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 /* Deal with the reporting of failure conditions on a filesystem such as
419 * inconsistencies detected or read IO failures.
421 * On ext2, we can store the error state of the filesystem in the
422 * superblock. That is not possible on ext4, because we may have other
423 * write ordering constraints on the superblock which prevent us from
424 * writing it out straight away; and given that the journal is about to
425 * be aborted, we can't rely on the current, or future, transactions to
426 * write out the superblock safely.
428 * We'll just use the jbd2_journal_abort() error code to record an error in
429 * the journal instead. On recovery, the journal will complain about
430 * that error until we've noted it down and cleared it.
433 static void ext4_handle_error(struct super_block
*sb
)
435 if (sb
->s_flags
& MS_RDONLY
)
438 if (!test_opt(sb
, ERRORS_CONT
)) {
439 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
441 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
443 jbd2_journal_abort(journal
, -EIO
);
445 if (test_opt(sb
, ERRORS_RO
)) {
446 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
447 sb
->s_flags
|= MS_RDONLY
;
449 if (test_opt(sb
, ERRORS_PANIC
))
450 panic("EXT4-fs (device %s): panic forced after error\n",
454 void __ext4_error(struct super_block
*sb
, const char *function
,
455 unsigned int line
, const char *fmt
, ...)
457 struct va_format vaf
;
463 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
464 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
467 ext4_handle_error(sb
);
470 void ext4_error_inode(struct inode
*inode
, const char *function
,
471 unsigned int line
, ext4_fsblk_t block
,
472 const char *fmt
, ...)
475 struct va_format vaf
;
476 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
478 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
479 es
->s_last_error_block
= cpu_to_le64(block
);
480 save_error_info(inode
->i_sb
, function
, line
);
484 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
485 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
487 printk(KERN_CONT
"block %llu: ", block
);
488 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
491 ext4_handle_error(inode
->i_sb
);
494 void ext4_error_file(struct file
*file
, const char *function
,
495 unsigned int line
, ext4_fsblk_t block
,
496 const char *fmt
, ...)
499 struct va_format vaf
;
500 struct ext4_super_block
*es
;
501 struct inode
*inode
= file
->f_dentry
->d_inode
;
502 char pathname
[80], *path
;
504 es
= EXT4_SB(inode
->i_sb
)->s_es
;
505 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
506 save_error_info(inode
->i_sb
, function
, line
);
507 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
512 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
514 printk(KERN_CONT
"block %llu: ", block
);
518 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
521 ext4_handle_error(inode
->i_sb
);
524 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
531 errstr
= "IO failure";
534 errstr
= "Out of memory";
537 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
538 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
539 errstr
= "Journal has aborted";
541 errstr
= "Readonly filesystem";
544 /* If the caller passed in an extra buffer for unknown
545 * errors, textualise them now. Else we just return
548 /* Check for truncated error codes... */
549 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
558 /* __ext4_std_error decodes expected errors from journaling functions
559 * automatically and invokes the appropriate error response. */
561 void __ext4_std_error(struct super_block
*sb
, const char *function
,
562 unsigned int line
, int errno
)
567 /* Special case: if the error is EROFS, and we're not already
568 * inside a transaction, then there's really no point in logging
570 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
571 (sb
->s_flags
& MS_RDONLY
))
574 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
575 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
576 sb
->s_id
, function
, line
, errstr
);
577 save_error_info(sb
, function
, line
);
579 ext4_handle_error(sb
);
583 * ext4_abort is a much stronger failure handler than ext4_error. The
584 * abort function may be used to deal with unrecoverable failures such
585 * as journal IO errors or ENOMEM at a critical moment in log management.
587 * We unconditionally force the filesystem into an ABORT|READONLY state,
588 * unless the error response on the fs has been set to panic in which
589 * case we take the easy way out and panic immediately.
592 void __ext4_abort(struct super_block
*sb
, const char *function
,
593 unsigned int line
, const char *fmt
, ...)
597 save_error_info(sb
, function
, line
);
599 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
605 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
606 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
607 sb
->s_flags
|= MS_RDONLY
;
608 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
609 if (EXT4_SB(sb
)->s_journal
)
610 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
611 save_error_info(sb
, function
, line
);
613 if (test_opt(sb
, ERRORS_PANIC
))
614 panic("EXT4-fs panic from previous error\n");
617 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
619 struct va_format vaf
;
625 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
629 void __ext4_warning(struct super_block
*sb
, const char *function
,
630 unsigned int line
, const char *fmt
, ...)
632 struct va_format vaf
;
638 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
639 sb
->s_id
, function
, line
, &vaf
);
643 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
644 struct super_block
*sb
, ext4_group_t grp
,
645 unsigned long ino
, ext4_fsblk_t block
,
646 const char *fmt
, ...)
650 struct va_format vaf
;
652 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
654 es
->s_last_error_ino
= cpu_to_le32(ino
);
655 es
->s_last_error_block
= cpu_to_le64(block
);
656 __save_error_info(sb
, function
, line
);
662 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
663 sb
->s_id
, function
, line
, grp
);
665 printk(KERN_CONT
"inode %lu: ", ino
);
667 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
668 printk(KERN_CONT
"%pV\n", &vaf
);
671 if (test_opt(sb
, ERRORS_CONT
)) {
672 ext4_commit_super(sb
, 0);
676 ext4_unlock_group(sb
, grp
);
677 ext4_handle_error(sb
);
679 * We only get here in the ERRORS_RO case; relocking the group
680 * may be dangerous, but nothing bad will happen since the
681 * filesystem will have already been marked read/only and the
682 * journal has been aborted. We return 1 as a hint to callers
683 * who might what to use the return value from
684 * ext4_grp_locked_error() to distinguish between the
685 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
686 * aggressively from the ext4 function in question, with a
687 * more appropriate error code.
689 ext4_lock_group(sb
, grp
);
693 void ext4_update_dynamic_rev(struct super_block
*sb
)
695 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
697 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
701 "updating to rev %d because of new feature flag, "
702 "running e2fsck is recommended",
705 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
706 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
707 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
708 /* leave es->s_feature_*compat flags alone */
709 /* es->s_uuid will be set by e2fsck if empty */
712 * The rest of the superblock fields should be zero, and if not it
713 * means they are likely already in use, so leave them alone. We
714 * can leave it up to e2fsck to clean up any inconsistencies there.
719 * Open the external journal device
721 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
723 struct block_device
*bdev
;
724 char b
[BDEVNAME_SIZE
];
726 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
732 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
733 __bdevname(dev
, b
), PTR_ERR(bdev
));
738 * Release the journal device
740 static int ext4_blkdev_put(struct block_device
*bdev
)
742 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
745 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
747 struct block_device
*bdev
;
750 bdev
= sbi
->journal_bdev
;
752 ret
= ext4_blkdev_put(bdev
);
753 sbi
->journal_bdev
= NULL
;
758 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
760 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
763 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
767 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
768 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
770 printk(KERN_ERR
"sb_info orphan list:\n");
771 list_for_each(l
, &sbi
->s_orphan
) {
772 struct inode
*inode
= orphan_list_entry(l
);
774 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
775 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
776 inode
->i_mode
, inode
->i_nlink
,
781 static void ext4_put_super(struct super_block
*sb
)
783 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
784 struct ext4_super_block
*es
= sbi
->s_es
;
787 ext4_unregister_li_request(sb
);
788 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
790 flush_workqueue(sbi
->dio_unwritten_wq
);
791 destroy_workqueue(sbi
->dio_unwritten_wq
);
795 ext4_commit_super(sb
, 1);
797 if (sbi
->s_journal
) {
798 err
= jbd2_journal_destroy(sbi
->s_journal
);
799 sbi
->s_journal
= NULL
;
801 ext4_abort(sb
, "Couldn't clean up the journal");
804 del_timer(&sbi
->s_err_report
);
805 ext4_release_system_zone(sb
);
807 ext4_ext_release(sb
);
808 ext4_xattr_put_super(sb
);
810 if (!(sb
->s_flags
& MS_RDONLY
)) {
811 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
812 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
813 ext4_commit_super(sb
, 1);
816 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
818 kobject_del(&sbi
->s_kobj
);
820 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
821 brelse(sbi
->s_group_desc
[i
]);
822 ext4_kvfree(sbi
->s_group_desc
);
823 ext4_kvfree(sbi
->s_flex_groups
);
824 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
825 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
826 percpu_counter_destroy(&sbi
->s_dirs_counter
);
827 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
830 for (i
= 0; i
< MAXQUOTAS
; i
++)
831 kfree(sbi
->s_qf_names
[i
]);
834 /* Debugging code just in case the in-memory inode orphan list
835 * isn't empty. The on-disk one can be non-empty if we've
836 * detected an error and taken the fs readonly, but the
837 * in-memory list had better be clean by this point. */
838 if (!list_empty(&sbi
->s_orphan
))
839 dump_orphan_list(sb
, sbi
);
840 J_ASSERT(list_empty(&sbi
->s_orphan
));
842 invalidate_bdev(sb
->s_bdev
);
843 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
845 * Invalidate the journal device's buffers. We don't want them
846 * floating about in memory - the physical journal device may
847 * hotswapped, and it breaks the `ro-after' testing code.
849 sync_blockdev(sbi
->journal_bdev
);
850 invalidate_bdev(sbi
->journal_bdev
);
851 ext4_blkdev_remove(sbi
);
854 kthread_stop(sbi
->s_mmp_tsk
);
855 sb
->s_fs_info
= NULL
;
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
861 kobject_put(&sbi
->s_kobj
);
862 wait_for_completion(&sbi
->s_kobj_unregister
);
863 kfree(sbi
->s_blockgroup_lock
);
867 static struct kmem_cache
*ext4_inode_cachep
;
870 * Called inside transaction, so use GFP_NOFS
872 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
874 struct ext4_inode_info
*ei
;
876 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
880 ei
->vfs_inode
.i_version
= 1;
881 ei
->vfs_inode
.i_data
.writeback_index
= 0;
882 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
883 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
884 spin_lock_init(&ei
->i_prealloc_lock
);
885 ei
->i_reserved_data_blocks
= 0;
886 ei
->i_reserved_meta_blocks
= 0;
887 ei
->i_allocated_meta_blocks
= 0;
888 ei
->i_da_metadata_calc_len
= 0;
889 spin_lock_init(&(ei
->i_block_reservation_lock
));
891 ei
->i_reserved_quota
= 0;
894 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
895 spin_lock_init(&ei
->i_completed_io_lock
);
896 ei
->cur_aio_dio
= NULL
;
898 ei
->i_datasync_tid
= 0;
899 atomic_set(&ei
->i_ioend_count
, 0);
900 atomic_set(&ei
->i_aiodio_unwritten
, 0);
902 return &ei
->vfs_inode
;
905 static int ext4_drop_inode(struct inode
*inode
)
907 int drop
= generic_drop_inode(inode
);
909 trace_ext4_drop_inode(inode
, drop
);
913 static void ext4_i_callback(struct rcu_head
*head
)
915 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
916 INIT_LIST_HEAD(&inode
->i_dentry
);
917 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
920 static void ext4_destroy_inode(struct inode
*inode
)
922 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
923 ext4_msg(inode
->i_sb
, KERN_ERR
,
924 "Inode %lu (%p): orphan list check failed!",
925 inode
->i_ino
, EXT4_I(inode
));
926 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
927 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
931 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
934 static void init_once(void *foo
)
936 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
938 INIT_LIST_HEAD(&ei
->i_orphan
);
939 #ifdef CONFIG_EXT4_FS_XATTR
940 init_rwsem(&ei
->xattr_sem
);
942 init_rwsem(&ei
->i_data_sem
);
943 inode_init_once(&ei
->vfs_inode
);
946 static int init_inodecache(void)
948 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
949 sizeof(struct ext4_inode_info
),
950 0, (SLAB_RECLAIM_ACCOUNT
|
953 if (ext4_inode_cachep
== NULL
)
958 static void destroy_inodecache(void)
960 kmem_cache_destroy(ext4_inode_cachep
);
963 void ext4_clear_inode(struct inode
*inode
)
965 invalidate_inode_buffers(inode
);
966 end_writeback(inode
);
968 ext4_discard_preallocations(inode
);
969 if (EXT4_I(inode
)->jinode
) {
970 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
971 EXT4_I(inode
)->jinode
);
972 jbd2_free_inode(EXT4_I(inode
)->jinode
);
973 EXT4_I(inode
)->jinode
= NULL
;
977 static inline void ext4_show_quota_options(struct seq_file
*seq
,
978 struct super_block
*sb
)
980 #if defined(CONFIG_QUOTA)
981 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
983 if (sbi
->s_jquota_fmt
) {
986 switch (sbi
->s_jquota_fmt
) {
997 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1000 if (sbi
->s_qf_names
[USRQUOTA
])
1001 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1003 if (sbi
->s_qf_names
[GRPQUOTA
])
1004 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1006 if (test_opt(sb
, USRQUOTA
))
1007 seq_puts(seq
, ",usrquota");
1009 if (test_opt(sb
, GRPQUOTA
))
1010 seq_puts(seq
, ",grpquota");
1016 * - it's set to a non-default value OR
1017 * - if the per-sb default is different from the global default
1019 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
1022 unsigned long def_mount_opts
;
1023 struct super_block
*sb
= vfs
->mnt_sb
;
1024 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1025 struct ext4_super_block
*es
= sbi
->s_es
;
1027 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
1028 def_errors
= le16_to_cpu(es
->s_errors
);
1030 if (sbi
->s_sb_block
!= 1)
