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
);
72 * nilfs_error() - report failure condition on a filesystem
74 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
75 * reporting an error message. It should be called when NILFS detects
76 * incoherences or defects of meta data on disk. As for sustainable
77 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
78 * function should be used instead.
80 * The segment constructor must not call this function because it can
83 void nilfs_error(struct super_block
*sb
, const char *function
,
86 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
90 printk(KERN_CRIT
"NILFS error (device %s): %s: ", sb
->s_id
, function
);
95 if (!(sb
->s_flags
& MS_RDONLY
)) {
96 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
98 if (!nilfs_test_opt(sbi
, ERRORS_CONT
))
99 nilfs_detach_segment_constructor(sbi
);
101 down_write(&nilfs
->ns_sem
);
102 if (!(nilfs
->ns_mount_state
& NILFS_ERROR_FS
)) {
103 nilfs
->ns_mount_state
|= NILFS_ERROR_FS
;
104 nilfs
->ns_sbp
[0]->s_state
|=
105 cpu_to_le16(NILFS_ERROR_FS
);
106 nilfs_commit_super(sbi
, 1);
108 up_write(&nilfs
->ns_sem
);
110 if (nilfs_test_opt(sbi
, ERRORS_RO
)) {
111 printk(KERN_CRIT
"Remounting filesystem read-only\n");
112 sb
->s_flags
|= MS_RDONLY
;
116 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
117 panic("NILFS (device %s): panic forced after error\n",
121 void nilfs_warning(struct super_block
*sb
, const char *function
,
122 const char *fmt
, ...)
127 printk(KERN_WARNING
"NILFS warning (device %s): %s: ",
134 static struct kmem_cache
*nilfs_inode_cachep
;
136 struct inode
*nilfs_alloc_inode(struct super_block
*sb
)
138 struct nilfs_inode_info
*ii
;
140 ii
= kmem_cache_alloc(nilfs_inode_cachep
, GFP_NOFS
);
145 ii
->vfs_inode
.i_version
= 1;
146 nilfs_btnode_cache_init(&ii
->i_btnode_cache
);
147 return &ii
->vfs_inode
;
150 void nilfs_destroy_inode(struct inode
*inode
)
152 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
155 static void init_once(void *obj
)
157 struct nilfs_inode_info
*ii
= obj
;
159 INIT_LIST_HEAD(&ii
->i_dirty
);
160 #ifdef CONFIG_NILFS_XATTR
161 init_rwsem(&ii
->xattr_sem
);
163 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
164 ii
->i_bmap
= (struct nilfs_bmap
*)&ii
->i_bmap_union
;
165 inode_init_once(&ii
->vfs_inode
);
168 static int nilfs_init_inode_cache(void)
170 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
171 sizeof(struct nilfs_inode_info
),
172 0, SLAB_RECLAIM_ACCOUNT
,
175 return (nilfs_inode_cachep
== NULL
) ? -ENOMEM
: 0;
178 static inline void nilfs_destroy_inode_cache(void)
180 kmem_cache_destroy(nilfs_inode_cachep
);
183 static void nilfs_clear_inode(struct inode
*inode
)
185 struct nilfs_inode_info
*ii
= NILFS_I(inode
);
187 #ifdef CONFIG_NILFS_POSIX_ACL
188 if (ii
->i_acl
&& ii
->i_acl
!= NILFS_ACL_NOT_CACHED
) {
189 posix_acl_release(ii
->i_acl
);
190 ii
->i_acl
= NILFS_ACL_NOT_CACHED
;
192 if (ii
->i_default_acl
&& ii
->i_default_acl
!= NILFS_ACL_NOT_CACHED
) {
193 posix_acl_release(ii
->i_default_acl
);
194 ii
->i_default_acl
= NILFS_ACL_NOT_CACHED
;
198 * Free resources allocated in nilfs_read_inode(), here.
