2 * super.c - NILFS module and super block management.
4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 * Written by Ryusuke Konishi <ryusuke@osrg.net>
23 * linux/fs/ext2/super.c
25 * Copyright (C) 1992, 1993, 1994, 1995
26 * Remy Card (card@masi.ibp.fr)
27 * Laboratoire MASI - Institut Blaise Pascal
28 * Universite Pierre et Marie Curie (Paris VI)
32 * linux/fs/minix/inode.c
34 * Copyright (C) 1991, 1992 Linus Torvalds
36 * Big-endian to little-endian byte-swapping/bitmaps by
37 * David S. Miller (davem@caip.rutgers.edu), 1995
40 #include <linux/module.h>
41 #include <linux/string.h>
42 #include <linux/slab.h>
43 #include <linux/init.h>
44 #include <linux/blkdev.h>
45 #include <linux/parser.h>
46 #include <linux/random.h>
47 #include <linux/crc32.h>
48 #include <linux/smp_lock.h>
49 #include <linux/vfs.h>
50 #include <linux/writeback.h>
51 #include <linux/kobject.h>
52 #include <linux/exportfs.h>
63 MODULE_AUTHOR("NTT Corp.");
64 MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
66 MODULE_VERSION(NILFS_VERSION
);
67 MODULE_LICENSE("GPL");
69 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
);
70 static int test_exclusive_mount(struct file_system_type
*fs_type
,
71 struct block_device
*bdev
, int flags
);
74 * nilfs_error() - report failure condition on a filesystem
76 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
77 * reporting an error message. It should be called when NILFS detects
78 * incoherences or defects of meta data on disk. As for sustainable
79 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
80 * function should be used instead.
82 * The segment constructor must not call this function because it can
85 void nilfs_error(struct super_block
*sb
, const char *function
,
88 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
92 printk(KERN_CRIT
"NILFS error (device %s): %s: ", sb
->s_id
, function
);
97 if (!(sb
->s_flags
& MS_RDONLY
)) {
98 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
100 if (!nilfs_test_opt(sbi
, ERRORS_CONT
))
101 nilfs_detach_segment_constructor(sbi
);
103 down_write(&nilfs
->ns_sem
);
104 if (!(nilfs
->ns_mount_state
& NILFS_ERROR_FS
)) {
105 nilfs
->ns_mount_state
|= NILFS_ERROR_FS
;
106 nilfs
->ns_sbp
[0]->s_state
|=
107 cpu_to_le16(NILFS_ERROR_FS
);
108 nilfs_commit_super(sbi
, 1);
110 up_write(&nilfs
->ns_sem
);
112 if (nilfs_test_opt(sbi
, ERRORS_RO
)) {
113 printk(KERN_CRIT
"Remounting filesystem read-only\n");
114 sb
->s_flags
|= MS_RDONLY
;
118 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
119 panic("NILFS (device %s): panic forced after error\n",
123 void nilfs_warning(struct super_block
*sb
, const char *function
,
124 const char *fmt
, ...)
129 printk(KERN_WARNING
"NILFS warning (device %s): %s: ",
136 static struct kmem_cache
*nilfs_inode_cachep
;
138 struct inode
*nilfs_alloc_inode(struct super_block
*sb
)
140 struct nilfs_inode_info
*ii
;
142 ii
= kmem_cache_alloc(nilfs_inode_cachep
, GFP_NOFS
);
147 ii
->vfs_inode
.i_version
= 1;
148 nilfs_btnode_cache_init(&ii
->i_btnode_cache
);
149 return &ii
->vfs_inode
;
152 void nilfs_destroy_inode(struct inode
*inode
)
154 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
157 static void init_once(void *obj
)
159 struct nilfs_inode_info
*ii
= obj
;
161 INIT_LIST_HEAD(&ii
->i_dirty
);
162 #ifdef CONFIG_NILFS_XATTR
163 init_rwsem(&ii
->xattr_sem
);
165 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
166 ii
->i_bmap
= (struct nilfs_bmap
*)&ii
->i_bmap_union
;
167 inode_init_once(&ii
->vfs_inode
);
170 static int nilfs_init_inode_cache(void)
172 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
173 sizeof(struct nilfs_inode_info
),
174 0, SLAB_RECLAIM_ACCOUNT
,
177 return (nilfs_inode_cachep
== NULL
) ? -ENOMEM
: 0;
180 static inline void nilfs_destroy_inode_cache(void)
182 kmem_cache_destroy(nilfs_inode_cachep
);
185 static void nilfs_clear_inode(struct inode
*inode
)
187 struct nilfs_inode_info
*ii
= NILFS_I(inode
);
189 #ifdef CONFIG_NILFS_POSIX_ACL
190 if (ii
->i_acl
&& ii
->i_acl
!= NILFS_ACL_NOT_CACHED
) {
191 posix_acl_release(ii
->i_acl
);
192 ii
->i_acl
= NILFS_ACL_NOT_CACHED
;
194 if (ii
->i_default_acl
&& ii
->i_default_acl
!= NILFS_ACL_NOT_CACHED
) {
195 posix_acl_release(ii
->i_default_acl
);
196 ii
->i_default_acl
= NILFS_ACL_NOT_CACHED
;
200 * Free resources allocated in nilfs_read_inode(), here.