1031 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
1032 if (test_opt(sb
, MINIX_DF
))
1033 seq_puts(seq
, ",minixdf");
1034 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1035 seq_puts(seq
, ",grpid");
1036 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
1037 seq_puts(seq
, ",nogrpid");
1038 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
1039 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
1040 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
1042 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
1043 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
1044 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
1046 if (test_opt(sb
, ERRORS_RO
)) {
1047 if (def_errors
== EXT4_ERRORS_PANIC
||
1048 def_errors
== EXT4_ERRORS_CONTINUE
) {
1049 seq_puts(seq
, ",errors=remount-ro");
1052 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1053 seq_puts(seq
, ",errors=continue");
1054 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1055 seq_puts(seq
, ",errors=panic");
1056 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
1057 seq_puts(seq
, ",nouid32");
1058 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
1059 seq_puts(seq
, ",debug");
1060 if (test_opt(sb
, OLDALLOC
))
1061 seq_puts(seq
, ",oldalloc");
1062 #ifdef CONFIG_EXT4_FS_XATTR
1063 if (test_opt(sb
, XATTR_USER
))
1064 seq_puts(seq
, ",user_xattr");
1065 if (!test_opt(sb
, XATTR_USER
))
1066 seq_puts(seq
, ",nouser_xattr");
1068 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1069 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
1070 seq_puts(seq
, ",acl");
1071 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
1072 seq_puts(seq
, ",noacl");
1074 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
1075 seq_printf(seq
, ",commit=%u",
1076 (unsigned) (sbi
->s_commit_interval
/ HZ
));
1078 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
1079 seq_printf(seq
, ",min_batch_time=%u",
1080 (unsigned) sbi
->s_min_batch_time
);
1082 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
1083 seq_printf(seq
, ",max_batch_time=%u",
1084 (unsigned) sbi
->s_min_batch_time
);
1088 * We're changing the default of barrier mount option, so
1089 * let's always display its mount state so it's clear what its
1092 seq_puts(seq
, ",barrier=");
1093 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1094 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1095 seq_puts(seq
, ",journal_async_commit");
1096 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1097 seq_puts(seq
, ",journal_checksum");
1098 if (test_opt(sb
, I_VERSION
))
1099 seq_puts(seq
, ",i_version");
1100 if (!test_opt(sb
, DELALLOC
) &&
1101 !(def_mount_opts
& EXT4_DEFM_NODELALLOC
))
1102 seq_puts(seq
, ",nodelalloc");
1104 if (!test_opt(sb
, MBLK_IO_SUBMIT
))
1105 seq_puts(seq
, ",nomblk_io_submit");
1107 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1109 * journal mode get enabled in different ways
1110 * So just print the value even if we didn't specify it
1112 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1113 seq_puts(seq
, ",data=journal");
1114 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1115 seq_puts(seq
, ",data=ordered");
1116 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1117 seq_puts(seq
, ",data=writeback");
1119 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1120 seq_printf(seq
, ",inode_readahead_blks=%u",
1121 sbi
->s_inode_readahead_blks
);
1123 if (test_opt(sb
, DATA_ERR_ABORT
))
1124 seq_puts(seq
, ",data_err=abort");
1126 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1127 seq_puts(seq
, ",noauto_da_alloc");
1129 if (test_opt(sb
, DISCARD
) && !(def_mount_opts
& EXT4_DEFM_DISCARD
))
1130 seq_puts(seq
, ",discard");
1132 if (test_opt(sb
, NOLOAD
))
1133 seq_puts(seq
, ",norecovery");
1135 if (test_opt(sb
, DIOREAD_NOLOCK
))
1136 seq_puts(seq
, ",dioread_nolock");
1138 if (test_opt(sb
, BLOCK_VALIDITY
) &&
1139 !(def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
))
1140 seq_puts(seq
, ",block_validity");
1142 if (!test_opt(sb
, INIT_INODE_TABLE
))
1143 seq_puts(seq
, ",noinit_inode_table");
1144 else if (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)
1145 seq_printf(seq
, ",init_inode_table=%u",
1146 (unsigned) sbi
->s_li_wait_mult
);
1148 ext4_show_quota_options(seq
, sb
);
1153 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1154 u64 ino
, u32 generation
)
1156 struct inode
*inode
;
1158 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1159 return ERR_PTR(-ESTALE
);
1160 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1161 return ERR_PTR(-ESTALE
);
1163 /* iget isn't really right if the inode is currently unallocated!!
1165 * ext4_read_inode will return a bad_inode if the inode had been
1166 * deleted, so we should be safe.
1168 * Currently we don't know the generation for parent directory, so
1169 * a generation of 0 means "accept any"
1171 inode
= ext4_iget(sb
, ino
);
1173 return ERR_CAST(inode
);
1174 if (generation
&& inode
->i_generation
!= generation
) {
1176 return ERR_PTR(-ESTALE
);
1182 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1183 int fh_len
, int fh_type
)
1185 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1186 ext4_nfs_get_inode
);
1189 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1190 int fh_len
, int fh_type
)
1192 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1193 ext4_nfs_get_inode
);
1197 * Try to release metadata pages (indirect blocks, directories) which are
1198 * mapped via the block device. Since these pages could have journal heads
1199 * which would prevent try_to_free_buffers() from freeing them, we must use
1200 * jbd2 layer's try_to_free_buffers() function to release them.
1202 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1205 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1207 WARN_ON(PageChecked(page
));
1208 if (!page_has_buffers(page
))
1211 return jbd2_journal_try_to_free_buffers(journal
, page
,
1212 wait
& ~__GFP_WAIT
);
1213 return try_to_free_buffers(page
);
1217 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1218 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1220 static int ext4_write_dquot(struct dquot
*dquot
);
1221 static int ext4_acquire_dquot(struct dquot
*dquot
);
1222 static int ext4_release_dquot(struct dquot
*dquot
);
1223 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1224 static int ext4_write_info(struct super_block
*sb
, int type
);
1225 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1227 static int ext4_quota_off(struct super_block
*sb
, int type
);
1228 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1229 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1230 size_t len
, loff_t off
);
1231 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1232 const char *data
, size_t len
, loff_t off
);
1234 static const struct dquot_operations ext4_quota_operations
= {
1235 .get_reserved_space
= ext4_get_reserved_space
,
1236 .write_dquot
= ext4_write_dquot
,
1237 .acquire_dquot
= ext4_acquire_dquot
,
1238 .release_dquot
= ext4_release_dquot
,
1239 .mark_dirty
= ext4_mark_dquot_dirty
,
1240 .write_info
= ext4_write_info
,
1241 .alloc_dquot
= dquot_alloc
,
1242 .destroy_dquot
= dquot_destroy
,
1245 static const struct quotactl_ops ext4_qctl_operations
= {
1246 .quota_on
= ext4_quota_on
,
1247 .quota_off
= ext4_quota_off
,
1248 .quota_sync
= dquot_quota_sync
,
1249 .get_info
= dquot_get_dqinfo
,
1250 .set_info
= dquot_set_dqinfo
,
1251 .get_dqblk
= dquot_get_dqblk
,
1252 .set_dqblk
= dquot_set_dqblk
1256 static const struct super_operations ext4_sops
= {
1257 .alloc_inode
= ext4_alloc_inode
,
1258 .destroy_inode
= ext4_destroy_inode
,
1259 .write_inode
= ext4_write_inode
,
1260 .dirty_inode
= ext4_dirty_inode
,
1261 .drop_inode
= ext4_drop_inode
,
1262 .evict_inode
= ext4_evict_inode
,
1263 .put_super
= ext4_put_super
,
1264 .sync_fs
= ext4_sync_fs
,
1265 .freeze_fs
= ext4_freeze
,
1266 .unfreeze_fs
= ext4_unfreeze
,
1267 .statfs
= ext4_statfs
,
1268 .remount_fs
= ext4_remount
,
1269 .show_options
= ext4_show_options
,
1271 .quota_read
= ext4_quota_read
,
1272 .quota_write
= ext4_quota_write
,
1274 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1277 static const struct super_operations ext4_nojournal_sops
= {
1278 .alloc_inode
= ext4_alloc_inode
,
1279 .destroy_inode
= ext4_destroy_inode
,
1280 .write_inode
= ext4_write_inode
,
1281 .dirty_inode
= ext4_dirty_inode
,
1282 .drop_inode
= ext4_drop_inode
,
1283 .evict_inode
= ext4_evict_inode
,
1284 .write_super
= ext4_write_super
,
1285 .put_super
= ext4_put_super
,
1286 .statfs
= ext4_statfs
,
1287 .remount_fs
= ext4_remount
,
1288 .show_options
= ext4_show_options
,
1290 .quota_read
= ext4_quota_read
,
1291 .quota_write
= ext4_quota_write
,
1293 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1296 static const struct export_operations ext4_export_ops
= {
1297 .fh_to_dentry
= ext4_fh_to_dentry
,
1298 .fh_to_parent
= ext4_fh_to_parent
,
1299 .get_parent
= ext4_get_parent
,
1303 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1304 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1305 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1306 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1307 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1308 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1309 Opt_journal_update
, Opt_journal_dev
,
1310 Opt_journal_checksum
, Opt_journal_async_commit
,
1311 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1312 Opt_data_err_abort
, Opt_data_err_ignore
,
1313 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1314 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1315 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1316 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1317 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1318 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1319 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1320 Opt_dioread_nolock
, Opt_dioread_lock
,
1321 Opt_discard
, Opt_nodiscard
,
1322 Opt_init_inode_table
, Opt_noinit_inode_table
,
1325 static const match_table_t tokens
= {
1326 {Opt_bsd_df
, "bsddf"},
1327 {Opt_minix_df
, "minixdf"},
1328 {Opt_grpid
, "grpid"},
1329 {Opt_grpid
, "bsdgroups"},
1330 {Opt_nogrpid
, "nogrpid"},
1331 {Opt_nogrpid
, "sysvgroups"},
1332 {Opt_resgid
, "resgid=%u"},
1333 {Opt_resuid
, "resuid=%u"},
1335 {Opt_err_cont
, "errors=continue"},
1336 {Opt_err_panic
, "errors=panic"},
1337 {Opt_err_ro
, "errors=remount-ro"},
1338 {Opt_nouid32
, "nouid32"},
1339 {Opt_debug
, "debug"},
1340 {Opt_oldalloc
, "oldalloc"},
1341 {Opt_orlov
, "orlov"},
1342 {Opt_user_xattr
, "user_xattr"},
1343 {Opt_nouser_xattr
, "nouser_xattr"},
1345 {Opt_noacl
, "noacl"},
1346 {Opt_noload
, "noload"},
1347 {Opt_noload
, "norecovery"},
1350 {Opt_commit
, "commit=%u"},
1351 {Opt_min_batch_time
, "min_batch_time=%u"},
1352 {Opt_max_batch_time
, "max_batch_time=%u"},
1353 {Opt_journal_update
, "journal=update"},
1354 {Opt_journal_dev
, "journal_dev=%u"},
1355 {Opt_journal_checksum
, "journal_checksum"},
1356 {Opt_journal_async_commit
, "journal_async_commit"},
1357 {Opt_abort
, "abort"},
1358 {Opt_data_journal
, "data=journal"},
1359 {Opt_data_ordered
, "data=ordered"},
1360 {Opt_data_writeback
, "data=writeback"},
1361 {Opt_data_err_abort
, "data_err=abort"},
1362 {Opt_data_err_ignore
, "data_err=ignore"},
1363 {Opt_offusrjquota
, "usrjquota="},
1364 {Opt_usrjquota
, "usrjquota=%s"},
1365 {Opt_offgrpjquota
, "grpjquota="},
1366 {Opt_grpjquota
, "grpjquota=%s"},
1367 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1368 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1369 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1370 {Opt_grpquota
, "grpquota"},
1371 {Opt_noquota
, "noquota"},
1372 {Opt_quota
, "quota"},
1373 {Opt_usrquota
, "usrquota"},
1374 {Opt_barrier
, "barrier=%u"},
1375 {Opt_barrier
, "barrier"},
1376 {Opt_nobarrier
, "nobarrier"},
1377 {Opt_i_version
, "i_version"},
1378 {Opt_stripe
, "stripe=%u"},
1379 {Opt_resize
, "resize"},
1380 {Opt_delalloc
, "delalloc"},
1381 {Opt_nodelalloc
, "nodelalloc"},
1382 {Opt_mblk_io_submit
, "mblk_io_submit"},
1383 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1384 {Opt_block_validity
, "block_validity"},
1385 {Opt_noblock_validity
, "noblock_validity"},
1386 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1387 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1388 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1389 {Opt_auto_da_alloc
, "auto_da_alloc"},
1390 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1391 {Opt_dioread_nolock
, "dioread_nolock"},
1392 {Opt_dioread_lock
, "dioread_lock"},
1393 {Opt_discard
, "discard"},
1394 {Opt_nodiscard
, "nodiscard"},
1395 {Opt_init_inode_table
, "init_itable=%u"},
1396 {Opt_init_inode_table
, "init_itable"},
1397 {Opt_noinit_inode_table
, "noinit_itable"},
1401 static ext4_fsblk_t
get_sb_block(void **data
)
1403 ext4_fsblk_t sb_block
;
1404 char *options
= (char *) *data
;
1406 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1407 return 1; /* Default location */
1410 /* TODO: use simple_strtoll with >32bit ext4 */
1411 sb_block
= simple_strtoul(options
, &options
, 0);
1412 if (*options
&& *options
!= ',') {
1413 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1417 if (*options
== ',')
1419 *data
= (void *) options
;
1424 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1425 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1426 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1429 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1431 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1434 if (sb_any_quota_loaded(sb
) &&
1435 !sbi
->s_qf_names
[qtype
]) {
1436 ext4_msg(sb
, KERN_ERR
,
1437 "Cannot change journaled "
1438 "quota options when quota turned on");
1441 qname
= match_strdup(args
);
1443 ext4_msg(sb
, KERN_ERR
,
1444 "Not enough memory for storing quotafile name");
1447 if (sbi
->s_qf_names
[qtype
] &&
1448 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1449 ext4_msg(sb
, KERN_ERR
,
1450 "%s quota file already specified", QTYPE2NAME(qtype
));
1454 sbi
->s_qf_names
[qtype
] = qname
;
1455 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1456 ext4_msg(sb
, KERN_ERR
,
1457 "quotafile must be on filesystem root");
1458 kfree(sbi
->s_qf_names
[qtype
]);
1459 sbi
->s_qf_names
[qtype
] = NULL
;
1466 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1469 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1471 if (sb_any_quota_loaded(sb
) &&
1472 sbi
->s_qf_names
[qtype
]) {
1473 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1474 " when quota turned on");
1478 * The space will be released later when all options are confirmed
1481 sbi
->s_qf_names
[qtype
] = NULL
;
1486 static int parse_options(char *options
, struct super_block
*sb
,
1487 unsigned long *journal_devnum
,
1488 unsigned int *journal_ioprio
,
1489 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1491 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1493 substring_t args
[MAX_OPT_ARGS
];
1503 while ((p
= strsep(&options
, ",")) != NULL
) {
1509 * Initialize args struct so we know whether arg was
1510 * found; some options take optional arguments.
