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
->s_state
|= cpu_to_le16(NILFS_ERROR_FS
);
107 nilfs_commit_super(sbi
);
109 up_write(&nilfs
->ns_sem
);
111 if (nilfs_test_opt(sbi
, ERRORS_RO
)) {
112 printk(KERN_CRIT
"Remounting filesystem read-only\n");
113 sb
->s_flags
|= MS_RDONLY
;
117 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
118 panic("NILFS (device %s): panic forced after error\n",
122 void nilfs_warning(struct super_block
*sb
, const char *function
,
123 const char *fmt
, ...)
128 printk(KERN_WARNING
"NILFS warning (device %s): %s: ",
135 static struct kmem_cache
*nilfs_inode_cachep
;
137 struct inode
*nilfs_alloc_inode(struct super_block
*sb
)
139 struct nilfs_inode_info
*ii
;
141 ii
= kmem_cache_alloc(nilfs_inode_cachep
, GFP_NOFS
);
146 ii
->vfs_inode
.i_version
= 1;
147 nilfs_btnode_cache_init(&ii
->i_btnode_cache
);
148 return &ii
->vfs_inode
;
151 void nilfs_destroy_inode(struct inode
*inode
)
153 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
156 static void init_once(void *obj
)
158 struct nilfs_inode_info
*ii
= obj
;
160 INIT_LIST_HEAD(&ii
->i_dirty
);
161 #ifdef CONFIG_NILFS_XATTR
162 init_rwsem(&ii
->xattr_sem
);
164 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
165 ii
->i_bmap
= (struct nilfs_bmap
*)&ii
->i_bmap_union
;
166 inode_init_once(&ii
->vfs_inode
);
169 static int nilfs_init_inode_cache(void)
171 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
172 sizeof(struct nilfs_inode_info
),
173 0, SLAB_RECLAIM_ACCOUNT
,
176 return (nilfs_inode_cachep
== NULL
) ? -ENOMEM
: 0;
179 static inline void nilfs_destroy_inode_cache(void)
181 kmem_cache_destroy(nilfs_inode_cachep
);
184 static void nilfs_clear_inode(struct inode
*inode
)
186 struct nilfs_inode_info
*ii
= NILFS_I(inode
);
188 #ifdef CONFIG_NILFS_POSIX_ACL
189 if (ii
->i_acl
&& ii
->i_acl
!= NILFS_ACL_NOT_CACHED
) {
190 posix_acl_release(ii
->i_acl
);
191 ii
->i_acl
= NILFS_ACL_NOT_CACHED
;
193 if (ii
->i_default_acl
&& ii
->i_default_acl
!= NILFS_ACL_NOT_CACHED
) {
194 posix_acl_release(ii
->i_default_acl
);
195 ii
->i_default_acl
= NILFS_ACL_NOT_CACHED
;
199 * Free resources allocated in nilfs_read_inode(), here.
201 BUG_ON(!list_empty(&ii
->i_dirty
));
205 if (test_bit(NILFS_I_BMAP
, &ii
->i_state
))
206 nilfs_bmap_clear(ii
->i_bmap
);
208 nilfs_btnode_cache_clear(&ii
->i_btnode_cache
);
212 * nilfs_update_last_segment - change pointer to the latest segment
213 * @sbi: nilfs_sb_info
214 * @update_cno: flag whether to update checkpoint number.
216 * nilfs_update_last_segment() changes information in the super block
217 * after a partial segment is written out successfully. The super
218 * block is marked dirty. It will be written out at the next VFS sync
219 * operations such as sync_supers() and generic_shutdown_super().
