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_extents.h"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry
*ext4_proc_root
;
58 static struct kset
*ext4_kset
;
59 static struct ext4_lazy_init
*ext4_li_info
;
60 static struct mutex ext4_li_mtx
;
61 static struct ext4_features
*ext4_feat
;
63 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
64 unsigned long journal_devnum
);
65 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
);
66 static int ext4_commit_super(struct super_block
*sb
, int sync
);
67 static void ext4_mark_recovery_complete(struct super_block
*sb
,
68 struct ext4_super_block
*es
);
69 static void ext4_clear_journal_err(struct super_block
*sb
,
70 struct ext4_super_block
*es
);
71 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
72 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
74 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
75 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
76 static int ext4_unfreeze(struct super_block
*sb
);
77 static void ext4_write_super(struct super_block
*sb
);
78 static int ext4_freeze(struct super_block
*sb
);
79 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
80 const char *dev_name
, void *data
);
81 static inline int ext2_feature_set_ok(struct super_block
*sb
);
82 static inline int ext3_feature_set_ok(struct super_block
*sb
);
83 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block
*sb
);
86 static void ext4_clear_request_list(void);
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type
= {
93 .kill_sb
= kill_block_super
,
94 .fs_flags
= FS_REQUIRES_DEV
,
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type
= {
104 .owner
= THIS_MODULE
,
107 .kill_sb
= kill_block_super
,
108 .fs_flags
= FS_REQUIRES_DEV
,
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
112 #define IS_EXT3_SB(sb) (0)
115 void *ext4_kvmalloc(size_t size
, gfp_t flags
)
119 ret
= kmalloc(size
, flags
);
121 ret
= __vmalloc(size
, flags
, PAGE_KERNEL
);
125 void *ext4_kvzalloc(size_t size
, gfp_t flags
)
129 ret
= kzalloc(size
, flags
);
131 ret
= __vmalloc(size
, flags
| __GFP_ZERO
, PAGE_KERNEL
);
135 void ext4_kvfree(void *ptr
)
137 if (is_vmalloc_addr(ptr
))
144 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
145 struct ext4_group_desc
*bg
)
147 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
148 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
149 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
152 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
153 struct ext4_group_desc
*bg
)
155 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
156 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
157 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
160 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
161 struct ext4_group_desc
*bg
)
163 return le32_to_cpu(bg
->bg_inode_table_lo
) |
164 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
165 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
168 __u32
ext4_free_group_clusters(struct super_block
*sb
,
169 struct ext4_group_desc
*bg
)
171 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
172 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
173 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
176 __u32
ext4_free_inodes_count(struct super_block
*sb
,
177 struct ext4_group_desc
*bg
)
179 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
180 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
181 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
184 __u32
ext4_used_dirs_count(struct super_block
*sb
,
185 struct ext4_group_desc
*bg
)
187 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
188 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
189 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
192 __u32
ext4_itable_unused_count(struct super_block
*sb
,
193 struct ext4_group_desc
*bg
)
195 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
196 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
197 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
200 void ext4_block_bitmap_set(struct super_block
*sb
,
201 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
203 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
204 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
205 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
208 void ext4_inode_bitmap_set(struct super_block
*sb
,
209 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
211 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
212 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
213 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
216 void ext4_inode_table_set(struct super_block
*sb
,
217 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
219 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
220 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
221 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
224 void ext4_free_group_clusters_set(struct super_block
*sb
,
225 struct ext4_group_desc
*bg
, __u32 count
)
227 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
228 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
229 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
232 void ext4_free_inodes_set(struct super_block
*sb
,
233 struct ext4_group_desc
*bg
, __u32 count
)
235 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
236 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
237 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
240 void ext4_used_dirs_set(struct super_block
*sb
,
241 struct ext4_group_desc
*bg
, __u32 count
)
243 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
244 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
245 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
248 void ext4_itable_unused_set(struct super_block
*sb
,
249 struct ext4_group_desc
*bg
, __u32 count
)
251 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
252 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
253 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
257 /* Just increment the non-pointer handle value */
258 static handle_t
*ext4_get_nojournal(void)
260 handle_t
*handle
= current
->journal_info
;
261 unsigned long ref_cnt
= (unsigned long)handle
;
263 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
266 handle
= (handle_t
*)ref_cnt
;
268 current
->journal_info
= handle
;
273 /* Decrement the non-pointer handle value */
274 static void ext4_put_nojournal(handle_t
*handle
)
276 unsigned long ref_cnt
= (unsigned long)handle
;
278 BUG_ON(ref_cnt
== 0);
281 handle
= (handle_t
*)ref_cnt
;
283 current
->journal_info
= handle
;
287 * Wrappers for jbd2_journal_start/end.
289 * The only special thing we need to do here is to make sure that all
290 * journal_end calls result in the superblock being marked dirty, so
291 * that sync() will call the filesystem's write_super callback if
294 * To avoid j_barrier hold in userspace when a user calls freeze(),
295 * ext4 prevents a new handle from being started by s_frozen, which
296 * is in an upper layer.
298 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
303 trace_ext4_journal_start(sb
, nblocks
, _RET_IP_
);
304 if (sb
->s_flags
& MS_RDONLY
)
305 return ERR_PTR(-EROFS
);
307 journal
= EXT4_SB(sb
)->s_journal
;
308 handle
= ext4_journal_current_handle();
311 * If a handle has been started, it should be allowed to
312 * finish, otherwise deadlock could happen between freeze
313 * and others(e.g. truncate) due to the restart of the
314 * journal handle if the filesystem is forzen and active
315 * handles are not stopped.
318 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
321 return ext4_get_nojournal();
323 * Special case here: if the journal has aborted behind our
324 * backs (eg. EIO in the commit thread), then we still need to
325 * take the FS itself readonly cleanly.
327 if (is_journal_aborted(journal
)) {
328 ext4_abort(sb
, "Detected aborted journal");
329 return ERR_PTR(-EROFS
);
331 return jbd2_journal_start(journal
, nblocks
);
335 * The only special thing we need to do here is to make sure that all
336 * jbd2_journal_stop calls result in the superblock being marked dirty, so
337 * that sync() will call the filesystem's write_super callback if
340 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
342 struct super_block
*sb
;
346 if (!ext4_handle_valid(handle
)) {
347 ext4_put_nojournal(handle
);
350 sb
= handle
->h_transaction
->t_journal
->j_private
;
352 rc
= jbd2_journal_stop(handle
);
357 __ext4_std_error(sb
, where
, line
, err
);
361 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
362 const char *err_fn
, struct buffer_head
*bh
,
363 handle_t
*handle
, int err
)
366 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
368 BUG_ON(!ext4_handle_valid(handle
));
371 BUFFER_TRACE(bh
, "abort");
376 if (is_handle_aborted(handle
))
379 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
380 caller
, line
, errstr
, err_fn
);
382 jbd2_journal_abort_handle(handle
);
385 static void __save_error_info(struct super_block
*sb
, const char *func
,
388 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
390 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
391 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
392 es
->s_last_error_time
= cpu_to_le32(get_seconds());
393 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
394 es
->s_last_error_line
= cpu_to_le32(line
);
395 if (!es
->s_first_error_time
) {
396 es
->s_first_error_time
= es
->s_last_error_time
;
397 strncpy(es
->s_first_error_func
, func
,
398 sizeof(es
->s_first_error_func
));
399 es
->s_first_error_line
= cpu_to_le32(line
);
400 es
->s_first_error_ino
= es
->s_last_error_ino
;
401 es
->s_first_error_block
= es
->s_last_error_block
;
404 * Start the daily error reporting function if it hasn't been
407 if (!es
->s_error_count
)
408 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
409 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
412 static void save_error_info(struct super_block
*sb
, const char *func
,
415 __save_error_info(sb
, func
, line
);
416 ext4_commit_super(sb
, 1);
420 * The del_gendisk() function uninitializes the disk-specific data
421 * structures, including the bdi structure, without telling anyone
422 * else. Once this happens, any attempt to call mark_buffer_dirty()
423 * (for example, by ext4_commit_super), will cause a kernel OOPS.
424 * This is a kludge to prevent these oops until we can put in a proper
425 * hook in del_gendisk() to inform the VFS and file system layers.
427 static int block_device_ejected(struct super_block
*sb
)
429 struct inode
*bd_inode
= sb
->s_bdev
->bd_inode
;
430 struct backing_dev_info
*bdi
= bd_inode
->i_mapping
->backing_dev_info
;
432 return bdi
->dev
== NULL
;
435 static void ext4_journal_commit_callback(journal_t
*journal
, transaction_t
*txn
)
437 struct super_block
*sb
= journal
->j_private
;
438 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
439 int error
= is_journal_aborted(journal
);
440 struct ext4_journal_cb_entry
*jce
, *tmp
;
442 spin_lock(&sbi
->s_md_lock
);
443 list_for_each_entry_safe(jce
, tmp
, &txn
->t_private_list
, jce_list
) {
444 list_del_init(&jce
->jce_list
);
445 spin_unlock(&sbi
->s_md_lock
);
446 jce
->jce_func(sb
, jce
, error
);
447 spin_lock(&sbi
->s_md_lock
);
449 spin_unlock(&sbi
->s_md_lock
);
452 /* Deal with the reporting of failure conditions on a filesystem such as
453 * inconsistencies detected or read IO failures.
455 * On ext2, we can store the error state of the filesystem in the
456 * superblock. That is not possible on ext4, because we may have other
457 * write ordering constraints on the superblock which prevent us from
458 * writing it out straight away; and given that the journal is about to
459 * be aborted, we can't rely on the current, or future, transactions to
460 * write out the superblock safely.
462 * We'll just use the jbd2_journal_abort() error code to record an error in
463 * the journal instead. On recovery, the journal will complain about
464 * that error until we've noted it down and cleared it.
467 static void ext4_handle_error(struct super_block
*sb
)
469 if (sb
->s_flags
& MS_RDONLY
)
472 if (!test_opt(sb
, ERRORS_CONT
)) {
473 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
475 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
477 jbd2_journal_abort(journal
, -EIO
);
479 if (test_opt(sb
, ERRORS_RO
)) {
480 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
481 sb
->s_flags
|= MS_RDONLY
;
483 if (test_opt(sb
, ERRORS_PANIC
))
484 panic("EXT4-fs (device %s): panic forced after error\n",
488 void __ext4_error(struct super_block
*sb
, const char *function
,
489 unsigned int line
, const char *fmt
, ...)
491 struct va_format vaf
;
497 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
498 sb
->s_id
, function
, line
, current
->comm
, &vaf
);
501 ext4_handle_error(sb
);
504 void ext4_error_inode(struct inode
*inode
, const char *function
,
505 unsigned int line
, ext4_fsblk_t block
,
506 const char *fmt
, ...)
509 struct va_format vaf
;
510 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
512 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
513 es
->s_last_error_block
= cpu_to_le64(block
);
514 save_error_info(inode
->i_sb
, function
, line
);
518 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
519 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
521 printk(KERN_CONT
"block %llu: ", block
);
522 printk(KERN_CONT
"comm %s: %pV\n", current
->comm
, &vaf
);
525 ext4_handle_error(inode
->i_sb
);
528 void ext4_error_file(struct file
*file
, const char *function
,
529 unsigned int line
, ext4_fsblk_t block
,
530 const char *fmt
, ...)
533 struct va_format vaf
;
534 struct ext4_super_block
*es
;
535 struct inode
*inode
= file
->f_dentry
->d_inode
;
536 char pathname
[80], *path
;
538 es
= EXT4_SB(inode
->i_sb
)->s_es
;
539 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
540 save_error_info(inode
->i_sb
, function
, line
);
541 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
545 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
546 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
548 printk(KERN_CONT
"block %llu: ", block
);
552 printk(KERN_CONT
"comm %s: path %s: %pV\n", current
->comm
, path
, &vaf
);
555 ext4_handle_error(inode
->i_sb
);
558 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
565 errstr
= "IO failure";
568 errstr
= "Out of memory";
571 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
572 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
573 errstr
= "Journal has aborted";
575 errstr
= "Readonly filesystem";
578 /* If the caller passed in an extra buffer for unknown
579 * errors, textualise them now. Else we just return
582 /* Check for truncated error codes... */
583 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
592 /* __ext4_std_error decodes expected errors from journaling functions
593 * automatically and invokes the appropriate error response. */
595 void __ext4_std_error(struct super_block
*sb
, const char *function
,
596 unsigned int line
, int errno
)
601 /* Special case: if the error is EROFS, and we're not already
602 * inside a transaction, then there's really no point in logging
604 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
605 (sb
->s_flags
& MS_RDONLY
))
608 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
609 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
610 sb
->s_id
, function
, line
, errstr
);
611 save_error_info(sb
, function
, line
);
613 ext4_handle_error(sb
);
617 * ext4_abort is a much stronger failure handler than ext4_error. The
618 * abort function may be used to deal with unrecoverable failures such
619 * as journal IO errors or ENOMEM at a critical moment in log management.
621 * We unconditionally force the filesystem into an ABORT|READONLY state,
622 * unless the error response on the fs has been set to panic in which
623 * case we take the easy way out and panic immediately.
626 void __ext4_abort(struct super_block
*sb
, const char *function
,
627 unsigned int line
, const char *fmt
, ...)
631 save_error_info(sb
, function
, line
);
633 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
639 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
640 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
641 sb
->s_flags
|= MS_RDONLY
;
642 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
643 if (EXT4_SB(sb
)->s_journal
)
644 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
645 save_error_info(sb
, function
, line
);
647 if (test_opt(sb
, ERRORS_PANIC
))
648 panic("EXT4-fs panic from previous error\n");
651 void ext4_msg(struct super_block
*sb
, const char *prefix
, const char *fmt
, ...)
653 struct va_format vaf
;
659 printk("%sEXT4-fs (%s): %pV\n", prefix
, sb
->s_id
, &vaf
);
663 void __ext4_warning(struct super_block
*sb
, const char *function
,
664 unsigned int line
, const char *fmt
, ...)
666 struct va_format vaf
;
672 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: %pV\n",
673 sb
->s_id
, function
, line
, &vaf
);
677 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
678 struct super_block
*sb
, ext4_group_t grp
,
679 unsigned long ino
, ext4_fsblk_t block
,
680 const char *fmt
, ...)
684 struct va_format vaf
;
686 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
688 es
->s_last_error_ino
= cpu_to_le32(ino
);
689 es
->s_last_error_block
= cpu_to_le64(block
);
690 __save_error_info(sb
, function
, line
);
696 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u, ",
697 sb
->s_id
, function
, line
, grp
);
699 printk(KERN_CONT
"inode %lu: ", ino
);
701 printk(KERN_CONT
"block %llu:", (unsigned long long) block
);
702 printk(KERN_CONT
"%pV\n", &vaf
);
705 if (test_opt(sb
, ERRORS_CONT
)) {
706 ext4_commit_super(sb
, 0);
710 ext4_unlock_group(sb
, grp
);
711 ext4_handle_error(sb
);
713 * We only get here in the ERRORS_RO case; relocking the group
714 * may be dangerous, but nothing bad will happen since the
715 * filesystem will have already been marked read/only and the
716 * journal has been aborted. We return 1 as a hint to callers
717 * who might what to use the return value from
718 * ext4_grp_locked_error() to distinguish between the
719 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
720 * aggressively from the ext4 function in question, with a
721 * more appropriate error code.
723 ext4_lock_group(sb
, grp
);
727 void ext4_update_dynamic_rev(struct super_block
*sb
)
729 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
731 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
735 "updating to rev %d because of new feature flag, "
736 "running e2fsck is recommended",
739 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
740 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
741 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
742 /* leave es->s_feature_*compat flags alone */
743 /* es->s_uuid will be set by e2fsck if empty */
746 * The rest of the superblock fields should be zero, and if not it
747 * means they are likely already in use, so leave them alone. We
748 * can leave it up to e2fsck to clean up any inconsistencies there.
