery
, has_huge_files
, has_bigalloc
;
3122 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3123 ext4_group_t first_not_zeroed
;
3125 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3129 sbi
->s_blockgroup_lock
=
3130 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3131 if (!sbi
->s_blockgroup_lock
) {
3135 sb
->s_fs_info
= sbi
;
3136 sbi
->s_mount_opt
= 0;
3137 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3138 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3139 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3140 sbi
->s_sb_block
= sb_block
;
3141 if (sb
->s_bdev
->bd_part
)
3142 sbi
->s_sectors_written_start
=
3143 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3145 /* Cleanup superblock name */
3146 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3150 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3152 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3157 * The ext4 superblock will not be buffer aligned for other than 1kB
3158 * block sizes. We need to calculate the offset from buffer start.
3160 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3161 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3162 offset
= do_div(logical_sb_block
, blocksize
);
3164 logical_sb_block
= sb_block
;
3167 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3168 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3172 * Note: s_es must be initialized as soon as possible because
3173 * some ext4 macro-instructions depend on its value
3175 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3177 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3178 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3180 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3182 /* Set defaults before we parse the mount options */
3183 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3184 set_opt(sb
, INIT_INODE_TABLE
);
3185 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3187 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3188 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3192 if (def_mount_opts
& EXT4_DEFM_UID16
)
3193 set_opt(sb
, NO_UID32
);
3194 /* xattr user namespace & acls are now defaulted on */
3195 #ifdef CONFIG_EXT4_FS_XATTR
3196 set_opt(sb
, XATTR_USER
);
3198 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3199 set_opt(sb
, POSIX_ACL
);
3201 set_opt(sb
, MBLK_IO_SUBMIT
);
3202 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3203 set_opt(sb
, JOURNAL_DATA
);
3204 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3205 set_opt(sb
, ORDERED_DATA
);
3206 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3207 set_opt(sb
, WRITEBACK_DATA
);
3209 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3210 set_opt(sb
, ERRORS_PANIC
);
3211 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3212 set_opt(sb
, ERRORS_CONT
);
3214 set_opt(sb
, ERRORS_RO
);
3215 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3216 set_opt(sb
, BLOCK_VALIDITY
);
3217 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3218 set_opt(sb
, DISCARD
);
3220 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3221 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3222 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3223 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3224 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3226 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3227 set_opt(sb
, BARRIER
);
3230 * enable delayed allocation by default
3231 * Use -o nodelalloc to turn it off
3233 if (!IS_EXT3_SB(sb
) &&
3234 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3235 set_opt(sb
, DELALLOC
);
3238 * set default s_li_wait_mult for lazyinit, for the case there is
3239 * no mount option specified.
3241 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3243 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3244 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3245 ext4_msg(sb
, KERN_WARNING
,
3246 "failed to parse options in superblock: %s",
3247 sbi
->s_es
->s_mount_opts
);
3249 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3250 &journal_ioprio
, NULL
, 0))
3253 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3254 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3255 "with data=journal disables delayed "
3256 "allocation and O_DIRECT support!\n");
3257 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3258 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3259 "both data=journal and delalloc");
3262 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3263 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3264 "both data=journal and delalloc");
3267 if (test_opt(sb
, DELALLOC
))
3268 clear_opt(sb
, DELALLOC
);
3271 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3272 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3273 if (blocksize
< PAGE_SIZE
) {
3274 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3275 "dioread_nolock if block size != PAGE_SIZE");
3280 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3281 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3283 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3284 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3285 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3286 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3287 ext4_msg(sb
, KERN_WARNING
,
3288 "feature flags set on rev 0 fs, "
3289 "running e2fsck is recommended");
3291 if (IS_EXT2_SB(sb
)) {
3292 if (ext2_feature_set_ok(sb
))
3293 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3294 "using the ext4 subsystem");
3296 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3297 "to feature incompatibilities");
3302 if (IS_EXT3_SB(sb
)) {
3303 if (ext3_feature_set_ok(sb
))
3304 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3305 "using the ext4 subsystem");
3307 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3308 "to feature incompatibilities");
3314 * Check feature flags regardless of the revision level, since we
3315 * previously didn't change the revision level when setting the flags,
3316 * so there is a chance incompat flags are set on a rev 0 filesystem.