202 BUG_ON(!list_empty(&ii
->i_dirty
));
206 if (test_bit(NILFS_I_BMAP
, &ii
->i_state
))
207 nilfs_bmap_clear(ii
->i_bmap
);
209 nilfs_btnode_cache_clear(&ii
->i_btnode_cache
);
212 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
, int dupsb
)
214 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
216 int barrier_done
= 0;
218 if (nilfs_test_opt(sbi
, BARRIER
)) {
219 set_buffer_ordered(nilfs
->ns_sbh
[0]);
223 set_buffer_dirty(nilfs
->ns_sbh
[0]);
224 err
= sync_dirty_buffer(nilfs
->ns_sbh
[0]);
225 if (err
== -EOPNOTSUPP
&& barrier_done
) {
226 nilfs_warning(sbi
->s_super
, __func__
,
227 "barrier-based sync failed. "
228 "disabling barriers\n");
229 nilfs_clear_opt(sbi
, BARRIER
);
231 clear_buffer_ordered(nilfs
->ns_sbh
[0]);
236 "NILFS: unable to write superblock (err=%d)\n", err
);
237 if (err
== -EIO
&& nilfs
->ns_sbh
[1]) {
238 nilfs_fall_back_super_block(nilfs
);
242 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
245 * The latest segment becomes trailable from the position
246 * written in superblock.
248 clear_nilfs_discontinued(nilfs
);
250 /* update GC protection for recent segments */
251 if (nilfs
->ns_sbh
[1]) {
254 set_buffer_dirty(nilfs
->ns_sbh
[1]);
255 if (!sync_dirty_buffer(nilfs
->ns_sbh
[1]))
256 sbp
= nilfs
->ns_sbp
[1];
260 spin_lock(&nilfs
->ns_last_segment_lock
);
261 nilfs
->ns_prot_seq
= le64_to_cpu(sbp
->s_last_seq
);
262 spin_unlock(&nilfs
->ns_last_segment_lock
);
269 int nilfs_commit_super(struct nilfs_sb_info
*sbi
, int dupsb
)
271 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
272 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
273 sector_t nfreeblocks
;
277 /* nilfs->sem must be locked by the caller. */
278 if (sbp
[0]->s_magic
!= NILFS_SUPER_MAGIC
) {
279 if (sbp
[1] && sbp
[1]->s_magic
== NILFS_SUPER_MAGIC
)
280 nilfs_swap_super_block(nilfs
);
282 printk(KERN_CRIT
"NILFS: superblock broke on dev %s\n",
287 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
289 printk(KERN_ERR
"NILFS: failed to count free blocks\n");
292 spin_lock(&nilfs
->ns_last_segment_lock
);
293 sbp
[0]->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
294 sbp
[0]->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
295 sbp
[0]->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
296 spin_unlock(&nilfs
->ns_last_segment_lock
);
299 nilfs
->ns_sbwtime
[0] = t
;
300 sbp
[0]->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
301 sbp
[0]->s_wtime
= cpu_to_le64(t
);
303 sbp
[0]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
304 (unsigned char *)sbp
[0],
306 if (dupsb
&& sbp
[1]) {
307 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
308 nilfs
->ns_sbwtime
[1] = t
;
310 sbi
->s_super
->s_dirt
= 0;
311 return nilfs_sync_super(sbi
, dupsb
);
314 static void nilfs_put_super(struct super_block
*sb
)
316 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
317 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
319 nilfs_detach_segment_constructor(sbi
);
321 if (!(sb
->s_flags
& MS_RDONLY
)) {
322 down_write(&nilfs
->ns_sem
);
323 nilfs
->ns_sbp
[0]->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
324 nilfs_commit_super(sbi
, 1);
325 up_write(&nilfs
->ns_sem
);
328 nilfs_detach_checkpoint(sbi
);
329 put_nilfs(sbi
->s_nilfs
);
331 sb
->s_fs_info
= NULL
;
336 * nilfs_write_super - write super block(s) of NILFS
339 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
340 * clears s_dirt. This function is called in the section protected by
343 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
344 * of the struct the_nilfs. Lock order must be as follows:
347 * 2. down_write(&nilfs->ns_sem)
349 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
350 * of the super block (nilfs->ns_sbp[]).
