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_LICENSE("GPL");
68 static void nilfs_write_super(struct super_block
*sb
);
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
;
322 nilfs_write_super(sb
);
324 nilfs_detach_segment_constructor(sbi
);
326 if (!(sb
->s_flags
& MS_RDONLY
)) {
327 down_write(&nilfs
->ns_sem
);
328 nilfs
->ns_sbp
[0]->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
329 nilfs_commit_super(sbi
, 1);
330 up_write(&nilfs
->ns_sem
);
333 nilfs_detach_checkpoint(sbi
);
334 put_nilfs(sbi
->s_nilfs
);
336 sb
->s_fs_info
= NULL
;
343 * nilfs_write_super - write super block(s) of NILFS
346 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
347 * clears s_dirt. This function is called in the section protected by
350 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
351 * of the struct the_nilfs. Lock order must be as follows:
354 * 2. down_write(&nilfs->ns_sem)
356 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
357 * of the super block (nilfs->ns_sbp[]).
359 * In most cases, VFS functions call lock_super() before calling these
360 * methods. So we must be careful not to bring on deadlocks when using
361 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
363 * Note that order of lock_kernel() and lock_super() depends on contexts
364 * of VFS. We should also note that lock_kernel() can be used in its
365 * protective section and only the outermost one has an effect.
367 static void nilfs_write_super(struct super_block
*sb
)
369 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
370 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
372 down_write(&nilfs
->ns_sem
);
373 if (!(sb
->s_flags
& MS_RDONLY
)) {
374 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
375 u64 t
= get_seconds();
378 if (!nilfs_discontinued(nilfs
) && t
>= nilfs
->ns_sbwtime
[0] &&
379 t
< nilfs
->ns_sbwtime
[0] + NILFS_SB_FREQ
) {
380 up_write(&nilfs
->ns_sem
);
383 dupsb
= sbp
[1] && t
> nilfs
->ns_sbwtime
[1] + NILFS_ALTSB_FREQ
;
384 nilfs_commit_super(sbi
, dupsb
);
387 up_write(&nilfs
->ns_sem
);
390 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
394 /* This function is called when super block should be written back */
396 err
= nilfs_construct_segment(sb
);
400 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
402 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
403 struct nilfs_checkpoint
*raw_cp
;
404 struct buffer_head
*bh_cp
;
407 down_write(&nilfs
->ns_sem
);
408 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
409 up_write(&nilfs
->ns_sem
);
411 sbi
->s_ifile
= nilfs_mdt_new(
412 nilfs
, sbi
->s_super
, NILFS_IFILE_INO
, NILFS_IFILE_GFP
);
416 err
= nilfs_palloc_init_blockgroup(sbi
->s_ifile
, nilfs
->ns_inode_size
);
420 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
423 if (err
== -ENOENT
|| err
== -EINVAL
) {
425 "NILFS: Invalid checkpoint "
426 "(checkpoint number=%llu)\n",
427 (unsigned long long)cno
);
432 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
435 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
436 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
438 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
442 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
444 nilfs_mdt_destroy(sbi
->s_ifile
);
447 down_write(&nilfs
->ns_sem
);
448 list_del_init(&sbi
->s_list
);
449 up_write(&nilfs
->ns_sem
);
454 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
456 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
458 nilfs_mdt_clear(sbi
->s_ifile
);
459 nilfs_mdt_destroy(sbi
->s_ifile
);
461 down_write(&nilfs
->ns_sem
);
462 list_del_init(&sbi
->s_list
);
463 up_write(&nilfs
->ns_sem
);
466 static int nilfs_mark_recovery_complete(struct nilfs_sb_info
*sbi
)
468 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
471 down_write(&nilfs
->ns_sem
);
472 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
473 nilfs
->ns_mount_state
|= NILFS_VALID_FS
;
474 err
= nilfs_commit_super(sbi
, 1);
476 printk(KERN_INFO
"NILFS: recovery complete.\n");
478 up_write(&nilfs
->ns_sem
);
482 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
484 struct super_block
*sb
= dentry
->d_sb
;
485 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
486 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
487 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
488 unsigned long long blocks
;
489 unsigned long overhead
;
490 unsigned long nrsvblocks
;
491 sector_t nfreeblocks
;
495 * Compute all of the segment blocks
497 * The blocks before first segment and after last segment
500 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
501 - nilfs
->ns_first_data_block
;
502 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
505 * Compute the overhead
507 * When distributing meta data blocks outside semgent structure,
508 * We must count them as the overhead.