200 BUG_ON(!list_empty(&ii
->i_dirty
));
204 if (test_bit(NILFS_I_BMAP
, &ii
->i_state
))
205 nilfs_bmap_clear(ii
->i_bmap
);
207 nilfs_btnode_cache_clear(&ii
->i_btnode_cache
);
210 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
, int dupsb
)
212 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
214 int barrier_done
= 0;
216 if (nilfs_test_opt(sbi
, BARRIER
)) {
217 set_buffer_ordered(nilfs
->ns_sbh
[0]);
221 set_buffer_dirty(nilfs
->ns_sbh
[0]);
222 err
= sync_dirty_buffer(nilfs
->ns_sbh
[0]);
223 if (err
== -EOPNOTSUPP
&& barrier_done
) {
224 nilfs_warning(sbi
->s_super
, __func__
,
225 "barrier-based sync failed. "
226 "disabling barriers\n");
227 nilfs_clear_opt(sbi
, BARRIER
);
229 clear_buffer_ordered(nilfs
->ns_sbh
[0]);
234 "NILFS: unable to write superblock (err=%d)\n", err
);
235 if (err
== -EIO
&& nilfs
->ns_sbh
[1]) {
236 nilfs_fall_back_super_block(nilfs
);
240 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
243 * The latest segment becomes trailable from the position
244 * written in superblock.
246 clear_nilfs_discontinued(nilfs
);
248 /* update GC protection for recent segments */
249 if (nilfs
->ns_sbh
[1]) {
252 set_buffer_dirty(nilfs
->ns_sbh
[1]);
253 if (!sync_dirty_buffer(nilfs
->ns_sbh
[1]))
254 sbp
= nilfs
->ns_sbp
[1];
258 spin_lock(&nilfs
->ns_last_segment_lock
);
259 nilfs
->ns_prot_seq
= le64_to_cpu(sbp
->s_last_seq
);
260 spin_unlock(&nilfs
->ns_last_segment_lock
);
267 int nilfs_commit_super(struct nilfs_sb_info
*sbi
, int dupsb
)
269 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
270 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
271 sector_t nfreeblocks
;
275 /* nilfs->sem must be locked by the caller. */
276 if (sbp
[0]->s_magic
!= NILFS_SUPER_MAGIC
) {
277 if (sbp
[1] && sbp
[1]->s_magic
== NILFS_SUPER_MAGIC
)
278 nilfs_swap_super_block(nilfs
);
280 printk(KERN_CRIT
"NILFS: superblock broke on dev %s\n",
285 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
287 printk(KERN_ERR
"NILFS: failed to count free blocks\n");
290 spin_lock(&nilfs
->ns_last_segment_lock
);
291 sbp
[0]->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
292 sbp
[0]->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
293 sbp
[0]->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
294 spin_unlock(&nilfs
->ns_last_segment_lock
);
297 nilfs
->ns_sbwtime
[0] = t
;
298 sbp
[0]->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
299 sbp
[0]->s_wtime
= cpu_to_le64(t
);
301 sbp
[0]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
302 (unsigned char *)sbp
[0],
304 if (dupsb
&& sbp
[1]) {
305 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
306 nilfs
->ns_sbwtime
[1] = t
;
308 sbi
->s_super
->s_dirt
= 0;
309 return nilfs_sync_super(sbi
, dupsb
);
312 static void nilfs_put_super(struct super_block
*sb
)
314 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
315 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
320 nilfs_write_super(sb
);
322 nilfs_detach_segment_constructor(sbi
);
324 if (!(sb
->s_flags
& MS_RDONLY
)) {
325 down_write(&nilfs
->ns_sem
);
326 nilfs
->ns_sbp
[0]->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
327 nilfs_commit_super(sbi
, 1);
328 up_write(&nilfs
->ns_sem
);
330 down_write(&nilfs
->ns_sem
);
331 if (nilfs
->ns_current
== sbi
)
332 nilfs
->ns_current
= NULL
;
333 up_write(&nilfs
->ns_sem
);
335 nilfs_detach_checkpoint(sbi
);
336 put_nilfs(sbi
->s_nilfs
);
338 sb
->s_fs_info
= NULL
;
339 nilfs_put_sbinfo(sbi
);
345 * nilfs_write_super - write super block(s) of NILFS
348 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
349 * clears s_dirt. This function is called in the section protected by
352 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
353 * of the struct the_nilfs. Lock order must be as follows:
356 * 2. down_write(&nilfs->ns_sem)
358 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
359 * of the super block (nilfs->ns_sbp[]).