753 * Open the external journal device
755 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
757 struct block_device
*bdev
;
758 char b
[BDEVNAME_SIZE
];
760 bdev
= blkdev_get_by_dev(dev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
, sb
);
766 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
767 __bdevname(dev
, b
), PTR_ERR(bdev
));
772 * Release the journal device
774 static int ext4_blkdev_put(struct block_device
*bdev
)
776 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
779 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
781 struct block_device
*bdev
;
784 bdev
= sbi
->journal_bdev
;
786 ret
= ext4_blkdev_put(bdev
);
787 sbi
->journal_bdev
= NULL
;
792 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
794 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
797 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
801 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
802 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
804 printk(KERN_ERR
"sb_info orphan list:\n");
805 list_for_each(l
, &sbi
->s_orphan
) {
806 struct inode
*inode
= orphan_list_entry(l
);
808 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
809 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
810 inode
->i_mode
, inode
->i_nlink
,
815 static void ext4_put_super(struct super_block
*sb
)
817 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
818 struct ext4_super_block
*es
= sbi
->s_es
;
821 ext4_unregister_li_request(sb
);
822 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
824 flush_workqueue(sbi
->dio_unwritten_wq
);
825 destroy_workqueue(sbi
->dio_unwritten_wq
);
829 ext4_commit_super(sb
, 1);
831 if (sbi
->s_journal
) {
832 err
= jbd2_journal_destroy(sbi
->s_journal
);
833 sbi
->s_journal
= NULL
;
835 ext4_abort(sb
, "Couldn't clean up the journal");
838 del_timer(&sbi
->s_err_report
);
839 ext4_release_system_zone(sb
);
841 ext4_ext_release(sb
);
842 ext4_xattr_put_super(sb
);
844 if (!(sb
->s_flags
& MS_RDONLY
)) {
845 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
846 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
847 ext4_commit_super(sb
, 1);
850 remove_proc_entry("options", sbi
->s_proc
);
851 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
853 kobject_del(&sbi
->s_kobj
);
855 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
856 brelse(sbi
->s_group_desc
[i
]);
857 ext4_kvfree(sbi
->s_group_desc
);
858 ext4_kvfree(sbi
->s_flex_groups
);
859 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
860 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
861 percpu_counter_destroy(&sbi
->s_dirs_counter
);
862 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
865 for (i
= 0; i
< MAXQUOTAS
; i
++)
866 kfree(sbi
->s_qf_names
[i
]);
869 /* Debugging code just in case the in-memory inode orphan list
870 * isn't empty. The on-disk one can be non-empty if we've
871 * detected an error and taken the fs readonly, but the
872 * in-memory list had better be clean by this point. */
873 if (!list_empty(&sbi
->s_orphan
))
874 dump_orphan_list(sb
, sbi
);
875 J_ASSERT(list_empty(&sbi
->s_orphan
));
877 invalidate_bdev(sb
->s_bdev
);
878 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
880 * Invalidate the journal device's buffers. We don't want them
881 * floating about in memory - the physical journal device may
882 * hotswapped, and it breaks the `ro-after' testing code.
884 sync_blockdev(sbi
->journal_bdev
);
885 invalidate_bdev(sbi
->journal_bdev
);
886 ext4_blkdev_remove(sbi
);
889 kthread_stop(sbi
->s_mmp_tsk
);
890 sb
->s_fs_info
= NULL
;
892 * Now that we are completely done shutting down the
893 * superblock, we need to actually destroy the kobject.
896 kobject_put(&sbi
->s_kobj
);
897 wait_for_completion(&sbi
->s_kobj_unregister
);
898 kfree(sbi
->s_blockgroup_lock
);
902 static struct kmem_cache
*ext4_inode_cachep
;
905 * Called inside transaction, so use GFP_NOFS
907 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
909 struct ext4_inode_info
*ei
;
911 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
915 ei
->vfs_inode
.i_version
= 1;
916 ei
->vfs_inode
.i_data
.writeback_index
= 0;
917 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
918 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
919 spin_lock_init(&ei
->i_prealloc_lock
);
920 ei
->i_reserved_data_blocks
= 0;
921 ei
->i_reserved_meta_blocks
= 0;
922 ei
->i_allocated_meta_blocks
= 0;
923 ei
->i_da_metadata_calc_len
= 0;
924 spin_lock_init(&(ei
->i_block_reservation_lock
));
926 ei
->i_reserved_quota
= 0;
929 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
930 spin_lock_init(&ei
->i_completed_io_lock
);
931 ei
->cur_aio_dio
= NULL
;
933 ei
->i_datasync_tid
= 0;
934 atomic_set(&ei
->i_ioend_count
, 0);
935 atomic_set(&ei
->i_aiodio_unwritten
, 0);
937 return &ei
->vfs_inode
;
940 static int ext4_drop_inode(struct inode
*inode
)
942 int drop
= generic_drop_inode(inode
);
944 trace_ext4_drop_inode(inode
, drop
);
948 static void ext4_i_callback(struct rcu_head
*head
)
950 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
951 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
954 static void ext4_destroy_inode(struct inode
*inode
)
956 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
957 ext4_msg(inode
->i_sb
, KERN_ERR
,
958 "Inode %lu (%p): orphan list check failed!",
959 inode
->i_ino
, EXT4_I(inode
));
960 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
961 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
965 call_rcu(&inode
->i_rcu
, ext4_i_callback
);
968 static void init_once(void *foo
)
970 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
972 INIT_LIST_HEAD(&ei
->i_orphan
);
973 #ifdef CONFIG_EXT4_FS_XATTR
974 init_rwsem(&ei
->xattr_sem
);
976 init_rwsem(&ei
->i_data_sem
);
977 inode_init_once(&ei
->vfs_inode
);
980 static int init_inodecache(void)
982 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
983 sizeof(struct ext4_inode_info
),
984 0, (SLAB_RECLAIM_ACCOUNT
|
987 if (ext4_inode_cachep
== NULL
)
992 static void destroy_inodecache(void)
994 kmem_cache_destroy(ext4_inode_cachep
);
997 void ext4_clear_inode(struct inode
*inode
)
999 invalidate_inode_buffers(inode
);
1000 end_writeback(inode
);
1002 ext4_discard_preallocations(inode
);
1003 if (EXT4_I(inode
)->jinode
) {
1004 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode
),
1005 EXT4_I(inode
)->jinode
);
1006 jbd2_free_inode(EXT4_I(inode
)->jinode
);
1007 EXT4_I(inode
)->jinode
= NULL
;
1011 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1012 u64 ino
, u32 generation
)
1014 struct inode
*inode
;
1016 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1017 return ERR_PTR(-ESTALE
);
1018 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1019 return ERR_PTR(-ESTALE
);
1021 /* iget isn't really right if the inode is currently unallocated!!
1023 * ext4_read_inode will return a bad_inode if the inode had been
1024 * deleted, so we should be safe.
1026 * Currently we don't know the generation for parent directory, so
1027 * a generation of 0 means "accept any"
1029 inode
= ext4_iget(sb
, ino
);
1031 return ERR_CAST(inode
);
1032 if (generation
&& inode
->i_generation
!= generation
) {
1034 return ERR_PTR(-ESTALE
);
1040 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1041 int fh_len
, int fh_type
)
1043 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1044 ext4_nfs_get_inode
);
1047 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1048 int fh_len
, int fh_type
)
1050 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1051 ext4_nfs_get_inode
);
1055 * Try to release metadata pages (indirect blocks, directories) which are
1056 * mapped via the block device. Since these pages could have journal heads
1057 * which would prevent try_to_free_buffers() from freeing them, we must use
1058 * jbd2 layer's try_to_free_buffers() function to release them.
1060 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1063 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1065 WARN_ON(PageChecked(page
));
1066 if (!page_has_buffers(page
))
1069 return jbd2_journal_try_to_free_buffers(journal
, page
,
1070 wait
& ~__GFP_WAIT
);
1071 return try_to_free_buffers(page
);
1075 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1076 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1078 static int ext4_write_dquot(struct dquot
*dquot
);
1079 static int ext4_acquire_dquot(struct dquot
*dquot
);
1080 static int ext4_release_dquot(struct dquot
*dquot
);
1081 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1082 static int ext4_write_info(struct super_block
*sb
, int type
);
1083 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1085 static int ext4_quota_off(struct super_block
*sb
, int type
);
1086 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1087 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1088 size_t len
, loff_t off
);
1089 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1090 const char *data
, size_t len
, loff_t off
);
1092 static const struct dquot_operations ext4_quota_operations
= {
1093 .get_reserved_space
= ext4_get_reserved_space
,
1094 .write_dquot
= ext4_write_dquot
,
1095 .acquire_dquot
= ext4_acquire_dquot
,
1096 .release_dquot
= ext4_release_dquot
,
1097 .mark_dirty
= ext4_mark_dquot_dirty
,
1098 .write_info
= ext4_write_info
,
1099 .alloc_dquot
= dquot_alloc
,
1100 .destroy_dquot
= dquot_destroy
,
1103 static const struct quotactl_ops ext4_qctl_operations
= {
1104 .quota_on
= ext4_quota_on
,
1105 .quota_off
= ext4_quota_off
,
1106 .quota_sync
= dquot_quota_sync
,
1107 .get_info
= dquot_get_dqinfo
,
1108 .set_info
= dquot_set_dqinfo
,
1109 .get_dqblk
= dquot_get_dqblk
,
1110 .set_dqblk
= dquot_set_dqblk
1114 static const struct super_operations ext4_sops
= {
1115 .alloc_inode
= ext4_alloc_inode
,
1116 .destroy_inode
= ext4_destroy_inode
,
1117 .write_inode
= ext4_write_inode
,
1118 .dirty_inode
= ext4_dirty_inode
,
1119 .drop_inode
= ext4_drop_inode
,
1120 .evict_inode
= ext4_evict_inode
,
1121 .put_super
= ext4_put_super
,
1122 .sync_fs
= ext4_sync_fs
,
1123 .freeze_fs
= ext4_freeze
,
1124 .unfreeze_fs
= ext4_unfreeze
,
1125 .statfs
= ext4_statfs
,
1126 .remount_fs
= ext4_remount
,
1127 .show_options
= ext4_show_options
,
1129 .quota_read
= ext4_quota_read
,
1130 .quota_write
= ext4_quota_write
,
1132 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1135 static const struct super_operations ext4_nojournal_sops
= {
1136 .alloc_inode
= ext4_alloc_inode
,
1137 .destroy_inode
= ext4_destroy_inode
,
1138 .write_inode
= ext4_write_inode
,
1139 .dirty_inode
= ext4_dirty_inode
,
1140 .drop_inode
= ext4_drop_inode
,
1141 .evict_inode
= ext4_evict_inode
,
1142 .write_super
= ext4_write_super
,
1143 .put_super
= ext4_put_super
,
1144 .statfs
= ext4_statfs
,
1145 .remount_fs
= ext4_remount
,
1146 .show_options
= ext4_show_options
,
1148 .quota_read
= ext4_quota_read
,
1149 .quota_write
= ext4_quota_write
,
1151 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1154 static const struct export_operations ext4_export_ops
= {
1155 .fh_to_dentry
= ext4_fh_to_dentry
,
1156 .fh_to_parent
= ext4_fh_to_parent
,
1157 .get_parent
= ext4_get_parent
,
1161 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1162 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1163 Opt_nouid32
, Opt_debug
, Opt_removed
,
1164 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1165 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
,
1166 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1167 Opt_journal_dev
, Opt_journal_checksum
, Opt_journal_async_commit
,
1168 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1169 Opt_data_err_abort
, Opt_data_err_ignore
,
1170 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1171 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1172 Opt_noquota
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1173 Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1174 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
, Opt_mblk_io_submit
,
1175 Opt_nomblk_io_submit
, Opt_block_validity
, Opt_noblock_validity
,
1176 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1177 Opt_dioread_nolock
, Opt_dioread_lock
,
1178 Opt_discard
, Opt_nodiscard
, Opt_init_itable
, Opt_noinit_itable
,
1181 static const match_table_t tokens
= {
1182 {Opt_bsd_df
, "bsddf"},
1183 {Opt_minix_df
, "minixdf"},
1184 {Opt_grpid
, "grpid"},
1185 {Opt_grpid
, "bsdgroups"},
1186 {Opt_nogrpid
, "nogrpid"},
1187 {Opt_nogrpid
, "sysvgroups"},
1188 {Opt_resgid
, "resgid=%u"},
1189 {Opt_resuid
, "resuid=%u"},
1191 {Opt_err_cont
, "errors=continue"},
1192 {Opt_err_panic