512 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
516 buf
->f_type
= NILFS_SUPER_MAGIC
;
517 buf
->f_bsize
= sb
->s_blocksize
;
518 buf
->f_blocks
= blocks
- overhead
;
519 buf
->f_bfree
= nfreeblocks
;
520 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
521 (buf
->f_bfree
- nrsvblocks
) : 0;
522 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
523 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
524 buf
->f_namelen
= NILFS_NAME_LEN
;
525 buf
->f_fsid
.val
[0] = (u32
)id
;
526 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
531 static struct super_operations nilfs_sops
= {
532 .alloc_inode
= nilfs_alloc_inode
,
533 .destroy_inode
= nilfs_destroy_inode
,
534 .dirty_inode
= nilfs_dirty_inode
,
535 /* .write_inode = nilfs_write_inode, */
536 /* .put_inode = nilfs_put_inode, */
537 /* .drop_inode = nilfs_drop_inode, */
538 .delete_inode
= nilfs_delete_inode
,
539 .put_super
= nilfs_put_super
,
540 .write_super
= nilfs_write_super
,
541 .sync_fs
= nilfs_sync_fs
,
542 /* .write_super_lockfs */
544 .statfs
= nilfs_statfs
,
545 .remount_fs
= nilfs_remount
,
546 .clear_inode
= nilfs_clear_inode
,
551 static struct inode
*
552 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
556 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
557 ino
!= NILFS_SKETCH_INO
)
558 return ERR_PTR(-ESTALE
);
560 inode
= nilfs_iget(sb
, ino
);
562 return ERR_CAST(inode
);
563 if (generation
&& inode
->i_generation
!= generation
) {
565 return ERR_PTR(-ESTALE
);
571 static struct dentry
*
572 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
575 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
576 nilfs_nfs_get_inode
);
579 static struct dentry
*
580 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
583 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
584 nilfs_nfs_get_inode
);
587 static struct export_operations nilfs_export_ops
= {
588 .fh_to_dentry
= nilfs_fh_to_dentry
,
589 .fh_to_parent
= nilfs_fh_to_parent
,
590 .get_parent
= nilfs_get_parent
,
594 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
595 Opt_barrier
, Opt_snapshot
, Opt_order
,
599 static match_table_t tokens
= {
600 {Opt_err_cont
, "errors=continue"},
601 {Opt_err_panic
, "errors=panic"},
602 {Opt_err_ro
, "errors=remount-ro"},
603 {Opt_barrier
, "barrier=%s"},
604 {Opt_snapshot
, "cp=%u"},
605 {Opt_order
, "order=%s"},
609 static int match_bool(substring_t
*s
, int *result
)
611 int len
= s
->to
- s
->from
;
613 if (strncmp(s
->from
, "on", len
) == 0)
615 else if (strncmp(s
->from
, "off", len
) == 0)
622 static int parse_options(char *options
, struct super_block
*sb
)
624 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
626 substring_t args
[MAX_OPT_ARGS
];
632 while ((p
= strsep(&options
, ",")) != NULL
) {
637 token
= match_token(p
, tokens
, args
);
640 if (match_bool(&args
[0], &option
))
643 nilfs_set_opt(sbi
, BARRIER
);
645 nilfs_clear_opt(sbi
, BARRIER
);
648 if (strcmp(args
[0].from
, "relaxed") == 0)
649 /* Ordered data semantics */
650 nilfs_clear_opt(sbi
, STRICT_ORDER
);
651 else if (strcmp(args
[0].from
, "strict") == 0)
652 /* Strict in-order semantics */
653 nilfs_set_opt(sbi
, STRICT_ORDER
);
658 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
661 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
664 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
667 if (match_int(&args
[0], &option
) || option
<= 0)
669 if (!(sb
->s_flags
& MS_RDONLY
))
671 sbi
->s_snapshot_cno
= option
;
672 nilfs_set_opt(sbi
, SNAPSHOT
);
676 "NILFS: Unrecognized mount option \"%s\"\n", p