352 * In most cases, VFS functions call lock_super() before calling these
353 * methods. So we must be careful not to bring on deadlocks when using
354 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
356 * Note that order of lock_kernel() and lock_super() depends on contexts
357 * of VFS. We should also note that lock_kernel() can be used in its
358 * protective section and only the outermost one has an effect.
360 static void nilfs_write_super(struct super_block
*sb
)
362 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
363 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
365 down_write(&nilfs
->ns_sem
);
366 if (!(sb
->s_flags
& MS_RDONLY
)) {
367 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
368 u64 t
= get_seconds();
371 if (!nilfs_discontinued(nilfs
) && t
>= nilfs
->ns_sbwtime
[0] &&
372 t
< nilfs
->ns_sbwtime
[0] + NILFS_SB_FREQ
) {
373 up_write(&nilfs
->ns_sem
);
376 dupsb
= sbp
[1] && t
> nilfs
->ns_sbwtime
[1] + NILFS_ALTSB_FREQ
;
377 nilfs_commit_super(sbi
, dupsb
);
380 up_write(&nilfs
->ns_sem
);
383 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
387 /* This function is called when super block should be written back */
389 err
= nilfs_construct_segment(sb
);
393 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
395 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
396 struct nilfs_checkpoint
*raw_cp
;
397 struct buffer_head
*bh_cp
;
400 down_write(&nilfs
->ns_sem
);
401 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
402 up_write(&nilfs
->ns_sem
);
404 sbi
->s_ifile
= nilfs_mdt_new(
405 nilfs
, sbi
->s_super
, NILFS_IFILE_INO
, NILFS_IFILE_GFP
);
409 err
= nilfs_palloc_init_blockgroup(sbi
->s_ifile
, nilfs
->ns_inode_size
);
413 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
416 if (err
== -ENOENT
|| err
== -EINVAL
) {
418 "NILFS: Invalid checkpoint "
419 "(checkpoint number=%llu)\n",
420 (unsigned long long)cno
);
425 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
428 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
429 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
431 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
435 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
437 nilfs_mdt_destroy(sbi
->s_ifile
);
440 down_write(&nilfs
->ns_sem
);
441 list_del_init(&sbi
->s_list
);
442 up_write(&nilfs
->ns_sem
);
447 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
449 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
451 nilfs_mdt_clear(sbi
->s_ifile
);
452 nilfs_mdt_destroy(sbi
->s_ifile
);
454 down_write(&nilfs
->ns_sem
);
455 list_del_init(&sbi
->s_list
);
456 up_write(&nilfs
->ns_sem
);
459 static int nilfs_mark_recovery_complete(struct nilfs_sb_info
*sbi
)
461 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
464 down_write(&nilfs
->ns_sem
);
465 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
466 nilfs
->ns_mount_state
|= NILFS_VALID_FS
;
467 err
= nilfs_commit_super(sbi
, 1);
469 printk(KERN_INFO
"NILFS: recovery complete.\n");
471 up_write(&nilfs
->ns_sem
);
475 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
477 struct super_block
*sb
= dentry
->d_sb
;
478 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
479 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
480 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
481 unsigned long long blocks
;
482 unsigned long overhead
;
483 unsigned long nrsvblocks
;
484 sector_t nfreeblocks
;
488 * Compute all of the segment blocks
490 * The blocks before first segment and after last segment
493 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
494 - nilfs
->ns_first_data_block
;
495 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
498 * Compute the overhead
500 * When distributing meta data blocks outside semgent structure,
501 * We must count them as the overhead.