361 * In most cases, VFS functions call lock_super() before calling these
362 * methods. So we must be careful not to bring on deadlocks when using
363 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
365 * Note that order of lock_kernel() and lock_super() depends on contexts
366 * of VFS. We should also note that lock_kernel() can be used in its
367 * protective section and only the outermost one has an effect.
369 static void nilfs_write_super(struct super_block
*sb
)
371 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
372 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
374 down_write(&nilfs
->ns_sem
);
375 if (!(sb
->s_flags
& MS_RDONLY
)) {
376 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
377 u64 t
= get_seconds();
380 if (!nilfs_discontinued(nilfs
) && t
>= nilfs
->ns_sbwtime
[0] &&
381 t
< nilfs
->ns_sbwtime
[0] + NILFS_SB_FREQ
) {
382 up_write(&nilfs
->ns_sem
);
385 dupsb
= sbp
[1] && t
> nilfs
->ns_sbwtime
[1] + NILFS_ALTSB_FREQ
;
386 nilfs_commit_super(sbi
, dupsb
);
389 up_write(&nilfs
->ns_sem
);
392 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
396 nilfs_write_super(sb
);
398 /* This function is called when super block should be written back */
400 err
= nilfs_construct_segment(sb
);
404 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
406 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
407 struct nilfs_checkpoint
*raw_cp
;
408 struct buffer_head
*bh_cp
;
411 down_write(&nilfs
->ns_sem
);
412 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
413 up_write(&nilfs
->ns_sem
);
415 sbi
->s_ifile
= nilfs_mdt_new(
416 nilfs
, sbi
->s_super
, NILFS_IFILE_INO
, NILFS_IFILE_GFP
);
420 err
= nilfs_palloc_init_blockgroup(sbi
->s_ifile
, nilfs
->ns_inode_size
);
424 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
427 if (err
== -ENOENT
|| err
== -EINVAL
) {
429 "NILFS: Invalid checkpoint "
430 "(checkpoint number=%llu)\n",
431 (unsigned long long)cno
);
436 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
439 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
440 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
442 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
446 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
448 nilfs_mdt_destroy(sbi
->s_ifile
);
451 down_write(&nilfs
->ns_sem
);
452 list_del_init(&sbi
->s_list
);
453 up_write(&nilfs
->ns_sem
);
458 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
460 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
462 nilfs_mdt_clear(sbi
->s_ifile
);
463 nilfs_mdt_destroy(sbi
->s_ifile
);
465 down_write(&nilfs
->ns_sem
);
466 list_del_init(&sbi
->s_list
);
467 up_write(&nilfs
->ns_sem
);
470 static int nilfs_mark_recovery_complete(struct nilfs_sb_info
*sbi
)
472 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
475 down_write(&nilfs
->ns_sem
);
476 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
477 nilfs
->ns_mount_state
|= NILFS_VALID_FS
;
478 err
= nilfs_commit_super(sbi
, 1);
480 printk(KERN_INFO
"NILFS: recovery complete.\n");
482 up_write(&nilfs
->ns_sem
);
486 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
488 struct super_block
*sb
= dentry
->d_sb
;
489 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
490 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
491 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
492 unsigned long long blocks
;
493 unsigned long overhead
;
494 unsigned long nrsvblocks
;
495 sector_t nfreeblocks
;
499 * Compute all of the segment blocks
501 * The blocks before first segment and after last segment
504 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
505 - nilfs
->ns_first_data_block
;
506 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
509 * Compute the overhead
511 * When distributing meta data blocks outside semgent structure,
512 * We must count them as the overhead.