3318 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3321 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3322 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3323 ext4_msg(sb
, KERN_ERR
,
3324 "Unsupported filesystem blocksize %d", blocksize
);
3328 if (sb
->s_blocksize
!= blocksize
) {
3329 /* Validate the filesystem blocksize */
3330 if (!sb_set_blocksize(sb
, blocksize
)) {
3331 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3337 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3338 offset
= do_div(logical_sb_block
, blocksize
);
3339 bh
= sb_bread(sb
, logical_sb_block
);
3341 ext4_msg(sb
, KERN_ERR
,
3342 "Can't read superblock on 2nd try");
3345 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3347 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3348 ext4_msg(sb
, KERN_ERR
,
3349 "Magic mismatch, very weird!");
3354 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3355 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3356 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3358 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3360 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3361 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3362 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3364 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3365 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3366 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3367 (!is_power_of_2(sbi
->s_inode_size
)) ||
3368 (sbi
->s_inode_size
> blocksize
)) {
3369 ext4_msg(sb
, KERN_ERR
,
3370 "unsupported inode size: %d",
3374 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3375 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3378 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3379 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3380 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3381 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3382 !is_power_of_2(sbi
->s_desc_size
)) {
3383 ext4_msg(sb
, KERN_ERR
,
3384 "unsupported descriptor size %lu",
3389 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3391 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3392 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3393 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3396 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3397 if (sbi
->s_inodes_per_block
== 0)
3399 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3400 sbi
->s_inodes_per_block
;
3401 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3403 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3404 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3405 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3407 for (i
= 0; i
< 4; i
++)
3408 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3409 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3410 i
= le32_to_cpu(es
->s_flags
);
3411 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3412 sbi
->s_hash_unsigned
= 3;
3413 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3414 #ifdef __CHAR_UNSIGNED__
3415 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3416 sbi
->s_hash_unsigned
= 3;
3418 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3423 /* Handle clustersize */
3424 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3425 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3426 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3428 if (clustersize
< blocksize
) {
3429 ext4_msg(sb
, KERN_ERR
,
3430 "cluster size (%d) smaller than "
3431 "block size (%d)", clustersize
, blocksize
);
3434 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3435 le32_to_cpu(es
->s_log_block_size
);
3436 sbi
->s_clusters_per_group
=
3437 le32_to_cpu(es
->s_clusters_per_group
);
3438 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3439 ext4_msg(sb
, KERN_ERR
,
3440 "#clusters per group too big: %lu",
3441 sbi
->s_clusters_per_group
);
3444 if (sbi
->s_blocks_per_group
!=
3445 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3446 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3447 "clusters per group (%lu) inconsistent",
3448 sbi
->s_blocks_per_group
,
3449 sbi
->s_clusters_per_group
);
3453 if (clustersize
!= blocksize
) {
3454 ext4_warning(sb
, "fragment/cluster size (%d) != "
3455 "block size (%d)", clustersize
,
3457 clustersize
= blocksize
;
3459 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3460 ext4_msg(sb
, KERN_ERR
,
3461 "#blocks per group too big: %lu",
3462 sbi
->s_blocks_per_group
);
3465 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3466 sbi
->s_cluster_bits
= 0;
3468 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3470 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3471 ext4_msg(sb
, KERN_ERR
,
3472 "#inodes per group too big: %lu",
3473 sbi
->s_inodes_per_group
);
3478 * Test whether we have more sectors than will fit in sector_t,
3479 * and whether the max offset is addressable by the page cache.
3481 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3482 ext4_blocks_count(es
));
3484 ext4_msg(sb
, KERN_ERR
, "filesystem"
3485 " too large to mount safely on this system");
3486 if (sizeof(sector_t
) < 8)
3487 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3492 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3495 /* check blocks count against device size */
3496 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3497 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3498 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3499 "exceeds size of device (%llu blocks)",
3500 ext4_blocks_count(es
), blocks_count
);
3505 * It makes no sense for the first data block to be beyond the end
3506 * of the filesystem.
3508 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3509 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3510 "block %u is beyond end of filesystem (%llu)",
3511 le32_to_cpu(es
->s_first_data_block
),
3512 ext4_blocks_count(es
));
3515 blocks_count
= (ext4_blocks_count(es
) -
3516 le32_to_cpu(es
->s_first_data_block
) +
3517 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3518 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3519 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3520 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3521 "(block count %llu, first data block %u, "
3522 "blocks per group %lu)", sbi
->s_groups_count
,
3523 ext4_blocks_count(es
),
3524 le32_to_cpu(es
->s_first_data_block
),
3525 EXT4_BLOCKS_PER_GROUP(sb
));
3528 sbi
->s_groups_count
= blocks_count
;
3529 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3530 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3531 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3532 EXT4_DESC_PER_BLOCK(sb
);
3533 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3534 sizeof(struct buffer_head
*),
3536 if (sbi
->s_group_desc
== NULL
) {
3537 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3541 #ifdef CONFIG_PROC_FS
3543 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3546 bgl_lock_init(sbi
->s_blockgroup_lock
);
3548 for (i
= 0; i
< db_count
; i
++) {
3549 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3550 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3551 if (!sbi
->s_group_desc
[i
]) {
3552 ext4_msg(sb
, KERN_ERR
,
3553 "can't read group descriptor %d", i
);
3558 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3559 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3563 if (!ext4_fill_flex_info(sb
)) {
3564 ext4_msg(sb
, KERN_ERR
,
3565 "unable to initialize "
3566 "flex_bg meta info!");
3570 sbi
->s_gdb_count
= db_count
;
3571 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3572 spin_lock_init(&sbi
->s_next_gen_lock
);
3574 init_timer(&sbi
->s_err_report
);
3575 sbi
->s_err_report
.function
= print_daily_error_info
;
3576 sbi
->s_err_report
.data
= (unsigned long) sb
;
3578 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3579 ext4_count_free_blocks(sb
));
3581 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3582 ext4_count_free_inodes(sb
));
3585 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3586 ext4_count_dirs(sb
));
3589 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3592 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3596 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3597 sbi
->s_max_writeback_mb_bump
= 128;
3600 * set up enough so that it can read an inode
3602 if (!test_opt(sb
, NOLOAD
) &&
3603 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3604 sb
->s_op
= &ext4_sops
;
3606 sb
->s_op
= &ext4_nojournal_sops
;
3607 sb
->s_export_op
= &ext4_export_ops
;
3608 sb
->s_xattr
= ext4_xattr_handlers
;
3610 sb
->s_qcop
= &ext4_qctl_operations
;
3611 sb
->dq_op
= &ext4_quota_operations
;
3613 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3615 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3616 mutex_init(&sbi
->s_orphan_lock
);
3617 sbi
->s_resize_flags
= 0;
3621 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3622 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3623 EXT4_FEATURE_INCOMPAT_RECOVER
));
3625 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3626 !(sb
->s_flags
& MS_RDONLY
))
3627 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3631 * The first inode we look at is the journal inode. Don't try
3632 * root first: it may be modified in the journal!