);
684 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
685 struct nilfs_super_block
*sbp
)
688 NILFS_MOUNT_ERRORS_CONT
| NILFS_MOUNT_BARRIER
;
691 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
693 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
694 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
695 int max_mnt_count
= le16_to_cpu(sbp
->s_max_mnt_count
);
696 int mnt_count
= le16_to_cpu(sbp
->s_mnt_count
);
698 /* nilfs->sem must be locked by the caller. */
699 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
700 printk(KERN_WARNING
"NILFS warning: mounting unchecked fs\n");
701 } else if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
703 "NILFS warning: mounting fs with errors\n");
705 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
707 "NILFS warning: maximal mount count reached\n");
711 sbp
->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
713 sbp
->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
714 sbp
->s_state
= cpu_to_le16(le16_to_cpu(sbp
->s_state
) & ~NILFS_VALID_FS
);
715 sbp
->s_mtime
= cpu_to_le64(get_seconds());
716 return nilfs_commit_super(sbi
, 1);
719 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
720 u64 pos
, int blocksize
,
721 struct buffer_head
**pbh
)
723 unsigned long long sb_index
= pos
;
724 unsigned long offset
;
726 offset
= do_div(sb_index
, blocksize
);
727 *pbh
= sb_bread(sb
, sb_index
);
730 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
733 int nilfs_store_magic_and_option(struct super_block
*sb
,
734 struct nilfs_super_block
*sbp
,
737 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
739 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
741 /* FS independent flags */
742 #ifdef NILFS_ATIME_DISABLE
743 sb
->s_flags
|= MS_NOATIME
;
746 nilfs_set_default_options(sbi
, sbp
);
748 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
749 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
750 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
751 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
753 return !parse_options(data
, sb
) ? -EINVAL
: 0 ;
757 * nilfs_fill_super() - initialize a super block instance
759 * @data: mount options
760 * @silent: silent mode flag
761 * @nilfs: the_nilfs struct
763 * This function is called exclusively by bd_mount_mutex.
764 * So, the recovery process is protected from other simultaneous mounts.
767 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
768 struct the_nilfs
*nilfs
)
770 struct nilfs_sb_info
*sbi
;
775 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
782 sbi
->s_nilfs
= nilfs
;
785 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
789 spin_lock_init(&sbi
->s_inode_lock
);
790 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
791 INIT_LIST_HEAD(&sbi
->s_list
);
794 * Following initialization is overlapped because
795 * nilfs_sb_info structure has been cleared at the beginning.
796 * But we reserve them to keep our interest and make ready
797 * for the future change.
799 get_random_bytes(&sbi
->s_next_generation
,
800 sizeof(sbi
->s_next_generation
));
801 spin_lock_init(&sbi
->s_next_gen_lock
);
803 sb
->s_op
= &nilfs_sops
;
804 sb
->s_export_op
= &nilfs_export_ops
;
808 if (!nilfs_loaded(nilfs
)) {
809 err
= load_nilfs(nilfs
, sbi
);
813 cno
= nilfs_last_cno(nilfs
);
815 if (sb
->s_flags
& MS_RDONLY
) {
816 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
817 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
818 sbi
->s_snapshot_cno
);
823 "NILFS: The specified checkpoint is "