1512 args
[0].to
= args
[0].from
= NULL
;
1513 token
= match_token(p
, tokens
, args
);
1516 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1517 clear_opt(sb
, MINIX_DF
);
1520 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1521 set_opt(sb
, MINIX_DF
);
1525 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1530 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1531 clear_opt(sb
, GRPID
);
1535 if (match_int(&args
[0], &option
))
1537 sbi
->s_resuid
= option
;
1540 if (match_int(&args
[0], &option
))
1542 sbi
->s_resgid
= option
;
1545 /* handled by get_sb_block() instead of here */
1546 /* *sb_block = match_int(&args[0]); */
1549 clear_opt(sb
, ERRORS_CONT
);
1550 clear_opt(sb
, ERRORS_RO
);
1551 set_opt(sb
, ERRORS_PANIC
);
1554 clear_opt(sb
, ERRORS_CONT
);
1555 clear_opt(sb
, ERRORS_PANIC
);
1556 set_opt(sb
, ERRORS_RO
);
1559 clear_opt(sb
, ERRORS_RO
);
1560 clear_opt(sb
, ERRORS_PANIC
);
1561 set_opt(sb
, ERRORS_CONT
);
1564 set_opt(sb
, NO_UID32
);
1570 set_opt(sb
, OLDALLOC
);
1573 clear_opt(sb
, OLDALLOC
);
1575 #ifdef CONFIG_EXT4_FS_XATTR
1576 case Opt_user_xattr
:
1577 set_opt(sb
, XATTR_USER
);
1579 case Opt_nouser_xattr
:
1580 clear_opt(sb
, XATTR_USER
);
1583 case Opt_user_xattr
:
1584 case Opt_nouser_xattr
:
1585 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1588 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1590 set_opt(sb
, POSIX_ACL
);
1593 clear_opt(sb
, POSIX_ACL
);
1598 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1601 case Opt_journal_update
:
1603 /* Eventually we will want to be able to create
1604 a journal file here. For now, only allow the
1605 user to specify an existing inode to be the
1608 ext4_msg(sb
, KERN_ERR
,
1609 "Cannot specify journal on remount");
1612 set_opt(sb
, UPDATE_JOURNAL
);
1614 case Opt_journal_dev
:
1616 ext4_msg(sb
, KERN_ERR
,
1617 "Cannot specify journal on remount");
1620 if (match_int(&args
[0], &option
))
1622 *journal_devnum
= option
;
1624 case Opt_journal_checksum
:
1625 set_opt(sb
, JOURNAL_CHECKSUM
);
1627 case Opt_journal_async_commit
:
1628 set_opt(sb
, JOURNAL_ASYNC_COMMIT
);
1629 set_opt(sb
, JOURNAL_CHECKSUM
);
1632 set_opt(sb
, NOLOAD
);
1635 if (match_int(&args
[0], &option
))
1640 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1641 sbi
->s_commit_interval
= HZ
* option
;
1643 case Opt_max_batch_time
:
1644 if (match_int(&args
[0], &option
))
1649 option
= EXT4_DEF_MAX_BATCH_TIME
;
1650 sbi
->s_max_batch_time
= option
;
1652 case Opt_min_batch_time
:
1653 if (match_int(&args
[0], &option
))
1657 sbi
->s_min_batch_time
= option
;
1659 case Opt_data_journal
:
1660 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1662 case Opt_data_ordered
:
1663 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1665 case Opt_data_writeback
:
1666 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1669 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1670 ext4_msg(sb
, KERN_ERR
,
1671 "Cannot change data mode on remount");
1675 clear_opt(sb
, DATA_FLAGS
);
1676 sbi
->s_mount_opt
|= data_opt
;
1679 case Opt_data_err_abort
:
1680 set_opt(sb
, DATA_ERR_ABORT
);
1682 case Opt_data_err_ignore
:
1683 clear_opt(sb
, DATA_ERR_ABORT
);
1687 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1691 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1694 case Opt_offusrjquota
:
1695 if (!clear_qf_name(sb
, USRQUOTA
))
1698 case Opt_offgrpjquota
:
1699 if (!clear_qf_name(sb
, GRPQUOTA
))
1703 case Opt_jqfmt_vfsold
:
1704 qfmt
= QFMT_VFS_OLD
;
1706 case Opt_jqfmt_vfsv0
:
1709 case Opt_jqfmt_vfsv1
:
1712 if (sb_any_quota_loaded(sb
) &&
1713 sbi
->s_jquota_fmt
!= qfmt
) {
1714 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1715 "journaled quota options when "
1719 sbi
->s_jquota_fmt
= qfmt
;
1724 set_opt(sb
, USRQUOTA
);
1728 set_opt(sb
, GRPQUOTA
);
1731 if (sb_any_quota_loaded(sb
)) {
1732 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1733 "options when quota turned on");
1736 clear_opt(sb
, QUOTA
);
1737 clear_opt(sb
, USRQUOTA
);
1738 clear_opt(sb
, GRPQUOTA
);
1744 ext4_msg(sb
, KERN_ERR
,
1745 "quota options not supported");
1749 case Opt_offusrjquota
:
1750 case Opt_offgrpjquota
:
1751 case Opt_jqfmt_vfsold
:
1752 case Opt_jqfmt_vfsv0
:
1753 case Opt_jqfmt_vfsv1
:
1754 ext4_msg(sb
, KERN_ERR
,
1755 "journaled quota options not supported");
1761 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1764 clear_opt(sb
, BARRIER
);
1768 if (match_int(&args
[0], &option
))
1771 option
= 1; /* No argument, default to 1 */
1773 set_opt(sb
, BARRIER
);
1775 clear_opt(sb
, BARRIER
);
1781 ext4_msg(sb
, KERN_ERR
,
1782 "resize option only available "
1786 if (match_int(&args
[0], &option
) != 0)
1788 *n_blocks_count
= option
;
1791 ext4_msg(sb
, KERN_WARNING
,
1792 "Ignoring deprecated nobh option");
1795 ext4_msg(sb
, KERN_WARNING
,
1796 "Ignoring deprecated bh option");
1799 set_opt(sb
, I_VERSION
);
1800 sb
->s_flags
|= MS_I_VERSION
;
1802 case Opt_nodelalloc
:
1803 clear_opt(sb
, DELALLOC
);
1804 clear_opt2(sb
, EXPLICIT_DELALLOC
);
1806 case Opt_mblk_io_submit
:
1807 set_opt(sb
, MBLK_IO_SUBMIT
);
1809 case Opt_nomblk_io_submit
:
1810 clear_opt(sb
, MBLK_IO_SUBMIT
);
1813 if (match_int(&args
[0], &option
))
1817 sbi
->s_stripe
= option
;
1820 set_opt(sb
, DELALLOC
);
1821 set_opt2(sb
, EXPLICIT_DELALLOC
);
1823 case Opt_block_validity
:
1824 set_opt(sb
, BLOCK_VALIDITY
);
1826 case Opt_noblock_validity
:
1827 clear_opt(sb
, BLOCK_VALIDITY
);
1829 case Opt_inode_readahead_blks
:
1830 if (match_int(&args
[0], &option
))
1832 if (option
< 0 || option
> (1 << 30))
1834 if (option
&& !is_power_of_2(option
)) {
1835 ext4_msg(sb
, KERN_ERR
,
1836 "EXT4-fs: inode_readahead_blks"
1837 " must be a power of 2");
1840 sbi
->s_inode_readahead_blks
= option
;
1842 case Opt_journal_ioprio
:
1843 if (match_int(&args
[0], &option
))
1845 if (option
< 0 || option
> 7)
1847 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1850 case Opt_noauto_da_alloc
:
1851 set_opt(sb
, NO_AUTO_DA_ALLOC
);
1853 case Opt_auto_da_alloc
:
1855 if (match_int(&args
[0], &option
))
1858 option
= 1; /* No argument, default to 1 */
1860 clear_opt(sb
, NO_AUTO_DA_ALLOC
);
1862 set_opt(sb
,NO_AUTO_DA_ALLOC
);
1865 set_opt(sb
, DISCARD
);
1868 clear_opt(sb
, DISCARD
);
1870 case Opt_dioread_nolock
:
1871 set_opt(sb
, DIOREAD_NOLOCK
);
1873 case Opt_dioread_lock
:
1874 clear_opt(sb
, DIOREAD_NOLOCK
);
1876 case Opt_init_inode_table
:
1877 set_opt(sb
, INIT_INODE_TABLE
);
1879 if (match_int(&args
[0], &option
))
1882 option
= EXT4_DEF_LI_WAIT_MULT
;
1885 sbi
->s_li_wait_mult
= option
;
1887 case Opt_noinit_inode_table
:
1888 clear_opt(sb
, INIT_INODE_TABLE
);
1891 ext4_msg(sb
, KERN_ERR
,
1892 "Unrecognized mount option \"%s\" "
1893 "or missing value", p
);
1898 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1899 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1900 clear_opt(sb
, USRQUOTA
);
1902 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1903 clear_opt(sb
, GRPQUOTA
);
1905 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1906 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1911 if (!sbi
->s_jquota_fmt
) {
1912 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1917 if (sbi
->s_jquota_fmt
) {
1918 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1919 "specified with no journaling "
1928 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1931 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1934 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1935 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1936 "forcing read-only mode");
1941 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1942 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1943 "running e2fsck is recommended");
1944 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1945 ext4_msg(sb
, KERN_WARNING
,
1946 "warning: mounting fs with errors, "
1947 "running e2fsck is recommended");
1948 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1949 le16_to_cpu(es
->s_mnt_count
) >=
1950 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1951 ext4_msg(sb
, KERN_WARNING
,
1952 "warning: maximal mount count reached, "
1953 "running e2fsck is recommended");
1954 else if (le32_to_cpu(es
->s_checkinterval
) &&
1955 (le32_to_cpu(es
->s_lastcheck
) +
1956 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1957 ext4_msg(sb
, KERN_WARNING
,
1958 "warning: checktime reached, "
1959 "running e2fsck is recommended");
1960 if (!sbi
->s_journal
)
1961 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1962 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1963 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1964 le16_add_cpu(&es
->s_mnt_count
, 1);
1965 es
->s_mtime
= cpu_to_le32(get_seconds());
1966 ext4_update_dynamic_rev(sb
);
1968 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1970 ext4_commit_super(sb
, 1);
1971 if (test_opt(sb
, DEBUG
))
1972 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1973 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1975 sbi
->s_groups_count
,
1976 EXT4_BLOCKS_PER_GROUP(sb
),
1977 EXT4_INODES_PER_GROUP(sb
),
1978 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1980 cleancache_init_fs(sb
);
1984 static int ext4_fill_flex_info(struct super_block
*sb
)
1986 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1987 struct ext4_group_desc
*gdp
= NULL
;
1988 ext4_group_t flex_group_count
;
1989 ext4_group_t flex_group
;
1990 int groups_per_flex
= 0;
1994 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1995 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1997 if (groups_per_flex
< 2) {
1998 sbi
->s_log_groups_per_flex
= 0;
2002 /* We allocate both existing and potentially added groups */
2003 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
2004 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
2005 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
2006 size
= flex_group_count
* sizeof(struct flex_groups
);
2007 sbi
->s_flex_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
2008 if (sbi
->s_flex_groups
== NULL
) {
2009 ext4_msg(sb
, KERN_ERR
, "not enough memory for %u flex groups",
2014 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2015 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2017 flex_group
= ext4_flex_group(sbi
, i
);
2018 atomic_add(ext4_free_inodes_count(sb
, gdp
),
2019 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
2020 atomic_add(ext4_free_blks_count(sb
, gdp
),
2021 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
2022 atomic_add(ext4_used_dirs_count(sb
, gdp
),
2023 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
2031 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
2032 struct ext4_group_desc
*gdp
)
2036 if (sbi
->s_es
->s_feature_ro_compat
&
2037 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
2038 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
2039 __le32 le_group
= cpu_to_le32(block_group
);
2041 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
2042 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
2043 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
2044 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
2045 /* for checksum of struct ext4_group_desc do the rest...*/
2046 if ((sbi
->s_es
->s_feature_incompat
&
2047 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
2048 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
2049 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
2050 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
2054 return cpu_to_le16(crc
);
2057 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
2058 struct ext4_group_desc
*gdp
)
2060 if ((sbi
->s_es
->s_feature_ro_compat
&
2061 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
2062 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
2068 /* Called at mount-time, super-block is locked */
2069 static int ext4_check_descriptors(struct super_block
*sb
,
2070 ext4_group_t
*first_not_zeroed
)
2072 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2073 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
2074 ext4_fsblk_t last_block
;
2075 ext4_fsblk_t block_bitmap
;
2076 ext4_fsblk_t inode_bitmap
;
2077 ext4_fsblk_t inode_table
;
2078 int flexbg_flag
= 0;
2079 ext4_group_t i
, grp
= sbi
->s_groups_count
;
2081 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2084 ext4_debug("Checking group descriptors");
2086 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
2087 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
2089 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
2090 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
2092 last_block
= first_block
+
2093 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2095 if ((grp
== sbi
->s_groups_count
) &&
2096 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2099 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
2100 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
2101 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2102 "Block bitmap for group %u not in group "
2103 "(block %llu)!", i
, block_bitmap
);
2106 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
2107 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
2108 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2109 "Inode bitmap for group %u not in group "
2110 "(block %llu)!", i
, inode_bitmap
);
2113 inode_table
= ext4_inode_table(sb
, gdp
);
2114 if (inode_table
< first_block
||
2115 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
2116 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2117 "Inode table for group %u not in group "
2118 "(block %llu)!", i
, inode_table
);
2121 ext4_lock_group(sb
, i
);
2122 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
2123 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
2124 "Checksum for group %u failed (%u!=%u)",
2125 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
2126 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
2127 if (!(sb
->s_flags
& MS_RDONLY
)) {
2128 ext4_unlock_group(sb
, i
);
2132 ext4_unlock_group(sb
, i
);
2134 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2136 if (NULL
!= first_not_zeroed
)
2137 *first_not_zeroed
= grp
;
2139 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2140 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2144 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2145 * the superblock) which were deleted from all directories, but held open by
2146 * a process at the time of a crash. We walk the list and try to delete these
2147 * inodes at recovery time (only with a read-write filesystem).