3634 if (!test_opt(sb
, NOLOAD
) &&
3635 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3636 if (ext4_load_journal(sb
, es
, journal_devnum
))
3638 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3639 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3640 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3641 "suppressed and not mounted read-only");
3642 goto failed_mount_wq
;
3644 clear_opt(sb
, DATA_FLAGS
);
3645 sbi
->s_journal
= NULL
;
3650 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3651 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3652 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3653 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3654 goto failed_mount_wq
;
3657 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3658 jbd2_journal_set_features(sbi
->s_journal
,
3659 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3660 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3661 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3662 jbd2_journal_set_features(sbi
->s_journal
,
3663 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3664 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3665 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3667 jbd2_journal_clear_features(sbi
->s_journal
,
3668 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3669 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3672 /* We have now updated the journal if required, so we can
3673 * validate the data journaling mode. */
3674 switch (test_opt(sb
, DATA_FLAGS
)) {
3676 /* No mode set, assume a default based on the journal
3677 * capabilities: ORDERED_DATA if the journal can
3678 * cope, else JOURNAL_DATA
3680 if (jbd2_journal_check_available_features
3681 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3682 set_opt(sb
, ORDERED_DATA
);
3684 set_opt(sb
, JOURNAL_DATA
);
3687 case EXT4_MOUNT_ORDERED_DATA
:
3688 case EXT4_MOUNT_WRITEBACK_DATA
:
3689 if (!jbd2_journal_check_available_features
3690 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3691 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3692 "requested data journaling mode");
3693 goto failed_mount_wq
;
3698 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3701 * The journal may have updated the bg summary counts, so we
3702 * need to update the global counters.
3704 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3705 ext4_count_free_blocks(sb
));
3706 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3707 ext4_count_free_inodes(sb
));
3708 percpu_counter_set(&sbi
->s_dirs_counter
,
3709 ext4_count_dirs(sb
));
3710 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3714 * The maximum number of concurrent works can be high and
3715 * concurrency isn't really necessary. Limit it to 1.
3717 EXT4_SB(sb
)->dio_unwritten_wq
=
3718 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3719 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3720 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3721 goto failed_mount_wq
;
3725 * The jbd2_journal_load will have done any necessary log recovery,
3726 * so we can safely mount the rest of the filesystem now.
3729 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3731 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3732 ret
= PTR_ERR(root
);
3736 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3737 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3740 sb
->s_root
= d_alloc_root(root
);
3742 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3747 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3749 /* determine the minimum size of new large inodes, if present */
3750 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3751 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3752 EXT4_GOOD_OLD_INODE_SIZE
;
3753 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3754 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3755 if (sbi
->s_want_extra_isize
<
3756 le16_to_cpu(es
->s_want_extra_isize
))
3757 sbi
->s_want_extra_isize
=
3758 le16_to_cpu(es
->s_want_extra_isize
);
3759 if (sbi
->s_want_extra_isize
<
3760 le16_to_cpu(es
->s_min_extra_isize
))
3761 sbi
->s_want_extra_isize
=
3762 le16_to_cpu(es
->s_min_extra_isize
);
3765 /* Check if enough inode space is available */
3766 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3767 sbi
->s_inode_size
) {
3768 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3769 EXT4_GOOD_OLD_INODE_SIZE
;
3770 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3774 err
= ext4_setup_system_zone(sb
);
3776 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3782 err
= ext4_mb_init(sb
, needs_recovery
);
3784 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3789 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3793 sbi
->s_kobj
.kset
= ext4_kset
;
3794 init_completion(&sbi
->s_kobj_unregister
);
3795 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3798 ext4_mb_release(sb
);
3799 ext4_ext_release(sb
);
3803 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3804 ext4_orphan_cleanup(sb
, es
);
3805 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3806 if (needs_recovery
) {
3807 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3808 ext4_mark_recovery_complete(sb
, es
);
3810 if (EXT4_SB(sb
)->s_journal
) {
3811 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3812 descr
= " journalled data mode";
3813 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3814 descr
= " ordered data mode";
3816 descr
= " writeback data mode";
3818 descr
= "out journal";
3820 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3821 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3822 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3824 if (es
->s_error_count
)
3825 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3832 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3838 ext4_msg(sb
, KERN_ERR
, "mount failed");
3839 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3841 ext4_release_system_zone(sb
);
3842 if (sbi
->s_journal
) {
3843 jbd2_journal_destroy(sbi
->s_journal
);
3844 sbi
->s_journal
= NULL
;
3847 del_timer(&sbi
->s_err_report
);
3848 if (sbi
->s_flex_groups
)
3849 ext4_kvfree(sbi
->s_flex_groups
);
3850 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
3851 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3852 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3853 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
3855 kthread_stop(sbi
->s_mmp_tsk
);
3857 for (i
= 0; i
< db_count
; i
++)
3858 brelse(sbi
->s_group_desc
[i
]);
3859 ext4_kvfree(sbi
->s_group_desc
);
3862 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3865 for (i
= 0; i
< MAXQUOTAS
; i
++)
3866 kfree(sbi
->s_qf_names
[i
]);
3868 ext4_blkdev_remove(sbi
);
3871 sb
->s_fs_info
= NULL
;
3872 kfree(sbi
->s_blockgroup_lock
);
3880 * Setup any per-fs journal parameters now. We'll do this both on
3881 * initial mount, once the journal has been initialised but before we've
3882 * done any recovery; and again on any subsequent remount.