825 "(checkpoint number=%llu).\n",
826 (unsigned long long)sbi
->s_snapshot_cno
);
830 cno
= sbi
->s_snapshot_cno
;
832 /* Read-only mount */
833 sbi
->s_snapshot_cno
= cno
;
836 err
= nilfs_attach_checkpoint(sbi
, cno
);
838 printk(KERN_ERR
"NILFS: error loading a checkpoint"
839 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
843 if (!(sb
->s_flags
& MS_RDONLY
)) {
844 err
= nilfs_attach_segment_constructor(sbi
);
846 goto failed_checkpoint
;
849 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
851 printk(KERN_ERR
"NILFS: get root inode failed\n");
855 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
857 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
861 sb
->s_root
= d_alloc_root(root
);
864 printk(KERN_ERR
"NILFS: get root dentry failed\n");
869 if (!(sb
->s_flags
& MS_RDONLY
)) {
870 down_write(&nilfs
->ns_sem
);
871 nilfs_setup_super(sbi
);
872 up_write(&nilfs
->ns_sem
);
875 err
= nilfs_mark_recovery_complete(sbi
);
877 printk(KERN_ERR
"NILFS: recovery failed.\n");
888 nilfs_detach_segment_constructor(sbi
);
891 nilfs_detach_checkpoint(sbi
);
895 sb
->s_fs_info
= NULL
;
900 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
902 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
903 struct nilfs_super_block
*sbp
;
904 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
905 unsigned long old_sb_flags
;
906 struct nilfs_mount_options old_opts
;
909 old_sb_flags
= sb
->s_flags
;
910 old_opts
.mount_opt
= sbi
->s_mount_opt
;
911 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
913 if (!parse_options(data
, sb
)) {
917 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
919 if ((*flags
& MS_RDONLY
) &&
920 sbi
->s_snapshot_cno
!= old_opts
.snapshot_cno
) {
921 printk(KERN_WARNING
"NILFS (device %s): couldn't "
922 "remount to a different snapshot. \n",
928 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
930 if (*flags
& MS_RDONLY
) {
931 /* Shutting down the segment constructor */
932 nilfs_detach_segment_constructor(sbi
);
933 sb
->s_flags
|= MS_RDONLY
;
935 sbi
->s_snapshot_cno
= nilfs_last_cno(nilfs
);
936 /* nilfs_set_opt(sbi, SNAPSHOT); */
939 * Remounting a valid RW partition RDONLY, so set
940 * the RDONLY flag and then mark the partition as valid again.
942 down_write(&nilfs
->ns_sem
);
943 sbp
= nilfs
->ns_sbp
[0];
944 if (!(sbp
->s_state
& le16_to_cpu(NILFS_VALID_FS
)) &&
945 (nilfs
->ns_mount_state
& NILFS_VALID_FS
))
946 sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
947 sbp
->s_mtime
= cpu_to_le64(get_seconds());
948 nilfs_commit_super(sbi
, 1);
949 up_write(&nilfs
->ns_sem
);
952 * Mounting a RDONLY partition read-write, so reread and
953 * store the current valid flag. (It may have been changed
954 * by fsck since we originally mounted the partition.)
956 down(&sb
->s_bdev
->bd_mount_sem
);
957 /* Check existing RW-mount */
958 if (test_exclusive_mount(sb
->s_type
, sb
->s_bdev
, 0)) {
959 printk(KERN_WARNING
"NILFS (device %s): couldn't "
960 "remount because a RW-mount exists.\n",
963 goto rw_remount_failed
;
965 if (sbi
->s_snapshot_cno
!= nilfs_last_cno(nilfs
)) {
966 printk(KERN_WARNING
"NILFS (device %s): couldn't "
967 "remount because the current RO-mount is not "
971 goto rw_remount_failed
;
973 sb
->s_flags
&= ~MS_RDONLY
;
974 nilfs_clear_opt(sbi
, SNAPSHOT
);
975 sbi
->s_snapshot_cno
= 0;
977 err
= nilfs_attach_segment_constructor(sbi
);
979 goto rw_remount_failed
;
981 down_write(&nilfs
->ns_sem
);
982 nilfs_setup_super(sbi
);
983 up_write(&nilfs
->ns_sem
);
985 up(&sb
->s_bdev
->bd_mount_sem
);
991 up(&sb
->s_bdev
