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
;
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
);
331 nilfs_detach_checkpoint(sbi
);
332 put_nilfs(sbi
->s_nilfs
);
334 sb
->s_fs_info
= NULL
;
339 * nilfs_write_super - write super block(s) of NILFS
342 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
343 * clears s_dirt. This function is called in the section protected by
346 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
347 * of the struct the_nilfs. Lock order must be as follows:
350 * 2. down_write(&nilfs->ns_sem)
352 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
353 * of the super block (nilfs->ns_sbp[]).
355 * In most cases, VFS functions call lock_super() before calling these
356 * methods. So we must be careful not to bring on deadlocks when using
357 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
359 * Note that order of lock_kernel() and lock_super() depends on contexts
360 * of VFS. We should also note that lock_kernel() can be used in its
361 * protective section and only the outermost one has an effect.
363 static void nilfs_write_super(struct super_block
*sb
)
365 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
366 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
368 down_write(&nilfs
->ns_sem
);
369 if (!(sb
->s_flags
& MS_RDONLY
)) {
370 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
371 u64 t
= get_seconds();
374 if (!nilfs_discontinued(nilfs
) && t
>= nilfs
->ns_sbwtime
[0] &&
375 t
< nilfs
->ns_sbwtime
[0] + NILFS_SB_FREQ
) {
376 up_write(&nilfs
->ns_sem
);
379 dupsb
= sbp
[1] && t
> nilfs
->ns_sbwtime
[1] + NILFS_ALTSB_FREQ
;
380 nilfs_commit_super(sbi
, dupsb
);
383 up_write(&nilfs
->ns_sem
);
386 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
390 /* This function is called when super block should be written back */
392 err
= nilfs_construct_segment(sb
);
396 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
398 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
399 struct nilfs_checkpoint
*raw_cp
;
400 struct buffer_head
*bh_cp
;
403 down_write(&nilfs
->ns_sem
);
404 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
405 up_write(&nilfs
->ns_sem
);
407 sbi
->s_ifile
= nilfs_mdt_new(
408 nilfs
, sbi
->s_super
, NILFS_IFILE_INO
, NILFS_IFILE_GFP
);
412 err
= nilfs_palloc_init_blockgroup(sbi
->s_ifile
, nilfs
->ns_inode_size
);
416 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
419 if (err
== -ENOENT
|| err
== -EINVAL
) {
421 "NILFS: Invalid checkpoint "
422 "(checkpoint number=%llu)\n",
423 (unsigned long long)cno
);
428 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
431 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
432 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
434 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
438 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
440 nilfs_mdt_destroy(sbi
->s_ifile
);
443 down_write(&nilfs
->ns_sem
);
444 list_del_init(&sbi
->s_list
);
445 up_write(&nilfs
->ns_sem
);
450 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
452 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
454 nilfs_mdt_clear(sbi
->s_ifile
);
455 nilfs_mdt_destroy(sbi
->s_ifile
);
457 down_write(&nilfs
->ns_sem
);
458 list_del_init(&sbi
->s_list
);
459 up_write(&nilfs
->ns_sem
);
462 static int nilfs_mark_recovery_complete(struct nilfs_sb_info
*sbi
)
464 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
467 down_write(&nilfs
->ns_sem
);
468 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
469 nilfs
->ns_mount_state
|= NILFS_VALID_FS
;
470 err
= nilfs_commit_super(sbi
, 1);
472 printk(KERN_INFO
"NILFS: recovery complete.\n");
474 up_write(&nilfs
->ns_sem
);
478 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
480 struct super_block
*sb
= dentry
->d_sb
;
481 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
482 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
483 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
484 unsigned long long blocks
;
485 unsigned long overhead
;
486 unsigned long nrsvblocks
;
487 sector_t nfreeblocks
;
491 * Compute all of the segment blocks
493 * The blocks before first segment and after last segment
496 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
497 - nilfs
->ns_first_data_block
;
498 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
501 * Compute the overhead
503 * When distributing meta data blocks outside semgent structure,
504 * We must count them as the overhead.