3884 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3886 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3888 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3889 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3890 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3892 write_lock(&journal
->j_state_lock
);
3893 if (test_opt(sb
, BARRIER
))
3894 journal
->j_flags
|= JBD2_BARRIER
;
3896 journal
->j_flags
&= ~JBD2_BARRIER
;
3897 if (test_opt(sb
, DATA_ERR_ABORT
))
3898 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3900 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3901 write_unlock(&journal
->j_state_lock
);
3904 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3905 unsigned int journal_inum
)
3907 struct inode
*journal_inode
;
3910 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3912 /* First, test for the existence of a valid inode on disk. Bad
3913 * things happen if we iget() an unused inode, as the subsequent
3914 * iput() will try to delete it. */
3916 journal_inode
= ext4_iget(sb
, journal_inum
);
3917 if (IS_ERR(journal_inode
)) {
3918 ext4_msg(sb
, KERN_ERR
, "no journal found");
3921 if (!journal_inode
->i_nlink
) {
3922 make_bad_inode(journal_inode
);
3923 iput(journal_inode
);
3924 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3928 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3929 journal_inode
, journal_inode
->i_size
);
3930 if (!S_ISREG(journal_inode
->i_mode
)) {
3931 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3932 iput(journal_inode
);
3936 journal
= jbd2_journal_init_inode(journal_inode
);
3938 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3939 iput(journal_inode
);
3942 journal
->j_private
= sb
;
3943 ext4_init_journal_params(sb
, journal
);
3947 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3950 struct buffer_head
*bh
;
3954 int hblock
, blocksize
;
3955 ext4_fsblk_t sb_block
;
3956 unsigned long offset
;
3957 struct ext4_super_block
*es
;
3958 struct block_device
*bdev
;
3960 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3962 bdev
= ext4_blkdev_get(j_dev
, sb
);
3966 blocksize
= sb
->s_blocksize
;
3967 hblock
= bdev_logical_block_size(bdev
);
3968 if (blocksize
< hblock
) {
3969 ext4_msg(sb
, KERN_ERR
,
3970 "blocksize too small for journal device");
3974 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3975 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3976 set_blocksize(bdev
, blocksize
);
3977 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3978 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3979 "external journal");
3983 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3984 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3985 !(le32_to_cpu(es
->s_feature_incompat
) &
3986 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3987 ext4_msg(sb
, KERN_ERR
, "external journal has "
3993 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3994 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3999 len
= ext4_blocks_count(es
);
4000 start
= sb_block
+ 1;
4001 brelse(bh
); /* we're done with the superblock */
4003 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
4004 start
, len
, blocksize
);
4006 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
4009 journal
->j_private
= sb
;
4010 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
4011 wait_on_buffer(journal
->j_sb_buffer
);
4012 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
4013 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
4016 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
4017 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
4018 "user (unsupported) - %d",
4019 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
4022 EXT4_SB(sb
)->journal_bdev
= bdev
;
4023 ext4_init_journal_params(sb
, journal
);
4027 jbd2_journal_destroy(journal
);
4029 ext4_blkdev_put(bdev
);
4033 static int ext4_load_journal(struct super_block
*sb
,
4034 struct ext4_super_block
*es
,
4035 unsigned long journal_devnum
)
4038 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
4041 int really_read_only
;
4043 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4045 if (journal_devnum
&&
4046 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4047 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
4048 "numbers have changed");
4049 journal_dev
= new_decode_dev(journal_devnum
);
4051 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
4053 really_read_only
= bdev_read_only(sb
->s_bdev
);
4056 * Are we loading a blank journal or performing recovery after a
4057 * crash? For recovery, we need to check in advance whether we
4058 * can get read-write access to the device.
4060 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
4061 if (sb
->s_flags
& MS_RDONLY
) {
4062 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
4063 "required on readonly filesystem");
4064 if (really_read_only
) {
4065 ext4_msg(sb
, KERN_ERR
, "write access "
4066 "unavailable, cannot proceed");
4069 ext4_msg(sb
, KERN_INFO
, "write access will "
4070 "be enabled during recovery");
4074 if (journal_inum
&& journal_dev
) {
4075 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4076 "and inode journals!");
4081 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4084 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4088 if (!(journal
->j_flags
& JBD2_BARRIER
))
4089 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4091 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
4092 err
= jbd2_journal_update_format(journal
);
4094 ext4_msg(sb
, KERN_ERR
, "error updating journal");
4095 jbd2_journal_destroy(journal
);
4100 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4101 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4103 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4105 memcpy(save
, ((char *) es
) +
4106 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4107 err
= jbd2_journal_load(journal
);
4109 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4110 save
, EXT4_S_ERR_LEN
);
4115 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4116 jbd2_journal_destroy(journal
);
4120 EXT4_SB(sb
)->s_journal
= journal
;
4121 ext4_clear_journal_err(sb
, es
);
4123 if (!really_read_only
&& journal_devnum
&&
4124 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4125 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4127 /* Make sure we flush the recovery flag to disk. */
4128 ext4_commit_super(sb
, 1);
4134 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4136 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4137 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4140 if (!sbh
|| block_device_ejected(sb
))
4142 if (buffer_write_io_error(sbh
)) {
4144 * Oh, dear. A previous attempt to write the
4145 * superblock failed. This could happen because the
4146 * USB device was yanked out. Or it could happen to
4147 * be a transient write error and maybe the block will
4148 * be remapped. Nothing we can do but to retry the
4149 * write and hope for the best.