516 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
520 buf
->f_type
= NILFS_SUPER_MAGIC
;
521 buf
->f_bsize
= sb
->s_blocksize
;
522 buf
->f_blocks
= blocks
- overhead
;
523 buf
->f_bfree
= nfreeblocks
;
524 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
525 (buf
->f_bfree
- nrsvblocks
) : 0;
526 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
527 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
528 buf
->f_namelen
= NILFS_NAME_LEN
;
529 buf
->f_fsid
.val
[0] = (u32
)id
;
530 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
535 static struct super_operations nilfs_sops
= {
536 .alloc_inode
= nilfs_alloc_inode
,
537 .destroy_inode
= nilfs_destroy_inode
,
538 .dirty_inode
= nilfs_dirty_inode
,
539 /* .write_inode = nilfs_write_inode, */
540 /* .put_inode = nilfs_put_inode, */
541 /* .drop_inode = nilfs_drop_inode, */
542 .delete_inode
= nilfs_delete_inode
,
543 .put_super
= nilfs_put_super
,
544 .write_super
= nilfs_write_super
,
545 .sync_fs
= nilfs_sync_fs
,
546 /* .write_super_lockfs */
548 .statfs
= nilfs_statfs
,
549 .remount_fs
= nilfs_remount
,
550 .clear_inode
= nilfs_clear_inode
,
555 static struct inode
*
556 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
560 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
561 ino
!= NILFS_SKETCH_INO
)
562 return ERR_PTR(-ESTALE
);
564 inode
= nilfs_iget(sb
, ino
);
566 return ERR_CAST(inode
);
567 if (generation
&& inode
->i_generation
!= generation
) {
569 return ERR_PTR(-ESTALE
);
575 static struct dentry
*
576 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
579 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
580 nilfs_nfs_get_inode
);
583 static struct dentry
*
584 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
587 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
588 nilfs_nfs_get_inode
);
591 static struct export_operations nilfs_export_ops
= {
592 .fh_to_dentry
= nilfs_fh_to_dentry
,
593 .fh_to_parent
= nilfs_fh_to_parent
,
594 .get_parent
= nilfs_get_parent
,
598 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
599 Opt_barrier
, Opt_snapshot
, Opt_order
,
603 static match_table_t tokens
= {
604 {Opt_err_cont
, "errors=continue"},
605 {Opt_err_panic
, "errors=panic"},
606 {Opt_err_ro
, "errors=remount-ro"},
607 {Opt_barrier
, "barrier=%s"},
608 {Opt_snapshot
, "cp=%u"},
609 {Opt_order
, "order=%s"},
613 static int match_bool(substring_t
*s
, int *result
)
615 int len
= s
->to
- s
->from
;
617 if (strncmp(s
->from
, "on", len
) == 0)
619 else if (strncmp(s
->from
, "off", len
) == 0)
626 static int parse_options(char *options
, struct super_block
*sb
)
628 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
630 substring_t args
[MAX_OPT_ARGS
];
636 while ((p
= strsep(&options
, ",")) != NULL
) {
641 token
= match_token(p
, tokens
, args
);
644 if (match_bool(&args
[0], &option
))
647 nilfs_set_opt(sbi
, BARRIER
);
649 nilfs_clear_opt(sbi
, BARRIER
);
652 if (strcmp(args
[0].from
, "relaxed") == 0)
653 /* Ordered data semantics */
654 nilfs_clear_opt(sbi
, STRICT_ORDER
);
655 else if (strcmp(args
[0].from
, "strict") == 0)
656 /* Strict in-order semantics */
657 nilfs_set_opt(sbi
, STRICT_ORDER
);
662 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
665 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
668 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
671 if (match_int(&args
[0], &option
) || option
<= 0)
673 if (!(sb
->s_flags
& MS_RDONLY
))
675 sbi
->s_snapshot_cno
= option
;
676 nilfs_set_opt(sbi
, SNAPSHOT
);
680 "NILFS: Unrecognized mount option \"%s\"\n", p
);
688 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
689 struct nilfs_super_block
*sbp
)
692 NILFS_MOUNT_ERRORS_CONT
| NILFS_MOUNT_BARRIER
;
695 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
697 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
698 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
699 int max_mnt_count
= le16_to_cpu(sbp
->s_max_mnt_count
);
700 int mnt_count
= le16_to_cpu(sbp
->s_mnt_count
);
702 /* nilfs->sem must be locked by the caller. */
703 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
704 printk(KERN_WARNING
"NILFS warning: mounting unchecked fs\n");
705 } else if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
707 "NILFS warning: mounting fs with errors\n");
709 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
711 "NILFS warning: maximal mount count reached\n");
715 sbp
->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
717 sbp
->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
718 sbp
->s_state
= cpu_to_le16(le16_to_cpu(sbp
->s_state
) & ~NILFS_VALID_FS
);
719 sbp
->s_mtime
= cpu_to_le64(get_seconds());
720 return nilfs_commit_super(sbi
, 1);
723 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
724 u64 pos
, int blocksize
,
725 struct buffer_head
**pbh
)
727 unsigned long long sb_index
= pos
;
728 unsigned long offset
;
730 offset
= do_div(sb_index
, blocksize
);
731 *pbh
= sb_bread(sb
, sb_index
);
734 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
737 int nilfs_store_magic_and_option(struct super_block
*sb
,
738 struct nilfs_super_block
*sbp
,
741 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
743 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
745 /* FS independent flags */
746 #ifdef NILFS_ATIME_DISABLE
747 sb
->s_flags
|= MS_NOATIME
;
750 nilfs_set_default_options(sbi
, sbp
);
752 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
753 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
754 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
755 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
757 return !parse_options(data
, sb
) ? -EINVAL
: 0 ;
761 * nilfs_fill_super() - initialize a super block instance
763 * @data: mount options
764 * @silent: silent mode flag
765 * @nilfs: the_nilfs struct
767 * This function is called exclusively by bd_mount_mutex.