2149 * In order to keep the orphan inode chain consistent during traversal (in
2150 * case of crash during recovery), we link each inode into the superblock
2151 * orphan list_head and handle it the same way as an inode deletion during
2152 * normal operation (which journals the operations for us).
2154 * We only do an iget() and an iput() on each inode, which is very safe if we
2155 * accidentally point at an in-use or already deleted inode. The worst that
2156 * can happen in this case is that we get a "bit already cleared" message from
2157 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2158 * e2fsck was run on this filesystem, and it must have already done the orphan
2159 * inode cleanup for us, so we can safely abort without any further action.
2161 static void ext4_orphan_cleanup(struct super_block
*sb
,
2162 struct ext4_super_block
*es
)
2164 unsigned int s_flags
= sb
->s_flags
;
2165 int nr_orphans
= 0, nr_truncates
= 0;
2169 if (!es
->s_last_orphan
) {
2170 jbd_debug(4, "no orphan inodes to clean up\n");
2174 if (bdev_read_only(sb
->s_bdev
)) {
2175 ext4_msg(sb
, KERN_ERR
, "write access "
2176 "unavailable, skipping orphan cleanup");
2180 /* Check if feature set would not allow a r/w mount */
2181 if (!ext4_feature_set_ok(sb
, 0)) {
2182 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2183 "unknown ROCOMPAT features");
2187 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2188 if (es
->s_last_orphan
)
2189 jbd_debug(1, "Errors on filesystem, "
2190 "clearing orphan list.\n");
2191 es
->s_last_orphan
= 0;
2192 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2196 if (s_flags
& MS_RDONLY
) {
2197 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2198 sb
->s_flags
&= ~MS_RDONLY
;
2201 /* Needed for iput() to work correctly and not trash data */
2202 sb
->s_flags
|= MS_ACTIVE
;
2203 /* Turn on quotas so that they are updated correctly */
2204 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2205 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2206 int ret
= ext4_quota_on_mount(sb
, i
);
2208 ext4_msg(sb
, KERN_ERR
,
2209 "Cannot turn on journaled "
2210 "quota: error %d", ret
);
2215 while (es
->s_last_orphan
) {
2216 struct inode
*inode
;
2218 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2219 if (IS_ERR(inode
)) {
2220 es
->s_last_orphan
= 0;
2224 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2225 dquot_initialize(inode
);
2226 if (inode
->i_nlink
) {
2227 ext4_msg(sb
, KERN_DEBUG
,
2228 "%s: truncating inode %lu to %lld bytes",
2229 __func__
, inode
->i_ino
, inode
->i_size
);
2230 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2231 inode
->i_ino
, inode
->i_size
);
2232 ext4_truncate(inode
);
2235 ext4_msg(sb
, KERN_DEBUG
,
2236 "%s: deleting unreferenced inode %lu",
2237 __func__
, inode
->i_ino
);
2238 jbd_debug(2, "deleting unreferenced inode %lu\n",
2242 iput(inode
); /* The delete magic happens here! */
2245 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2248 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2249 PLURAL(nr_orphans
));
2251 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2252 PLURAL(nr_truncates
));
2254 /* Turn quotas off */
2255 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2256 if (sb_dqopt(sb
)->files
[i
])
2257 dquot_quota_off(sb
, i
);
2260 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2264 * Maximal extent format file size.
2265 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2266 * extent format containers, within a sector_t, and within i_blocks
2267 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2268 * so that won't be a limiting factor.
2270 * However there is other limiting factor. We do store extents in the form
2271 * of starting block and length, hence the resulting length of the extent
2272 * covering maximum file size must fit into on-disk format containers as
2273 * well. Given that length is always by 1 unit bigger than max unit (because
2274 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2276 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2278 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2281 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2283 /* small i_blocks in vfs inode? */
2284 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2286 * CONFIG_LBDAF is not enabled implies the inode
2287 * i_block represent total blocks in 512 bytes
2288 * 32 == size of vfs inode i_blocks * 8
2290 upper_limit
= (1LL << 32) - 1;
2292 /* total blocks in file system block size */
2293 upper_limit
>>= (blkbits
- 9);
2294 upper_limit
<<= blkbits
;
2298 * 32-bit extent-start container, ee_block. We lower the maxbytes
2299 * by one fs block, so ee_len can cover the extent of maximum file
2302 res
= (1LL << 32) - 1;
2305 /* Sanity check against vm- & vfs- imposed limits */
2306 if (res
> upper_limit
)
2313 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2314 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2315 * We need to be 1 filesystem block less than the 2^48 sector limit.
2317 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2319 loff_t res
= EXT4_NDIR_BLOCKS
;
2322 /* This is calculated to be the largest file size for a dense, block
2323 * mapped file such that the file's total number of 512-byte sectors,
2324 * including data and all indirect blocks, does not exceed (2^48 - 1).
2326 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2327 * number of 512-byte sectors of the file.
2330 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2332 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2333 * the inode i_block field represents total file blocks in
2334 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2336 upper_limit
= (1LL << 32) - 1;
2338 /* total blocks in file system block size */
2339 upper_limit
>>= (bits
- 9);
2343 * We use 48 bit ext4_inode i_blocks
2344 * With EXT4_HUGE_FILE_FL set the i_blocks
2345 * represent total number of blocks in
2346 * file system block size
2348 upper_limit
= (1LL << 48) - 1;
2352 /* indirect blocks */
2354 /* double indirect blocks */
2355 meta_blocks
+= 1 + (1LL << (bits
-2));
2356 /* tripple indirect blocks */
2357 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2359 upper_limit
-= meta_blocks
;
2360 upper_limit
<<= bits
;
2362 res
+= 1LL << (bits
-2);
2363 res
+= 1LL << (2*(bits
-2));
2364 res
+= 1LL << (3*(bits
-2));
2366 if (res
> upper_limit
)
2369 if (res
> MAX_LFS_FILESIZE
)
2370 res
= MAX_LFS_FILESIZE
;
2375 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2376 ext4_fsblk_t logical_sb_block
, int nr
)
2378 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2379 ext4_group_t bg
, first_meta_bg
;
2382 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2384 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2386 return logical_sb_block
+ nr
+ 1;
2387 bg
= sbi
->s_desc_per_block
* nr
;
2388 if (ext4_bg_has_super(sb
, bg
))
2391 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2395 * ext4_get_stripe_size: Get the stripe size.
2396 * @sbi: In memory super block info
2398 * If we have specified it via mount option, then
2399 * use the mount option value. If the value specified at mount time is
2400 * greater than the blocks per group use the super block value.
2401 * If the super block value is greater than blocks per group return 0.
2402 * Allocator needs it be less than blocks per group.
2405 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2407 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2408 unsigned long stripe_width
=
2409 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2412 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2413 ret
= sbi
->s_stripe
;
2414 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2416 else if (stride
<= sbi
->s_blocks_per_group
)
2422 * If the stripe width is 1, this makes no sense and
2423 * we set it to 0 to turn off stripe handling code.
2434 struct attribute attr
;
2435 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2436 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2437 const char *, size_t);
2441 static int parse_strtoul(const char *buf
,
2442 unsigned long max
, unsigned long *value
)
2446 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2447 endp
= skip_spaces(endp
);
2448 if (*endp
|| *value
> max
)
2454 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2455 struct ext4_sb_info
*sbi
,
2458 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2459 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2462 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2463 struct ext4_sb_info
*sbi
, char *buf
)
2465 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2467 if (!sb
->s_bdev
->bd_part
)
2468 return snprintf(buf
, PAGE_SIZE
, "0\n");
2469 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2470 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2471 sbi
->s_sectors_written_start
) >> 1);
2474 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2475 struct ext4_sb_info
*sbi
, char *buf
)
2477 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2479 if (!sb
->s_bdev
->bd_part
)
2480 return snprintf(buf
, PAGE_SIZE
, "0\n");
2481 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2482 (unsigned long long)(sbi
->s_kbytes_written
+
2483 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2484 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2487 static ssize_t
extent_cache_hits_show(struct ext4_attr
*a
,
2488 struct ext4_sb_info
*sbi
, char *buf
)
2490 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_hits
);
2493 static ssize_t
extent_cache_misses_show(struct ext4_attr
*a
,
2494 struct ext4_sb_info
*sbi
, char *buf
)
2496 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_misses
);
2499 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2500 struct ext4_sb_info
*sbi
,
2501 const char *buf
, size_t count
)
2505 if (parse_strtoul(buf
, 0x40000000, &t
))
2508 if (t
&& !is_power_of_2(t
))
2511 sbi
->s_inode_readahead_blks
= t
;
2515 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2516 struct ext4_sb_info
*sbi
, char *buf
)
2518 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2520 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2523 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2524 struct ext4_sb_info
*sbi
,
2525 const char *buf
, size_t count
)
2527 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2530 if (parse_strtoul(buf
, 0xffffffff, &t
))
2536 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2537 static struct ext4_attr ext4_attr_##_name = { \
2538 .attr = {.name = __stringify(_name), .mode = _mode }, \
2541 .offset = offsetof(struct ext4_sb_info, _elname), \
2543 #define EXT4_ATTR(name, mode, show, store) \
2544 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2546 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2547 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2548 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2549 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2550 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2551 #define ATTR_LIST(name) &ext4_attr_##name.attr
2553 EXT4_RO_ATTR(delayed_allocation_blocks
);
2554 EXT4_RO_ATTR(session_write_kbytes
);
2555 EXT4_RO_ATTR(lifetime_write_kbytes
);
2556 EXT4_RO_ATTR(extent_cache_hits
);
2557 EXT4_RO_ATTR(extent_cache_misses
);
2558 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2559 inode_readahead_blks_store
, s_inode_readahead_blks
);
2560 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2561 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2562 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2563 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2564 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2565 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2566 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2567 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2569 static struct attribute
*ext4_attrs
[] = {
2570 ATTR_LIST(delayed_allocation_blocks
),
2571 ATTR_LIST(session_write_kbytes
),
2572 ATTR_LIST(lifetime_write_kbytes
),
2573 ATTR_LIST(extent_cache_hits
),
2574 ATTR_LIST(extent_cache_misses
),
2575 ATTR_LIST(inode_readahead_blks
),
2576 ATTR_LIST(inode_goal
),
2577 ATTR_LIST(mb_stats
),
2578 ATTR_LIST(mb_max_to_scan
),
2579 ATTR_LIST(mb_min_to_scan
),
2580 ATTR_LIST(mb_order2_req
),
2581 ATTR_LIST(mb_stream_req
),
2582 ATTR_LIST(mb_group_prealloc
),
2583 ATTR_LIST(max_writeback_mb_bump
),
2587 /* Features this copy of ext4 supports */
2588 EXT4_INFO_ATTR(lazy_itable_init
);
2589 EXT4_INFO_ATTR(batched_discard
);
2591 static struct attribute
*ext4_feat_attrs
[] = {
2592 ATTR_LIST(lazy_itable_init
),
2593 ATTR_LIST(batched_discard
),
2597 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2598 struct attribute
*attr
, char *buf
)
2600 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2602 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2604 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2607 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2608 struct attribute
*attr
,
2609 const char *buf
, size_t len
)
2611 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2613 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2615 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2618 static void ext4_sb_release(struct kobject
*kobj
)
2620 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2622 complete(&sbi
->s_kobj_unregister
);
2625 static const struct sysfs_ops ext4_attr_ops
= {
2626 .show
= ext4_attr_show
,
2627 .store
= ext4_attr_store
,
2630 static struct kobj_type ext4_ktype
= {
2631 .default_attrs
= ext4_attrs
,
2632 .sysfs_ops
= &ext4_attr_ops
,
2633 .release
= ext4_sb_release
,
2636 static void ext4_feat_release(struct kobject
*kobj
)
2638 complete(&ext4_feat
->f_kobj_unregister
);
2641 static struct kobj_type ext4_feat_ktype
= {
2642 .default_attrs
= ext4_feat_attrs
,
2643 .sysfs_ops
= &ext4_attr_ops
,
2644 .release
= ext4_feat_release
,
2648 * Check whether this filesystem can be mounted based on
2649 * the features present and the RDONLY/RDWR mount requested.