508 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
512 buf
->f_type
= NILFS_SUPER_MAGIC
;
513 buf
->f_bsize
= sb
->s_blocksize
;
514 buf
->f_blocks
= blocks
- overhead
;
515 buf
->f_bfree
= nfreeblocks
;
516 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
517 (buf
->f_bfree
- nrsvblocks
) : 0;
518 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
519 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
520 buf
->f_namelen
= NILFS_NAME_LEN
;
521 buf
->f_fsid
.val
[0] = (u32
)id
;
522 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
527 static struct super_operations nilfs_sops
= {
528 .alloc_inode
= nilfs_alloc_inode
,
529 .destroy_inode
= nilfs_destroy_inode
,
530 .dirty_inode
= nilfs_dirty_inode
,
531 /* .write_inode = nilfs_write_inode, */
532 /* .put_inode = nilfs_put_inode, */
533 /* .drop_inode = nilfs_drop_inode, */
534 .delete_inode
= nilfs_delete_inode
,
535 .put_super
= nilfs_put_super
,
536 .write_super
= nilfs_write_super
,
537 .sync_fs
= nilfs_sync_fs
,
538 /* .write_super_lockfs */
540 .statfs
= nilfs_statfs
,
541 .remount_fs
= nilfs_remount
,
542 .clear_inode
= nilfs_clear_inode
,
547 static struct inode
*
548 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
552 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
553 ino
!= NILFS_SKETCH_INO
)
554 return ERR_PTR(-ESTALE
);
556 inode
= nilfs_iget(sb
, ino
);
558 return ERR_CAST(inode
);
559 if (generation
&& inode
->i_generation
!= generation
) {
561 return ERR_PTR(-ESTALE
);
567 static struct dentry
*
568 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
571 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
572 nilfs_nfs_get_inode
);
575 static struct dentry
*
576 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
579 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
580 nilfs_nfs_get_inode
);
583 static struct export_operations nilfs_export_ops
= {
584 .fh_to_dentry
= nilfs_fh_to_dentry
,
585 .fh_to_parent
= nilfs_fh_to_parent
,
586 .get_parent
= nilfs_get_parent
,
590 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
591 Opt_barrier
, Opt_snapshot
, Opt_order
,
595 static match_table_t tokens
= {
596 {Opt_err_cont
, "errors=continue"},
597 {Opt_err_panic
, "errors=panic"},
598 {Opt_err_ro
, "errors=remount-ro"},
599 {Opt_barrier
, "barrier=%s"},
600 {Opt_snapshot
, "cp=%u"},
601 {Opt_order
, "order=%s"},
605 static int match_bool(substring_t
*s
, int *result
)
607 int len
= s
->to
- s
->from
;
609 if (strncmp(s
->from
, "on", len
) == 0)
611 else if (strncmp(s
->from
, "off", len
) == 0)
618 static int parse_options(char *options
, struct super_block
*sb
)
620 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
622 substring_t args
[MAX_OPT_ARGS
];
628 while ((p
= strsep(&options
, ",")) != NULL
) {
633 token
= match_token(p
, tokens
, args
);
636 if (match_bool(&args
[0], &option
))
639 nilfs_set_opt(sbi
, BARRIER
);
641 nilfs_clear_opt(sbi
, BARRIER
);
644 if (strcmp(args
[0].from
, "relaxed") == 0)
645 /* Ordered data semantics */
646 nilfs_clear_opt(sbi
, STRICT_ORDER
);
647 else if (strcmp(args
[0].from
, "strict") == 0)
648 /* Strict in-order semantics */
649 nilfs_set_opt(sbi
, STRICT_ORDER
);
654 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
657 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
660 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
663 if (match_int(&args
[0], &option
) || option
<= 0)
665 if (!(sb
->s_flags
& MS_RDONLY
))
667 sbi
->s_snapshot_cno
= option
;
668 nilfs_set_opt(sbi
, SNAPSHOT
);
672 "NILFS: Unrecognized mount option \"%s\"\n", p