->bd_mount_sem
);
993 sb
->s_flags
= old_sb_flags
;
994 sbi
->s_mount_opt
= old_opts
.mount_opt
;
995 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
999 struct nilfs_super_data
{
1000 struct block_device
*bdev
;
1006 * nilfs_identify - pre-read mount options needed to identify mount instance
1007 * @data: mount options
1008 * @sd: nilfs_super_data
1010 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1012 char *p
, *options
= data
;
1013 substring_t args
[MAX_OPT_ARGS
];
1018 p
= strsep(&options
, ",");
1019 if (p
!= NULL
&& *p
) {
1020 token
= match_token(p
, tokens
, args
);
1021 if (token
== Opt_snapshot
) {
1022 if (!(sd
->flags
& MS_RDONLY
))
1025 ret
= match_int(&args
[0], &option
);
1036 "NILFS: invalid mount option: %s\n", p
);
1040 BUG_ON(options
== data
);
1041 *(options
- 1) = ',';
1046 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1048 struct nilfs_super_data
*sd
= data
;
1050 s
->s_bdev
= sd
->bdev
;
1051 s
->s_dev
= s
->s_bdev
->bd_dev
;
1055 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1057 struct nilfs_super_data
*sd
= data
;
1059 return s
->s_bdev
== sd
->bdev
;
1062 static int nilfs_test_bdev_super2(struct super_block
*s
, void *data
)
1064 struct nilfs_super_data
*sd
= data
;
1067 if (s
->s_bdev
!= sd
->bdev
)
1070 if (!((s
->s_flags
| sd
->flags
) & MS_RDONLY
))
1071 return 1; /* Reuse an old R/W-mode super_block */
1073 if (s
->s_flags
& sd
->flags
& MS_RDONLY
) {
1074 if (down_read_trylock(&s
->s_umount
)) {
1076 (sd
->cno
== NILFS_SB(s
)->s_snapshot_cno
);
1077 up_read(&s
->s_umount
);
1079 * This path is locked with sb_lock by sget().
1080 * So, drop_super() causes deadlock.
1089 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1090 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1092 struct nilfs_super_data sd
;
1093 struct super_block
*s
, *s2
;
1094 struct the_nilfs
*nilfs
= NULL
;
1095 int err
, need_to_close
= 1;
1097 sd
.bdev
= open_bdev_exclusive(dev_name
, flags
, fs_type
);
1098 if (IS_ERR(sd
.bdev
))
1099 return PTR_ERR(sd
.bdev
);
1102 * To get mount instance using sget() vfs-routine, NILFS needs
1103 * much more information than normal filesystems to identify mount
1104 * instance. For snapshot mounts, not only a mount type (ro-mount
1105 * or rw-mount) but also a checkpoint number is required.
1106 * The results are passed in sget() using nilfs_super_data.
1110 if (nilfs_identify((char *)data
, &sd
)) {
1116 * once the super is inserted into the list by sget, s_umount
1117 * will protect the lockfs code from trying to start a snapshot
1118 * while we are mounting
1120 down(&sd
.bdev
->bd_mount_sem
);
1122 (err
= test_exclusive_mount(fs_type
, sd
.bdev
, flags
^ MS_RDONLY
))) {
1123 err
= (err
< 0) ? : -EBUSY
;
1128 * Phase-1: search any existent instance and get the_nilfs
1130 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1136 nilfs
= alloc_nilfs(sd
.bdev
);
1140 struct nilfs_sb_info
*sbi
= NILFS_SB(s
);
1143 * s_umount protects super_block from unmount process;
1144 * It covers pointers of nilfs_sb_info and the_nilfs.
1146 nilfs
= sbi
->s_nilfs
;
1148 up_write(&s
->s_umount
);
1151 * Phase-2: search specified snapshot or R/W mode super_block
1154 /* trying to get the latest checkpoint. */
1155 sd
.cno
= nilfs_last_cno(nilfs
);
1157 s2
= sget(fs_type
, nilfs_test_bdev_super2
,
1158 nilfs_set_bdev_super
, &sd
);
1159 deactivate_super(s
);
1161 * Although deactivate_super() invokes close_bdev_exclusive() at
1162 * kill_block_super(). Here, s is an existent mount; we need
1163 * one more close_bdev_exclusive() call.