2650 * Returns 1 if this filesystem can be mounted as requested,
2651 * 0 if it cannot be.
2653 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2655 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2656 ext4_msg(sb
, KERN_ERR
,
2657 "Couldn't mount because of "
2658 "unsupported optional features (%x)",
2659 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2660 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2667 /* Check that feature set is OK for a read-write mount */
2668 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2669 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2670 "unsupported optional features (%x)",
2671 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2672 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2676 * Large file size enabled file system can only be mounted
2677 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2679 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2680 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2681 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2682 "cannot be mounted RDWR without "
2691 * This function is called once a day if we have errors logged
2692 * on the file system
2694 static void print_daily_error_info(unsigned long arg
)
2696 struct super_block
*sb
= (struct super_block
*) arg
;
2697 struct ext4_sb_info
*sbi
;
2698 struct ext4_super_block
*es
;
2703 if (es
->s_error_count
)
2704 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2705 le32_to_cpu(es
->s_error_count
));
2706 if (es
->s_first_error_time
) {
2707 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2708 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2709 (int) sizeof(es
->s_first_error_func
),
2710 es
->s_first_error_func
,
2711 le32_to_cpu(es
->s_first_error_line
));
2712 if (es
->s_first_error_ino
)
2713 printk(": inode %u",
2714 le32_to_cpu(es
->s_first_error_ino
));
2715 if (es
->s_first_error_block
)
2716 printk(": block %llu", (unsigned long long)
2717 le64_to_cpu(es
->s_first_error_block
));
2720 if (es
->s_last_error_time
) {
2721 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2722 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2723 (int) sizeof(es
->s_last_error_func
),
2724 es
->s_last_error_func
,
2725 le32_to_cpu(es
->s_last_error_line
));
2726 if (es
->s_last_error_ino
)
2727 printk(": inode %u",
2728 le32_to_cpu(es
->s_last_error_ino
));
2729 if (es
->s_last_error_block
)
2730 printk(": block %llu", (unsigned long long)
2731 le64_to_cpu(es
->s_last_error_block
));
2734 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2737 /* Find next suitable group and run ext4_init_inode_table */
2738 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2740 struct ext4_group_desc
*gdp
= NULL
;
2741 ext4_group_t group
, ngroups
;
2742 struct super_block
*sb
;
2743 unsigned long timeout
= 0;
2747 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2749 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2750 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2756 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2760 if (group
== ngroups
)
2765 ret
= ext4_init_inode_table(sb
, group
,
2766 elr
->lr_timeout
? 0 : 1);
2767 if (elr
->lr_timeout
== 0) {
2768 timeout
= (jiffies
- timeout
) *
2769 elr
->lr_sbi
->s_li_wait_mult
;
2770 elr
->lr_timeout
= timeout
;
2772 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2773 elr
->lr_next_group
= group
+ 1;
2780 * Remove lr_request from the list_request and free the
2781 * request structure. Should be called with li_list_mtx held
2783 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2785 struct ext4_sb_info
*sbi
;
2792 list_del(&elr
->lr_request
);
2793 sbi
->s_li_request
= NULL
;
2797 static void ext4_unregister_li_request(struct super_block
*sb
)
2799 mutex_lock(&ext4_li_mtx
);
2800 if (!ext4_li_info
) {
2801 mutex_unlock(&ext4_li_mtx
);
2805 mutex_lock(&ext4_li_info
->li_list_mtx
);
2806 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2807 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2808 mutex_unlock(&ext4_li_mtx
);
2811 static struct task_struct
*ext4_lazyinit_task
;
2814 * This is the function where ext4lazyinit thread lives. It walks
2815 * through the request list searching for next scheduled filesystem.
2816 * When such a fs is found, run the lazy initialization request
2817 * (ext4_rn_li_request) and keep track of the time spend in this
2818 * function. Based on that time we compute next schedule time of
2819 * the request. When walking through the list is complete, compute
2820 * next waking time and put itself into sleep.
2822 static int ext4_lazyinit_thread(void *arg
)
2824 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2825 struct list_head
*pos
, *n
;
2826 struct ext4_li_request
*elr
;
2827 unsigned long next_wakeup
, cur
;
2829 BUG_ON(NULL
== eli
);
2833 next_wakeup
= MAX_JIFFY_OFFSET
;
2835 mutex_lock(&eli
->li_list_mtx
);
2836 if (list_empty(&eli
->li_request_list
)) {
2837 mutex_unlock(&eli
->li_list_mtx
);
2841 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2842 elr
= list_entry(pos
, struct ext4_li_request
,
2845 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2846 if (ext4_run_li_request(elr
) != 0) {
2847 /* error, remove the lazy_init job */
2848 ext4_remove_li_request(elr
);
2853 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2854 next_wakeup
= elr
->lr_next_sched
;
2856 mutex_unlock(&eli
->li_list_mtx
);
2858 if (freezing(current
))
2862 if ((time_after_eq(cur
, next_wakeup
)) ||
2863 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2868 schedule_timeout_interruptible(next_wakeup
- cur
);
2870 if (kthread_should_stop()) {
2871 ext4_clear_request_list();
2878 * It looks like the request list is empty, but we need
2879 * to check it under the li_list_mtx lock, to prevent any
2880 * additions into it, and of course we should lock ext4_li_mtx
2881 * to atomically free the list and ext4_li_info, because at
2882 * this point another ext4 filesystem could be registering
2885 mutex_lock(&ext4_li_mtx
);
2886 mutex_lock(&eli
->li_list_mtx
);
2887 if (!list_empty(&eli
->li_request_list
)) {
2888 mutex_unlock(&eli
->li_list_mtx
);
2889 mutex_unlock(&ext4_li_mtx
);
2892 mutex_unlock(&eli
->li_list_mtx
);
2893 kfree(ext4_li_info
);
2894 ext4_li_info
= NULL
;
2895 mutex_unlock(&ext4_li_mtx
);
2900 static void ext4_clear_request_list(void)
2902 struct list_head
*pos
, *n
;
2903 struct ext4_li_request
*elr
;
2905 mutex_lock(&ext4_li_info
->li_list_mtx
);
2906 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2907 elr
= list_entry(pos
, struct ext4_li_request
,
2909 ext4_remove_li_request(elr
);
2911 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2914 static int ext4_run_lazyinit_thread(void)
2916 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2917 ext4_li_info
, "ext4lazyinit");
2918 if (IS_ERR(ext4_lazyinit_task
)) {
2919 int err
= PTR_ERR(ext4_lazyinit_task
);
2920 ext4_clear_request_list();
2921 kfree(ext4_li_info
);
2922 ext4_li_info
= NULL
;
2923 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2924 "initialization thread\n",
2928 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2933 * Check whether it make sense to run itable init. thread or not.
2934 * If there is at least one uninitialized inode table, return
2935 * corresponding group number, else the loop goes through all
2936 * groups and return total number of groups.
2938 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2940 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2941 struct ext4_group_desc
*gdp
= NULL
;
2943 for (group
= 0; group
< ngroups
; group
++) {
2944 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2948 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2955 static int ext4_li_info_new(void)
2957 struct ext4_lazy_init
*eli
= NULL
;
2959 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2963 INIT_LIST_HEAD(&eli
->li_request_list
);
2964 mutex_init(&eli
->li_list_mtx
);
2966 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2973 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2976 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2977 struct ext4_li_request
*elr
;
2980 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2986 elr
->lr_next_group
= start
;
2989 * Randomize first schedule time of the request to
2990 * spread the inode table initialization requests
2993 get_random_bytes(&rnd
, sizeof(rnd
));
2994 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2995 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
3000 static int ext4_register_li_request(struct super_block
*sb
,
3001 ext4_group_t first_not_zeroed
)
3003 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3004 struct ext4_li_request
*elr
;
3005 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
3008 if (sbi
->s_li_request
!= NULL
) {
3010 * Reset timeout so it can be computed again, because
3011 * s_li_wait_mult might have changed.
3013 sbi
->s_li_request
->lr_timeout
= 0;
3017 if (first_not_zeroed
== ngroups
||
3018 (sb
->s_flags
& MS_RDONLY
) ||
3019 !test_opt(sb
, INIT_INODE_TABLE
))
3022 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
3026 mutex_lock(&ext4_li_mtx
);
3028 if (NULL
== ext4_li_info
) {
3029 ret
= ext4_li_info_new();
3034 mutex_lock(&ext4_li_info
->li_list_mtx
);
3035 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
3036 mutex_unlock(&ext4_li_info
->li_list_mtx
);
3038 sbi
->s_li_request
= elr
;
3040 * set elr to NULL here since it has been inserted to
3041 * the request_list and the removal and free of it is
3042 * handled by ext4_clear_request_list from now on.
3046 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
3047 ret
= ext4_run_lazyinit_thread();
3052 mutex_unlock(&ext4_li_mtx
);
3059 * We do not need to lock anything since this is called on
3062 static void ext4_destroy_lazyinit_thread(void)
3065 * If thread exited earlier
3066 * there's nothing to be done.
3068 if (!ext4_li_info
|| !ext4_lazyinit_task
)
3071 kthread_stop(ext4_lazyinit_task
);
3074 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
3075 __releases(kernel_lock
)
3076 __acquires(kernel_lock
)
3078 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3079 struct buffer_head
*bh
;
3080 struct ext4_super_block
*es
= NULL
;
3081 struct ext4_sb_info
*sbi
;
3083 ext4_fsblk_t sb_block
= get_sb_block(&data
);
3084 ext4_fsblk_t logical_sb_block
;
3085 unsigned long offset
= 0;
3086 unsigned long journal_devnum
= 0;
3087 unsigned long def_mount_opts
;
3093 unsigned int db_count
;
3095 int needs_recovery
, has_huge_files
;
3098 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3099 ext4_group_t first_not_zeroed
;
3101 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
3105 sbi
->s_blockgroup_lock
=
3106 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
3107 if (!sbi
->s_blockgroup_lock
) {
3111 sb
->s_fs_info
= sbi
;
3112 sbi
->s_mount_opt
= 0;
3113 sbi
->s_resuid
= EXT4_DEF_RESUID
;
3114 sbi
->s_resgid
= EXT4_DEF_RESGID
;
3115 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
3116 sbi
->s_sb_block
= sb_block
;
3117 if (sb
->s_bdev
->bd_part
)
3118 sbi
->s_sectors_written_start
=
3119 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
3121 /* Cleanup superblock name */
3122 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
3126 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3128 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3133 * The ext4 superblock will not be buffer aligned for other than 1kB
3134 * block sizes. We need to calculate the offset from buffer start.