, "errors=panic"},
1193 {Opt_err_ro
, "errors=remount-ro"},
1194 {Opt_nouid32
, "nouid32"},
1195 {Opt_debug
, "debug"},
1196 {Opt_removed
, "oldalloc"},
1197 {Opt_removed
, "orlov"},
1198 {Opt_user_xattr
, "user_xattr"},
1199 {Opt_nouser_xattr
, "nouser_xattr"},
1201 {Opt_noacl
, "noacl"},
1202 {Opt_noload
, "norecovery"},
1203 {Opt_noload
, "noload"},
1204 {Opt_removed
, "nobh"},
1205 {Opt_removed
, "bh"},
1206 {Opt_commit
, "commit=%u"},
1207 {Opt_min_batch_time
, "min_batch_time=%u"},
1208 {Opt_max_batch_time
, "max_batch_time=%u"},
1209 {Opt_journal_dev
, "journal_dev=%u"},
1210 {Opt_journal_checksum
, "journal_checksum"},
1211 {Opt_journal_async_commit
, "journal_async_commit"},
1212 {Opt_abort
, "abort"},
1213 {Opt_data_journal
, "data=journal"},
1214 {Opt_data_ordered
, "data=ordered"},
1215 {Opt_data_writeback
, "data=writeback"},
1216 {Opt_data_err_abort
, "data_err=abort"},
1217 {Opt_data_err_ignore
, "data_err=ignore"},
1218 {Opt_offusrjquota
, "usrjquota="},
1219 {Opt_usrjquota
, "usrjquota=%s"},
1220 {Opt_offgrpjquota
, "grpjquota="},
1221 {Opt_grpjquota
, "grpjquota=%s"},
1222 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1223 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1224 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1225 {Opt_grpquota
, "grpquota"},
1226 {Opt_noquota
, "noquota"},
1227 {Opt_quota
, "quota"},
1228 {Opt_usrquota
, "usrquota"},
1229 {Opt_barrier
, "barrier=%u"},
1230 {Opt_barrier
, "barrier"},
1231 {Opt_nobarrier
, "nobarrier"},
1232 {Opt_i_version
, "i_version"},
1233 {Opt_stripe
, "stripe=%u"},
1234 {Opt_delalloc
, "delalloc"},
1235 {Opt_nodelalloc
, "nodelalloc"},
1236 {Opt_mblk_io_submit
, "mblk_io_submit"},
1237 {Opt_nomblk_io_submit
, "nomblk_io_submit"},
1238 {Opt_block_validity
, "block_validity"},
1239 {Opt_noblock_validity
, "noblock_validity"},
1240 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1241 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1242 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1243 {Opt_auto_da_alloc
, "auto_da_alloc"},
1244 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1245 {Opt_dioread_nolock
, "dioread_nolock"},
1246 {Opt_dioread_lock
, "dioread_lock"},
1247 {Opt_discard
, "discard"},
1248 {Opt_nodiscard
, "nodiscard"},
1249 {Opt_init_itable
, "init_itable=%u"},
1250 {Opt_init_itable
, "init_itable"},
1251 {Opt_noinit_itable
, "noinit_itable"},
1252 {Opt_removed
, "check=none"}, /* mount option from ext2/3 */
1253 {Opt_removed
, "nocheck"}, /* mount option from ext2/3 */
1254 {Opt_removed
, "reservation"}, /* mount option from ext2/3 */
1255 {Opt_removed
, "noreservation"}, /* mount option from ext2/3 */
1256 {Opt_removed
, "journal=%u"}, /* mount option from ext2/3 */
1260 static ext4_fsblk_t
get_sb_block(void **data
)
1262 ext4_fsblk_t sb_block
;
1263 char *options
= (char *) *data
;
1265 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1266 return 1; /* Default location */
1269 /* TODO: use simple_strtoll with >32bit ext4 */
1270 sb_block
= simple_strtoul(options
, &options
, 0);
1271 if (*options
&& *options
!= ',') {
1272 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1276 if (*options
== ',')
1278 *data
= (void *) options
;
1283 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1284 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1285 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1288 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1290 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1293 if (sb_any_quota_loaded(sb
) &&
1294 !sbi
->s_qf_names
[qtype
]) {
1295 ext4_msg(sb
, KERN_ERR
,
1296 "Cannot change journaled "
1297 "quota options when quota turned on");
1300 qname
= match_strdup(args
);
1302 ext4_msg(sb
, KERN_ERR
,
1303 "Not enough memory for storing quotafile name");
1306 if (sbi
->s_qf_names
[qtype
] &&
1307 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1308 ext4_msg(sb
, KERN_ERR
,
1309 "%s quota file already specified", QTYPE2NAME(qtype
));
1313 sbi
->s_qf_names
[qtype
] = qname
;
1314 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1315 ext4_msg(sb
, KERN_ERR
,
1316 "quotafile must be on filesystem root");
1317 kfree(sbi
->s_qf_names
[qtype
]);
1318 sbi
->s_qf_names
[qtype
] = NULL
;
1325 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1328 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1330 if (sb_any_quota_loaded(sb
) &&
1331 sbi
->s_qf_names
[qtype
]) {
1332 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1333 " when quota turned on");
1337 * The space will be released later when all options are confirmed
1340 sbi
->s_qf_names
[qtype
] = NULL
;
1345 #define MOPT_SET 0x0001
1346 #define MOPT_CLEAR 0x0002
1347 #define MOPT_NOSUPPORT 0x0004
1348 #define MOPT_EXPLICIT 0x0008
1349 #define MOPT_CLEAR_ERR 0x0010
1350 #define MOPT_GTE0 0x0020
1353 #define MOPT_QFMT 0x0040
1355 #define MOPT_Q MOPT_NOSUPPORT
1356 #define MOPT_QFMT MOPT_NOSUPPORT
1358 #define MOPT_DATAJ 0x0080
1360 static const struct mount_opts
{
1364 } ext4_mount_opts
[] = {
1365 {Opt_minix_df
, EXT4_MOUNT_MINIX_DF
, MOPT_SET
},
1366 {Opt_bsd_df
, EXT4_MOUNT_MINIX_DF
, MOPT_CLEAR
},
1367 {Opt_grpid
, EXT4_MOUNT_GRPID
, MOPT_SET
},
1368 {Opt_nogrpid
, EXT4_MOUNT_GRPID
, MOPT_CLEAR
},
1369 {Opt_mblk_io_submit
, EXT4_MOUNT_MBLK_IO_SUBMIT
, MOPT_SET
},
1370 {Opt_nomblk_io_submit
, EXT4_MOUNT_MBLK_IO_SUBMIT
, MOPT_CLEAR
},
1371 {Opt_block_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_SET
},
1372 {Opt_noblock_validity
, EXT4_MOUNT_BLOCK_VALIDITY
, MOPT_CLEAR
},
1373 {Opt_dioread_nolock
, EXT4_MOUNT_DIOREAD_NOLOCK
, MOPT_SET
},
1374 {Opt_dioread_lock
, EXT4_MOUNT_DIOREAD_NOLOCK
, MOPT_CLEAR
},
1375 {Opt_discard
, EXT4_MOUNT_DISCARD
, MOPT_SET
},
1376 {Opt_nodiscard
, EXT4_MOUNT_DISCARD
, MOPT_CLEAR
},
1377 {Opt_delalloc
, EXT4_MOUNT_DELALLOC
, MOPT_SET
| MOPT_EXPLICIT
},
1378 {Opt_nodelalloc
, EXT4_MOUNT_DELALLOC
, MOPT_CLEAR
| MOPT_EXPLICIT
},
1379 {Opt_journal_checksum
, EXT4_MOUNT_JOURNAL_CHECKSUM
, MOPT_SET
},
1380 {Opt_journal_async_commit
, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT
|
1381 EXT4_MOUNT_JOURNAL_CHECKSUM
), MOPT_SET
},
1382 {Opt_noload
, EXT4_MOUNT_NOLOAD
, MOPT_SET
},
1383 {Opt_err_panic
, EXT4_MOUNT_ERRORS_PANIC
, MOPT_SET
| MOPT_CLEAR_ERR
},
1384 {Opt_err_ro
, EXT4_MOUNT_ERRORS_RO
, MOPT_SET
| MOPT_CLEAR_ERR
},
1385 {Opt_err_cont
, EXT4_MOUNT_ERRORS_CONT
, MOPT_SET
| MOPT_CLEAR_ERR
},
1386 {Opt_data_err_abort
, EXT4_MOUNT_DATA_ERR_ABORT
, MOPT_SET
},
1387 {Opt_data_err_ignore
, EXT4_MOUNT_DATA_ERR_ABORT
, MOPT_CLEAR
},
1388 {Opt_barrier
, EXT4_MOUNT_BARRIER
, MOPT_SET
},
1389 {Opt_nobarrier
, EXT4_MOUNT_BARRIER
, MOPT_CLEAR
},
1390 {Opt_noauto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_SET
},
1391 {Opt_auto_da_alloc
, EXT4_MOUNT_NO_AUTO_DA_ALLOC
, MOPT_CLEAR
},
1392 {Opt_noinit_itable
, EXT4_MOUNT_INIT_INODE_TABLE
, MOPT_CLEAR
},
1393 {Opt_commit
, 0, MOPT_GTE0
},
1394 {Opt_max_batch_time
, 0, MOPT_GTE0
},
1395 {Opt_min_batch_time
, 0, MOPT_GTE0
},
1396 {Opt_inode_readahead_blks
, 0, MOPT_GTE0
},
1397 {Opt_init_itable
, 0, MOPT_GTE0
},
1398 {Opt_stripe
, 0, MOPT_GTE0
},
1399 {Opt_data_journal
, EXT4_MOUNT_JOURNAL_DATA
, MOPT_DATAJ
},
1400 {Opt_data_ordered
, EXT4_MOUNT_ORDERED_DATA
, MOPT_DATAJ
},
1401 {Opt_data_writeback
, EXT4_MOUNT_WRITEBACK_DATA
, MOPT_DATAJ
},
1402 #ifdef CONFIG_EXT4_FS_XATTR
1403 {Opt_user_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_SET
},
1404 {Opt_nouser_xattr
, EXT4_MOUNT_XATTR_USER
, MOPT_CLEAR
},
1406 {Opt_user_xattr
, 0, MOPT_NOSUPPORT
},
1407 {Opt_nouser_xattr
, 0, MOPT_NOSUPPORT
},
1409 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1410 {Opt_acl
, EXT4_MOUNT_POSIX_ACL
, MOPT_SET
},
1411 {Opt_noacl
, EXT4_MOUNT_POSIX_ACL
, MOPT_CLEAR
},
1413 {Opt_acl
, 0, MOPT_NOSUPPORT
},
1414 {Opt_noacl
, 0, MOPT_NOSUPPORT
},
1416 {Opt_nouid32
, EXT4_MOUNT_NO_UID32
, MOPT_SET
},
1417 {Opt_debug
, EXT4_MOUNT_DEBUG
, MOPT_SET
},
1418 {Opt_quota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
, MOPT_SET
| MOPT_Q
},
1419 {Opt_usrquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
,
1421 {Opt_grpquota
, EXT4_MOUNT_QUOTA
| EXT4_MOUNT_GRPQUOTA
,
1423 {Opt_noquota
, (EXT4_MOUNT_QUOTA
| EXT4_MOUNT_USRQUOTA
|
1424 EXT4_MOUNT_GRPQUOTA
), MOPT_CLEAR
| MOPT_Q
},
1425 {Opt_usrjquota
, 0, MOPT_Q
},
1426 {Opt_grpjquota
, 0, MOPT_Q
},
1427 {Opt_offusrjquota
, 0, MOPT_Q
},
1428 {Opt_offgrpjquota
, 0, MOPT_Q
},
1429 {Opt_jqfmt_vfsold
, QFMT_VFS_OLD
, MOPT_QFMT
},
1430 {Opt_jqfmt_vfsv0
, QFMT_VFS_V0
, MOPT_QFMT
},
1431 {Opt_jqfmt_vfsv1
, QFMT_VFS_V1
, MOPT_QFMT
},
1435 static int handle_mount_opt(struct super_block
*sb
, char *opt
, int token
,
1436 substring_t
*args
, unsigned long *journal_devnum
,
1437 unsigned int *journal_ioprio
, int is_remount
)
1439 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1440 const struct mount_opts
*m
;
1443 if (args
->from
&& match_int(args
, &arg
))
1447 return 1; /* handled by get_sb_block() */
1449 ext4_msg(sb
, KERN_WARNING
,
1450 "Ignoring removed %s option", opt
);
1453 sbi
->s_resuid
= arg
;
1456 sbi
->s_resgid
= arg
;
1459 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1462 sb
->s_flags
|= MS_I_VERSION
;
1464 case Opt_journal_dev
:
1466 ext4_msg(sb
, KERN_ERR
,
1467 "Cannot specify journal on remount");
1470 *journal_devnum
= arg
;
1472 case Opt_journal_ioprio
:
1473 if (arg
< 0 || arg
> 7)
1475 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
, arg
);
1479 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1480 if (token
!= m
->token
)
1482 if (args
->from
&& (m
->flags
& MOPT_GTE0
) && (arg
< 0))
1484 if (m
->flags
& MOPT_EXPLICIT
)
1485 set_opt2(sb
, EXPLICIT_DELALLOC
);
1486 if (m
->flags
& MOPT_CLEAR_ERR
)
1487 clear_opt(sb
, ERRORS_MASK
);
1488 if (token
== Opt_noquota
&& sb_any_quota_loaded(sb
)) {
1489 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1490 "options when quota turned on");
1494 if (m
->flags
& MOPT_NOSUPPORT
) {
1495 ext4_msg(sb
, KERN_ERR
, "%s option not supported", opt
);
1496 } else if (token
== Opt_commit
) {
1498 arg
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1499 sbi
->s_commit_interval
= HZ
* arg
;
1500 } else if (token
== Opt_max_batch_time
) {
1502 arg
= EXT4_DEF_MAX_BATCH_TIME
;
1503 sbi
->s_max_batch_time
= arg
;
1504 } else if (token
== Opt_min_batch_time
) {
1505 sbi
->s_min_batch_time
= arg
;
1506 } else if (token
== Opt_inode_readahead_blks
) {
1507 if (arg
> (1 << 30))
1509 if (arg
&& !is_power_of_2(arg
)) {
1510 ext4_msg(sb
, KERN_ERR
,
1511 "EXT4-fs: inode_readahead_blks"
1512 " must be a power of 2");
1515 sbi
->s_inode_readahead_blks
= arg
;
1516 } else if (token
== Opt_init_itable
) {
1517 set_opt(sb
, INIT_INODE_TABLE
);
1519 arg
= EXT4_DEF_LI_WAIT_MULT
;
1520 sbi
->s_li_wait_mult
= arg
;
1521 } else if (token
== Opt_stripe
) {
1522 sbi
->s_stripe
= arg
;
1523 } else if (m
->flags
& MOPT_DATAJ
) {
1525 if (!sbi
->s_journal
)
1526 ext4_msg(sb
, KERN_WARNING
, "Remounting file system with no journal so ignoring journalled data option");
1527 else if (test_opt(sb
, DATA_FLAGS
) !=
1529 ext4_msg(sb
, KERN_ERR
,
1530 "Cannot change data mode on remount");
1534 clear_opt(sb
, DATA_FLAGS
);
1535 sbi
->s_mount_opt
|= m
->mount_opt
;
1538 } else if (token
== Opt_usrjquota
) {
1539 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1541 } else if (token
== Opt_grpjquota
) {
1542 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1544 } else if (token
== Opt_offusrjquota
) {
1545 if (!clear_qf_name(sb
, USRQUOTA
))
1547 } else if (token
== Opt_offgrpjquota
) {
1548 if (!clear_qf_name(sb
, GRPQUOTA
))
1550 } else if (m
->flags
& MOPT_QFMT
) {
1551 if (sb_any_quota_loaded(sb
) &&
1552 sbi
->s_jquota_fmt
!= m
->mount_opt
) {
1553 ext4_msg(sb
, KERN_ERR
, "Cannot "
1554 "change journaled quota options "
1555 "when quota turned on");
1558 sbi
->s_jquota_fmt
= m
->mount_opt
;
1563 if (m
->flags
& MOPT_CLEAR
)
1565 else if (unlikely(!(m
->flags
& MOPT_SET
))) {
1566 ext4_msg(sb
, KERN_WARNING
,
1567 "buggy handling of option %s", opt
);
1572 sbi
->s_mount_opt
|= m
->mount_opt
;
1574 sbi
->s_mount_opt
&= ~m
->mount_opt
;
1578 ext4_msg(sb
, KERN_ERR
, "Unrecognized mount option \"%s\" "
1579 "or missing value", opt
);
1583 static int parse_options(char *options
, struct super_block
*sb
,
1584 unsigned long *journal_devnum
,
1585 unsigned int *journal_ioprio
,
1588 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1590 substring_t args
[MAX_OPT_ARGS
];
1596 while ((p
= strsep(&options
, ",")) != NULL
) {
1600 * Initialize args struct so we know whether arg was
1601 * found; some options take optional arguments.