);
680 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
681 struct nilfs_super_block
*sbp
)
684 NILFS_MOUNT_ERRORS_CONT
| NILFS_MOUNT_BARRIER
;
687 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
689 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
690 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
691 int max_mnt_count
= le16_to_cpu(sbp
->s_max_mnt_count
);
692 int mnt_count
= le16_to_cpu(sbp
->s_mnt_count
);
694 /* nilfs->sem must be locked by the caller. */
695 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
696 printk(KERN_WARNING
"NILFS warning: mounting unchecked fs\n");
697 } else if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
699 "NILFS warning: mounting fs with errors\n");
701 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
703 "NILFS warning: maximal mount count reached\n");
707 sbp
->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
709 sbp
->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
710 sbp
->s_state
= cpu_to_le16(le16_to_cpu(sbp
->s_state
) & ~NILFS_VALID_FS
);
711 sbp
->s_mtime
= cpu_to_le64(get_seconds());
712 return nilfs_commit_super(sbi
, 1);
715 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
716 u64 pos
, int blocksize
,
717 struct buffer_head
**pbh
)
719 unsigned long long sb_index
= pos
;
720 unsigned long offset
;
722 offset
= do_div(sb_index
, blocksize
);
723 *pbh
= sb_bread(sb
, sb_index
);
726 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
729 int nilfs_store_magic_and_option(struct super_block
*sb
,
730 struct nilfs_super_block
*sbp
,
733 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
735 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
737 /* FS independent flags */
738 #ifdef NILFS_ATIME_DISABLE
739 sb
->s_flags
|= MS_NOATIME
;
742 nilfs_set_default_options(sbi
, sbp
);
744 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
745 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
746 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
747 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
749 return !parse_options(data
, sb
) ? -EINVAL
: 0 ;
753 * nilfs_fill_super() - initialize a super block instance
755 * @data: mount options
756 * @silent: silent mode flag
757 * @nilfs: the_nilfs struct
759 * This function is called exclusively by bd_mount_mutex.
760 * So, the recovery process is protected from other simultaneous mounts.
763 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
764 struct the_nilfs
*nilfs
)
766 struct nilfs_sb_info
*sbi
;
771 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
778 sbi
->s_nilfs
= nilfs
;
781 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
785 spin_lock_init(&sbi
->s_inode_lock
);
786 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
787 INIT_LIST_HEAD(&sbi
->s_list
);
790 * Following initialization is overlapped because
791 * nilfs_sb_info structure has been cleared at the beginning.
792 * But we reserve them to keep our interest and make ready
793 * for the future change.
795 get_random_bytes(&sbi
->s_next_generation
,
796 sizeof(sbi
->s_next_generation
));
797 spin_lock_init(&sbi
->s_next_gen_lock
);
799 sb
->s_op
= &nilfs_sops
;
800 sb
->s_export_op
= &nilfs_export_ops
;
804 if (!nilfs_loaded(nilfs
)) {
805 err
= load_nilfs(nilfs
, sbi
);
809 cno
= nilfs_last_cno(nilfs
);
811 if (sb
->s_flags
& MS_RDONLY
) {
812 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
813 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
814 sbi
->s_snapshot_cno
);
819 "NILFS: The specified checkpoint is "