768 * So, the recovery process is protected from other simultaneous mounts.
771 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
772 struct the_nilfs
*nilfs
)
774 struct nilfs_sb_info
*sbi
;
779 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
786 sbi
->s_nilfs
= nilfs
;
788 atomic_set(&sbi
->s_count
, 1);
790 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
794 spin_lock_init(&sbi
->s_inode_lock
);
795 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
796 INIT_LIST_HEAD(&sbi
->s_list
);
799 * Following initialization is overlapped because
800 * nilfs_sb_info structure has been cleared at the beginning.
801 * But we reserve them to keep our interest and make ready
802 * for the future change.
804 get_random_bytes(&sbi
->s_next_generation
,
805 sizeof(sbi
->s_next_generation
));
806 spin_lock_init(&sbi
->s_next_gen_lock
);
808 sb
->s_op
= &nilfs_sops
;
809 sb
->s_export_op
= &nilfs_export_ops
;
813 if (!nilfs_loaded(nilfs
)) {
814 err
= load_nilfs(nilfs
, sbi
);
818 cno
= nilfs_last_cno(nilfs
);
820 if (sb
->s_flags
& MS_RDONLY
) {
821 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
822 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
823 sbi
->s_snapshot_cno
);
828 "NILFS: The specified checkpoint is "
830 "(checkpoint number=%llu).\n",
831 (unsigned long long)sbi
->s_snapshot_cno
);
835 cno
= sbi
->s_snapshot_cno
;
837 /* Read-only mount */
838 sbi
->s_snapshot_cno
= cno
;
841 err
= nilfs_attach_checkpoint(sbi
, cno
);
843 printk(KERN_ERR
"NILFS: error loading a checkpoint"
844 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
848 if (!(sb
->s_flags
& MS_RDONLY
)) {
849 err
= nilfs_attach_segment_constructor(sbi
);
851 goto failed_checkpoint
;
854 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
856 printk(KERN_ERR
"NILFS: get root inode failed\n");
860 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
862 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
866 sb
->s_root
= d_alloc_root(root
);
869 printk(KERN_ERR
"NILFS: get root dentry failed\n");
874 if (!(sb
->s_flags
& MS_RDONLY
)) {
875 down_write(&nilfs
->ns_sem
);
876 nilfs_setup_super(sbi
);
877 up_write(&nilfs
->ns_sem
);
880 err
= nilfs_mark_recovery_complete(sbi
);
882 printk(KERN_ERR
"NILFS: recovery failed.\n");
886 down_write(&nilfs
->ns_sem
);
887 if (!nilfs_test_opt(sbi
, SNAPSHOT
))
888 nilfs
->ns_current
= sbi
;
889 up_write(&nilfs
->ns_sem
);
898 nilfs_detach_segment_constructor(sbi
);
901 nilfs_detach_checkpoint(sbi
);
905 sb
->s_fs_info
= NULL
;
906 nilfs_put_sbinfo(sbi
);
910 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
912 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
913 struct nilfs_super_block
*sbp
;
914 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
915 unsigned long old_sb_flags
;
916 struct nilfs_mount_options old_opts
;
921 old_sb_flags
= sb
->s_flags
;
922 old_opts
.mount_opt
= sbi
->s_mount_opt
;
923 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
925 if (!parse_options(data
, sb
)) {
929 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
931 if ((*flags
& MS_RDONLY
) &&
932 sbi
->s_snapshot_cno
!= old_opts
.snapshot_cno
) {
933 printk(KERN_WARNING
"NILFS (device %s): couldn't "
934 "remount to a different snapshot. \n",
940 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
942 if (*flags
& MS_RDONLY
) {
943 /* Shutting down the segment constructor */
944 nilfs_detach_segment_constructor(sbi
);
945 sb
->s_flags
|= MS_RDONLY
;
947 sbi
->s_snapshot_cno
= nilfs_last_cno(nilfs
);
948 /* nilfs_set_opt(sbi, SNAPSHOT); */
951 * Remounting a valid RW partition RDONLY, so set
952 * the RDONLY flag and then mark the partition as valid again.