1603 args
[0].to
= args
[0].from
= 0;
1604 token
= match_token(p
, tokens
, args
);
1605 if (handle_mount_opt(sb
, p
, token
, args
, journal_devnum
,
1606 journal_ioprio
, is_remount
) < 0)
1610 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1611 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1612 clear_opt(sb
, USRQUOTA
);
1614 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1615 clear_opt(sb
, GRPQUOTA
);
1617 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1618 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1623 if (!sbi
->s_jquota_fmt
) {
1624 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1629 if (sbi
->s_jquota_fmt
) {
1630 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1631 "specified with no journaling "
1640 static inline void ext4_show_quota_options(struct seq_file
*seq
,
1641 struct super_block
*sb
)
1643 #if defined(CONFIG_QUOTA)
1644 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1646 if (sbi
->s_jquota_fmt
) {
1649 switch (sbi
->s_jquota_fmt
) {
1660 seq_printf(seq
, ",jqfmt=%s", fmtname
);
1663 if (sbi
->s_qf_names
[USRQUOTA
])
1664 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
1666 if (sbi
->s_qf_names
[GRPQUOTA
])
1667 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
1669 if (test_opt(sb
, USRQUOTA
))
1670 seq_puts(seq
, ",usrquota");
1672 if (test_opt(sb
, GRPQUOTA
))
1673 seq_puts(seq
, ",grpquota");
1677 static const char *token2str(int token
)
1679 static const struct match_token
*t
;
1681 for (t
= tokens
; t
->token
!= Opt_err
; t
++)
1682 if (t
->token
== token
&& !strchr(t
->pattern
, '='))
1689 * - it's set to a non-default value OR
1690 * - if the per-sb default is different from the global default
1692 static int _ext4_show_options(struct seq_file
*seq
, struct super_block
*sb
,
1695 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1696 struct ext4_super_block
*es
= sbi
->s_es
;
1697 int def_errors
, def_mount_opt
= nodefs
? 0 : sbi
->s_def_mount_opt
;
1698 const struct mount_opts
*m
;
1699 char sep
= nodefs
? '\n' : ',';
1701 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1702 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1704 if (sbi
->s_sb_block
!= 1)
1705 SEQ_OPTS_PRINT("sb=%llu", sbi
->s_sb_block
);
1707 for (m
= ext4_mount_opts
; m
->token
!= Opt_err
; m
++) {
1708 int want_set
= m
->flags
& MOPT_SET
;
1709 if (((m
->flags
& (MOPT_SET
|MOPT_CLEAR
)) == 0) ||
1710 (m
->flags
& MOPT_CLEAR_ERR
))
1712 if (!(m
->mount_opt
& (sbi
->s_mount_opt
^ def_mount_opt
)))
1713 continue; /* skip if same as the default */
1715 (sbi
->s_mount_opt
& m
->mount_opt
) != m
->mount_opt
) ||
1716 (!want_set
&& (sbi
->s_mount_opt
& m
->mount_opt
)))
1717 continue; /* select Opt_noFoo vs Opt_Foo */
1718 SEQ_OPTS_PRINT("%s", token2str(m
->token
));
1721 if (nodefs
|| sbi
->s_resuid
!= EXT4_DEF_RESUID
||
1722 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
)
1723 SEQ_OPTS_PRINT("resuid=%u", sbi
->s_resuid
);
1724 if (nodefs
|| sbi
->s_resgid
!= EXT4_DEF_RESGID
||
1725 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
)
1726 SEQ_OPTS_PRINT("resgid=%u", sbi
->s_resgid
);
1727 def_errors
= nodefs
? -1 : le16_to_cpu(es
->s_errors
);
1728 if (test_opt(sb
, ERRORS_RO
) && def_errors
!= EXT4_ERRORS_RO
)
1729 SEQ_OPTS_PUTS("errors=remount-ro");
1730 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
1731 SEQ_OPTS_PUTS("errors=continue");
1732 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
1733 SEQ_OPTS_PUTS("errors=panic");
1734 if (nodefs
|| sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
)
1735 SEQ_OPTS_PRINT("commit=%lu", sbi
->s_commit_interval
/ HZ
);
1736 if (nodefs
|| sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
)
1737 SEQ_OPTS_PRINT("min_batch_time=%u", sbi
->s_min_batch_time
);
1738 if (nodefs
|| sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
)
1739 SEQ_OPTS_PRINT("max_batch_time=%u", sbi
->s_max_batch_time
);
1740 if (sb
->s_flags
& MS_I_VERSION
)
1741 SEQ_OPTS_PUTS("i_version");
1742 if (nodefs
|| sbi
->s_stripe
)
1743 SEQ_OPTS_PRINT("stripe=%lu", sbi
->s_stripe
);
1744 if (EXT4_MOUNT_DATA_FLAGS
& (sbi
->s_mount_opt
^ def_mount_opt
)) {
1745 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1746 SEQ_OPTS_PUTS("data=journal");
1747 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1748 SEQ_OPTS_PUTS("data=ordered");
1749 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1750 SEQ_OPTS_PUTS("data=writeback");
1753 sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1754 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1755 sbi
->s_inode_readahead_blks
);
1757 if (nodefs
|| (test_opt(sb
, INIT_INODE_TABLE
) &&
1758 (sbi
->s_li_wait_mult
!= EXT4_DEF_LI_WAIT_MULT
)))
1759 SEQ_OPTS_PRINT("init_itable=%u", sbi
->s_li_wait_mult
);
1761 ext4_show_quota_options(seq
, sb
);
1765 static int ext4_show_options(struct seq_file
*seq
, struct dentry
*root
)
1767 return _ext4_show_options(seq
, root
->d_sb
, 0);
1770 static int options_seq_show(struct seq_file
*seq
, void *offset
)
1772 struct super_block
*sb
= seq
->private;
1775 seq_puts(seq
, (sb
->s_flags
& MS_RDONLY
) ? "ro" : "rw");
1776 rc
= _ext4_show_options(seq
, sb
, 1);
1777 seq_puts(seq
, "\n");
1781 static int options_open_fs(struct inode
*inode
, struct file
*file
)
1783 return single_open(file
, options_seq_show
, PDE(inode
)->data
);
1786 static const struct file_operations ext4_seq_options_fops
= {
1787 .owner
= THIS_MODULE
,
1788 .open
= options_open_fs
,
1790 .llseek
= seq_lseek
,
1791 .release
= single_release
,
1794 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1797 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1800 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1801 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1802 "forcing read-only mode");
1807 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1808 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1809 "running e2fsck is recommended");
1810 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1811 ext4_msg(sb
, KERN_WARNING
,
1812 "warning: mounting fs with errors, "
1813 "running e2fsck is recommended");
1814 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) > 0 &&
1815 le16_to_cpu(es
->s_mnt_count
) >=
1816 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1817 ext4_msg(sb
, KERN_WARNING
,
1818 "warning: maximal mount count reached, "
1819 "running e2fsck is recommended");
1820 else if (le32_to_cpu(es
->s_checkinterval
) &&
1821 (le32_to_cpu(es
->s_lastcheck
) +
1822 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1823 ext4_msg(sb
, KERN_WARNING
,
1824 "warning: checktime reached, "
1825 "running e2fsck is recommended");
1826 if (!sbi
->s_journal
)
1827 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1828 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1829 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1830 le16_add_cpu(&es
->s_mnt_count
, 1);
1831 es
->s_mtime
= cpu_to_le32(get_seconds());
1832 ext4_update_dynamic_rev(sb
);
1834 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1836 ext4_commit_super(sb
, 1);
1838 if (test_opt(sb
, DEBUG
))
1839 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1840 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1842 sbi
->s_groups_count
,
1843 EXT4_BLOCKS_PER_GROUP(sb
),
1844 EXT4_INODES_PER_GROUP(sb
),
1845 sbi
->s_mount_opt
, sbi
->s_mount_opt2
);
1847 cleancache_init_fs(sb
);
1851 static int ext4_fill_flex_info(struct super_block
*sb
)
1853 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1854 struct ext4_group_desc
*gdp
= NULL
;
1855 ext4_group_t flex_group_count
;
1856 ext4_group_t flex_group
;
1857 unsigned int groups_per_flex
= 0;
1861 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1862 if (sbi
->s_log_groups_per_flex
< 1 || sbi
->s_log_groups_per_flex
> 31) {
1863 sbi
->s_log_groups_per_flex
= 0;
1866 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1868 /* We allocate both existing and potentially added groups */
1869 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1870 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1871 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1872 size
= flex_group_count
* sizeof(struct flex_groups
);
1873 sbi
->s_flex_groups
= ext4_kvzalloc(size
, GFP_KERNEL
);
1874 if (sbi
->s_flex_groups
== NULL
) {
1875 ext4_msg(sb
, KERN_ERR
, "not enough memory for %u flex groups",
1880 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1881 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1883 flex_group
= ext4_flex_group(sbi
, i
);
1884 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1885 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1886 atomic_add(ext4_free_group_clusters(sb
, gdp
),
1887 &sbi
->s_flex_groups
[flex_group
].free_clusters
);
1888 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1889 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1897 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1898 struct ext4_group_desc
*gdp
)
1902 if (sbi
->s_es
->s_feature_ro_compat
&
1903 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1904 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1905 __le32 le_group
= cpu_to_le32(block_group
);
1907 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1908 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1909 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1910 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1911 /* for checksum of struct ext4_group_desc do the rest...*/
1912 if ((sbi
->s_es
->s_feature_incompat
&
1913 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1914 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1915 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1916 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1920 return cpu_to_le16(crc
);
1923 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1924 struct ext4_group_desc
*gdp
)
1926 if ((sbi
->s_es
->s_feature_ro_compat
&
1927 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1928 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1934 /* Called at mount-time, super-block is locked */
1935 static int ext4_check_descriptors(struct super_block
*sb
,
1936 ext4_group_t
*first_not_zeroed
)
1938 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1939 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1940 ext4_fsblk_t last_block
;
1941 ext4_fsblk_t block_bitmap
;
1942 ext4_fsblk_t inode_bitmap
;
1943 ext4_fsblk_t inode_table
;
1944 int flexbg_flag
= 0;
1945 ext4_group_t i
, grp
= sbi
->s_groups_count
;
1947 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1950 ext4_debug("Checking group descriptors");
1952 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1953 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1955 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1956 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1958 last_block
= first_block
+
1959 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1961 if ((grp
== sbi
->s_groups_count
) &&
1962 !(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
1965 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1966 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1967 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1968 "Block bitmap for group %u not in group "
1969 "(block %llu)!", i
, block_bitmap
);
1972 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1973 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1974 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1975 "Inode bitmap for group %u not in group "
1976 "(block %llu)!", i
, inode_bitmap
);
1979 inode_table
= ext4_inode_table(sb
, gdp
);
1980 if (inode_table
< first_block
||
1981 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1982 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1983 "Inode table for group %u not in group "
1984 "(block %llu)!", i
, inode_table
);
1987 ext4_lock_group(sb
, i
);
1988 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1989 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1990 "Checksum for group %u failed (%u!=%u)",
1991 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1992 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1993 if (!(sb
->s_flags
& MS_RDONLY
)) {
1994 ext4_unlock_group(sb
, i
);
1998 ext4_unlock_group(sb
, i
);
2000 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2002 if (NULL
!= first_not_zeroed
)
2003 *first_not_zeroed
= grp
;
2005 ext4_free_blocks_count_set(sbi
->s_es
,
2006 EXT4_C2B(sbi
, ext4_count_free_clusters(sb
)));
2007 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2011 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2012 * the superblock) which were deleted from all directories, but held open by
2013 * a process at the time of a crash. We walk the list and try to delete these
2014 * inodes at recovery time (only with a read-write filesystem).
2016 * In order to keep the orphan inode chain consistent during traversal (in
2017 * case of crash during recovery), we link each inode into the superblock
2018 * orphan list_head and handle it the same way as an inode deletion during
2019 * normal operation (which journals the operations for us).
2021 * We only do an iget() and an iput() on each inode, which is very safe if we
2022 * accidentally point at an in-use or already deleted inode. The worst that
2023 * can happen in this case is that we get a "bit already cleared" message from
2024 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2025 * e2fsck was run on this filesystem, and it must have already done the orphan
2026 * inode cleanup for us, so we can safely abort without any further action.
2028 static void ext4_orphan_cleanup(struct super_block
*sb
,
2029 struct ext4_super_block
*es
)
2031 unsigned int s_flags
= sb
->s_flags
;
2032 int nr_orphans
= 0, nr_truncates
= 0;
2036 if (!es
->s_last_orphan
) {
2037 jbd_debug(4, "no orphan inodes to clean up\n");
2041 if (bdev_read_only(sb
->s_bdev
)) {
2042 ext4_msg(sb
, KERN_ERR
, "write access "
2043 "unavailable, skipping orphan cleanup");
2047 /* Check if feature set would not allow a r/w mount */
2048 if (!ext4_feature_set_ok(sb
, 0)) {
2049 ext4_msg(sb
, KERN_INFO
, "Skipping orphan cleanup due to "
2050 "unknown ROCOMPAT features");
2054 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2055 if (es
->s_last_orphan
)
2056 jbd_debug(1, "Errors on filesystem, "
2057 "clearing orphan list.\n");
2058 es
->s_last_orphan
= 0;
2059 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2063 if (s_flags
& MS_RDONLY
) {
2064 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2065 sb
->s_flags
&= ~MS_RDONLY
;
2068 /* Needed for iput() to work correctly and not trash data */
2069 sb
->s_flags
|= MS_ACTIVE
;
2070 /* Turn on quotas so that they are updated correctly */
2071 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2072 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2073 int ret
= ext4_quota_on_mount(sb
, i
);
2075 ext4_msg(sb
, KERN_ERR
,
2076 "Cannot turn on journaled "
2077 "quota: error %d", ret
);
2082 while (es
->s_last_orphan
) {
2083 struct inode
*inode
;
2085 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2086 if (IS_ERR(inode
)) {
2087 es
->s_last_orphan
= 0;
2091 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2092 dquot_initialize(inode
);
2093 if (inode
->i_nlink
) {
2094 ext4_msg(sb
, KERN_DEBUG
,
2095 "%s: truncating inode %lu to %lld bytes",
2096 __func__
, inode
->i_ino
, inode
->i_size
);
2097 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2098 inode
->i_ino
, inode
->i_size
);
2099 ext4_truncate(inode
);
2102 ext4_msg(sb
, KERN_DEBUG
,
2103 "%s: deleting unreferenced inode %lu",
2104 __func__
, inode
->i_ino
);
2105 jbd_debug(2, "deleting unreferenced inode %lu\n",
2109 iput(inode
); /* The delete magic happens here! */
2112 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2115 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2116 PLURAL(nr_orphans
));
2118 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2119 PLURAL(nr_truncates
));
2121 /* Turn quotas off */
2122 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2123 if (sb_dqopt(sb
)->files
[i
])
2124 dquot_quota_off(sb
, i
);
2127 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2131 * Maximal extent format file size.
2132 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2133 * extent format containers, within a sector_t, and within i_blocks
2134 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2135 * so that won't be a limiting factor.
2137 * However there is other limiting factor. We do store extents in the form
2138 * of starting block and length, hence the resulting length of the extent
2139 * covering maximum file size must fit into on-disk format containers as
2140 * well. Given that length is always by 1 unit bigger than max unit (because
2141 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2143 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2145 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2148 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2150 /* small i_blocks in vfs inode? */
2151 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2153 * CONFIG_LBDAF is not enabled implies the inode
2154 * i_block represent total blocks in 512 bytes
2155 * 32 == size of vfs inode i_blocks * 8
2157 upper_limit
= (1LL << 32) - 1;
2159 /* total blocks in file system block size */
2160 upper_limit
>>= (blkbits
- 9);
2161 upper_limit
<<= blkbits
;
2165 * 32-bit extent-start container, ee_block. We lower the maxbytes
2166 * by one fs block, so ee_len can cover the extent of maximum file
2169 res
= (1LL << 32) - 1;
2172 /* Sanity check against vm- & vfs- imposed limits */
2173 if (res
> upper_limit
)
2180 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2181 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2182 * We need to be 1 filesystem block less than the 2^48 sector limit.
2184 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2186 loff_t res
= EXT4_NDIR_BLOCKS
;
2189 /* This is calculated to be the largest file size for a dense, block
2190 * mapped file such that the file's total number of 512-byte sectors,
2191 * including data and all indirect blocks, does not exceed (2^48 - 1).
2193 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2194 * number of 512-byte sectors of the file.
2197 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2199 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2200 * the inode i_block field represents total file blocks in
2201 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2203 upper_limit
= (1LL << 32) - 1;
2205 /* total blocks in file system block size */
2206 upper_limit
>>= (bits
- 9);
2210 * We use 48 bit ext4_inode i_blocks
2211 * With EXT4_HUGE_FILE_FL set the i_blocks
2212 * represent total number of blocks in
2213 * file system block size
2215 upper_limit
= (1LL << 48) - 1;
2219 /* indirect blocks */
2221 /* double indirect blocks */
2222 meta_blocks
+= 1 + (1LL << (bits
-2));
2223 /* tripple indirect blocks */
2224 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2226 upper_limit
-= meta_blocks
;
2227 upper_limit
<<= bits
;
2229 res
+= 1LL << (bits
-2);
2230 res
+= 1LL << (2*(bits
-2));
2231 res
+= 1LL << (3*(bits
-2));
2233 if (res
> upper_limit
)
2236 if (res
> MAX_LFS_FILESIZE
)
2237 res
= MAX_LFS_FILESIZE
;
2242 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2243 ext4_fsblk_t logical_sb_block
, int nr
)
2245 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2246 ext4_group_t bg
, first_meta_bg
;
2249 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2251 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2253 return logical_sb_block
+ nr
+ 1;
2254 bg
= sbi
->s_desc_per_block
* nr
;
2255 if (ext4_bg_has_super(sb
, bg
))
2258 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2262 * ext4_get_stripe_size: Get the stripe size.
2263 * @sbi: In memory super block info
2265 * If we have specified it via mount option, then
2266 * use the mount option value. If the value specified at mount time is
2267 * greater than the blocks per group use the super block value.
2268 * If the super block value is greater than blocks per group return 0.
2269 * Allocator needs it be less than blocks per group.
2272 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2274 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2275 unsigned long stripe_width
=
2276 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2279 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2280 ret
= sbi
->s_stripe
;
2281 else if (stripe_width
<= sbi
->s_blocks_per_group
)
2283 else if (stride
<= sbi
->s_blocks_per_group
)
2289 * If the stripe width is 1, this makes no sense and
2290 * we set it to 0 to turn off stripe handling code.