3136 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3137 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3138 offset
= do_div(logical_sb_block
, blocksize
);
3140 logical_sb_block
= sb_block
;
3143 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3144 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3148 * Note: s_es must be initialized as soon as possible because
3149 * some ext4 macro-instructions depend on its value
3151 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3153 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3154 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3156 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3158 /* Set defaults before we parse the mount options */
3159 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3160 set_opt(sb
, INIT_INODE_TABLE
);
3161 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3163 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
3164 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
3168 if (def_mount_opts
& EXT4_DEFM_UID16
)
3169 set_opt(sb
, NO_UID32
);
3170 /* xattr user namespace & acls are now defaulted on */
3171 #ifdef CONFIG_EXT4_FS_XATTR
3172 set_opt(sb
, XATTR_USER
);
3174 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3175 set_opt(sb
, POSIX_ACL
);
3177 set_opt(sb
, MBLK_IO_SUBMIT
);
3178 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3179 set_opt(sb
, JOURNAL_DATA
);
3180 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3181 set_opt(sb
, ORDERED_DATA
);
3182 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3183 set_opt(sb
, WRITEBACK_DATA
);
3185 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3186 set_opt(sb
, ERRORS_PANIC
);
3187 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3188 set_opt(sb
, ERRORS_CONT
);
3190 set_opt(sb
, ERRORS_RO
);
3191 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3192 set_opt(sb
, BLOCK_VALIDITY
);
3193 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3194 set_opt(sb
, DISCARD
);
3196 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3197 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3198 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3199 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3200 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3202 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3203 set_opt(sb
, BARRIER
);
3206 * enable delayed allocation by default
3207 * Use -o nodelalloc to turn it off
3209 if (!IS_EXT3_SB(sb
) &&
3210 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3211 set_opt(sb
, DELALLOC
);
3214 * set default s_li_wait_mult for lazyinit, for the case there is
3215 * no mount option specified.
3217 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3219 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3220 &journal_devnum
, &journal_ioprio
, NULL
, 0)) {
3221 ext4_msg(sb
, KERN_WARNING
,
3222 "failed to parse options in superblock: %s",
3223 sbi
->s_es
->s_mount_opts
);
3225 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3226 &journal_ioprio
, NULL
, 0))
3229 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3230 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3231 "with data=journal disables delayed "
3232 "allocation and O_DIRECT support!\n");
3233 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3234 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3235 "both data=journal and delalloc");
3238 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3239 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3240 "both data=journal and delalloc");
3243 if (test_opt(sb
, DELALLOC
))
3244 clear_opt(sb
, DELALLOC
);
3247 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3248 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3249 if (blocksize
< PAGE_SIZE
) {
3250 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3251 "dioread_nolock if block size != PAGE_SIZE");
3256 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3257 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3259 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3260 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3261 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3262 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3263 ext4_msg(sb
, KERN_WARNING
,
3264 "feature flags set on rev 0 fs, "
3265 "running e2fsck is recommended");
3267 if (IS_EXT2_SB(sb
)) {
3268 if (ext2_feature_set_ok(sb
))
3269 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3270 "using the ext4 subsystem");
3272 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3273 "to feature incompatibilities");
3278 if (IS_EXT3_SB(sb
)) {
3279 if (ext3_feature_set_ok(sb
))
3280 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3281 "using the ext4 subsystem");
3283 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3284 "to feature incompatibilities");
3290 * Check feature flags regardless of the revision level, since we
3291 * previously didn't change the revision level when setting the flags,
3292 * so there is a chance incompat flags are set on a rev 0 filesystem.
3294 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3297 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3298 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3299 ext4_msg(sb
, KERN_ERR
,
3300 "Unsupported filesystem blocksize %d", blocksize
);
3304 if (sb
->s_blocksize
!= blocksize
) {
3305 /* Validate the filesystem blocksize */
3306 if (!sb_set_blocksize(sb
, blocksize
)) {
3307 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3313 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3314 offset
= do_div(logical_sb_block
, blocksize
);
3315 bh
= sb_bread(sb
, logical_sb_block
);
3317 ext4_msg(sb
, KERN_ERR
,
3318 "Can't read superblock on 2nd try");
3321 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3323 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3324 ext4_msg(sb
, KERN_ERR
,
3325 "Magic mismatch, very weird!");
3330 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3331 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3332 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3334 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3336 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3337 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3338 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3340 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3341 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3342 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3343 (!is_power_of_2(sbi
->s_inode_size
)) ||
3344 (sbi
->s_inode_size
> blocksize
)) {
3345 ext4_msg(sb
, KERN_ERR
,
3346 "unsupported inode size: %d",
3350 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3351 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3354 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3355 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3356 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3357 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3358 !is_power_of_2(sbi
->s_desc_size
)) {
3359 ext4_msg(sb
, KERN_ERR
,
3360 "unsupported descriptor size %lu",
3365 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3367 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3368 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3369 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3372 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3373 if (sbi
->s_inodes_per_block
== 0)
3375 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3376 sbi
->s_inodes_per_block
;
3377 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3379 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3380 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3381 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3383 for (i
= 0; i
< 4; i
++)
3384 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3385 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3386 i
= le32_to_cpu(es
->s_flags
);
3387 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3388 sbi
->s_hash_unsigned
= 3;
3389 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3390 #ifdef __CHAR_UNSIGNED__
3391 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3392 sbi
->s_hash_unsigned
= 3;
3394 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3399 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3400 ext4_msg(sb
, KERN_ERR
,
3401 "#blocks per group too big: %lu",
3402 sbi
->s_blocks_per_group
);
3405 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3406 ext4_msg(sb
, KERN_ERR
,
3407 "#inodes per group too big: %lu",
3408 sbi
->s_inodes_per_group
);
3413 * Test whether we have more sectors than will fit in sector_t,
3414 * and whether the max offset is addressable by the page cache.
3416 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3417 ext4_blocks_count(es
));
3419 ext4_msg(sb
, KERN_ERR
, "filesystem"
3420 " too large to mount safely on this system");
3421 if (sizeof(sector_t
) < 8)
3422 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3427 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3430 /* check blocks count against device size */
3431 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3432 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3433 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3434 "exceeds size of device (%llu blocks)",
3435 ext4_blocks_count(es
), blocks_count
);
3440 * It makes no sense for the first data block to be beyond the end
3441 * of the filesystem.
3443 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3444 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
3445 "block %u is beyond end of filesystem (%llu)",
3446 le32_to_cpu(es
->s_first_data_block
),
3447 ext4_blocks_count(es
));
3450 blocks_count
= (ext4_blocks_count(es
) -
3451 le32_to_cpu(es
->s_first_data_block
) +
3452 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3453 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3454 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3455 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3456 "(block count %llu, first data block %u, "
3457 "blocks per group %lu)", sbi
->s_groups_count
,
3458 ext4_blocks_count(es
),
3459 le32_to_cpu(es
->s_first_data_block
),
3460 EXT4_BLOCKS_PER_GROUP(sb
));
3463 sbi
->s_groups_count
= blocks_count
;
3464 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3465 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3466 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3467 EXT4_DESC_PER_BLOCK(sb
);
3468 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3469 sizeof(struct buffer_head
*),
3471 if (sbi
->s_group_desc
== NULL
) {
3472 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3476 #ifdef CONFIG_PROC_FS
3478 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3481 bgl_lock_init(sbi
->s_blockgroup_lock
);
3483 for (i
= 0; i
< db_count
; i
++) {
3484 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3485 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3486 if (!sbi
->s_group_desc
[i
]) {
3487 ext4_msg(sb
, KERN_ERR
,
3488 "can't read group descriptor %d", i
);
3493 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3494 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3497 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3498 if (!ext4_fill_flex_info(sb
)) {
3499 ext4_msg(sb
, KERN_ERR
,
3500 "unable to initialize "
3501 "flex_bg meta info!");
3505 sbi
->s_gdb_count
= db_count
;
3506 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3507 spin_lock_init(&sbi
->s_next_gen_lock
);
3509 init_timer(&sbi
->s_err_report
);
3510 sbi
->s_err_report
.function
= print_daily_error_info
;
3511 sbi
->s_err_report
.data
= (unsigned long) sb
;
3513 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3514 ext4_count_free_blocks(sb
));
3516 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3517 ext4_count_free_inodes(sb
));
3520 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3521 ext4_count_dirs(sb
));
3524 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3527 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3531 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3532 sbi
->s_max_writeback_mb_bump
= 128;
3535 * set up enough so that it can read an inode
3537 if (!test_opt(sb
, NOLOAD
) &&
3538 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3539 sb
->s_op
= &ext4_sops
;
3541 sb
->s_op
= &ext4_nojournal_sops
;
3542 sb
->s_export_op
= &ext4_export_ops
;
3543 sb
->s_xattr
= ext4_xattr_handlers
;
3545 sb
->s_qcop
= &ext4_qctl_operations
;
3546 sb
->dq_op
= &ext4_quota_operations
;
3548 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3550 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3551 mutex_init(&sbi
->s_orphan_lock
);
3552 sbi
->s_resize_flags
= 0;
3556 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3557 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3558 EXT4_FEATURE_INCOMPAT_RECOVER
));
3560 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3561 !(sb
->s_flags
& MS_RDONLY
))
3562 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3566 * The first inode we look at is the journal inode. Don't try
3567 * root first: it may be modified in the journal!
3569 if (!test_opt(sb
, NOLOAD
) &&
3570 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3571 if (ext4_load_journal(sb
, es
, journal_devnum
))
3573 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3574 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3575 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3576 "suppressed and not mounted read-only");
3577 goto failed_mount_wq
;
3579 clear_opt(sb
, DATA_FLAGS
);
3580 sbi
->s_journal
= NULL
;
3585 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3586 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3587 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3588 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3589 goto failed_mount_wq
;
3592 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3593 jbd2_journal_set_features(sbi
->s_journal
,
3594 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3595 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3596 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3597 jbd2_journal_set_features(sbi
->s_journal
,
3598 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3599 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3600 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3602 jbd2_journal_clear_features(sbi
->s_journal
,
3603 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3604 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3607 /* We have now updated the journal if required, so we can
3608 * validate the data journaling mode. */
3609 switch (test_opt(sb
, DATA_FLAGS
)) {
3611 /* No mode set, assume a default based on the journal
3612 * capabilities: ORDERED_DATA if the journal can
3613 * cope, else JOURNAL_DATA
3615 if (jbd2_journal_check_available_features
3616 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3617 set_opt(sb
, ORDERED_DATA
);
3619 set_opt(sb
, JOURNAL_DATA
);
3622 case EXT4_MOUNT_ORDERED_DATA
:
3623 case EXT4_MOUNT_WRITEBACK_DATA
:
3624 if (!jbd2_journal_check_available_features
3625 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3626 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3627 "requested data journaling mode");
3628 goto failed_mount_wq
;
3633 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3636 * The journal may have updated the bg summary counts, so we
3637 * need to update the global counters.
3639 percpu_counter_set(&sbi
->s_freeblocks_counter
,
3640 ext4_count_free_blocks(sb
));
3641 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3642 ext4_count_free_inodes(sb
));
3643 percpu_counter_set(&sbi
->s_dirs_counter
,
3644 ext4_count_dirs(sb
));
3645 percpu_counter_set(&sbi
->s_dirtyblocks_counter
, 0);
3649 * The maximum number of concurrent works can be high and
3650 * concurrency isn't really necessary. Limit it to 1.
3652 EXT4_SB(sb
)->dio_unwritten_wq
=
3653 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3654 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3655 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3656 goto failed_mount_wq
;
3660 * The jbd2_journal_load will have done any necessary log recovery,
3661 * so we can safely mount the rest of the filesystem now.
3664 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3666 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3667 ret
= PTR_ERR(root
);
3671 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3672 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3675 sb
->s_root
= d_alloc_root(root
);
3677 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3682 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3684 /* determine the minimum size of new large inodes, if present */
3685 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3686 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3687 EXT4_GOOD_OLD_INODE_SIZE
;
3688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3689 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3690 if (sbi
->s_want_extra_isize
<
3691 le16_to_cpu(es
->s_want_extra_isize
))
3692 sbi
->s_want_extra_isize
=
3693 le16_to_cpu(es
->s_want_extra_isize
);
3694 if (sbi
->s_want_extra_isize
<
3695 le16_to_cpu(es
->s_min_extra_isize
))
3696 sbi
->s_want_extra_isize
=
3697 le16_to_cpu(es
->s_min_extra_isize
);
3700 /* Check if enough inode space is available */
3701 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3702 sbi
->s_inode_size
) {
3703 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3704 EXT4_GOOD_OLD_INODE_SIZE
;
3705 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3709 err
= ext4_setup_system_zone(sb
);
3711 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3717 err
= ext4_mb_init(sb
, needs_recovery
);
3719 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3724 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3728 sbi
->s_kobj
.kset
= ext4_kset
;
3729 init_completion(&sbi
->s_kobj_unregister
);
3730 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3733 ext4_mb_release(sb
);
3734 ext4_ext_release(sb
);
3738 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3739 ext4_orphan_cleanup(sb
, es
);
3740 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3741 if (needs_recovery
) {
3742 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3743 ext4_mark_recovery_complete(sb
, es
);
3745 if (EXT4_SB(sb
)->s_journal
) {
3746 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3747 descr
= " journalled data mode";
3748 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3749 descr
= " ordered data mode";
3751 descr
= " writeback data mode";
3753 descr
= "out journal";
3755 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3756 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3757 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3759 if (es
->s_error_count
)
3760 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3767 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3773 ext4_msg(sb
, KERN_ERR
, "mount failed");
3774 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3776 ext4_release_system_zone(sb
);
3777 if (sbi
->s_journal
) {
3778 jbd2_journal_destroy(sbi
->s_journal
);
3779 sbi
->s_journal
= NULL
;
3782 del_timer(&sbi
->s_err_report
);
3783 if (sbi
->s_flex_groups
)
3784 ext4_kvfree(sbi
->s_flex_groups
);
3785 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3786 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3787 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3788 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3790 kthread_stop(sbi
->s_mmp_tsk
);
3792 for (i
= 0; i
< db_count
; i
++)
3793 brelse(sbi
->s_group_desc
[i
]);
3794 ext4_kvfree(sbi
->s_group_desc
);
3797 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3800 for (i
= 0; i
< MAXQUOTAS
; i
++)
3801 kfree(sbi
->s_qf_names
[i
]);
3803 ext4_blkdev_remove(sbi
);
3806 sb
->s_fs_info
= NULL
;
3807 kfree(sbi
->s_blockgroup_lock
);
3815 * Setup any per-fs journal parameters now. We'll do this both on
3816 * initial mount, once the journal has been initialised but before we've
3817 * done any recovery; and again on any subsequent remount.