821 "(checkpoint number=%llu).\n",
822 (unsigned long long)sbi
->s_snapshot_cno
);
826 cno
= sbi
->s_snapshot_cno
;
828 /* Read-only mount */
829 sbi
->s_snapshot_cno
= cno
;
832 err
= nilfs_attach_checkpoint(sbi
, cno
);
834 printk(KERN_ERR
"NILFS: error loading a checkpoint"
835 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
839 if (!(sb
->s_flags
& MS_RDONLY
)) {
840 err
= nilfs_attach_segment_constructor(sbi
);
842 goto failed_checkpoint
;
845 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
847 printk(KERN_ERR
"NILFS: get root inode failed\n");
851 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
853 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
857 sb
->s_root
= d_alloc_root(root
);
860 printk(KERN_ERR
"NILFS: get root dentry failed\n");
865 if (!(sb
->s_flags
& MS_RDONLY
)) {
866 down_write(&nilfs
->ns_sem
);
867 nilfs_setup_super(sbi
);
868 up_write(&nilfs
->ns_sem
);
871 err
= nilfs_mark_recovery_complete(sbi
);
873 printk(KERN_ERR
"NILFS: recovery failed.\n");
884 nilfs_detach_segment_constructor(sbi
);
887 nilfs_detach_checkpoint(sbi
);
891 sb
->s_fs_info
= NULL
;
896 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
898 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
899 struct nilfs_super_block
*sbp
;
900 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
901 unsigned long old_sb_flags
;
902 struct nilfs_mount_options old_opts
;
905 old_sb_flags
= sb
->s_flags
;
906 old_opts
.mount_opt
= sbi
->s_mount_opt
;
907 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
909 if (!parse_options(data
, sb
)) {
913 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
915 if ((*flags
& MS_RDONLY
) &&
916 sbi
->s_snapshot_cno
!= old_opts
.snapshot_cno
) {
917 printk(KERN_WARNING
"NILFS (device %s): couldn't "
918 "remount to a different snapshot. \n",
924 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
926 if (*flags
& MS_RDONLY
) {
927 /* Shutting down the segment constructor */
928 nilfs_detach_segment_constructor(sbi
);
929 sb
->s_flags
|= MS_RDONLY
;
931 sbi
->s_snapshot_cno
= nilfs_last_cno(nilfs
);
932 /* nilfs_set_opt(sbi, SNAPSHOT); */
935 * Remounting a valid RW partition RDONLY, so set
936 * the RDONLY flag and then mark the partition as valid again.
938 down_write(&nilfs
->ns_sem
);
939 sbp
= nilfs
->ns_sbp
[0];
940 if (!(sbp
->s_state
& le16_to_cpu(NILFS_VALID_FS
)) &&
941 (nilfs
->ns_mount_state
& NILFS_VALID_FS
))
942 sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
943 sbp
->s_mtime
= cpu_to_le64(get_seconds());
944 nilfs_commit_super(sbi
, 1);
945 up_write(&nilfs
->ns_sem
);
948 * Mounting a RDONLY partition read-write, so reread and
949 * store the current valid flag. (It may have been changed
950 * by fsck since we originally mounted the partition.)
952 down(&sb
->s_bdev
->bd_mount_sem
);
953 /* Check existing RW-mount */
954 if (test_exclusive_mount(sb
->s_type
, sb
->s_bdev
, 0)) {
955 printk(KERN_WARNING
"NILFS (device %s): couldn't "
956 "remount because a RW-mount exists.\n",
959 goto rw_remount_failed
;
961 if (sbi
->s_snapshot_cno
!= nilfs_last_cno(nilfs
)) {
962 printk(KERN_WARNING
"NILFS (device %s): couldn't "
963 "remount because the current RO-mount is not "
967 goto rw_remount_failed
;
969 sb
->s_flags
&= ~MS_RDONLY
;
970 nilfs_clear_opt(sbi
, SNAPSHOT
);
971 sbi
->s_snapshot_cno
= 0;
973 err
= nilfs_attach_segment_constructor(sbi
);
975 goto rw_remount_failed
;
977 down_write(&nilfs
->ns_sem
);
978 nilfs_setup_super(sbi
);
979 up_write(&nilfs
->ns_sem
);
981 up(&sb
->s_bdev
->bd_mount_sem
);
987 up(&sb
->s_bdev