2301 struct attribute attr
;
2302 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2303 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2304 const char *, size_t);
2308 static int parse_strtoul(const char *buf
,
2309 unsigned long max
, unsigned long *value
)
2313 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2314 endp
= skip_spaces(endp
);
2315 if (*endp
|| *value
> max
)
2321 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2322 struct ext4_sb_info
*sbi
,
2325 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2327 percpu_counter_sum(&sbi
->s_dirtyclusters_counter
)));
2330 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2331 struct ext4_sb_info
*sbi
, char *buf
)
2333 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2335 if (!sb
->s_bdev
->bd_part
)
2336 return snprintf(buf
, PAGE_SIZE
, "0\n");
2337 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2338 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2339 sbi
->s_sectors_written_start
) >> 1);
2342 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2343 struct ext4_sb_info
*sbi
, char *buf
)
2345 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2347 if (!sb
->s_bdev
->bd_part
)
2348 return snprintf(buf
, PAGE_SIZE
, "0\n");
2349 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2350 (unsigned long long)(sbi
->s_kbytes_written
+
2351 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2352 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2355 static ssize_t
extent_cache_hits_show(struct ext4_attr
*a
,
2356 struct ext4_sb_info
*sbi
, char *buf
)
2358 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_hits
);
2361 static ssize_t
extent_cache_misses_show(struct ext4_attr
*a
,
2362 struct ext4_sb_info
*sbi
, char *buf
)
2364 return snprintf(buf
, PAGE_SIZE
, "%lu\n", sbi
->extent_cache_misses
);
2367 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2368 struct ext4_sb_info
*sbi
,
2369 const char *buf
, size_t count
)
2373 if (parse_strtoul(buf
, 0x40000000, &t
))
2376 if (t
&& !is_power_of_2(t
))
2379 sbi
->s_inode_readahead_blks
= t
;
2383 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2384 struct ext4_sb_info
*sbi
, char *buf
)
2386 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2388 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2391 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2392 struct ext4_sb_info
*sbi
,
2393 const char *buf
, size_t count
)
2395 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2398 if (parse_strtoul(buf
, 0xffffffff, &t
))
2404 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2405 static struct ext4_attr ext4_attr_##_name = { \
2406 .attr = {.name = __stringify(_name), .mode = _mode }, \
2409 .offset = offsetof(struct ext4_sb_info, _elname), \
2411 #define EXT4_ATTR(name, mode, show, store) \
2412 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2414 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2415 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2416 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2417 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2418 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2419 #define ATTR_LIST(name) &ext4_attr_##name.attr
2421 EXT4_RO_ATTR(delayed_allocation_blocks
);
2422 EXT4_RO_ATTR(session_write_kbytes
);
2423 EXT4_RO_ATTR(lifetime_write_kbytes
);
2424 EXT4_RO_ATTR(extent_cache_hits
);
2425 EXT4_RO_ATTR(extent_cache_misses
);
2426 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2427 inode_readahead_blks_store
, s_inode_readahead_blks
);
2428 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2429 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2430 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2431 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2432 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2433 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2434 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2435 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2437 static struct attribute
*ext4_attrs
[] = {
2438 ATTR_LIST(delayed_allocation_blocks
),
2439 ATTR_LIST(session_write_kbytes
),
2440 ATTR_LIST(lifetime_write_kbytes
),
2441 ATTR_LIST(extent_cache_hits
),
2442 ATTR_LIST(extent_cache_misses
),
2443 ATTR_LIST(inode_readahead_blks
),
2444 ATTR_LIST(inode_goal
),
2445 ATTR_LIST(mb_stats
),
2446 ATTR_LIST(mb_max_to_scan
),
2447 ATTR_LIST(mb_min_to_scan
),
2448 ATTR_LIST(mb_order2_req
),
2449 ATTR_LIST(mb_stream_req
),
2450 ATTR_LIST(mb_group_prealloc
),
2451 ATTR_LIST(max_writeback_mb_bump
),
2455 /* Features this copy of ext4 supports */
2456 EXT4_INFO_ATTR(lazy_itable_init
);
2457 EXT4_INFO_ATTR(batched_discard
);
2459 static struct attribute
*ext4_feat_attrs
[] = {
2460 ATTR_LIST(lazy_itable_init
),
2461 ATTR_LIST(batched_discard
),
2465 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2466 struct attribute
*attr
, char *buf
)
2468 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2470 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2472 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2475 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2476 struct attribute
*attr
,
2477 const char *buf
, size_t len
)
2479 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2481 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2483 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2486 static void ext4_sb_release(struct kobject
*kobj
)
2488 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2490 complete(&sbi
->s_kobj_unregister
);
2493 static const struct sysfs_ops ext4_attr_ops
= {
2494 .show
= ext4_attr_show
,
2495 .store
= ext4_attr_store
,
2498 static struct kobj_type ext4_ktype
= {
2499 .default_attrs
= ext4_attrs
,
2500 .sysfs_ops
= &ext4_attr_ops
,
2501 .release
= ext4_sb_release
,
2504 static void ext4_feat_release(struct kobject
*kobj
)
2506 complete(&ext4_feat
->f_kobj_unregister
);
2509 static struct kobj_type ext4_feat_ktype
= {
2510 .default_attrs
= ext4_feat_attrs
,
2511 .sysfs_ops
= &ext4_attr_ops
,
2512 .release
= ext4_feat_release
,
2516 * Check whether this filesystem can be mounted based on
2517 * the features present and the RDONLY/RDWR mount requested.
2518 * Returns 1 if this filesystem can be mounted as requested,
2519 * 0 if it cannot be.
2521 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2523 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2524 ext4_msg(sb
, KERN_ERR
,
2525 "Couldn't mount because of "
2526 "unsupported optional features (%x)",
2527 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2528 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2535 /* Check that feature set is OK for a read-write mount */
2536 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2537 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2538 "unsupported optional features (%x)",
2539 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2540 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2544 * Large file size enabled file system can only be mounted
2545 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2548 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2549 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2550 "cannot be mounted RDWR without "
2555 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_BIGALLOC
) &&
2556 !EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_EXTENTS
)) {
2557 ext4_msg(sb
, KERN_ERR
,
2558 "Can't support bigalloc feature without "
2559 "extents feature\n");
2566 * This function is called once a day if we have errors logged
2567 * on the file system
2569 static void print_daily_error_info(unsigned long arg
)
2571 struct super_block
*sb
= (struct super_block
*) arg
;
2572 struct ext4_sb_info
*sbi
;
2573 struct ext4_super_block
*es
;
2578 if (es
->s_error_count
)
2579 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2580 le32_to_cpu(es
->s_error_count
));
2581 if (es
->s_first_error_time
) {
2582 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2583 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2584 (int) sizeof(es
->s_first_error_func
),
2585 es
->s_first_error_func
,
2586 le32_to_cpu(es
->s_first_error_line
));
2587 if (es
->s_first_error_ino
)
2588 printk(": inode %u",
2589 le32_to_cpu(es
->s_first_error_ino
));
2590 if (es
->s_first_error_block
)
2591 printk(": block %llu", (unsigned long long)
2592 le64_to_cpu(es
->s_first_error_block
));
2595 if (es
->s_last_error_time
) {
2596 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2597 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2598 (int) sizeof(es
->s_last_error_func
),
2599 es
->s_last_error_func
,
2600 le32_to_cpu(es
->s_last_error_line
));
2601 if (es
->s_last_error_ino
)
2602 printk(": inode %u",
2603 le32_to_cpu(es
->s_last_error_ino
));
2604 if (es
->s_last_error_block
)
2605 printk(": block %llu", (unsigned long long)
2606 le64_to_cpu(es
->s_last_error_block
));
2609 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2612 /* Find next suitable group and run ext4_init_inode_table */
2613 static int ext4_run_li_request(struct ext4_li_request
*elr
)
2615 struct ext4_group_desc
*gdp
= NULL
;
2616 ext4_group_t group
, ngroups
;
2617 struct super_block
*sb
;
2618 unsigned long timeout
= 0;
2622 ngroups
= EXT4_SB(sb
)->s_groups_count
;
2624 for (group
= elr
->lr_next_group
; group
< ngroups
; group
++) {
2625 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2631 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2635 if (group
== ngroups
)
2640 ret
= ext4_init_inode_table(sb
, group
,
2641 elr
->lr_timeout
? 0 : 1);
2642 if (elr
->lr_timeout
== 0) {
2643 timeout
= (jiffies
- timeout
) *
2644 elr
->lr_sbi
->s_li_wait_mult
;
2645 elr
->lr_timeout
= timeout
;
2647 elr
->lr_next_sched
= jiffies
+ elr
->lr_timeout
;
2648 elr
->lr_next_group
= group
+ 1;
2655 * Remove lr_request from the list_request and free the
2656 * request structure. Should be called with li_list_mtx held
2658 static void ext4_remove_li_request(struct ext4_li_request
*elr
)
2660 struct ext4_sb_info
*sbi
;
2667 list_del(&elr
->lr_request
);
2668 sbi
->s_li_request
= NULL
;
2672 static void ext4_unregister_li_request(struct super_block
*sb
)
2674 mutex_lock(&ext4_li_mtx
);
2675 if (!ext4_li_info
) {
2676 mutex_unlock(&ext4_li_mtx
);
2680 mutex_lock(&ext4_li_info
->li_list_mtx
);
2681 ext4_remove_li_request(EXT4_SB(sb
)->s_li_request
);
2682 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2683 mutex_unlock(&ext4_li_mtx
);
2686 static struct task_struct
*ext4_lazyinit_task
;
2689 * This is the function where ext4lazyinit thread lives. It walks
2690 * through the request list searching for next scheduled filesystem.
2691 * When such a fs is found, run the lazy initialization request
2692 * (ext4_rn_li_request) and keep track of the time spend in this
2693 * function. Based on that time we compute next schedule time of
2694 * the request. When walking through the list is complete, compute
2695 * next waking time and put itself into sleep.
2697 static int ext4_lazyinit_thread(void *arg
)
2699 struct ext4_lazy_init
*eli
= (struct ext4_lazy_init
*)arg
;
2700 struct list_head
*pos
, *n
;
2701 struct ext4_li_request
*elr
;
2702 unsigned long next_wakeup
, cur
;
2704 BUG_ON(NULL
== eli
);
2708 next_wakeup
= MAX_JIFFY_OFFSET
;
2710 mutex_lock(&eli
->li_list_mtx
);
2711 if (list_empty(&eli
->li_request_list
)) {
2712 mutex_unlock(&eli
->li_list_mtx
);
2716 list_for_each_safe(pos
, n
, &eli
->li_request_list
) {
2717 elr
= list_entry(pos
, struct ext4_li_request
,
2720 if (time_after_eq(jiffies
, elr
->lr_next_sched
)) {
2721 if (ext4_run_li_request(elr
) != 0) {
2722 /* error, remove the lazy_init job */
2723 ext4_remove_li_request(elr
);
2728 if (time_before(elr
->lr_next_sched
, next_wakeup
))
2729 next_wakeup
= elr
->lr_next_sched
;
2731 mutex_unlock(&eli
->li_list_mtx
);
2736 if ((time_after_eq(cur
, next_wakeup
)) ||
2737 (MAX_JIFFY_OFFSET
== next_wakeup
)) {
2742 schedule_timeout_interruptible(next_wakeup
- cur
);
2744 if (kthread_should_stop()) {
2745 ext4_clear_request_list();
2752 * It looks like the request list is empty, but we need
2753 * to check it under the li_list_mtx lock, to prevent any
2754 * additions into it, and of course we should lock ext4_li_mtx
2755 * to atomically free the list and ext4_li_info, because at
2756 * this point another ext4 filesystem could be registering
2759 mutex_lock(&ext4_li_mtx
);
2760 mutex_lock(&eli
->li_list_mtx
);
2761 if (!list_empty(&eli
->li_request_list
)) {
2762 mutex_unlock(&eli
->li_list_mtx
);
2763 mutex_unlock(&ext4_li_mtx
);
2766 mutex_unlock(&eli
->li_list_mtx
);
2767 kfree(ext4_li_info
);
2768 ext4_li_info
= NULL
;
2769 mutex_unlock(&ext4_li_mtx
);
2774 static void ext4_clear_request_list(void)
2776 struct list_head
*pos
, *n
;
2777 struct ext4_li_request
*elr
;
2779 mutex_lock(&ext4_li_info
->li_list_mtx
);
2780 list_for_each_safe(pos
, n
, &ext4_li_info
->li_request_list
) {
2781 elr
= list_entry(pos
, struct ext4_li_request
,
2783 ext4_remove_li_request(elr
);
2785 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2788 static int ext4_run_lazyinit_thread(void)
2790 ext4_lazyinit_task
= kthread_run(ext4_lazyinit_thread
,
2791 ext4_li_info
, "ext4lazyinit");
2792 if (IS_ERR(ext4_lazyinit_task
)) {
2793 int err
= PTR_ERR(ext4_lazyinit_task
);
2794 ext4_clear_request_list();
2795 kfree(ext4_li_info
);
2796 ext4_li_info
= NULL
;
2797 printk(KERN_CRIT
"EXT4: error %d creating inode table "
2798 "initialization thread\n",
2802 ext4_li_info
->li_state
|= EXT4_LAZYINIT_RUNNING
;
2807 * Check whether it make sense to run itable init. thread or not.
2808 * If there is at least one uninitialized inode table, return
2809 * corresponding group number, else the loop goes through all
2810 * groups and return total number of groups.
2812 static ext4_group_t
ext4_has_uninit_itable(struct super_block
*sb
)
2814 ext4_group_t group
, ngroups
= EXT4_SB(sb
)->s_groups_count
;
2815 struct ext4_group_desc
*gdp
= NULL
;
2817 for (group
= 0; group
< ngroups
; group
++) {
2818 gdp
= ext4_get_group_desc(sb
, group
, NULL
);
2822 if (!(gdp
->bg_flags
& cpu_to_le16(EXT4_BG_INODE_ZEROED
)))
2829 static int ext4_li_info_new(void)
2831 struct ext4_lazy_init
*eli
= NULL
;
2833 eli
= kzalloc(sizeof(*eli
), GFP_KERNEL
);
2837 INIT_LIST_HEAD(&eli
->li_request_list
);
2838 mutex_init(&eli
->li_list_mtx
);
2840 eli
->li_state
|= EXT4_LAZYINIT_QUIT
;
2847 static struct ext4_li_request
*ext4_li_request_new(struct super_block
*sb
,
2850 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2851 struct ext4_li_request
*elr
;
2854 elr
= kzalloc(sizeof(*elr
), GFP_KERNEL
);
2860 elr
->lr_next_group
= start
;
2863 * Randomize first schedule time of the request to
2864 * spread the inode table initialization requests
2867 get_random_bytes(&rnd
, sizeof(rnd
));
2868 elr
->lr_next_sched
= jiffies
+ (unsigned long)rnd
%
2869 (EXT4_DEF_LI_MAX_START_DELAY
* HZ
);
2874 static int ext4_register_li_request(struct super_block
*sb
,
2875 ext4_group_t first_not_zeroed
)
2877 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2878 struct ext4_li_request
*elr
;
2879 ext4_group_t ngroups
= EXT4_SB(sb
)->s_groups_count
;
2882 if (sbi
->s_li_request
!= NULL
) {
2884 * Reset timeout so it can be computed again, because
2885 * s_li_wait_mult might have changed.
2887 sbi
->s_li_request
->lr_timeout
= 0;
2891 if (first_not_zeroed
== ngroups
||
2892 (sb
->s_flags
& MS_RDONLY
) ||
2893 !test_opt(sb
, INIT_INODE_TABLE
))
2896 elr
= ext4_li_request_new(sb
, first_not_zeroed
);
2900 mutex_lock(&ext4_li_mtx
);
2902 if (NULL
== ext4_li_info
) {
2903 ret
= ext4_li_info_new();
2908 mutex_lock(&ext4_li_info
->li_list_mtx
);
2909 list_add(&elr
->lr_request
, &ext4_li_info
->li_request_list
);
2910 mutex_unlock(&ext4_li_info
->li_list_mtx
);
2912 sbi
->s_li_request
= elr
;
2914 * set elr to NULL here since it has been inserted to
2915 * the request_list and the removal and free of it is
2916 * handled by ext4_clear_request_list from now on.
2920 if (!(ext4_li_info
->li_state
& EXT4_LAZYINIT_RUNNING
)) {
2921 ret
= ext4_run_lazyinit_thread();
2926 mutex_unlock(&ext4_li_mtx
);
2933 * We do not need to lock anything since this is called on
2936 static void ext4_destroy_lazyinit_thread(void)
2939 * If thread exited earlier
2940 * there's nothing to be done.