4151 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4152 "superblock detected");
4153 clear_buffer_write_io_error(sbh
);
4154 set_buffer_uptodate(sbh
);
4157 * If the file system is mounted read-only, don't update the
4158 * superblock write time. This avoids updating the superblock
4159 * write time when we are mounting the root file system
4160 * read/only but we need to replay the journal; at that point,
4161 * for people who are east of GMT and who make their clock
4162 * tick in localtime for Windows bug-for-bug compatibility,
4163 * the clock is set in the future, and this will cause e2fsck
4164 * to complain and force a full file system check.
4166 if (!(sb
->s_flags
& MS_RDONLY
))
4167 es
->s_wtime
= cpu_to_le32(get_seconds());
4168 if (sb
->s_bdev
->bd_part
)
4169 es
->s_kbytes_written
=
4170 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4171 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4172 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4174 es
->s_kbytes_written
=
4175 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4176 ext4_free_blocks_count_set(es
,
4177 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4178 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4179 es
->s_free_inodes_count
=
4180 cpu_to_le32(percpu_counter_sum_positive(
4181 &EXT4_SB(sb
)->s_freeinodes_counter
));
4183 BUFFER_TRACE(sbh
, "marking dirty");
4184 mark_buffer_dirty(sbh
);
4186 error
= sync_dirty_buffer(sbh
);
4190 error
= buffer_write_io_error(sbh
);
4192 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4194 clear_buffer_write_io_error(sbh
);
4195 set_buffer_uptodate(sbh
);
4202 * Have we just finished recovery? If so, and if we are mounting (or
4203 * remounting) the filesystem readonly, then we will end up with a
4204 * consistent fs on disk. Record that fact.
4206 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4207 struct ext4_super_block
*es
)
4209 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4211 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4212 BUG_ON(journal
!= NULL
);
4215 jbd2_journal_lock_updates(journal
);
4216 if (jbd2_journal_flush(journal
) < 0)
4219 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4220 sb
->s_flags
& MS_RDONLY
) {
4221 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4222 ext4_commit_super(sb
, 1);
4226 jbd2_journal_unlock_updates(journal
);
4230 * If we are mounting (or read-write remounting) a filesystem whose journal
4231 * has recorded an error from a previous lifetime, move that error to the
4232 * main filesystem now.
4234 static void ext4_clear_journal_err(struct super_block
*sb
,
4235 struct ext4_super_block
*es
)
4241 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4243 journal
= EXT4_SB(sb
)->s_journal
;
4246 * Now check for any error status which may have been recorded in the
4247 * journal by a prior ext4_error() or ext4_abort()
4250 j_errno
= jbd2_journal_errno(journal
);
4254 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4255 ext4_warning(sb
, "Filesystem error recorded "
4256 "from previous mount: %s", errstr
);
4257 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4259 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4260 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4261 ext4_commit_super(sb
, 1);
4263 jbd2_journal_clear_err(journal
);
4268 * Force the running and committing transactions to commit,
4269 * and wait on the commit.
4271 int ext4_force_commit(struct super_block
*sb
)
4276 if (sb
->s_flags
& MS_RDONLY
)
4279 journal
= EXT4_SB(sb
)->s_journal
;
4281 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4282 ret
= ext4_journal_force_commit(journal
);
4288 static void ext4_write_super(struct super_block
*sb
)
4291 ext4_commit_super(sb
, 1);
4295 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4299 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4301 trace_ext4_sync_fs(sb
, wait
);
4302 flush_workqueue(sbi
->dio_unwritten_wq
);
4303 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4305 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4311 * LVM calls this function before a (read-only) snapshot is created. This
4312 * gives us a chance to flush the journal completely and mark the fs clean.
4314 * Note that only this function cannot bring a filesystem to be in a clean
4315 * state independently, because ext4 prevents a new handle from being started
4316 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4319 static int ext4_freeze(struct super_block
*sb
)
4324 if (sb
->s_flags
& MS_RDONLY
)
4327 journal
= EXT4_SB(sb
)->s_journal
;
4329 /* Now we set up the journal barrier. */
4330 jbd2_journal_lock_updates(journal
);
4333 * Don't clear the needs_recovery flag if we failed to flush
4336 error
= jbd2_journal_flush(journal
);
4340 /* Journal blocked and flushed, clear needs_recovery flag. */
4341 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4342 error
= ext4_commit_super(sb
, 1);
4344 /* we rely on s_frozen to stop further updates */
4345 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4350 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4351 * flag here, even though the filesystem is not technically dirty yet.