->bd_mount_sem
);
989 sb
->s_flags
= old_sb_flags
;
990 sbi
->s_mount_opt
= old_opts
.mount_opt
;
991 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
995 struct nilfs_super_data
{
996 struct block_device
*bdev
;
1002 * nilfs_identify - pre-read mount options needed to identify mount instance
1003 * @data: mount options
1004 * @sd: nilfs_super_data
1006 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1008 char *p
, *options
= data
;
1009 substring_t args
[MAX_OPT_ARGS
];
1014 p
= strsep(&options
, ",");
1015 if (p
!= NULL
&& *p
) {
1016 token
= match_token(p
, tokens
, args
);
1017 if (token
== Opt_snapshot
) {
1018 if (!(sd
->flags
& MS_RDONLY
))
1021 ret
= match_int(&args
[0], &option
);
1032 "NILFS: invalid mount option: %s\n", p
);
1036 BUG_ON(options
== data
);
1037 *(options
- 1) = ',';
1042 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1044 struct nilfs_super_data
*sd
= data
;
1046 s
->s_bdev
= sd
->bdev
;
1047 s
->s_dev
= s
->s_bdev
->bd_dev
;
1051 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1053 struct nilfs_super_data
*sd
= data
;
1055 return s
->s_bdev
== sd
->bdev
;
1058 static int nilfs_test_bdev_super2(struct super_block
*s
, void *data
)
1060 struct nilfs_super_data
*sd
= data
;
1063 if (s
->s_bdev
!= sd
->bdev
)
1066 if (!((s
->s_flags
| sd
->flags
) & MS_RDONLY
))
1067 return 1; /* Reuse an old R/W-mode super_block */
1069 if (s
->s_flags
& sd
->flags
& MS_RDONLY
) {
1070 if (down_read_trylock(&s
->s_umount
)) {
1072 (sd
->cno
== NILFS_SB(s
)->s_snapshot_cno
);
1073 up_read(&s
->s_umount
);
1075 * This path is locked with sb_lock by sget().
1076 * So, drop_super() causes deadlock.
1085 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1086 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1088 struct nilfs_super_data sd
;
1089 struct super_block
*s
, *s2
;
1090 struct the_nilfs
*nilfs
= NULL
;
1091 int err
, need_to_close
= 1;
1093 sd
.bdev
= open_bdev_exclusive(dev_name
, flags
, fs_type
);
1094 if (IS_ERR(sd
.bdev
))
1095 return PTR_ERR(sd
.bdev
);
1098 * To get mount instance using sget() vfs-routine, NILFS needs
1099 * much more information than normal filesystems to identify mount
1100 * instance. For snapshot mounts, not only a mount type (ro-mount
1101 * or rw-mount) but also a checkpoint number is required.
1102 * The results are passed in sget() using nilfs_super_data.
1106 if (nilfs_identify((char *)data
, &sd
)) {
1112 * once the super is inserted into the list by sget, s_umount
1113 * will protect the lockfs code from trying to start a snapshot
1114 * while we are mounting
1116 down(&sd
.bdev
->bd_mount_sem
);
1118 (err
= test_exclusive_mount(fs_type
, sd
.bdev
, flags
^ MS_RDONLY
))) {
1119 err
= (err
< 0) ? : -EBUSY
;
1124 * Phase-1: search any existent instance and get the_nilfs
1126 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1132 nilfs
= alloc_nilfs(sd
.bdev
);
1136 struct nilfs_sb_info
*sbi
= NILFS_SB(s
);
1139 * s_umount protects super_block from unmount process;
1140 * It covers pointers of nilfs_sb_info and the_nilfs.
1142 nilfs
= sbi
->s_nilfs
;
1144 up_write(&s
->s_umount
);
1147 * Phase-2: search specified snapshot or R/W mode super_block
1150 /* trying to get the latest checkpoint. */
1151 sd
.cno
= nilfs_last_cno(nilfs
);
1153 s2
= sget(fs_type
, nilfs_test_bdev_super2
,
1154 nilfs_set_bdev_super
, &sd
);
1155 deactivate_super(s
);
1157 * Although deactivate_super() invokes close_bdev_exclusive() at
1158 * kill_block_super(). Here, s is an existent mount; we need
1159 * one more close_bdev_exclusive() call.