954 down_write(&nilfs
->ns_sem
);
955 sbp
= nilfs
->ns_sbp
[0];
956 if (!(sbp
->s_state
& le16_to_cpu(NILFS_VALID_FS
)) &&
957 (nilfs
->ns_mount_state
& NILFS_VALID_FS
))
958 sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
959 sbp
->s_mtime
= cpu_to_le64(get_seconds());
960 nilfs_commit_super(sbi
, 1);
961 up_write(&nilfs
->ns_sem
);
964 * Mounting a RDONLY partition read-write, so reread and
965 * store the current valid flag. (It may have been changed
966 * by fsck since we originally mounted the partition.)
968 down(&sb
->s_bdev
->bd_mount_sem
);
969 down_read(&nilfs
->ns_sem
);
970 if (nilfs
->ns_current
&& nilfs
->ns_current
!= sbi
) {
971 printk(KERN_WARNING
"NILFS (device %s): couldn't "
972 "remount because an RW-mount exists.\n",
974 up_read(&nilfs
->ns_sem
);
976 goto rw_remount_failed
;
978 up_read(&nilfs
->ns_sem
);
979 if (sbi
->s_snapshot_cno
!= nilfs_last_cno(nilfs
)) {
980 printk(KERN_WARNING
"NILFS (device %s): couldn't "
981 "remount because the current RO-mount is not "
985 goto rw_remount_failed
;
987 sb
->s_flags
&= ~MS_RDONLY
;
988 nilfs_clear_opt(sbi
, SNAPSHOT
);
989 sbi
->s_snapshot_cno
= 0;
991 err
= nilfs_attach_segment_constructor(sbi
);
993 goto rw_remount_failed
;
995 down_write(&nilfs
->ns_sem
);
996 nilfs_setup_super(sbi
);
997 nilfs
->ns_current
= sbi
;
998 up_write(&nilfs
->ns_sem
);
1000 up(&sb
->s_bdev
->bd_mount_sem
);
1007 up(&sb
->s_bdev
->bd_mount_sem
);
1009 sb
->s_flags
= old_sb_flags
;
1010 sbi
->s_mount_opt
= old_opts
.mount_opt
;
1011 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
1016 struct nilfs_super_data
{
1017 struct block_device
*bdev
;
1018 struct nilfs_sb_info
*sbi
;
1024 * nilfs_identify - pre-read mount options needed to identify mount instance
1025 * @data: mount options
1026 * @sd: nilfs_super_data
1028 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1030 char *p
, *options
= data
;
1031 substring_t args
[MAX_OPT_ARGS
];
1036 p
= strsep(&options
, ",");
1037 if (p
!= NULL
&& *p
) {
1038 token
= match_token(p
, tokens
, args
);
1039 if (token
== Opt_snapshot
) {
1040 if (!(sd
->flags
& MS_RDONLY
))
1043 ret
= match_int(&args
[0], &option
);
1054 "NILFS: invalid mount option: %s\n", p
);
1058 BUG_ON(options
== data
);
1059 *(options
- 1) = ',';
1064 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1066 struct nilfs_super_data
*sd
= data
;
1068 s
->s_bdev
= sd
->bdev
;
1069 s
->s_dev
= s
->s_bdev
->bd_dev
;
1073 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1075 struct nilfs_super_data
*sd
= data
;
1077 return sd
->sbi
&& s
->s_fs_info
== (void *)sd
->sbi
;
1081 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1082 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1084 struct nilfs_super_data sd
;
1085 struct super_block
*s
;
1086 struct the_nilfs
*nilfs
;
1087 int err
, need_to_close
= 1;
1089 sd
.bdev
= open_bdev_exclusive(dev_name
, flags
, fs_type
);
1090 if (IS_ERR(sd
.bdev
))
1091 return PTR_ERR(sd
.bdev
);
1094 * To get mount instance using sget() vfs-routine, NILFS needs
1095 * much more information than normal filesystems to identify mount
1096 * instance. For snapshot mounts, not only a mount type (ro-mount
1097 * or rw-mount) but also a checkpoint number is required.