4353 static int ext4_unfreeze(struct super_block
*sb
)
4355 if (sb
->s_flags
& MS_RDONLY
)
4359 /* Reset the needs_recovery flag before the fs is unlocked. */
4360 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4361 ext4_commit_super(sb
, 1);
4367 * Structure to save mount options for ext4_remount's benefit
4369 struct ext4_mount_options
{
4370 unsigned long s_mount_opt
;
4371 unsigned long s_mount_opt2
;
4374 unsigned long s_commit_interval
;
4375 u32 s_min_batch_time
, s_max_batch_time
;
4378 char *s_qf_names
[MAXQUOTAS
];
4382 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4384 struct ext4_super_block
*es
;
4385 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4386 ext4_fsblk_t n_blocks_count
= 0;
4387 unsigned long old_sb_flags
;
4388 struct ext4_mount_options old_opts
;
4389 int enable_quota
= 0;
4391 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4396 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4398 /* Store the original options */
4400 old_sb_flags
= sb
->s_flags
;
4401 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4402 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4403 old_opts
.s_resuid
= sbi
->s_resuid
;
4404 old_opts
.s_resgid
= sbi
->s_resgid
;
4405 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4406 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4407 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4409 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4410 for (i
= 0; i
< MAXQUOTAS
; i
++)
4411 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4413 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4414 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4417 * Allow the "check" option to be passed as a remount option.
4419 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4420 &n_blocks_count
, 1)) {
4425 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4426 ext4_abort(sb
, "Abort forced by user");
4428 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4429 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4433 if (sbi
->s_journal
) {
4434 ext4_init_journal_params(sb
, sbi
->s_journal
);
4435 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4438 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4439 n_blocks_count
> ext4_blocks_count(es
)) {
4440 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4445 if (*flags
& MS_RDONLY
) {
4446 err
= dquot_suspend(sb
, -1);
4451 * First of all, the unconditional stuff we have to do
4452 * to disable replay of the journal when we next remount
4454 sb
->s_flags
|= MS_RDONLY
;
4457 * OK, test if we are remounting a valid rw partition
4458 * readonly, and if so set the rdonly flag and then
4459 * mark the partition as valid again.
4461 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4462 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4463 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4466 ext4_mark_recovery_complete(sb
, es
);
4468 /* Make sure we can mount this feature set readwrite */
4469 if (!ext4_feature_set_ok(sb
, 0)) {
4474 * Make sure the group descriptor checksums
4475 * are sane. If they aren't, refuse to remount r/w.
4477 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4478 struct ext4_group_desc
*gdp
=
4479 ext4_get_group_desc(sb
, g
, NULL
);
4481 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4482 ext4_msg(sb
, KERN_ERR
,
4483 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4484 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4485 le16_to_cpu(gdp
->bg_checksum
));
4492 * If we have an unprocessed orphan list hanging
4493 * around from a previously readonly bdev mount,
4494 * require a full umount/remount for now.
4496 if (es
->s_last_orphan
) {
4497 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4498 "remount RDWR because of unprocessed "
4499 "orphan inode list. Please "
4500 "umount/remount instead");
4506 * Mounting a RDONLY partition read-write, so reread
4507 * and store the current valid flag. (It may have
4508 * been changed by e2fsck since we originally mounted
4512 ext4_clear_journal_err(sb
, es
);
4513 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4514 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4516 if (!ext4_setup_super(sb
, es
, 0))
4517 sb
->s_flags
&= ~MS_RDONLY
;
4518 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4519 EXT4_FEATURE_INCOMPAT_MMP
))
4520 if (ext4_multi_mount_protect(sb
,
4521 le64_to_cpu(es
->s_mmp_block
))) {
4530 * Reinitialize lazy itable initialization thread based on
4533 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4534 ext4_unregister_li_request(sb
);
4536 ext4_group_t first_not_zeroed
;
4537 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4538 ext4_register_li_request(sb
, first_not_zeroed
);
4541 ext4_setup_system_zone(sb
);
4542 if (sbi
->s_journal
== NULL
)
4543 ext4_commit_super(sb
, 1);
4546 /* Release old quota file names */
4547 for (i
= 0; i
< MAXQUOTAS
; i
++)
4548 if (old_opts
.s_qf_names
[i
] &&
4549 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4550 kfree(old_opts
.s_qf_names
[i
]);
4554 dquot_resume(sb
, -1);
4556 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4561 sb
->s_flags
= old_sb_flags
;
4562 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4563 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4564 sbi
->s_resuid
= old_opts
.s_resuid
;
4565 sbi
->s_resgid
= old_opts
.s_resgid
;
4566 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4567 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4568 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4570 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4571 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4572 if (sbi
->s_qf_names
[i
] &&
4573 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4574 kfree(sbi
->s_qf_names
[i
]);
4575 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4584 * Note: calculating the overhead so we can be compatible with
4585 * historical BSD practice is quite difficult in the face of
4586 * clusters/bigalloc. This is because multiple metadata blocks from
4587 * different block group can end up in the same allocation cluster.
4588 * Calculating the exact overhead in the face of clustered allocation
4589 * requires either O(all block bitmaps) in memory or O(number of block
4590 * groups**2) in time. We will still calculate the superblock for
4591 * older file systems --- and if we come across with a bigalloc file
4592 * system with zero in s_overhead_clusters the estimate will be close to
4593 * correct especially for very large cluster sizes --- but for newer
4594 * file systems, it's better to calculate this figure once at mkfs
4595 * time, and store it in the superblock. If the superblock value is
4596 * present (even for non-bigalloc file systems), we will use it.