1167 char b
[BDEVNAME_SIZE
];
1170 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1171 sb_set_blocksize(s
, block_size(sd
.bdev
));
1173 err
= nilfs_fill_super(s
, data
, flags
& MS_VERBOSE
, nilfs
);
1177 s
->s_flags
|= MS_ACTIVE
;
1179 } else if (!(s
->s_flags
& MS_RDONLY
)) {
1183 up(&sd
.bdev
->bd_mount_sem
);
1186 close_bdev_exclusive(sd
.bdev
, flags
);
1187 simple_set_mnt(mnt
, s
);
1191 up(&sd
.bdev
->bd_mount_sem
);
1194 close_bdev_exclusive(sd
.bdev
, flags
);
1198 up(&sd
.bdev
->bd_mount_sem
);
1200 close_bdev_exclusive(sd
.bdev
, flags
);
1205 /* Abandoning the newly allocated superblock */
1206 up(&sd
.bdev
->bd_mount_sem
);
1209 up_write(&s
->s_umount
);
1210 deactivate_super(s
);
1212 * deactivate_super() invokes close_bdev_exclusive().
1213 * We must finish all post-cleaning before this call;
1214 * put_nilfs() and unlocking bd_mount_sem need the block device.
1219 static int nilfs_test_bdev_super3(struct super_block
*s
, void *data
)
1221 struct nilfs_super_data
*sd
= data
;
1224 if (s
->s_bdev
!= sd
->bdev
)
1226 if (down_read_trylock(&s
->s_umount
)) {
1227 ret
= (s
->s_flags
& MS_RDONLY
) && s
->s_root
&&
1228 nilfs_test_opt(NILFS_SB(s
), SNAPSHOT
);
1229 up_read(&s
->s_umount
);
1231 return 0; /* ignore snapshot mounts */
1233 return !((sd
->flags
^ s
->s_flags
) & MS_RDONLY
);
1236 static int __false_bdev_super(struct super_block
*s
, void *data
)
1238 #if 0 /* XXX: workaround for lock debug. This is not good idea */
1239 up_write(&s
->s_umount
);
1245 * test_exclusive_mount - check whether an exclusive RW/RO mount exists or not.
1246 * fs_type: filesystem type
1247 * bdev: block device
1248 * flag: 0 (check rw-mount) or MS_RDONLY (check ro-mount)
1249 * res: pointer to an integer to store result
1251 * This function must be called within a section protected by bd_mount_mutex.
1253 static int test_exclusive_mount(struct file_system_type
*fs_type
,
1254 struct block_device
*bdev
, int flags
)
1256 struct super_block
*s
;
1257 struct nilfs_super_data sd
= { .flags
= flags
, .bdev
= bdev
};
1259 s
= sget(fs_type
, nilfs_test_bdev_super3
, __false_bdev_super
, &sd
);
1261 if (PTR_ERR(s
) != -EFAULT
)
1263 return 0; /* Not found */
1265 up_write(&s
->s_umount
);
1266 deactivate_super(s
);
1267 return 1; /* Found */
1270 struct file_system_type nilfs_fs_type
= {
1271 .owner
= THIS_MODULE
,
1273 .get_sb
= nilfs_get_sb
,
1274 .kill_sb
= kill_block_super
,
1275 .fs_flags
= FS_REQUIRES_DEV
,
1278 static int __init
init_nilfs_fs(void)
1282 err
= nilfs_init_inode_cache();
1286 err
= nilfs_init_transaction_cache();
1288 goto failed_inode_cache
;
1290 err
= nilfs_init_segbuf_cache();
1292 goto failed_transaction_cache
;
1294 err
= nilfs_btree_path_cache_init();
1296 goto failed_segbuf_cache
;
1298 err
= register_filesystem(&nilfs_fs_type
);
1300 goto failed_btree_path_cache
;
1304 failed_btree_path_cache
:
1305 nilfs_btree_path_cache_destroy();
1307 failed_segbuf_cache
:
1308 nilfs_destroy_segbuf_cache();
1310 failed_transaction_cache
:
1311 nilfs_destroy_transaction_cache();
1314 nilfs_destroy_inode_cache();
1320 static void __exit
exit_nilfs_fs(void)
1322 nilfs_destroy_segbuf_cache();
1323 nilfs_destroy_transaction_cache();
1324 nilfs_destroy_inode_cache();
1325 nilfs_btree_path_cache_destroy();
1326 unregister_filesystem(&nilfs_fs_type
);
1329 module_init(init_nilfs_fs
)
1330 module_exit(exit_nilfs_fs
)