2942 if (!ext4_li_info
|| !ext4_lazyinit_task
)
2945 kthread_stop(ext4_lazyinit_task
);
2948 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2950 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
2951 struct buffer_head
*bh
;
2952 struct ext4_super_block
*es
= NULL
;
2953 struct ext4_sb_info
*sbi
;
2955 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2956 ext4_fsblk_t logical_sb_block
;
2957 unsigned long offset
= 0;
2958 unsigned long journal_devnum
= 0;
2959 unsigned long def_mount_opts
;
2964 int blocksize
, clustersize
;
2965 unsigned int db_count
;
2967 int needs_recovery
, has_huge_files
, has_bigalloc
;
2970 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2971 ext4_group_t first_not_zeroed
;
2973 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2977 sbi
->s_blockgroup_lock
=
2978 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2979 if (!sbi
->s_blockgroup_lock
) {
2983 sb
->s_fs_info
= sbi
;
2984 sbi
->s_mount_opt
= 0;
2985 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2986 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2987 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2988 sbi
->s_sb_block
= sb_block
;
2989 if (sb
->s_bdev
->bd_part
)
2990 sbi
->s_sectors_written_start
=
2991 part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
2993 /* Cleanup superblock name */
2994 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2998 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
3000 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
3005 * The ext4 superblock will not be buffer aligned for other than 1kB
3006 * block sizes. We need to calculate the offset from buffer start.
3008 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
3009 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3010 offset
= do_div(logical_sb_block
, blocksize
);
3012 logical_sb_block
= sb_block
;
3015 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
3016 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
3020 * Note: s_es must be initialized as soon as possible because
3021 * some ext4 macro-instructions depend on its value
3023 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3025 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
3026 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
3028 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
3030 /* Set defaults before we parse the mount options */
3031 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
3032 set_opt(sb
, INIT_INODE_TABLE
);
3033 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
3035 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
)
3037 if (def_mount_opts
& EXT4_DEFM_UID16
)
3038 set_opt(sb
, NO_UID32
);
3039 /* xattr user namespace & acls are now defaulted on */
3040 #ifdef CONFIG_EXT4_FS_XATTR
3041 set_opt(sb
, XATTR_USER
);
3043 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3044 set_opt(sb
, POSIX_ACL
);
3046 set_opt(sb
, MBLK_IO_SUBMIT
);
3047 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
3048 set_opt(sb
, JOURNAL_DATA
);
3049 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
3050 set_opt(sb
, ORDERED_DATA
);
3051 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
3052 set_opt(sb
, WRITEBACK_DATA
);
3054 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
3055 set_opt(sb
, ERRORS_PANIC
);
3056 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
3057 set_opt(sb
, ERRORS_CONT
);
3059 set_opt(sb
, ERRORS_RO
);
3060 if (def_mount_opts
& EXT4_DEFM_BLOCK_VALIDITY
)
3061 set_opt(sb
, BLOCK_VALIDITY
);
3062 if (def_mount_opts
& EXT4_DEFM_DISCARD
)
3063 set_opt(sb
, DISCARD
);
3065 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
3066 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
3067 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
3068 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
3069 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
3071 if ((def_mount_opts
& EXT4_DEFM_NOBARRIER
) == 0)
3072 set_opt(sb
, BARRIER
);
3075 * enable delayed allocation by default
3076 * Use -o nodelalloc to turn it off
3078 if (!IS_EXT3_SB(sb
) &&
3079 ((def_mount_opts
& EXT4_DEFM_NODELALLOC
) == 0))
3080 set_opt(sb
, DELALLOC
);
3083 * set default s_li_wait_mult for lazyinit, for the case there is
3084 * no mount option specified.
3086 sbi
->s_li_wait_mult
= EXT4_DEF_LI_WAIT_MULT
;
3088 if (!parse_options((char *) sbi
->s_es
->s_mount_opts
, sb
,
3089 &journal_devnum
, &journal_ioprio
, 0)) {
3090 ext4_msg(sb
, KERN_WARNING
,
3091 "failed to parse options in superblock: %s",
3092 sbi
->s_es
->s_mount_opts
);
3094 sbi
->s_def_mount_opt
= sbi
->s_mount_opt
;
3095 if (!parse_options((char *) data
, sb
, &journal_devnum
,
3096 &journal_ioprio
, 0))
3099 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3100 printk_once(KERN_WARNING
"EXT4-fs: Warning: mounting "
3101 "with data=journal disables delayed "
3102 "allocation and O_DIRECT support!\n");
3103 if (test_opt2(sb
, EXPLICIT_DELALLOC
)) {
3104 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3105 "both data=journal and delalloc");
3108 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3109 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3110 "both data=journal and delalloc");
3113 if (test_opt(sb
, DELALLOC
))
3114 clear_opt(sb
, DELALLOC
);
3117 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
3118 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3119 if (blocksize
< PAGE_SIZE
) {
3120 ext4_msg(sb
, KERN_ERR
, "can't mount with "
3121 "dioread_nolock if block size != PAGE_SIZE");
3126 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3127 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3129 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
3130 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
3131 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
3132 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
3133 ext4_msg(sb
, KERN_WARNING
,
3134 "feature flags set on rev 0 fs, "
3135 "running e2fsck is recommended");
3137 if (IS_EXT2_SB(sb
)) {
3138 if (ext2_feature_set_ok(sb
))
3139 ext4_msg(sb
, KERN_INFO
, "mounting ext2 file system "
3140 "using the ext4 subsystem");
3142 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext2 due "
3143 "to feature incompatibilities");
3148 if (IS_EXT3_SB(sb
)) {
3149 if (ext3_feature_set_ok(sb
))
3150 ext4_msg(sb
, KERN_INFO
, "mounting ext3 file system "
3151 "using the ext4 subsystem");
3153 ext4_msg(sb
, KERN_ERR
, "couldn't mount as ext3 due "
3154 "to feature incompatibilities");
3160 * Check feature flags regardless of the revision level, since we
3161 * previously didn't change the revision level when setting the flags,
3162 * so there is a chance incompat flags are set on a rev 0 filesystem.
3164 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
3167 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
3168 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
3169 ext4_msg(sb
, KERN_ERR
,
3170 "Unsupported filesystem blocksize %d", blocksize
);
3174 if (sb
->s_blocksize
!= blocksize
) {
3175 /* Validate the filesystem blocksize */
3176 if (!sb_set_blocksize(sb
, blocksize
)) {
3177 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
3183 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
3184 offset
= do_div(logical_sb_block
, blocksize
);
3185 bh
= sb_bread(sb
, logical_sb_block
);
3187 ext4_msg(sb
, KERN_ERR
,
3188 "Can't read superblock on 2nd try");
3191 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
3193 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
3194 ext4_msg(sb
, KERN_ERR
,
3195 "Magic mismatch, very weird!");
3200 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3201 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
3202 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
3204 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
3206 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
3207 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
3208 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
3210 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
3211 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
3212 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
3213 (!is_power_of_2(sbi
->s_inode_size
)) ||
3214 (sbi
->s_inode_size
> blocksize
)) {
3215 ext4_msg(sb
, KERN_ERR
,
3216 "unsupported inode size: %d",
3220 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
3221 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
3224 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
3225 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
3226 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
3227 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
3228 !is_power_of_2(sbi
->s_desc_size
)) {
3229 ext4_msg(sb
, KERN_ERR
,
3230 "unsupported descriptor size %lu",
3235 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
3237 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
3238 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
3239 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
3242 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
3243 if (sbi
->s_inodes_per_block
== 0)
3245 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
3246 sbi
->s_inodes_per_block
;
3247 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
3249 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3250 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
3251 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
3253 for (i
= 0; i
< 4; i
++)
3254 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
3255 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
3256 i
= le32_to_cpu(es
->s_flags
);
3257 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
3258 sbi
->s_hash_unsigned
= 3;
3259 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
3260 #ifdef __CHAR_UNSIGNED__
3261 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
3262 sbi
->s_hash_unsigned
= 3;
3264 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
3269 /* Handle clustersize */
3270 clustersize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_cluster_size
);
3271 has_bigalloc
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3272 EXT4_FEATURE_RO_COMPAT_BIGALLOC
);
3274 if (clustersize
< blocksize
) {
3275 ext4_msg(sb
, KERN_ERR
,
3276 "cluster size (%d) smaller than "
3277 "block size (%d)", clustersize
, blocksize
);
3280 sbi
->s_cluster_bits
= le32_to_cpu(es
->s_log_cluster_size
) -
3281 le32_to_cpu(es
->s_log_block_size
);
3282 sbi
->s_clusters_per_group
=
3283 le32_to_cpu(es
->s_clusters_per_group
);
3284 if (sbi
->s_clusters_per_group
> blocksize
* 8) {
3285 ext4_msg(sb
, KERN_ERR
,
3286 "#clusters per group too big: %lu",
3287 sbi
->s_clusters_per_group
);
3290 if (sbi
->s_blocks_per_group
!=
3291 (sbi
->s_clusters_per_group
* (clustersize
/ blocksize
))) {
3292 ext4_msg(sb
, KERN_ERR
, "blocks per group (%lu) and "
3293 "clusters per group (%lu) inconsistent",
3294 sbi
->s_blocks_per_group
,
3295 sbi
->s_clusters_per_group
);
3299 if (clustersize
!= blocksize
) {
3300 ext4_warning(sb
, "fragment/cluster size (%d) != "
3301 "block size (%d)", clustersize
,
3303 clustersize
= blocksize
;
3305 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
3306 ext4_msg(sb
, KERN_ERR
,
3307 "#blocks per group too big: %lu",
3308 sbi
->s_blocks_per_group
);
3311 sbi
->s_clusters_per_group
= sbi
->s_blocks_per_group
;
3312 sbi
->s_cluster_bits
= 0;
3314 sbi
->s_cluster_ratio
= clustersize
/ blocksize
;
3316 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
3317 ext4_msg(sb
, KERN_ERR
,
3318 "#inodes per group too big: %lu",
3319 sbi
->s_inodes_per_group
);
3324 * Test whether we have more sectors than will fit in sector_t,
3325 * and whether the max offset is addressable by the page cache.
3327 err
= generic_check_addressable(sb
->s_blocksize_bits
,
3328 ext4_blocks_count(es
));
3330 ext4_msg(sb
, KERN_ERR
, "filesystem"
3331 " too large to mount safely on this system");
3332 if (sizeof(sector_t
) < 8)
3333 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
3338 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
3341 /* check blocks count against device size */
3342 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
3343 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
3344 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
3345 "exceeds size of device (%llu blocks)",
3346 ext4_blocks_count(es
), blocks_count
);
3351 * It makes no sense for the first data block to be beyond the end
3352 * of the filesystem.
3354 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
3355 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data "
3356 "block %u is beyond end of filesystem (%llu)",
3357 le32_to_cpu(es
->s_first_data_block
),
3358 ext4_blocks_count(es
));
3361 blocks_count
= (ext4_blocks_count(es
) -
3362 le32_to_cpu(es
->s_first_data_block
) +
3363 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
3364 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
3365 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
3366 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
3367 "(block count %llu, first data block %u, "
3368 "blocks per group %lu)", sbi
->s_groups_count
,
3369 ext4_blocks_count(es
),
3370 le32_to_cpu(es
->s_first_data_block
),
3371 EXT4_BLOCKS_PER_GROUP(sb
));
3374 sbi
->s_groups_count
= blocks_count
;
3375 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
3376 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
3377 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
3378 EXT4_DESC_PER_BLOCK(sb
);
3379 sbi
->s_group_desc
= ext4_kvmalloc(db_count
*
3380 sizeof(struct buffer_head
*),
3382 if (sbi
->s_group_desc
== NULL
) {
3383 ext4_msg(sb
, KERN_ERR
, "not enough memory");
3388 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
3391 proc_create_data("options", S_IRUGO
, sbi
->s_proc
,
3392 &ext4_seq_options_fops
, sb
);
3394 bgl_lock_init(sbi
->s_blockgroup_lock
);
3396 for (i
= 0; i
< db_count
; i
++) {
3397 block
= descriptor_loc(sb
, logical_sb_block
, i
);
3398 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
3399 if (!sbi
->s_group_desc
[i
]) {
3400 ext4_msg(sb
, KERN_ERR
,
3401 "can't read group descriptor %d", i
);
3406 if (!ext4_check_descriptors(sb
, &first_not_zeroed
)) {
3407 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
3410 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
3411 if (!ext4_fill_flex_info(sb
)) {
3412 ext4_msg(sb
, KERN_ERR
,
3413 "unable to initialize "
3414 "flex_bg meta info!");
3418 sbi
->s_gdb_count
= db_count
;
3419 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
3420 spin_lock_init(&sbi
->s_next_gen_lock
);
3422 init_timer(&sbi
->s_err_report
);
3423 sbi
->s_err_report
.function
= print_daily_error_info
;
3424 sbi
->s_err_report
.data
= (unsigned long) sb
;
3426 err
= percpu_counter_init(&sbi
->s_freeclusters_counter
,
3427 ext4_count_free_clusters(sb
));
3429 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3430 ext4_count_free_inodes(sb
));
3433 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3434 ext4_count_dirs(sb
));
3437 err
= percpu_counter_init(&sbi
->s_dirtyclusters_counter
, 0);
3440 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3444 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
3445 sbi
->s_max_writeback_mb_bump
= 128;
3448 * set up enough so that it can read an inode
3450 if (!test_opt(sb
, NOLOAD
) &&
3451 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
3452 sb
->s_op
= &ext4_sops
;
3454 sb
->s_op
= &ext4_nojournal_sops
;
3455 sb
->s_export_op
= &ext4_export_ops
;
3456 sb
->s_xattr
= ext4_xattr_handlers
;
3458 sb
->s_qcop
= &ext4_qctl_operations
;
3459 sb
->dq_op
= &ext4_quota_operations
;
3461 memcpy(sb
->s_uuid
, es
->s_uuid
, sizeof(es
->s_uuid
));
3463 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
3464 mutex_init(&sbi
->s_orphan_lock
);
3465 sbi
->s_resize_flags
= 0;
3469 needs_recovery
= (es
->s_last_orphan
!= 0 ||
3470 EXT4_HAS_INCOMPAT_FEATURE(sb
,
3471 EXT4_FEATURE_INCOMPAT_RECOVER
));
3473 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_MMP
) &&
3474 !(sb
->s_flags
& MS_RDONLY
))
3475 if (ext4_multi_mount_protect(sb
, le64_to_cpu(es
->s_mmp_block
)))
3479 * The first inode we look at is the journal inode. Don't try
3480 * root first: it may be modified in the journal!
3482 if (!test_opt(sb
, NOLOAD
) &&
3483 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3484 if (ext4_load_journal(sb
, es
, journal_devnum
))
3486 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
3487 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3488 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
3489 "suppressed and not mounted read-only");
3490 goto failed_mount_wq
;
3492 clear_opt(sb
, DATA_FLAGS
);
3493 sbi
->s_journal
= NULL
;
3498 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
3499 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
3500 JBD2_FEATURE_INCOMPAT_64BIT
)) {
3501 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
3502 goto failed_mount_wq
;
3505 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
3506 jbd2_journal_set_features(sbi
->s_journal
,
3507 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3508 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3509 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
3510 jbd2_journal_set_features(sbi
->s_journal
,
3511 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
3512 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
3513 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3515 jbd2_journal_clear_features(sbi
->s_journal
,
3516 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
3517 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
3520 /* We have now updated the journal if required, so we can
3521 * validate the data journaling mode. */
3522 switch (test_opt(sb
, DATA_FLAGS
)) {
3524 /* No mode set, assume a default based on the journal
3525 * capabilities: ORDERED_DATA if the journal can
3526 * cope, else JOURNAL_DATA
3528 if (jbd2_journal_check_available_features
3529 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
3530 set_opt(sb
, ORDERED_DATA
);
3532 set_opt(sb
, JOURNAL_DATA
);
3535 case EXT4_MOUNT_ORDERED_DATA
:
3536 case EXT4_MOUNT_WRITEBACK_DATA
:
3537 if (!jbd2_journal_check_available_features
3538 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
3539 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
3540 "requested data journaling mode");
3541 goto failed_mount_wq
;
3546 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3548 sbi
->s_journal
->j_commit_callback
= ext4_journal_commit_callback
;
3551 * The journal may have updated the bg summary counts, so we
3552 * need to update the global counters.