505 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
509 buf
->f_type
= NILFS_SUPER_MAGIC
;
510 buf
->f_bsize
= sb
->s_blocksize
;
511 buf
->f_blocks
= blocks
- overhead
;
512 buf
->f_bfree
= nfreeblocks
;
513 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
514 (buf
->f_bfree
- nrsvblocks
) : 0;
515 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
516 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
517 buf
->f_namelen
= NILFS_NAME_LEN
;
518 buf
->f_fsid
.val
[0] = (u32
)id
;
519 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
524 static struct super_operations nilfs_sops
= {
525 .alloc_inode
= nilfs_alloc_inode
,
526 .destroy_inode
= nilfs_destroy_inode
,
527 .dirty_inode
= nilfs_dirty_inode
,
528 /* .write_inode = nilfs_write_inode, */
529 /* .put_inode = nilfs_put_inode, */
530 /* .drop_inode = nilfs_drop_inode, */
531 .delete_inode
= nilfs_delete_inode
,
532 .put_super
= nilfs_put_super
,
533 .write_super
= nilfs_write_super
,
534 .sync_fs
= nilfs_sync_fs
,
535 /* .write_super_lockfs */
537 .statfs
= nilfs_statfs
,
538 .remount_fs
= nilfs_remount
,
539 .clear_inode
= nilfs_clear_inode
,
544 static struct inode
*
545 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
549 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
550 ino
!= NILFS_SKETCH_INO
)
551 return ERR_PTR(-ESTALE
);
553 inode
= nilfs_iget(sb
, ino
);
555 return ERR_CAST(inode
);
556 if (generation
&& inode
->i_generation
!= generation
) {
558 return ERR_PTR(-ESTALE
);
564 static struct dentry
*
565 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
568 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
569 nilfs_nfs_get_inode
);
572 static struct dentry
*
573 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
576 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
577 nilfs_nfs_get_inode
);
580 static struct export_operations nilfs_export_ops
= {
581 .fh_to_dentry
= nilfs_fh_to_dentry
,
582 .fh_to_parent
= nilfs_fh_to_parent
,
583 .get_parent
= nilfs_get_parent
,
587 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
588 Opt_barrier
, Opt_snapshot
, Opt_order
,
592 static match_table_t tokens
= {
593 {Opt_err_cont
, "errors=continue"},
594 {Opt_err_panic
, "errors=panic"},
595 {Opt_err_ro
, "errors=remount-ro"},
596 {Opt_barrier
, "barrier=%s"},
597 {Opt_snapshot
, "cp=%u"},
598 {Opt_order
, "order=%s"},
602 static int match_bool(substring_t
*s
, int *result
)
604 int len
= s
->to
- s
->from
;
606 if (strncmp(s
->from
, "on", len
) == 0)
608 else if (strncmp(s
->from
, "off", len
) == 0)
615 static int parse_options(char *options
, struct super_block
*sb
)
617 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
619 substring_t args
[MAX_OPT_ARGS
];
625 while ((p
= strsep(&options
, ",")) != NULL
) {
630 token
= match_token(p
, tokens
, args
);
633 if (match_bool(&args
[0], &option
))
636 nilfs_set_opt(sbi
, BARRIER
);
638 nilfs_clear_opt(sbi
, BARRIER
);
641 if (strcmp(args
[0].from
, "relaxed") == 0)
642 /* Ordered data semantics */
643 nilfs_clear_opt(sbi
, STRICT_ORDER
);
644 else if (strcmp(args
[0].from
, "strict") == 0)
645 /* Strict in-order semantics */
646 nilfs_set_opt(sbi
, STRICT_ORDER
);
651 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
654 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
657 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
660 if (match_int(&args
[0], &option
) || option
<= 0)
662 if (!(sb
->s_flags
& MS_RDONLY
))
664 sbi
->s_snapshot_cno
= option
;
665 nilfs_set_opt(sbi
, SNAPSHOT
);
669 "NILFS: Unrecognized mount option \"%s\"\n", p