1171 char b
[BDEVNAME_SIZE
];
1174 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1175 sb_set_blocksize(s
, block_size(sd
.bdev
));
1177 err
= nilfs_fill_super(s
, data
, flags
& MS_VERBOSE
, nilfs
);
1181 s
->s_flags
|= MS_ACTIVE
;
1183 } else if (!(s
->s_flags
& MS_RDONLY
)) {
1187 up(&sd
.bdev
->bd_mount_sem
);
1190 close_bdev_exclusive(sd
.bdev
, flags
);
1191 simple_set_mnt(mnt
, s
);
1195 up(&sd
.bdev
->bd_mount_sem
);
1198 close_bdev_exclusive(sd
.bdev
, flags
);
1202 up(&sd
.bdev
->bd_mount_sem
);
1204 close_bdev_exclusive(sd
.bdev
, flags
);
1209 /* Abandoning the newly allocated superblock */
1210 up(&sd
.bdev
->bd_mount_sem
);
1213 up_write(&s
->s_umount
);
1214 deactivate_super(s
);
1216 * deactivate_super() invokes close_bdev_exclusive().
1217 * We must finish all post-cleaning before this call;
1218 * put_nilfs() and unlocking bd_mount_sem need the block device.
1223 static int nilfs_test_bdev_super3(struct super_block
*s
, void *data
)
1225 struct nilfs_super_data
*sd
= data
;
1228 if (s
->s_bdev
!= sd
->bdev
)
1230 if (down_read_trylock(&s
->s_umount
)) {
1231 ret
= (s
->s_flags
& MS_RDONLY
) && s
->s_root
&&
1232 nilfs_test_opt(NILFS_SB(s
), SNAPSHOT
);
1233 up_read(&s
->s_umount
);
1235 return 0; /* ignore snapshot mounts */
1237 return !((sd
->flags
^ s
->s_flags
) & MS_RDONLY
);
1240 static int __false_bdev_super(struct super_block
*s
, void *data
)
1242 #if 0 /* XXX: workaround for lock debug. This is not good idea */
1243 up_write(&s
->s_umount
);
1249 * test_exclusive_mount - check whether an exclusive RW/RO mount exists or not.
1250 * fs_type: filesystem type
1251 * bdev: block device
1252 * flag: 0 (check rw-mount) or MS_RDONLY (check ro-mount)
1253 * res: pointer to an integer to store result
1255 * This function must be called within a section protected by bd_mount_mutex.
1257 static int test_exclusive_mount(struct file_system_type
*fs_type
,
1258 struct block_device
*bdev
, int flags
)
1260 struct super_block
*s
;
1261 struct nilfs_super_data sd
= { .flags
= flags
, .bdev
= bdev
};
1263 s
= sget(fs_type
, nilfs_test_bdev_super3
, __false_bdev_super
, &sd
);
1265 if (PTR_ERR(s
) != -EFAULT
)
1267 return 0; /* Not found */
1269 up_write(&s
->s_umount
);
1270 deactivate_super(s
);
1271 return 1; /* Found */
1274 struct file_system_type nilfs_fs_type
= {
1275 .owner
= THIS_MODULE
,
1277 .get_sb
= nilfs_get_sb
,
1278 .kill_sb
= kill_block_super
,
1279 .fs_flags
= FS_REQUIRES_DEV
,
1282 static int __init
init_nilfs_fs(void)
1286 err
= nilfs_init_inode_cache();
1290 err
= nilfs_init_transaction_cache();
1292 goto failed_inode_cache
;
1294 err
= nilfs_init_segbuf_cache();
1296 goto failed_transaction_cache
;
1298 err
= nilfs_btree_path_cache_init();
1300 goto failed_segbuf_cache
;
1302 err
= register_filesystem(&nilfs_fs_type
);
1304 goto failed_btree_path_cache
;
1308 failed_btree_path_cache
:
1309 nilfs_btree_path_cache_destroy();
1311 failed_segbuf_cache
:
1312 nilfs_destroy_segbuf_cache();
1314 failed_transaction_cache
:
1315 nilfs_destroy_transaction_cache();
1318 nilfs_destroy_inode_cache();
1324 static void __exit
exit_nilfs_fs(void)
1326 nilfs_destroy_segbuf_cache();
1327 nilfs_destroy_transaction_cache();
1328 nilfs_destroy_inode_cache();
1329 nilfs_btree_path_cache_destroy();
1330 unregister_filesystem(&nilfs_fs_type
);
1333 module_init(init_nilfs_fs
)
1334 module_exit(exit_nilfs_fs
)