3554 percpu_counter_set(&sbi
->s_freeclusters_counter
,
3555 ext4_count_free_clusters(sb
));
3556 percpu_counter_set(&sbi
->s_freeinodes_counter
,
3557 ext4_count_free_inodes(sb
));
3558 percpu_counter_set(&sbi
->s_dirs_counter
,
3559 ext4_count_dirs(sb
));
3560 percpu_counter_set(&sbi
->s_dirtyclusters_counter
, 0);
3564 * The maximum number of concurrent works can be high and
3565 * concurrency isn't really necessary. Limit it to 1.
3567 EXT4_SB(sb
)->dio_unwritten_wq
=
3568 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM
| WQ_UNBOUND
, 1);
3569 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3570 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3571 goto failed_mount_wq
;
3575 * The jbd2_journal_load will have done any necessary log recovery,
3576 * so we can safely mount the rest of the filesystem now.
3579 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3581 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3582 ret
= PTR_ERR(root
);
3586 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3587 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3591 sb
->s_root
= d_alloc_root(root
);
3594 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3599 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3601 /* determine the minimum size of new large inodes, if present */
3602 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3603 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3604 EXT4_GOOD_OLD_INODE_SIZE
;
3605 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3606 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3607 if (sbi
->s_want_extra_isize
<
3608 le16_to_cpu(es
->s_want_extra_isize
))
3609 sbi
->s_want_extra_isize
=
3610 le16_to_cpu(es
->s_want_extra_isize
);
3611 if (sbi
->s_want_extra_isize
<
3612 le16_to_cpu(es
->s_min_extra_isize
))
3613 sbi
->s_want_extra_isize
=
3614 le16_to_cpu(es
->s_min_extra_isize
);
3617 /* Check if enough inode space is available */
3618 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3619 sbi
->s_inode_size
) {
3620 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3621 EXT4_GOOD_OLD_INODE_SIZE
;
3622 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3626 err
= ext4_setup_system_zone(sb
);
3628 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3630 goto failed_mount4a
;
3634 err
= ext4_mb_init(sb
, needs_recovery
);
3636 ext4_msg(sb
, KERN_ERR
, "failed to initialize mballoc (%d)",
3641 err
= ext4_register_li_request(sb
, first_not_zeroed
);
3645 sbi
->s_kobj
.kset
= ext4_kset
;
3646 init_completion(&sbi
->s_kobj_unregister
);
3647 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3652 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3653 ext4_orphan_cleanup(sb
, es
);
3654 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3655 if (needs_recovery
) {
3656 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3657 ext4_mark_recovery_complete(sb
, es
);
3659 if (EXT4_SB(sb
)->s_journal
) {
3660 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3661 descr
= " journalled data mode";
3662 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3663 descr
= " ordered data mode";
3665 descr
= " writeback data mode";
3667 descr
= "out journal";
3669 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3670 "Opts: %s%s%s", descr
, sbi
->s_es
->s_mount_opts
,
3671 *sbi
->s_es
->s_mount_opts
? "; " : "", orig_data
);
3673 if (es
->s_error_count
)
3674 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3681 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3685 ext4_unregister_li_request(sb
);
3687 ext4_mb_release(sb
);
3689 ext4_ext_release(sb
);
3690 ext4_release_system_zone(sb
);
3695 ext4_msg(sb
, KERN_ERR
, "mount failed");
3696 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3698 if (sbi
->s_journal
) {
3699 jbd2_journal_destroy(sbi
->s_journal
);
3700 sbi
->s_journal
= NULL
;
3703 del_timer(&sbi
->s_err_report
);
3704 if (sbi
->s_flex_groups
)
3705 ext4_kvfree(sbi
->s_flex_groups
);
3706 percpu_counter_destroy(&sbi
->s_freeclusters_counter
);
3707 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3708 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3709 percpu_counter_destroy(&sbi
->s_dirtyclusters_counter
);
3711 kthread_stop(sbi
->s_mmp_tsk
);
3713 for (i
= 0; i
< db_count
; i
++)
3714 brelse(sbi
->s_group_desc
[i
]);
3715 ext4_kvfree(sbi
->s_group_desc
);
3718 remove_proc_entry("options", sbi
->s_proc
);
3719 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3722 for (i
= 0; i
< MAXQUOTAS
; i
++)
3723 kfree(sbi
->s_qf_names
[i
]);
3725 ext4_blkdev_remove(sbi
);
3728 sb
->s_fs_info
= NULL
;
3729 kfree(sbi
->s_blockgroup_lock
);
3737 * Setup any per-fs journal parameters now. We'll do this both on
3738 * initial mount, once the journal has been initialised but before we've
3739 * done any recovery; and again on any subsequent remount.
3741 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3743 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3745 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3746 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3747 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3749 write_lock(&journal
->j_state_lock
);
3750 if (test_opt(sb
, BARRIER
))
3751 journal
->j_flags
|= JBD2_BARRIER
;
3753 journal
->j_flags
&= ~JBD2_BARRIER
;
3754 if (test_opt(sb
, DATA_ERR_ABORT
))
3755 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3757 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3758 write_unlock(&journal
->j_state_lock
);
3761 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3762 unsigned int journal_inum
)
3764 struct inode
*journal_inode
;
3767 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3769 /* First, test for the existence of a valid inode on disk. Bad
3770 * things happen if we iget() an unused inode, as the subsequent
3771 * iput() will try to delete it. */
3773 journal_inode
= ext4_iget(sb
, journal_inum
);
3774 if (IS_ERR(journal_inode
)) {
3775 ext4_msg(sb
, KERN_ERR
, "no journal found");
3778 if (!journal_inode
->i_nlink
) {
3779 make_bad_inode(journal_inode
);
3780 iput(journal_inode
);
3781 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3785 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3786 journal_inode
, journal_inode
->i_size
);
3787 if (!S_ISREG(journal_inode
->i_mode
)) {
3788 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3789 iput(journal_inode
);
3793 journal
= jbd2_journal_init_inode(journal_inode
);
3795 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3796 iput(journal_inode
);
3799 journal
->j_private
= sb
;
3800 ext4_init_journal_params(sb
, journal
);
3804 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3807 struct buffer_head
*bh
;
3811 int hblock
, blocksize
;
3812 ext4_fsblk_t sb_block
;
3813 unsigned long offset
;
3814 struct ext4_super_block
*es
;
3815 struct block_device
*bdev
;
3817 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3819 bdev
= ext4_blkdev_get(j_dev
, sb
);
3823 blocksize
= sb
->s_blocksize
;
3824 hblock
= bdev_logical_block_size(bdev
);
3825 if (blocksize
< hblock
) {
3826 ext4_msg(sb
, KERN_ERR
,
3827 "blocksize too small for journal device");
3831 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3832 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3833 set_blocksize(bdev
, blocksize
);
3834 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3835 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3836 "external journal");
3840 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3841 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3842 !(le32_to_cpu(es
->s_feature_incompat
) &
3843 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3844 ext4_msg(sb
, KERN_ERR
, "external journal has "
3850 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3851 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3856 len
= ext4_blocks_count(es
);
3857 start
= sb_block
+ 1;
3858 brelse(bh
); /* we're done with the superblock */
3860 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3861 start
, len
, blocksize
);
3863 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3866 journal
->j_private
= sb
;
3867 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3868 wait_on_buffer(journal
->j_sb_buffer
);
3869 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3870 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3873 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3874 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3875 "user (unsupported) - %d",
3876 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3879 EXT4_SB(sb
)->journal_bdev
= bdev
;
3880 ext4_init_journal_params(sb
, journal
);
3884 jbd2_journal_destroy(journal
);
3886 ext4_blkdev_put(bdev
);
3890 static int ext4_load_journal(struct super_block
*sb
,
3891 struct ext4_super_block
*es
,
3892 unsigned long journal_devnum
)
3895 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3898 int really_read_only
;
3900 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3902 if (journal_devnum
&&
3903 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3904 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3905 "numbers have changed");
3906 journal_dev
= new_decode_dev(journal_devnum
);
3908 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3910 really_read_only
= bdev_read_only(sb
->s_bdev
);
3913 * Are we loading a blank journal or performing recovery after a
3914 * crash? For recovery, we need to check in advance whether we
3915 * can get read-write access to the device.
3917 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3918 if (sb
->s_flags
& MS_RDONLY
) {
3919 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3920 "required on readonly filesystem");
3921 if (really_read_only
) {
3922 ext4_msg(sb
, KERN_ERR
, "write access "
3923 "unavailable, cannot proceed");
3926 ext4_msg(sb
, KERN_INFO
, "write access will "
3927 "be enabled during recovery");
3931 if (journal_inum
&& journal_dev
) {
3932 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3933 "and inode journals!");
3938 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3941 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3945 if (!(journal
->j_flags
& JBD2_BARRIER
))
3946 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3948 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3949 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3951 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
3953 memcpy(save
, ((char *) es
) +
3954 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
3955 err
= jbd2_journal_load(journal
);
3957 memcpy(((char *) es
) + EXT4_S_ERR_START
,
3958 save
, EXT4_S_ERR_LEN
);
3963 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3964 jbd2_journal_destroy(journal
);
3968 EXT4_SB(sb
)->s_journal
= journal
;
3969 ext4_clear_journal_err(sb
, es
);
3971 if (!really_read_only
&& journal_devnum
&&
3972 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3973 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3975 /* Make sure we flush the recovery flag to disk. */
3976 ext4_commit_super(sb
, 1);
3982 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3984 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3985 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3988 if (!sbh
|| block_device_ejected(sb
))
3990 if (buffer_write_io_error(sbh
)) {
3992 * Oh, dear. A previous attempt to write the
3993 * superblock failed. This could happen because the
3994 * USB device was yanked out. Or it could happen to
3995 * be a transient write error and maybe the block will
3996 * be remapped. Nothing we can do but to retry the
3997 * write and hope for the best.
3999 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
4000 "superblock detected");
4001 clear_buffer_write_io_error(sbh
);
4002 set_buffer_uptodate(sbh
);
4005 * If the file system is mounted read-only, don't update the
4006 * superblock write time. This avoids updating the superblock
4007 * write time when we are mounting the root file system
4008 * read/only but we need to replay the journal; at that point,
4009 * for people who are east of GMT and who make their clock
4010 * tick in localtime for Windows bug-for-bug compatibility,
4011 * the clock is set in the future, and this will cause e2fsck
4012 * to complain and force a full file system check.
4014 if (!(sb
->s_flags
& MS_RDONLY
))
4015 es
->s_wtime
= cpu_to_le32(get_seconds());
4016 if (sb
->s_bdev
->bd_part
)
4017 es
->s_kbytes_written
=
4018 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
4019 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
4020 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
4022 es
->s_kbytes_written
=
4023 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
);
4024 ext4_free_blocks_count_set(es
,
4025 EXT4_C2B(EXT4_SB(sb
), percpu_counter_sum_positive(
4026 &EXT4_SB(sb
)->s_freeclusters_counter
)));
4027 es
->s_free_inodes_count
=
4028 cpu_to_le32(percpu_counter_sum_positive(
4029 &EXT4_SB(sb
)->s_freeinodes_counter
));
4031 BUFFER_TRACE(sbh
, "marking dirty");
4032 mark_buffer_dirty(sbh
);
4034 error
= sync_dirty_buffer(sbh
);
4038 error
= buffer_write_io_error(sbh
);
4040 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
4042 clear_buffer_write_io_error(sbh
);
4043 set_buffer_uptodate(sbh
);
4050 * Have we just finished recovery? If so, and if we are mounting (or
4051 * remounting) the filesystem readonly, then we will end up with a
4052 * consistent fs on disk. Record that fact.
4054 static void ext4_mark_recovery_complete(struct super_block
*sb
,
4055 struct ext4_super_block
*es
)
4057 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
4059 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
4060 BUG_ON(journal
!= NULL
);
4063 jbd2_journal_lock_updates(journal
);
4064 if (jbd2_journal_flush(journal
) < 0)
4067 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
4068 sb
->s_flags
& MS_RDONLY
) {
4069 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4070 ext4_commit_super(sb
, 1);
4074 jbd2_journal_unlock_updates(journal
);
4078 * If we are mounting (or read-write remounting) a filesystem whose journal
4079 * has recorded an error from a previous lifetime, move that error to the
4080 * main filesystem now.
4082 static void ext4_clear_journal_err(struct super_block
*sb
,
4083 struct ext4_super_block
*es
)
4089 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
4091 journal
= EXT4_SB(sb
)->s_journal
;
4094 * Now check for any error status which may have been recorded in the
4095 * journal by a prior ext4_error() or ext4_abort()
4098 j_errno
= jbd2_journal_errno(journal
);
4102 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
4103 ext4_warning(sb
, "Filesystem error recorded "
4104 "from previous mount: %s", errstr
);
4105 ext4_warning(sb
, "Marking fs in need of filesystem check.");
4107 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
4108 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
4109 ext4_commit_super(sb
, 1);
4111 jbd2_journal_clear_err(journal
);
4116 * Force the running and committing transactions to commit,
4117 * and wait on the commit.
4119 int ext4_force_commit(struct super_block
*sb
)
4124 if (sb
->s_flags
& MS_RDONLY
)
4127 journal
= EXT4_SB(sb
)->s_journal
;
4129 vfs_check_frozen(sb
, SB_FREEZE_TRANS
);
4130 ret
= ext4_journal_force_commit(journal
);
4136 static void ext4_write_super(struct super_block
*sb
)
4139 ext4_commit_super(sb
, 1);
4143 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
4147 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4149 trace_ext4_sync_fs(sb
, wait
);
4150 flush_workqueue(sbi
->dio_unwritten_wq
);
4151 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
4153 jbd2_log_wait_commit(sbi
->s_journal
, target
);
4159 * LVM calls this function before a (read-only) snapshot is created. This
4160 * gives us a chance to flush the journal completely and mark the fs clean.
4162 * Note that only this function cannot bring a filesystem to be in a clean
4163 * state independently, because ext4 prevents a new handle from being started
4164 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4167 static int ext4_freeze(struct super_block
*sb
)
4172 if (sb
->s_flags
& MS_RDONLY
)
4175 journal
= EXT4_SB(sb
)->s_journal
;
4177 /* Now we set up the journal barrier. */
4178 jbd2_journal_lock_updates(journal
);
4181 * Don't clear the needs_recovery flag if we failed to flush
4184 error
= jbd2_journal_flush(journal
);
4188 /* Journal blocked and flushed, clear needs_recovery flag. */
4189 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4190 error
= ext4_commit_super(sb
, 1);
4192 /* we rely on s_frozen to stop further updates */
4193 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4198 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4199 * flag here, even though the filesystem is not technically dirty yet.
4201 static int ext4_unfreeze(struct super_block
*sb
)
4203 if (sb
->s_flags
& MS_RDONLY
)
4207 /* Reset the needs_recovery flag before the fs is unlocked. */
4208 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
4209 ext4_commit_super(sb
, 1);
4215 * Structure to save mount options for ext4_remount's benefit
4217 struct ext4_mount_options
{
4218 unsigned long s_mount_opt
;
4219 unsigned long s_mount_opt2
;
4222 unsigned long s_commit_interval
;
4223 u32 s_min_batch_time
, s_max_batch_time
;
4226 char *s_qf_names
[MAXQUOTAS
];
4230 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
4232 struct ext4_super_block
*es
;
4233 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4234 unsigned long old_sb_flags
;
4235 struct ext4_mount_options old_opts
;
4236 int enable_quota
= 0;
4238 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
4243 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
4245 /* Store the original options */
4247 old_sb_flags
= sb
->s_flags
;
4248 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
4249 old_opts
.s_mount_opt2
= sbi
->s_mount_opt2
;
4250 old_opts
.s_resuid
= sbi
->s_resuid
;
4251 old_opts
.s_resgid
= sbi
->s_resgid
;
4252 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
4253 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
4254 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
4256 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
4257 for (i
= 0; i
< MAXQUOTAS
; i
++)
4258 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
4260 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
4261 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
4264 * Allow the "check" option to be passed as a remount option.
4266 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
, 1)) {
4271 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
4272 ext4_abort(sb
, "Abort forced by user");
4274 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
4275 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
4279 if (sbi
->s_journal
) {
4280 ext4_init_journal_params(sb
, sbi
->s_journal
);
4281 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
4284 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
)) {
4285 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
4290 if (*flags
& MS_RDONLY
) {
4291 err
= dquot_suspend(sb
, -1);
4296 * First of all, the unconditional stuff we have to do
4297 * to disable replay of the journal when we next remount
4299 sb
->s_flags
|= MS_RDONLY
;
4302 * OK, test if we are remounting a valid rw partition
4303 * readonly, and if so set the rdonly flag and then
4304 * mark the partition as valid again.