);
677 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
678 struct nilfs_super_block
*sbp
)
681 NILFS_MOUNT_ERRORS_CONT
| NILFS_MOUNT_BARRIER
;
684 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
686 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
687 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
688 int max_mnt_count
= le16_to_cpu(sbp
->s_max_mnt_count
);
689 int mnt_count
= le16_to_cpu(sbp
->s_mnt_count
);
691 /* nilfs->sem must be locked by the caller. */
692 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
693 printk(KERN_WARNING
"NILFS warning: mounting unchecked fs\n");
694 } else if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
696 "NILFS warning: mounting fs with errors\n");
698 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
700 "NILFS warning: maximal mount count reached\n");
704 sbp
->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
706 sbp
->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
707 sbp
->s_state
= cpu_to_le16(le16_to_cpu(sbp
->s_state
) & ~NILFS_VALID_FS
);
708 sbp
->s_mtime
= cpu_to_le64(get_seconds());
709 return nilfs_commit_super(sbi
, 1);
712 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
713 u64 pos
, int blocksize
,
714 struct buffer_head
**pbh
)
716 unsigned long long sb_index
= pos
;
717 unsigned long offset
;
719 offset
= do_div(sb_index
, blocksize
);
720 *pbh
= sb_bread(sb
, sb_index
);
723 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
726 int nilfs_store_magic_and_option(struct super_block
*sb
,
727 struct nilfs_super_block
*sbp
,
730 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
732 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
734 /* FS independent flags */
735 #ifdef NILFS_ATIME_DISABLE
736 sb
->s_flags
|= MS_NOATIME
;
739 nilfs_set_default_options(sbi
, sbp
);
741 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
742 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
743 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
744 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
746 return !parse_options(data
, sb
) ? -EINVAL
: 0 ;
750 * nilfs_fill_super() - initialize a super block instance
752 * @data: mount options
753 * @silent: silent mode flag
754 * @nilfs: the_nilfs struct
756 * This function is called exclusively by bd_mount_mutex.
757 * So, the recovery process is protected from other simultaneous mounts.
760 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
761 struct the_nilfs
*nilfs
)
763 struct nilfs_sb_info
*sbi
;
768 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
775 sbi
->s_nilfs
= nilfs
;
778 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
782 spin_lock_init(&sbi
->s_inode_lock
);
783 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
784 INIT_LIST_HEAD(&sbi
->s_list
);
787 * Following initialization is overlapped because
788 * nilfs_sb_info structure has been cleared at the beginning.
789 * But we reserve them to keep our interest and make ready
790 * for the future change.
792 get_random_bytes(&sbi
->s_next_generation
,
793 sizeof(sbi
->s_next_generation
));
794 spin_lock_init(&sbi
->s_next_gen_lock
);
796 sb
->s_op
= &nilfs_sops
;
797 sb
->s_export_op
= &nilfs_export_ops
;
801 if (!nilfs_loaded(nilfs
)) {
802 err
= load_nilfs(nilfs
, sbi
);
806 cno
= nilfs_last_cno(nilfs
);
808 if (sb
->s_flags
& MS_RDONLY
) {
809 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
810 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
811 sbi
->s_snapshot_cno
);
816 "NILFS: The specified checkpoint is "