1101 if (nilfs_identify((char *)data
, &sd
)) {
1106 nilfs
= find_or_create_nilfs(sd
.bdev
);
1112 down(&sd
.bdev
->bd_mount_sem
);
1116 * Check if an exclusive mount exists or not.
1117 * Snapshot mounts coexist with a current mount
1118 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1119 * ro-mount are mutually exclusive.
1121 down_read(&nilfs
->ns_sem
);
1122 if (nilfs
->ns_current
&&
1123 ((nilfs
->ns_current
->s_super
->s_flags
^ flags
)
1125 up_read(&nilfs
->ns_sem
);
1129 up_read(&nilfs
->ns_sem
);
1133 * Find existing nilfs_sb_info struct
1135 sd
.sbi
= nilfs_find_sbinfo(nilfs
, !(flags
& MS_RDONLY
), sd
.cno
);
1138 /* trying to get the latest checkpoint. */
1139 sd
.cno
= nilfs_last_cno(nilfs
);
1142 * Get super block instance holding the nilfs_sb_info struct.
1143 * A new instance is allocated if no existing mount is present or
1144 * existing instance has been unmounted.
1146 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1148 nilfs_put_sbinfo(sd
.sbi
);
1156 char b
[BDEVNAME_SIZE
];
1158 /* New superblock instance created */
1160 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1161 sb_set_blocksize(s
, block_size(sd
.bdev
));
1163 err
= nilfs_fill_super(s
, data
, flags
& MS_VERBOSE
, nilfs
);
1167 s
->s_flags
|= MS_ACTIVE
;
1171 up(&sd
.bdev
->bd_mount_sem
);
1174 close_bdev_exclusive(sd
.bdev
, flags
);
1175 simple_set_mnt(mnt
, s
);
1179 up(&sd
.bdev
->bd_mount_sem
);
1182 close_bdev_exclusive(sd
.bdev
, flags
);
1187 /* Abandoning the newly allocated superblock */
1188 up(&sd
.bdev
->bd_mount_sem
);
1190 up_write(&s
->s_umount
);
1191 deactivate_super(s
);
1193 * deactivate_super() invokes close_bdev_exclusive().
1194 * We must finish all post-cleaning before this call;
1195 * put_nilfs() and unlocking bd_mount_sem need the block device.
1200 struct file_system_type nilfs_fs_type
= {
1201 .owner
= THIS_MODULE
,
1203 .get_sb
= nilfs_get_sb
,
1204 .kill_sb
= kill_block_super
,
1205 .fs_flags
= FS_REQUIRES_DEV
,
1208 static int __init
init_nilfs_fs(void)
1212 err
= nilfs_init_inode_cache();
1216 err
= nilfs_init_transaction_cache();
1218 goto failed_inode_cache
;
1220 err
= nilfs_init_segbuf_cache();
1222 goto failed_transaction_cache
;
1224 err
= nilfs_btree_path_cache_init();
1226 goto failed_segbuf_cache
;
1228 err
= register_filesystem(&nilfs_fs_type
);
1230 goto failed_btree_path_cache
;
1234 failed_btree_path_cache
:
1235 nilfs_btree_path_cache_destroy();
1237 failed_segbuf_cache
:
1238 nilfs_destroy_segbuf_cache();
1240 failed_transaction_cache
:
1241 nilfs_destroy_transaction_cache();
1244 nilfs_destroy_inode_cache();
1250 static void __exit
exit_nilfs_fs(void)
1252 nilfs_destroy_segbuf_cache();
1253 nilfs_destroy_transaction_cache();
1254 nilfs_destroy_inode_cache();
1255 nilfs_btree_path_cache_destroy();
1256 unregister_filesystem(&nilfs_fs_type
);
1259 module_init(init_nilfs_fs
)
1260 module_exit(exit_nilfs_fs
)