3819 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3821 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3823 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3824 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3825 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3827 write_lock(&journal
->j_state_lock
);
3828 if (test_opt(sb
, BARRIER
))
3829 journal
->j_flags
|= JBD2_BARRIER
;
3831 journal
->j_flags
&= ~JBD2_BARRIER
;
3832 if (test_opt(sb
, DATA_ERR_ABORT
))
3833 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3835 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3836 write_unlock(&journal
->j_state_lock
);
3839 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3840 unsigned int journal_inum
)
3842 struct inode
*journal_inode
;
3845 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3847 /* First, test for the existence of a valid inode on disk. Bad
3848 * things happen if we iget() an unused inode, as the subsequent
3849 * iput() will try to delete it. */
3851 journal_inode
= ext4_iget(sb
, journal_inum
);
3852 if (IS_ERR(journal_inode
)) {
3853 ext4_msg(sb
, KERN_ERR
, "no journal found");
3856 if (!journal_inode
->i_nlink
) {
3857 make_bad_inode(journal_inode
);
3858 iput(journal_inode
);
3859 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3863 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3864 journal_inode
, journal_inode
->i_size
);
3865 if (!S_ISREG(journal_inode
->i_mode
)) {
3866 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3867 iput(journal_inode
);
3871 journal
= jbd2_journal_init_inode(journal_inode
);
3873 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3874 iput(journal_inode
);
3877 journal
->j_private
= sb
;
3878 ext4_init_journal_params(sb
, journal
);
3882 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3885 struct buffer_head
*bh
;
3889 int hblock
, blocksize
;
3890 ext4_fsblk_t sb_block
;
3891 unsigned long offset
;
3892 struct ext4_super_block
*es
;
3893 struct block_device
*bdev
;
3895 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3897 bdev
= ext4_blkdev_get(j_dev
, sb
);
3901 blocksize
= sb
->s_blocksize
;
3902 hblock
= bdev_logical_block_size(bdev
);
3903 if (blocksize
< hblock
) {
3904 ext4_msg(sb
, KERN_ERR
,
3905 "blocksize too small for journal device");
3909 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3910 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3911 set_blocksize(bdev
, blocksize
);
3912 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3913 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3914 "external journal");
3918 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3919 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3920 !(le32_to_cpu(es
->s_feature_incompat
) &
3921 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3922 ext4_msg(sb
, KERN_ERR
, "external journal has "
3928 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3929 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3934 len
= ext4_blocks_count(es
);
3935 start
= sb_block
+ 1;
3936 brelse(bh
); /* we're done with the superblock */
3938 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3939 start
, len
, blocksize
);
3941 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3944 journal
->j_private
= sb
;
3945 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3946 wait_on_buffer(journal
->j_sb_buffer
);
3947 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3948 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3951 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3952 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3953 "user (unsupported) - %d",
3954 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3957 EXT4_SB(sb
)->journal_bdev
= bdev
;
3958 ext4_init_journal_params(sb
, journal
);
3962 jbd2_journal_destroy(journal
);
3964 ext4_blkdev_put(bdev
);
3968 static int ext4_load_journal(struct super_block
*sb
,
3969 struct ext4_super_block
*es
,
3970 unsigned long journal_devnum
)
3973 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3976 int really_read_only
;
3978 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3980 if (journal_devnum
&&
3981 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3982 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3983 "numbers have changed");
3984 journal_dev
= new_decode_dev(journal_devnum
);
3986 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3988 really_read_only
= bdev_read_only(sb
->s_bdev
);
3991 * Are we loading a blank journal or performing recovery after a
3992 * crash? For recovery, we need to check in advance whether we
3993 * can get read-write access to the device.
3995 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3996 if (sb
->s_flags
& MS_RDONLY
) {
3997 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3998 "required on readonly filesystem");
3999 if (really_read_only
) {
4000 ext4_msg(sb
, KERN_ERR
, "write access "
4001 "unavailable, cannot proceed");
4004 ext4_msg(sb
, KERN_INFO
, "write access will "
4005 "be enabled during recovery");
4009 if (journal_inum
&& journal_dev
) {
4010 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
4011 "and inode journals!");
4016 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
4019 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
4023 if (!(journal
->j_flags
& JBD2_BARRIER
))
4024 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
4026 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
4027 err
= jbd2_journal_update_format(journal
);
4029 ext4_msg(sb
, KERN_ERR
, "error updating journal");
4030 jbd2_journal_destroy(journal
);
4035 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
4036 err
= jbd2_journal_wipe(journal
, !really_read_only
);
4038 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
4040 memcpy(save
, ((char *) es
) +
4041 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
4042 err
= jbd2_journal_load(journal
);
4044 memcpy(((char *) es
) + EXT4_S_ERR_START
,
4045 save
, EXT4_S_ERR_LEN
);
4050 ext4_msg(sb
, KERN_ERR
, "error loading journal");
4051 jbd2_journal_destroy(journal
);
4055 EXT4_SB(sb
)->s_journal
= journal
;
4056 ext4_clear_journal_err(sb
, es
);
4058 if (!really_read_only
&& journal_devnum
&&
4059 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
4060 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
4062 /* Make sure we flush the recovery flag to disk. */
4063 ext4_commit_super(sb
, 1);
4069 static int ext4_commit_super(struct super_block
*sb
, int sync
)
4071 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
4072 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
4077 if (buffer_write_io_error(sbh
)) {
4079 * Oh, dear. A previous attempt to write the
4080 * superblock failed. This could happen because the
4081 * USB device was yanked out. Or it could happen to
4082 * be a transient write error and maybe the block will
4083 * be remapped. Nothing we can do but to retry the
4084 * write and hope for the best.
4086 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4087 "superblock detected");
4088 clear_buffer_write_io_error(sbh
);
4089 set_buffer_uptodate(sbh
);
4092 * If the file system is mounted read-only, don't update the
4093 * superblock write time. This avoids updating the superblock
4094 * write time when we are mounting the root file system
4095 * read/only but we need to replay the journal; at that point,
4096 * for people who are east of GMT and who make their clock
4097 * tick in localtime for Windows bug-for-bug compatibility,
4098 * the clock is set in the future, and this will cause e2fsck
4099 * to complain and force a full file system check.
4101 if (!(sb
->s_flags
& MS_RDONLY
))
4102 es
->s_wtime
= cpu_to_le32(get_seconds());
4103 if (sb
->s_bdev
->bd_part
)
4104 es
->s_kbytes_written
=
4105 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4106 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4107 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4109 es
->s_kbytes_written
=
4110 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4111 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
4112 &EXT4_SB(sb
)->s_freeblocks_counter
));
4113 es
->s_free_inodes_count
=
4114 cpu_to_le32(percpu_counter_sum_positive(
4115 &EXT4_SB(sb
)->s_freeinodes_counter
));
4117 BUFFER_TRACE(sbh
, "marking dirty");
4118 mark_buffer_dirty(sbh
);
4120 error
= sync_dirty_buffer(sbh
);
4124 error
= buffer_write_io_error(sbh
);
4126 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4128 clear_buffer_write_io_error(sbh
);
4129 set_buffer_uptodate(sbh
);
4136 * Have we just finished recovery? If so, and if we are mounting (or
4137 * remounting) the filesystem readonly, then we will end up with a
4138 * consistent fs on disk. Record that fact.
4140 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4141 struct ext4_super_block
*es
)
4143 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4145 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4146 BUG_ON(journal
!= NULL
);
4149 jbd2_journal_lock_updates(journal
);
4150 if (jbd2_journal_flush(journal
) < 0)
4153 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4154 sb
->s_flags
& MS_RDONLY
) {
4155 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4156 ext4_commit_super(sb
, 1);
4160 jbd2_journal_unlock_updates(journal
);
4164 * If we are mounting (or read-write remounting) a filesystem whose journal
4165 * has recorded an error from a previous lifetime, move that error to the
4166 * main filesystem now.
4168 static void ext4_clear_journal_err(struct super_block
*sb
,
4169 struct ext4_super_block
*es
)
4175 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4177 journal
= EXT4_SB(sb
)->s_journal
;
4180 * Now check for any error status which may have been recorded in the
4181 * journal by a prior ext4_error() or ext4_abort()
4184 j_errno
= jbd2_journal_errno(journal
);
4188 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4189 ext4_warning(sb
, "Filesystem error recorded "
4190 "from previous mount: %s", errstr
);
4191 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4193 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4194 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4195 ext4_commit_super(sb
, 1);
4197 jbd2_journal_clear_err(journal
);
4202 * Force the running and committing transactions to commit,
4203 * and wait on the commit.
4205 int ext4_force_commit(struct super_block
*sb
)
4210 if (sb
->s_flags
& MS_RDONLY
)
4213 journal
= EXT4_SB(sb
)->s_journal
;
4215 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4216 ret
= ext4_journal_force_commit(journal
);
4222 static void ext4_write_super(struct super_block
*sb
)
4225 ext4_commit_super(sb
, 1);
4229 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4233 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4235 trace_ext4_sync_fs(sb
, wait
);
4236 flush_workqueue(sbi
->dio_unwritten_wq
);
4237 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4239 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4245 * LVM calls this function before a (read-only) snapshot is created. This
4246 * gives us a chance to flush the journal completely and mark the fs clean.
4248 * Note that only this function cannot bring a filesystem to be in a clean
4249 * state independently, because ext4 prevents a new handle from being started
4250 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4253 static int ext4_freeze(struct super_block
*sb
)
4258 if (sb
->s_flags
& MS_RDONLY
)
4261 journal
= EXT4_SB(sb
)->s_journal
;
4263 /* Now we set up the journal barrier. */
4264 jbd2_journal_lock_updates(journal
);
4267 * Don't clear the needs_recovery flag if we failed to flush
4270 error
= jbd2_journal_flush(journal
);
4274 /* Journal blocked and flushed, clear needs_recovery flag. */
4275 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4276 error
= ext4_commit_super(sb
, 1);
4278 /* we rely on s_frozen to stop further updates */
4279 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4284 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4285 * flag here, even though the filesystem is not technically dirty yet.
4287 static int ext4_unfreeze(struct super_block
*sb
)
4289 if (sb
->s_flags
& MS_RDONLY
)
4293 /* Reset the needs_recovery flag before the fs is unlocked. */
4294 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4295 ext4_commit_super(sb
, 1);
4301 * Structure to save mount options for ext4_remount's benefit
4303 struct ext4_mount_options
{
4304 unsigned long s_mount_opt
;
4305 unsigned long s_mount_opt2
;
4308 unsigned long s_commit_interval
;
4309 u32 s_min_batch_time
, s_max_batch_time
;
4312 char *s_qf_names
[MAXQUOTAS
];
4316 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4318 struct ext4_super_block
*es
;
4319 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4320 ext4_fsblk_t n_blocks_count
= 0;
4321 unsigned long old_sb_flags
;
4322 struct ext4_mount_options old_opts
;
4323 int enable_quota
= 0;
4325 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4330 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4332 /* Store the original options */
4334 old_sb_flags
= sb
->s_flags
;
4335 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4336 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4337 old_opts
.s_resuid
= sbi
->s_resuid
;
4338 old_opts
.s_resgid
= sbi
->s_resgid
;
4339 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4340 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4341 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4343 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4344 for (i
= 0; i
< MAXQUOTAS
; i
++)
4345 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4347 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4348 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4351 * Allow the "check" option to be passed as a remount option.
4353 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
4354 &n_blocks_count
, 1)) {
4359 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4360 ext4_abort(sb
, "Abort forced by user");
4362 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4363 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4367 if (sbi
->s_journal
) {
4368 ext4_init_journal_params(sb
, sbi
->s_journal
);
4369 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4372 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
4373 n_blocks_count
> ext4_blocks_count(es
)) {
4374 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4379 if (*flags
& MS_RDONLY
) {
4380 err
= dquot_suspend(sb
, -1);
4385 * First of all, the unconditional stuff we have to do
4386 * to disable replay of the journal when we next remount
4388 sb
->s_flags
|= MS_RDONLY
;
4391 * OK, test if we are remounting a valid rw partition
4392 * readonly, and if so set the rdonly flag and then
4393 * mark the partition as valid again.