4306 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
4307 (sbi
->s_mount_state
& EXT4_VALID_FS
))
4308 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
4311 ext4_mark_recovery_complete(sb
, es
);
4313 /* Make sure we can mount this feature set readwrite */
4314 if (!ext4_feature_set_ok(sb
, 0)) {
4319 * Make sure the group descriptor checksums
4320 * are sane. If they aren't, refuse to remount r/w.
4322 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
4323 struct ext4_group_desc
*gdp
=
4324 ext4_get_group_desc(sb
, g
, NULL
);
4326 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
4327 ext4_msg(sb
, KERN_ERR
,
4328 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4329 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
4330 le16_to_cpu(gdp
->bg_checksum
));
4337 * If we have an unprocessed orphan list hanging
4338 * around from a previously readonly bdev mount,
4339 * require a full umount/remount for now.
4341 if (es
->s_last_orphan
) {
4342 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
4343 "remount RDWR because of unprocessed "
4344 "orphan inode list. Please "
4345 "umount/remount instead");
4351 * Mounting a RDONLY partition read-write, so reread
4352 * and store the current valid flag. (It may have
4353 * been changed by e2fsck since we originally mounted
4357 ext4_clear_journal_err(sb
, es
);
4358 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
4359 if (!ext4_setup_super(sb
, es
, 0))
4360 sb
->s_flags
&= ~MS_RDONLY
;
4361 if (EXT4_HAS_INCOMPAT_FEATURE(sb
,
4362 EXT4_FEATURE_INCOMPAT_MMP
))
4363 if (ext4_multi_mount_protect(sb
,
4364 le64_to_cpu(es
->s_mmp_block
))) {
4373 * Reinitialize lazy itable initialization thread based on
4376 if ((sb
->s_flags
& MS_RDONLY
) || !test_opt(sb
, INIT_INODE_TABLE
))
4377 ext4_unregister_li_request(sb
);
4379 ext4_group_t first_not_zeroed
;
4380 first_not_zeroed
= ext4_has_uninit_itable(sb
);
4381 ext4_register_li_request(sb
, first_not_zeroed
);
4384 ext4_setup_system_zone(sb
);
4385 if (sbi
->s_journal
== NULL
)
4386 ext4_commit_super(sb
, 1);
4389 /* Release old quota file names */
4390 for (i
= 0; i
< MAXQUOTAS
; i
++)
4391 if (old_opts
.s_qf_names
[i
] &&
4392 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4393 kfree(old_opts
.s_qf_names
[i
]);
4397 dquot_resume(sb
, -1);
4399 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
4404 sb
->s_flags
= old_sb_flags
;
4405 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
4406 sbi
->s_mount_opt2
= old_opts
.s_mount_opt2
;
4407 sbi
->s_resuid
= old_opts
.s_resuid
;
4408 sbi
->s_resgid
= old_opts
.s_resgid
;
4409 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
4410 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
4411 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
4413 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
4414 for (i
= 0; i
< MAXQUOTAS
; i
++) {
4415 if (sbi
->s_qf_names
[i
] &&
4416 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
4417 kfree(sbi
->s_qf_names
[i
]);
4418 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
4427 * Note: calculating the overhead so we can be compatible with
4428 * historical BSD practice is quite difficult in the face of
4429 * clusters/bigalloc. This is because multiple metadata blocks from
4430 * different block group can end up in the same allocation cluster.
4431 * Calculating the exact overhead in the face of clustered allocation
4432 * requires either O(all block bitmaps) in memory or O(number of block
4433 * groups**2) in time. We will still calculate the superblock for
4434 * older file systems --- and if we come across with a bigalloc file
4435 * system with zero in s_overhead_clusters the estimate will be close to
4436 * correct especially for very large cluster sizes --- but for newer
4437 * file systems, it's better to calculate this figure once at mkfs
4438 * time, and store it in the superblock. If the superblock value is
4439 * present (even for non-bigalloc file systems), we will use it.
4441 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
4443 struct super_block
*sb
= dentry
->d_sb
;
4444 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
4445 struct ext4_super_block
*es
= sbi
->s_es
;
4446 struct ext4_group_desc
*gdp
;
4450 if (test_opt(sb
, MINIX_DF
)) {
4451 sbi
->s_overhead_last
= 0;
4452 } else if (es
->s_overhead_clusters
) {
4453 sbi
->s_overhead_last
= le32_to_cpu(es
->s_overhead_clusters
);
4454 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
4455 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
4456 ext4_fsblk_t overhead
= 0;
4459 * Compute the overhead (FS structures). This is constant
4460 * for a given filesystem unless the number of block groups
4461 * changes so we cache the previous value until it does.
4465 * All of the blocks before first_data_block are
4468 overhead
= EXT4_B2C(sbi
, le32_to_cpu(es
->s_first_data_block
));
4471 * Add the overhead found in each block group
4473 for (i
= 0; i
< ngroups
; i
++) {
4474 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
4475 overhead
+= ext4_num_overhead_clusters(sb
, i
, gdp
);
4478 sbi
->s_overhead_last
= overhead
;
4480 sbi
->s_blocks_last
= ext4_blocks_count(es
);
4483 buf
->f_type
= EXT4_SUPER_MAGIC
;
4484 buf
->f_bsize
= sb
->s_blocksize
;
4485 buf
->f_blocks
= (ext4_blocks_count(es
) -
4486 EXT4_C2B(sbi
, sbi
->s_overhead_last
));
4487 bfree
= percpu_counter_sum_positive(&sbi
->s_freeclusters_counter
) -
4488 percpu_counter_sum_positive(&sbi
->s_dirtyclusters_counter
);
4489 /* prevent underflow in case that few free space is available */
4490 buf
->f_bfree
= EXT4_C2B(sbi
, max_t(s64
, bfree
, 0));
4491 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
4492 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
4494 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
4495 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
4496 buf
->f_namelen
= EXT4_NAME_LEN
;
4497 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
4498 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
4499 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
4500 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
4505 /* Helper function for writing quotas on sync - we need to start transaction
4506 * before quota file is locked for write. Otherwise the are possible deadlocks:
4507 * Process 1 Process 2
4508 * ext4_create() quota_sync()
4509 * jbd2_journal_start() write_dquot()
4510 * dquot_initialize() down(dqio_mutex)
4511 * down(dqio_mutex) jbd2_journal_start()
4517 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
4519 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
4522 static int ext4_write_dquot(struct dquot
*dquot
)
4526 struct inode
*inode
;
4528 inode
= dquot_to_inode(dquot
);
4529 handle
= ext4_journal_start(inode
,
4530 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
4532 return PTR_ERR(handle
);
4533 ret
= dquot_commit(dquot
);
4534 err
= ext4_journal_stop(handle
);
4540 static int ext4_acquire_dquot(struct dquot
*dquot
)
4545 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4546 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
4548 return PTR_ERR(handle
);
4549 ret
= dquot_acquire(dquot
);
4550 err
= ext4_journal_stop(handle
);
4556 static int ext4_release_dquot(struct dquot
*dquot
)
4561 handle
= ext4_journal_start(dquot_to_inode(dquot
),
4562 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
4563 if (IS_ERR(handle
)) {
4564 /* Release dquot anyway to avoid endless cycle in dqput() */
4565 dquot_release(dquot
);
4566 return PTR_ERR(handle
);
4568 ret
= dquot_release(dquot
);
4569 err
= ext4_journal_stop(handle
);
4575 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4577 /* Are we journaling quotas? */
4578 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4579 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4580 dquot_mark_dquot_dirty(dquot
);
4581 return ext4_write_dquot(dquot
);
4583 return dquot_mark_dquot_dirty(dquot
);
4587 static int ext4_write_info(struct super_block
*sb
, int type
)
4592 /* Data block + inode block */
4593 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4595 return PTR_ERR(handle
);
4596 ret
= dquot_commit_info(sb
, type
);
4597 err
= ext4_journal_stop(handle
);
4604 * Turn on quotas during mount time - we need to find
4605 * the quota file and such...
4607 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4609 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4610 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4614 * Standard function to be called on quota_on
4616 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4621 if (!test_opt(sb
, QUOTA
))
4624 /* Quotafile not on the same filesystem? */
4625 if (path
->dentry
->d_sb
!= sb
)
4627 /* Journaling quota? */
4628 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4629 /* Quotafile not in fs root? */
4630 if (path
->dentry
->d_parent
!= sb
->s_root
)
4631 ext4_msg(sb
, KERN_WARNING
,
4632 "Quota file not on filesystem root. "
4633 "Journaled quota will not work");
4637 * When we journal data on quota file, we have to flush journal to see
4638 * all updates to the file when we bypass pagecache...
4640 if (EXT4_SB(sb
)->s_journal
&&
4641 ext4_should_journal_data(path
->dentry
->d_inode
)) {
4643 * We don't need to lock updates but journal_flush() could
4644 * otherwise be livelocked...
4646 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4647 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4648 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4653 return dquot_quota_on(sb
, type
, format_id
, path
);
4656 static int ext4_quota_off(struct super_block
*sb
, int type
)
4658 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4661 /* Force all delayed allocation blocks to be allocated.
4662 * Caller already holds s_umount sem */
4663 if (test_opt(sb
, DELALLOC
))
4664 sync_filesystem(sb
);
4669 /* Update modification times of quota files when userspace can
4670 * start looking at them */
4671 handle
= ext4_journal_start(inode
, 1);
4674 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4675 ext4_mark_inode_dirty(handle
, inode
);
4676 ext4_journal_stop(handle
);
4679 return dquot_quota_off(sb
, type
);
4682 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4683 * acquiring the locks... As quota files are never truncated and quota code
4684 * itself serializes the operations (and no one else should touch the files)
4685 * we don't have to be afraid of races */
4686 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4687 size_t len
, loff_t off
)
4689 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4690 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4692 int offset
= off
& (sb
->s_blocksize
- 1);
4695 struct buffer_head
*bh
;
4696 loff_t i_size
= i_size_read(inode
);
4700 if (off
+len
> i_size
)
4703 while (toread
> 0) {
4704 tocopy
= sb
->s_blocksize
- offset
< toread
?
4705 sb
->s_blocksize
- offset
: toread
;
4706 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4709 if (!bh
) /* A hole? */
4710 memset(data
, 0, tocopy
);
4712 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4722 /* Write to quotafile (we know the transaction is already started and has
4723 * enough credits) */
4724 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4725 const char *data
, size_t len
, loff_t off
)
4727 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4728 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4730 int offset
= off
& (sb
->s_blocksize
- 1);
4731 struct buffer_head
*bh
;
4732 handle_t
*handle
= journal_current_handle();
4734 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4735 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4736 " cancelled because transaction is not started",
4737 (unsigned long long)off
, (unsigned long long)len
);
4741 * Since we account only one data block in transaction credits,
4742 * then it is impossible to cross a block boundary.
4744 if (sb
->s_blocksize
- offset
< len
) {
4745 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4746 " cancelled because not block aligned",
4747 (unsigned long long)off
, (unsigned long long)len
);
4751 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4752 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4755 err
= ext4_journal_get_write_access(handle
, bh
);
4761 memcpy(bh
->b_data
+offset
, data
, len
);
4762 flush_dcache_page(bh
->b_page
);
4764 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4768 mutex_unlock(&inode
->i_mutex
);
4771 if (inode
->i_size
< off
+ len
) {
4772 i_size_write(inode
, off
+ len
);
4773 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4774 ext4_mark_inode_dirty(handle
, inode
);
4776 mutex_unlock(&inode
->i_mutex
);
4782 static struct dentry
*ext4_mount(struct file_system_type
*fs_type
, int flags
,
4783 const char *dev_name
, void *data
)
4785 return mount_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
);
4788 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4789 static inline void register_as_ext2(void)
4791 int err
= register_filesystem(&ext2_fs_type
);
4794 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4797 static inline void unregister_as_ext2(void)
4799 unregister_filesystem(&ext2_fs_type
);
4802 static inline int ext2_feature_set_ok(struct super_block
*sb
)
4804 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT2_FEATURE_INCOMPAT_SUPP
))
4806 if (sb
->s_flags
& MS_RDONLY
)
4808 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT2_FEATURE_RO_COMPAT_SUPP
))
4812 MODULE_ALIAS("ext2");
4814 static inline void register_as_ext2(void) { }
4815 static inline void unregister_as_ext2(void) { }
4816 static inline int ext2_feature_set_ok(struct super_block
*sb
) { return 0; }
4819 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4820 static inline void register_as_ext3(void)
4822 int err
= register_filesystem(&ext3_fs_type
);
4825 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4828 static inline void unregister_as_ext3(void)
4830 unregister_filesystem(&ext3_fs_type
);
4833 static inline int ext3_feature_set_ok(struct super_block
*sb
)
4835 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT3_FEATURE_INCOMPAT_SUPP
))
4837 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
4839 if (sb
->s_flags
& MS_RDONLY
)
4841 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT3_FEATURE_RO_COMPAT_SUPP
))
4845 MODULE_ALIAS("ext3");
4847 static inline void register_as_ext3(void) { }
4848 static inline void unregister_as_ext3(void) { }
4849 static inline int ext3_feature_set_ok(struct super_block
*sb
) { return 0; }
4852 static struct file_system_type ext4_fs_type
= {
4853 .owner
= THIS_MODULE
,
4855 .mount
= ext4_mount
,
4856 .kill_sb
= kill_block_super
,
4857 .fs_flags
= FS_REQUIRES_DEV
,
4860 static int __init
ext4_init_feat_adverts(void)
4862 struct ext4_features
*ef
;
4865 ef
= kzalloc(sizeof(struct ext4_features
), GFP_KERNEL
);
4869 ef
->f_kobj
.kset
= ext4_kset
;
4870 init_completion(&ef
->f_kobj_unregister
);
4871 ret
= kobject_init_and_add(&ef
->f_kobj
, &ext4_feat_ktype
, NULL
,
4884 static void ext4_exit_feat_adverts(void)
4886 kobject_put(&ext4_feat
->f_kobj
);
4887 wait_for_completion(&ext4_feat
->f_kobj_unregister
);
4891 /* Shared across all ext4 file systems */
4892 wait_queue_head_t ext4__ioend_wq
[EXT4_WQ_HASH_SZ
];
4893 struct mutex ext4__aio_mutex
[EXT4_WQ_HASH_SZ
];
4895 static int __init
ext4_init_fs(void)
4899 ext4_check_flag_values();
4901 for (i
= 0; i
< EXT4_WQ_HASH_SZ
; i
++) {
4902 mutex_init(&ext4__aio_mutex
[i
]);
4903 init_waitqueue_head(&ext4__ioend_wq
[i
]);
4906 err
= ext4_init_pageio();
4909 err
= ext4_init_system_zone();
4912 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4915 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4917 err
= ext4_init_feat_adverts();
4921 err
= ext4_init_mballoc();
4925 err
= ext4_init_xattr();
4928 err
= init_inodecache();
4933 err
= register_filesystem(&ext4_fs_type
);
4937 ext4_li_info
= NULL
;
4938 mutex_init(&ext4_li_mtx
);
4941 unregister_as_ext2();
4942 unregister_as_ext3();
4943 destroy_inodecache();
4947 ext4_exit_mballoc();
4949 ext4_exit_feat_adverts();
4952 remove_proc_entry("fs/ext4", NULL
);
4953 kset_unregister(ext4_kset
);
4955 ext4_exit_system_zone();
4961 static void __exit
ext4_exit_fs(void)
4963 ext4_destroy_lazyinit_thread();
4964 unregister_as_ext2();
4965 unregister_as_ext3();
4966 unregister_filesystem(&ext4_fs_type
);
4967 destroy_inodecache();
4969 ext4_exit_mballoc();
4970 ext4_exit_feat_adverts();
4971 remove_proc_entry("fs/ext4", NULL
);
4972 kset_unregister(ext4_kset
);
4973 ext4_exit_system_zone();
4977 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4978 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4979 MODULE_LICENSE("GPL");
4980 module_init(ext4_init_fs
)
4981 module_exit(ext4_exit_fs
)