818 "(checkpoint number=%llu).\n",
819 (unsigned long long)sbi
->s_snapshot_cno
);
823 cno
= sbi
->s_snapshot_cno
;
825 /* Read-only mount */
826 sbi
->s_snapshot_cno
= cno
;
829 err
= nilfs_attach_checkpoint(sbi
, cno
);
831 printk(KERN_ERR
"NILFS: error loading a checkpoint"
832 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
836 if (!(sb
->s_flags
& MS_RDONLY
)) {
837 err
= nilfs_attach_segment_constructor(sbi
);
839 goto failed_checkpoint
;
842 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
844 printk(KERN_ERR
"NILFS: get root inode failed\n");
848 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
850 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
854 sb
->s_root
= d_alloc_root(root
);
857 printk(KERN_ERR
"NILFS: get root dentry failed\n");
862 if (!(sb
->s_flags
& MS_RDONLY
)) {
863 down_write(&nilfs
->ns_sem
);
864 nilfs_setup_super(sbi
);
865 up_write(&nilfs
->ns_sem
);
868 err
= nilfs_mark_recovery_complete(sbi
);
870 printk(KERN_ERR
"NILFS: recovery failed.\n");
881 nilfs_detach_segment_constructor(sbi
);
884 nilfs_detach_checkpoint(sbi
);
888 sb
->s_fs_info
= NULL
;
893 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
895 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
896 struct nilfs_super_block
*sbp
;
897 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
898 unsigned long old_sb_flags
;
899 struct nilfs_mount_options old_opts
;
902 old_sb_flags
= sb
->s_flags
;
903 old_opts
.mount_opt
= sbi
->s_mount_opt
;
904 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
906 if (!parse_options(data
, sb
)) {
910 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
912 if ((*flags
& MS_RDONLY
) &&
913 sbi
->s_snapshot_cno
!= old_opts
.snapshot_cno
) {
914 printk(KERN_WARNING
"NILFS (device %s): couldn't "
915 "remount to a different snapshot. \n",
921 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
923 if (*flags
& MS_RDONLY
) {
924 /* Shutting down the segment constructor */
925 nilfs_detach_segment_constructor(sbi
);
926 sb
->s_flags
|= MS_RDONLY
;
928 sbi
->s_snapshot_cno
= nilfs_last_cno(nilfs
);
929 /* nilfs_set_opt(sbi, SNAPSHOT); */
932 * Remounting a valid RW partition RDONLY, so set
933 * the RDONLY flag and then mark the partition as valid again.
935 down_write(&nilfs
->ns_sem
);
936 sbp
= nilfs
->ns_sbp
[0];
937 if (!(sbp
->s_state
& le16_to_cpu(NILFS_VALID_FS
)) &&
938 (nilfs
->ns_mount_state
& NILFS_VALID_FS
))
939 sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
940 sbp
->s_mtime
= cpu_to_le64(get_seconds());
941 nilfs_commit_super(sbi
, 1);
942 up_write(&nilfs
->ns_sem
);
945 * Mounting a RDONLY partition read-write, so reread and
946 * store the current valid flag. (It may have been changed
947 * by fsck since we originally mounted the partition.)
949 down(&sb
->s_bdev
->bd_mount_sem
);
950 /* Check existing RW-mount */
951 if (test_exclusive_mount(sb
->s_type
, sb
->s_bdev
, 0)) {
952 printk(KERN_WARNING
"NILFS (device %s): couldn't "
953 "remount because a RW-mount exists.\n",
956 goto rw_remount_failed
;
958 if (sbi
->s_snapshot_cno
!= nilfs_last_cno(nilfs
)) {
959 printk(KERN_WARNING
"NILFS (device %s): couldn't "
960 "remount because the current RO-mount is not "
964 goto rw_remount_failed
;
966 sb
->s_flags
&= ~MS_RDONLY
;
967 nilfs_clear_opt(sbi
, SNAPSHOT
);
968 sbi
->s_snapshot_cno
= 0;
970 err
= nilfs_attach_segment_constructor(sbi
);
972 goto rw_remount_failed
;
974 down_write(&nilfs
->ns_sem
);
975 nilfs_setup_super(sbi
);
976 up_write(&nilfs
->ns_sem
);
978 up(&sb
->s_bdev
->bd_mount_sem
);
984 up(&sb
->s_bdev