4598 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4600 struct super_block
*sb
= dentry
->d_sb
;
4601 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4602 struct ext4_super_block
*es
= sbi
->s_es
;
4603 struct ext4_group_desc
*gdp
;
4607 if (test_opt(sb
, MINIX_DF
)) {
4608 sbi
->s_overhead_last
= 0;
4609 } else if (es
->s_overhead_clusters
) {
4610 sbi
->s_overhead_last
= le32_to_cpu(es
->s_overhead_clusters
);
4611 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4612 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4613 ext4_fsblk_t overhead
= 0;
4616 * Compute the overhead (FS structures). This is constant
4617 * for a given filesystem unless the number of block groups
4618 * changes so we cache the previous value until it does.
4622 * All of the blocks before first_data_block are
4625 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
4628 * Add the overhead found in each block group
4630 for (i
= 0; i
< ngroups
; i
++) {
4631 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
4632 overhead
+= ext4_num_overhead_clusters(sb
, i
, gdp
);
4635 sbi
->s_overhead_last
= overhead
;
4637 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4640 buf
->f_type
= EXT4_SUPER_MAGIC
;
4641 buf
->f_bsize
= sb
->s_blocksize
;
4642 buf
->f_blocks
= (ext4_blocks_count(es
) -
4643 EXT4_C2B(sbi
, sbi
->s_overhead_last
));
4644 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4645 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4646 /* prevent underflow in case that few free space is available */
4647 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4648 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4649 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4651 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4652 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4653 buf
->f_namelen
= EXT4_NAME_LEN
;
4654 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4655 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4656 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4657 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4662 /* Helper function for writing quotas on sync - we need to start transaction
4663 * before quota file is locked for write. Otherwise the are possible deadlocks:
4664 * Process 1 Process 2
4665 * ext4_create() quota_sync()
4666 * jbd2_journal_start() write_dquot()
4667 * dquot_initialize() down(dqio_mutex)
4668 * down(dqio_mutex) jbd2_journal_start()
4674 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4676 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4679 static int ext4_write_dquot(struct dquot
*dquot
)
4683 struct inode
*inode
;
4685 inode
= dquot_to_inode(dquot
);
4686 handle
= ext4_journal_start(inode
,
4687 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4689 return PTR_ERR(handle
);
4690 ret
= dquot_commit(dquot
);
4691 err
= ext4_journal_stop(handle
);
4697 static int ext4_acquire_dquot(struct dquot
*dquot
)
4702 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4703 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4705 return PTR_ERR(handle
);
4706 ret
= dquot_acquire(dquot
);
4707 err
= ext4_journal_stop(handle
);
4713 static int ext4_release_dquot(struct dquot
*dquot
)
4718 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4719 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4720 if (IS_ERR(handle
)) {
4721 /* Release dquot anyway to avoid endless cycle in dqput() */
4722 dquot_release(dquot
);
4723 return PTR_ERR(handle
);
4725 ret
= dquot_release(dquot
);
4726 err
= ext4_journal_stop(handle
);
4732 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4734 /* Are we journaling quotas? */
4735 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4736 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4737 dquot_mark_dquot_dirty(dquot
);
4738 return ext4_write_dquot(dquot
);
4740 return dquot_mark_dquot_dirty(dquot
);
4744 static int ext4_write_info(struct super_block
*sb
, int type
)
4749 /* Data block + inode block */
4750 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4752 return PTR_ERR(handle
);
4753 ret
= dquot_commit_info(sb
, type
);
4754 err
= ext4_journal_stop(handle
);
4761 * Turn on quotas during mount time - we need to find
4762 * the quota file and such...
4764 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4766 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4767 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4771 * Standard function to be called on quota_on
4773 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4778 if (!test_opt(sb
, QUOTA
))
4781 /* Quotafile not on the same filesystem? */
4782 if (path
->mnt
->mnt_sb
!= sb
)
4784 /* Journaling quota? */
4785 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4786 /* Quotafile not in fs root? */
4787 if (path
->dentry
->d_parent
!= sb
->s_root
)
4788 ext4_msg(sb
, KERN_WARNING
,
4789 "Quota file not on filesystem root. "
4790 "Journaled quota will not work");
4794 * When we journal data on quota file, we have to flush journal to see
4795 * all updates to the file when we bypass pagecache...
4797 if (EXT4_SB(sb
)->s_journal
&&
4798 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4800 * We don't need to lock updates but journal_flush() could
4801 * otherwise be livelocked...
4803 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4804 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4805 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4810 return dquot_quota_on(sb
, type
, format_id
, path
);
4813 static int ext4_quota_off(struct super_block
*sb
, int type
)
4815 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4818 /* Force all delayed allocation blocks to be allocated.
4819 * Caller already holds s_umount sem */
4820 if (test_opt(sb
, DELALLOC
))
4821 sync_filesystem(sb
);
4826 /* Update modification times of quota files when userspace can
4827 * start looking at them */
4828 handle
= ext4_journal_start(inode
, 1);
4831 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4832 ext4_mark_inode_dirty(handle
, inode
);
4833 ext4_journal_stop(handle
);
4836 return dquot_quota_off(sb
, type
);
4839 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4840 * acquiring the locks... As quota files are never truncated and quota code
4841 * itself serializes the operations (and no one else should touch the files)
4842 * we don't have to be afraid of races */
4843 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4844 size_t len
, loff_t off
)
4846 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4847 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4849 int offset
= off
& (sb
->s_blocksize
- 1);
4852 struct buffer_head
*bh
;
4853 loff_t i_size
= i_size_read(inode
);
4857 if (off
+len
> i_size
)
4860 while (toread
> 0) {
4861 tocopy
= sb
->s_blocksize
- offset
< toread
?