4395 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4396 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4397 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4400 ext4_mark_recovery_complete(sb
, es
);
4402 /* Make sure we can mount this feature set readwrite */
4403 if (!ext4_feature_set_ok(sb
, 0)) {
4408 * Make sure the group descriptor checksums
4409 * are sane. If they aren't, refuse to remount r/w.
4411 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4412 struct ext4_group_desc
*gdp
=
4413 ext4_get_group_desc(sb
, g
, NULL
);
4415 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4416 ext4_msg(sb
, KERN_ERR
,
4417 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4418 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4419 le16_to_cpu(gdp
->bg_checksum
));
4426 * If we have an unprocessed orphan list hanging
4427 * around from a previously readonly bdev mount,
4428 * require a full umount/remount for now.
4430 if (es
->s_last_orphan
) {
4431 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4432 "remount RDWR because of unprocessed "
4433 "orphan inode list. Please "
4434 "umount/remount instead");
4440 * Mounting a RDONLY partition read-write, so reread
4441 * and store the current valid flag. (It may have
4442 * been changed by e2fsck since we originally mounted
4446 ext4_clear_journal_err(sb
, es
);
4447 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4448 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
4450 if (!ext4_setup_super(sb
, es
, 0))
4451 sb
->s_flags
&= ~MS_RDONLY
;
4452 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4453 EXT4_FEATURE_INCOMPAT_MMP
))
4454 if (ext4_multi_mount_protect(sb
,
4455 le64_to_cpu(es
->s_mmp_block
))) {
4464 * Reinitialize lazy itable initialization thread based on
4467 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4468 ext4_unregister_li_request(sb
);
4470 ext4_group_t first_not_zeroed
;
4471 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4472 ext4_register_li_request(sb
, first_not_zeroed
);
4475 ext4_setup_system_zone(sb
);
4476 if (sbi
->s_journal
== NULL
)
4477 ext4_commit_super(sb
, 1);
4480 /* Release old quota file names */
4481 for (i
= 0; i
< MAXQUOTAS
; i
++)
4482 if (old_opts
.s_qf_names
[i
] &&
4483 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4484 kfree(old_opts
.s_qf_names
[i
]);
4488 dquot_resume(sb
, -1);
4490 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4495 sb
->s_flags
= old_sb_flags
;
4496 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4497 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4498 sbi
->s_resuid
= old_opts
.s_resuid
;
4499 sbi
->s_resgid
= old_opts
.s_resgid
;
4500 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4501 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4502 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4504 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4505 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4506 if (sbi
->s_qf_names
[i
] &&
4507 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4508 kfree(sbi
->s_qf_names
[i
]);
4509 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4517 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4519 struct super_block
*sb
= dentry
->d_sb
;
4520 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4521 struct ext4_super_block
*es
= sbi
->s_es
;
4525 if (test_opt(sb
, MINIX_DF
)) {
4526 sbi
->s_overhead_last
= 0;
4527 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4528 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4529 ext4_fsblk_t overhead
= 0;
4532 * Compute the overhead (FS structures). This is constant
4533 * for a given filesystem unless the number of block groups
4534 * changes so we cache the previous value until it does.
4538 * All of the blocks before first_data_block are
4541 overhead
= le32_to_cpu(es
->s_first_data_block
);
4544 * Add the overhead attributed to the superblock and
4545 * block group descriptors. If the sparse superblocks
4546 * feature is turned on, then not all groups have this.
4548 for (i
= 0; i
< ngroups
; i
++) {
4549 overhead
+= ext4_bg_has_super(sb
, i
) +
4550 ext4_bg_num_gdb(sb
, i
);
4555 * Every block group has an inode bitmap, a block
4556 * bitmap, and an inode table.
4558 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
4559 sbi
->s_overhead_last
= overhead
;
4561 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4564 buf
->f_type
= EXT4_SUPER_MAGIC
;
4565 buf
->f_bsize
= sb
->s_blocksize
;
4566 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
4567 bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
4568 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
4569 /* prevent underflow in case that few free space is available */
4570 buf
->f_bfree
= max_t(s64
, bfree
, 0);
4571 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4572 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4574 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4575 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4576 buf
->f_namelen
= EXT4_NAME_LEN
;
4577 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4578 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4579 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4580 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4585 /* Helper function for writing quotas on sync - we need to start transaction
4586 * before quota file is locked for write. Otherwise the are possible deadlocks:
4587 * Process 1 Process 2
4588 * ext4_create() quota_sync()
4589 * jbd2_journal_start() write_dquot()
4590 * dquot_initialize() down(dqio_mutex)
4591 * down(dqio_mutex) jbd2_journal_start()
4597 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4599 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4602 static int ext4_write_dquot(struct dquot
*dquot
)
4606 struct inode
*inode
;
4608 inode
= dquot_to_inode(dquot
);
4609 handle
= ext4_journal_start(inode
,
4610 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4612 return PTR_ERR(handle
);
4613 ret
= dquot_commit(dquot
);
4614 err
= ext4_journal_stop(handle
);
4620 static int ext4_acquire_dquot(struct dquot
*dquot
)
4625 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4626 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4628 return PTR_ERR(handle
);
4629 ret
= dquot_acquire(dquot
);
4630 err
= ext4_journal_stop(handle
);
4636 static int ext4_release_dquot(struct dquot
*dquot
)
4641 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4642 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4643 if (IS_ERR(handle
)) {
4644 /* Release dquot anyway to avoid endless cycle in dqput() */
4645 dquot_release(dquot
);
4646 return PTR_ERR(handle
);
4648 ret
= dquot_release(dquot
);
4649 err
= ext4_journal_stop(handle
);
4655 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4657 /* Are we journaling quotas? */
4658 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4659 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4660 dquot_mark_dquot_dirty(dquot
);
4661 return ext4_write_dquot(dquot
);
4663 return dquot_mark_dquot_dirty(dquot
);
4667 static int ext4_write_info(struct super_block
*sb
, int type
)
4672 /* Data block + inode block */
4673 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4675 return PTR_ERR(handle
);
4676 ret
= dquot_commit_info(sb
, type
);
4677 err
= ext4_journal_stop(handle
);
4684 * Turn on quotas during mount time - we need to find
4685 * the quota file and such...
4687 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4689 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4690 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4694 * Standard function to be called on quota_on
4696 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4701 if (!test_opt(sb
, QUOTA
))
4704 /* Quotafile not on the same filesystem? */
4705 if (path
->mnt
->mnt_sb
!= sb
)
4707 /* Journaling quota? */
4708 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4709 /* Quotafile not in fs root? */
4710 if (path
->dentry
->d_parent
!= sb
->s_root
)
4711 ext4_msg(sb
, KERN_WARNING
,
4712 "Quota file not on filesystem root. "
4713 "Journaled quota will not work");
4717 * When we journal data on quota file, we have to flush journal to see
4718 * all updates to the file when we bypass pagecache...
4720 if (EXT4_SB(sb
)->s_journal
&&
4721 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4723 * We don't need to lock updates but journal_flush() could
4724 * otherwise be livelocked...
4726 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4727 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4728 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4733 return dquot_quota_on(sb
, type
, format_id
, path
);
4736 static int ext4_quota_off(struct super_block
*sb
, int type
)
4738 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4741 /* Force all delayed allocation blocks to be allocated.
4742 * Caller already holds s_umount sem */
4743 if (test_opt(sb
, DELALLOC
))
4744 sync_filesystem(sb
);
4749 /* Update modification times of quota files when userspace can
4750 * start looking at them */
4751 handle
= ext4_journal_start(inode
, 1);
4754 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4755 ext4_mark_inode_dirty(handle
, inode
);
4756 ext4_journal_stop(handle
);
4759 return dquot_quota_off(sb
, type
);
4762 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4763 * acquiring the locks... As quota files are never truncated and quota code
4764 * itself serializes the operations (and no one else should touch the files)
4765 * we don't have to be afraid of races */
4766 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4767 size_t len
, loff_t off
)
4769 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4770 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4772 int offset
= off
& (sb
->s_blocksize
- 1);
4775 struct buffer_head
*bh
;
4776 loff_t i_size
= i_size_read(inode
);
4780 if (off
+len
> i_size
)
4783 while (toread
> 0) {
4784 tocopy
= sb
->s_blocksize
- offset
< toread
?
4785 sb
->s_blocksize
- offset
: toread
;
4786 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4789 if (!bh
) /* A hole? */
4790 memset(data
, 0, tocopy
);
4792 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4802 /* Write to quotafile (we know the transaction is already started and has
4803 * enough credits) */
4804 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4805 const char *data
, size_t len
, loff_t off
)
4807 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4808 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4810 int offset
= off
& (sb
->s_blocksize
- 1);
4811 struct buffer_head
*bh
;
4812 handle_t
*handle
= journal_current_handle();
4814 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4815 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4816 " cancelled because transaction is not started",
4817 (unsigned long long)off
, (unsigned long long)len
);
4821 * Since we account only one data block in transaction credits,
4822 * then it is impossible to cross a block boundary.
4824 if (sb
->s_blocksize
- offset
< len
) {
4825 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4826 " cancelled because not block aligned",
4827 (unsigned long long)off
, (unsigned long long)len
);
4831 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4832 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4835 err
= ext4_journal_get_write_access(handle
, bh
);
4841 memcpy(bh
->b_data
+offset
, data
, len
);
4842 flush_dcache_page(bh
->b_page
);
4844 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4848 mutex_unlock(&inode
->i_mutex
);
4851 if (inode
->i_size
< off
+ len
) {
4852 i_size_write(inode
, off
+ len
);
4853 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4854 ext4_mark_inode_dirty(handle
, inode
);
4856 mutex_unlock(&inode
->i_mutex
);
4862 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4863 const char *dev_name
, void *data
)
4865 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4868 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4869 static inline void register_as_ext2(void)
4871 int err
= register_filesystem(&ext2_fs_type
);
4874 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4877 static inline void unregister_as_ext2(void)
4879 unregister_filesystem(&ext2_fs_type
);
4882 static inline int ext2_feature_set_ok(struct super_block
*sb
)
4884 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
4886 if (sb
->s_flags
& MS_RDONLY
)
4888 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
4892 MODULE_ALIAS("ext2");
4894 static inline void register_as_ext2(void) { }
4895 static inline void unregister_as_ext2(void) { }
4896 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
4899 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4900 static inline void register_as_ext3(void)
4902 int err
= register_filesystem(&ext3_fs_type
);
4905 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4908 static inline void unregister_as_ext3(void)
4910 unregister_filesystem(&ext3_fs_type
);
4913 static inline int ext3_feature_set_ok(struct super_block
*sb
)
4915 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
4917 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
4919 if (sb
->s_flags
& MS_RDONLY
)
4921 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
4925 MODULE_ALIAS("ext3");
4927 static inline void register_as_ext3(void) { }
4928 static inline void unregister_as_ext3(void) { }
4929 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
4932 static struct file_system_type ext4_fs_type
= {
4933 .owner
= THIS_MODULE
,
4935 .mount
= ext4_mount
,
4936 .kill_sb
= kill_block_super
,
4937 .fs_flags
= FS_REQUIRES_DEV
,
4940 static int __init
ext4_init_feat_adverts(void)
4942 struct ext4_features
*ef
;
4945 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4949 ef
->f_kobj
.kset
= ext4_kset
;
4950 init_completion(&ef
->f_kobj_unregister
);
4951 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4964 static void ext4_exit_feat_adverts(void)
4966 kobject_put(&ext4_feat
->f_kobj
);
4967 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
4971 /* Shared across all ext4 file systems */
4972 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
4973 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
4975 static int __init
ext4_init_fs(void)
4979 ext4_check_flag_values();
4981 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
4982 mutex_init(&ext4__aio_mutex
[i
]);
4983 init_waitqueue_head(&ext4__ioend_wq
[i
]);
4986 err
= ext4_init_pageio();
4989 err
= ext4_init_system_zone();
4992 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4995 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4996 if (!ext4_proc_root
)
4999 err
= ext4_init_feat_adverts();
5003 err
= ext4_init_mballoc();
5007 err
= ext4_init_xattr();
5010 err
= init_inodecache();
5015 err
= register_filesystem(&ext4_fs_type
);
5019 ext4_li_info
= NULL
;
5020 mutex_init(&ext4_li_mtx
);
5023 unregister_as_ext2();
5024 unregister_as_ext3();
5025 destroy_inodecache();
5029 ext4_exit_mballoc();
5031 ext4_exit_feat_adverts();
5033 remove_proc_entry("fs/ext4", NULL
);
5035 kset_unregister(ext4_kset
);
5037 ext4_exit_system_zone();
5043 static void __exit
ext4_exit_fs(void)
5045 ext4_destroy_lazyinit_thread();
5046 unregister_as_ext2();
5047 unregister_as_ext3();
5048 unregister_filesystem(&ext4_fs_type
);
5049 destroy_inodecache();
5051 ext4_exit_mballoc();
5052 ext4_exit_feat_adverts();
5053 remove_proc_entry("fs/ext4", NULL
);
5054 kset_unregister(ext4_kset
);
5055 ext4_exit_system_zone();
5059 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5060 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5061 MODULE_LICENSE("GPL");
5062 module_init(ext4_init_fs
)
5063 module_exit(ext4_exit_fs
)