->bd_mount_sem
);
986 sb
->s_flags
= old_sb_flags
;
987 sbi
->s_mount_opt
= old_opts
.mount_opt
;
988 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
992 struct nilfs_super_data
{
993 struct block_device
*bdev
;
999 * nilfs_identify - pre-read mount options needed to identify mount instance
1000 * @data: mount options
1001 * @sd: nilfs_super_data
1003 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1005 char *p
, *options
= data
;
1006 substring_t args
[MAX_OPT_ARGS
];
1011 p
= strsep(&options
, ",");
1012 if (p
!= NULL
&& *p
) {
1013 token
= match_token(p
, tokens
, args
);
1014 if (token
== Opt_snapshot
) {
1015 if (!(sd
->flags
& MS_RDONLY
))
1018 ret
= match_int(&args
[0], &option
);
1029 "NILFS: invalid mount option: %s\n", p
);
1033 BUG_ON(options
== data
);
1034 *(options
- 1) = ',';
1039 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1041 struct nilfs_super_data
*sd
= data
;
1043 s
->s_bdev
= sd
->bdev
;
1044 s
->s_dev
= s
->s_bdev
->bd_dev
;
1048 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1050 struct nilfs_super_data
*sd
= data
;
1052 return s
->s_bdev
== sd
->bdev
;
1055 static int nilfs_test_bdev_super2(struct super_block
*s
, void *data
)
1057 struct nilfs_super_data
*sd
= data
;
1060 if (s
->s_bdev
!= sd
->bdev
)
1063 if (!((s
->s_flags
| sd
->flags
) & MS_RDONLY
))
1064 return 1; /* Reuse an old R/W-mode super_block */
1066 if (s
->s_flags
& sd
->flags
& MS_RDONLY
) {
1067 if (down_read_trylock(&s
->s_umount
)) {
1069 (sd
->cno
== NILFS_SB(s
)->s_snapshot_cno
);
1070 up_read(&s
->s_umount
);
1072 * This path is locked with sb_lock by sget().
1073 * So, drop_super() causes deadlock.
1082 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1083 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1085 struct nilfs_super_data sd
;
1086 struct super_block
*s
, *s2
;
1087 struct the_nilfs
*nilfs
= NULL
;
1088 int err
, need_to_close
= 1;
1090 sd
.bdev
= open_bdev_exclusive(dev_name
, flags
, fs_type
);
1091 if (IS_ERR(sd
.bdev
))
1092 return PTR_ERR(sd
.bdev
);
1095 * To get mount instance using sget() vfs-routine, NILFS needs
1096 * much more information than normal filesystems to identify mount
1097 * instance. For snapshot mounts, not only a mount type (ro-mount
1098 * or rw-mount) but also a checkpoint number is required.
1099 * The results are passed in sget() using nilfs_super_data.
1103 if (nilfs_identify((char *)data
, &sd
)) {
1109 * once the super is inserted into the list by sget, s_umount
1110 * will protect the lockfs code from trying to start a snapshot
1111 * while we are mounting
1113 down(&sd
.bdev
->bd_mount_sem
);
1115 (err
= test_exclusive_mount(fs_type
, sd
.bdev
, flags
^ MS_RDONLY
))) {
1116 err
= (err
< 0) ? : -EBUSY
;
1121 * Phase-1: search any existent instance and get the_nilfs
1123 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1129 nilfs
= alloc_nilfs(sd
.bdev
);
1133 struct nilfs_sb_info
*sbi
= NILFS_SB(s
);
1136 * s_umount protects super_block from unmount process;
1137 * It covers pointers of nilfs_sb_info and the_nilfs.
1139 nilfs
= sbi
->s_nilfs
;
1141 up_write(&s
->s_umount
);
1144 * Phase-2: search specified snapshot or R/W mode super_block
1147 /* trying to get the latest checkpoint. */
1148 sd
.cno
= nilfs_last_cno(nilfs
);
1150 s2
= sget(fs_type
, nilfs_test_bdev_super2
,
1151 nilfs_set_bdev_super
, &sd
);
1152 deactivate_super(s
);
1154 * Although deactivate_super() invokes close_bdev_exclusive() at
1155 * kill_block_super(). Here, s is an existent mount; we need
1156 * one more close_bdev_exclusive() call.