4862 sb
->s_blocksize
- offset
: toread
;
4863 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4866 if (!bh
) /* A hole? */
4867 memset(data
, 0, tocopy
);
4869 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4879 /* Write to quotafile (we know the transaction is already started and has
4880 * enough credits) */
4881 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4882 const char *data
, size_t len
, loff_t off
)
4884 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4885 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4887 int offset
= off
& (sb
->s_blocksize
- 1);
4888 struct buffer_head
*bh
;
4889 handle_t
*handle
= journal_current_handle();
4891 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4892 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4893 " cancelled because transaction is not started",
4894 (unsigned long long)off
, (unsigned long long)len
);
4898 * Since we account only one data block in transaction credits,
4899 * then it is impossible to cross a block boundary.
4901 if (sb
->s_blocksize
- offset
< len
) {
4902 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4903 " cancelled because not block aligned",
4904 (unsigned long long)off
, (unsigned long long)len
);
4908 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4909 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4912 err
= ext4_journal_get_write_access(handle
, bh
);
4918 memcpy(bh
->b_data
+offset
, data
, len
);
4919 flush_dcache_page(bh
->b_page
);
4921 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4925 mutex_unlock(&inode
->i_mutex
);
4928 if (inode
->i_size
< off
+ len
) {
4929 i_size_write(inode
, off
+ len
);
4930 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4931 ext4_mark_inode_dirty(handle
, inode
);
4933 mutex_unlock(&inode
->i_mutex
);
4939 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4940 const char *dev_name
, void *data
)
4942 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4945 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4946 static inline void register_as_ext2(void)
4948 int err
= register_filesystem(&ext2_fs_type
);
4951 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4954 static inline void unregister_as_ext2(void)
4956 unregister_filesystem(&ext2_fs_type
);
4959 static inline int ext2_feature_set_ok(struct super_block
*sb
)
4961 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
4963 if (sb
->s_flags
& MS_RDONLY
)
4965 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
4969 MODULE_ALIAS("ext2");
4971 static inline void register_as_ext2(void) { }
4972 static inline void unregister_as_ext2(void) { }
4973 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
4976 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4977 static inline void register_as_ext3(void)
4979 int err
= register_filesystem(&ext3_fs_type
);
4982 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4985 static inline void unregister_as_ext3(void)
4987 unregister_filesystem(&ext3_fs_type
);
4990 static inline int ext3_feature_set_ok(struct super_block
*sb
)
4992 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
4994 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
4996 if (sb
->s_flags
& MS_RDONLY
)
4998 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
5002 MODULE_ALIAS("ext3");
5004 static inline void register_as_ext3(void) { }
5005 static inline void unregister_as_ext3(void) { }
5006 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
5009 static struct file_system_type ext4_fs_type
= {
5010 .owner
= THIS_MODULE
,
5012 .mount
= ext4_mount
,
5013 .kill_sb
= kill_block_super
,
5014 .fs_flags
= FS_REQUIRES_DEV
,
5017 static int __init
ext4_init_feat_adverts(void)
5019 struct ext4_features
*ef
;
5022 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
5026 ef
->f_kobj
.kset
= ext4_kset
;
5027 init_completion(&ef
->f_kobj_unregister
);
5028 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
5041 static void ext4_exit_feat_adverts(void)
5043 kobject_put(&ext4_feat
->f_kobj
);
5044 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
5048 /* Shared across all ext4 file systems */
5049 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
5050 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
5052 static int __init
ext4_init_fs(void)
5056 ext4_check_flag_values();
5058 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
5059 mutex_init(&ext4__aio_mutex
[i
]);
5060 init_waitqueue_head(&ext4__ioend_wq
[i
]);
5063 err
= ext4_init_pageio();
5066 err
= ext4_init_system_zone();
5069 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
5072 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
5073 if (!ext4_proc_root
)
5076 err
= ext4_init_feat_adverts();
5080 err
= ext4_init_mballoc();
5084 err
= ext4_init_xattr();
5087 err
= init_inodecache();
5092 err
= register_filesystem(&ext4_fs_type
);
5096 ext4_li_info
= NULL
;
5097 mutex_init(&ext4_li_mtx
);
5100 unregister_as_ext2();
5101 unregister_as_ext3();
5102 destroy_inodecache();
5106 ext4_exit_mballoc();
5108 ext4_exit_feat_adverts();
5110 remove_proc_entry("fs/ext4", NULL
);
5112 kset_unregister(ext4_kset
);
5114 ext4_exit_system_zone();
5120 static void __exit
ext4_exit_fs(void)
5122 ext4_destroy_lazyinit_thread();
5123 unregister_as_ext2();
5124 unregister_as_ext3();
5125 unregister_filesystem(&ext4_fs_type
);
5126 destroy_inodecache();
5128 ext4_exit_mballoc();
5129 ext4_exit_feat_adverts();
5130 remove_proc_entry("fs/ext4", NULL
);
5131 kset_unregister(ext4_kset
);
5132 ext4_exit_system_zone();
5136 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5137 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5138 MODULE_LICENSE("GPL");
5139 module_init(ext4_init_fs
)
5140 module_exit(ext4_exit_fs
)