1164 char b
[BDEVNAME_SIZE
];
1167 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1168 sb_set_blocksize(s
, block_size(sd
.bdev
));
1170 err
= nilfs_fill_super(s
, data
, flags
& MS_VERBOSE
, nilfs
);
1174 s
->s_flags
|= MS_ACTIVE
;
1176 } else if (!(s
->s_flags
& MS_RDONLY
)) {
1180 up(&sd
.bdev
->bd_mount_sem
);
1183 close_bdev_exclusive(sd
.bdev
, flags
);
1184 simple_set_mnt(mnt
, s
);
1188 up(&sd
.bdev
->bd_mount_sem
);
1191 close_bdev_exclusive(sd
.bdev
, flags
);
1195 up(&sd
.bdev
->bd_mount_sem
);
1197 close_bdev_exclusive(sd
.bdev
, flags
);
1202 /* Abandoning the newly allocated superblock */
1203 up(&sd
.bdev
->bd_mount_sem
);
1206 up_write(&s
->s_umount
);
1207 deactivate_super(s
);
1209 * deactivate_super() invokes close_bdev_exclusive().
1210 * We must finish all post-cleaning before this call;
1211 * put_nilfs() and unlocking bd_mount_sem need the block device.
1216 static int nilfs_test_bdev_super3(struct super_block
*s
, void *data
)
1218 struct nilfs_super_data
*sd
= data
;
1221 if (s
->s_bdev
!= sd
->bdev
)
1223 if (down_read_trylock(&s
->s_umount
)) {
1224 ret
= (s
->s_flags
& MS_RDONLY
) && s
->s_root
&&
1225 nilfs_test_opt(NILFS_SB(s
), SNAPSHOT
);
1226 up_read(&s
->s_umount
);
1228 return 0; /* ignore snapshot mounts */
1230 return !((sd
->flags
^ s
->s_flags
) & MS_RDONLY
);
1233 static int __false_bdev_super(struct super_block
*s
, void *data
)
1235 #if 0 /* XXX: workaround for lock debug. This is not good idea */
1236 up_write(&s
->s_umount
);
1242 * test_exclusive_mount - check whether an exclusive RW/RO mount exists or not.
1243 * fs_type: filesystem type
1244 * bdev: block device
1245 * flag: 0 (check rw-mount) or MS_RDONLY (check ro-mount)
1246 * res: pointer to an integer to store result
1248 * This function must be called within a section protected by bd_mount_mutex.
1250 static int test_exclusive_mount(struct file_system_type
*fs_type
,
1251 struct block_device
*bdev
, int flags
)
1253 struct super_block
*s
;
1254 struct nilfs_super_data sd
= { .flags
= flags
, .bdev
= bdev
};
1256 s
= sget(fs_type
, nilfs_test_bdev_super3
, __false_bdev_super
, &sd
);
1258 if (PTR_ERR(s
) != -EFAULT
)
1260 return 0; /* Not found */
1262 up_write(&s
->s_umount
);
1263 deactivate_super(s
);
1264 return 1; /* Found */
1267 struct file_system_type nilfs_fs_type
= {
1268 .owner
= THIS_MODULE
,
1270 .get_sb
= nilfs_get_sb
,
1271 .kill_sb
= kill_block_super
,
1272 .fs_flags
= FS_REQUIRES_DEV
,
1275 static int __init
init_nilfs_fs(void)
1279 err
= nilfs_init_inode_cache();
1283 err
= nilfs_init_transaction_cache();
1285 goto failed_inode_cache
;
1287 err
= nilfs_init_segbuf_cache();
1289 goto failed_transaction_cache
;
1291 err
= nilfs_btree_path_cache_init();
1293 goto failed_segbuf_cache
;
1295 err
= register_filesystem(&nilfs_fs_type
);
1297 goto failed_btree_path_cache
;
1301 failed_btree_path_cache
:
1302 nilfs_btree_path_cache_destroy();
1304 failed_segbuf_cache
:
1305 nilfs_destroy_segbuf_cache();
1307 failed_transaction_cache
:
1308 nilfs_destroy_transaction_cache();
1311 nilfs_destroy_inode_cache();
1317 static void __exit
exit_nilfs_fs(void)
1319 nilfs_destroy_segbuf_cache();
1320 nilfs_destroy_transaction_cache();
1321 nilfs_destroy_inode_cache();
1322 nilfs_btree_path_cache_destroy();
1323 unregister_filesystem(&nilfs_fs_type
);
1326 module_init(init_nilfs_fs
)
1327 module_exit(exit_nilfs_fs
)