221 void nilfs_update_last_segment(struct nilfs_sb_info
*sbi
, int update_cno
)
223 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
224 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
;
226 /* nilfs->sem must be locked by the caller. */
227 spin_lock(&nilfs
->ns_last_segment_lock
);
229 nilfs
->ns_last_cno
= nilfs
->ns_cno
++;
230 sbp
->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
231 sbp
->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
232 sbp
->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
233 spin_unlock(&nilfs
->ns_last_segment_lock
);
235 sbi
->s_super
->s_dirt
= 1; /* must be set if delaying the call of
236 nilfs_commit_super() */
239 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
)
241 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
243 int barrier_done
= 0;
245 if (nilfs_test_opt(sbi
, BARRIER
)) {
246 set_buffer_ordered(nilfs
->ns_sbh
);
250 set_buffer_dirty(nilfs
->ns_sbh
);
251 err
= sync_dirty_buffer(nilfs
->ns_sbh
);
252 if (err
== -EOPNOTSUPP
&& barrier_done
) {
253 nilfs_warning(sbi
->s_super
, __func__
,
254 "barrier-based sync failed. "
255 "disabling barriers\n");
256 nilfs_clear_opt(sbi
, BARRIER
);
258 clear_buffer_ordered(nilfs
->ns_sbh
);
263 "NILFS: unable to write superblock (err=%d)\n", err
);
265 clear_nilfs_discontinued(nilfs
);
266 spin_lock(&nilfs
->ns_last_segment_lock
);
267 nilfs
->ns_prot_seq
= le64_to_cpu(nilfs
->ns_sbp
->s_last_seq
);
268 spin_unlock(&nilfs
->ns_last_segment_lock
);
274 int nilfs_commit_super(struct nilfs_sb_info
*sbi
)
276 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
277 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
;
278 sector_t nfreeblocks
;
281 /* nilfs->sem must be locked by the caller. */
282 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
284 printk(KERN_ERR
"NILFS: failed to count free blocks\n");
287 sbp
->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
288 sbp
->s_wtime
= cpu_to_le64(get_seconds());
290 sbp
->s_sum
= crc32_le(nilfs
->ns_crc_seed
, (unsigned char *)sbp
,
291 le16_to_cpu(sbp
->s_bytes
));
293 sbi
->s_super
->s_dirt
= 0;
294 return nilfs_sync_super(sbi
);
297 static void nilfs_put_super(struct super_block
*sb
)
299 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
300 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
302 nilfs_detach_segment_constructor(sbi
);
304 if (!(sb
->s_flags
& MS_RDONLY
)) {
305 down_write(&nilfs
->ns_sem
);
306 nilfs
->ns_sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
307 nilfs_commit_super(sbi
);
308 up_write(&nilfs
->ns_sem
);
311 nilfs_detach_checkpoint(sbi
);
312 put_nilfs(sbi
->s_nilfs
);
314 sb
->s_fs_info
= NULL
;
319 * nilfs_write_super - write super block(s) of NILFS
322 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
323 * clears s_dirt. This function is called in the section protected by
326 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
327 * of the struct the_nilfs. Lock order must be as follows:
330 * 2. down_write(&nilfs->ns_sem)
332 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
333 * of the super block (nilfs->ns_sbp).
335 * In most cases, VFS functions call lock_super() before calling these
336 * methods. So we must be careful not to bring on deadlocks when using
337 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
339 * Note that order of lock_kernel() and lock_super() depends on contexts
340 * of VFS. We should also note that lock_kernel() can be used in its
341 * protective section and only the outermost one has an effect.
343 static void nilfs_write_super(struct super_block
*sb
)
345 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
346 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
348 down_write(&nilfs
->ns_sem
);
349 if (!(sb
->s_flags
& MS_RDONLY
))
350 nilfs_commit_super(sbi
);
352 up_write(&nilfs
->ns_sem
);
355 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
359 /* This function is called when super block should be written back */
361 err
= nilfs_construct_segment(sb
);
365 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
367 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
368 struct nilfs_checkpoint
*raw_cp
;
369 struct buffer_head
*bh_cp
;
372 down_write(&nilfs
->ns_sem
);
373 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
374 up_write(&nilfs
->ns_sem
);
376 sbi
->s_ifile
= nilfs_mdt_new(
377 nilfs
, sbi
->s_super
, NILFS_IFILE_INO
, NILFS_IFILE_GFP
);
381 err
= nilfs_palloc_init_blockgroup(sbi
->s_ifile
, nilfs
->ns_inode_size
);
385 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
388 if (err
== -ENOENT
|| err
== -EINVAL
) {
390 "NILFS: Invalid checkpoint "
391 "(checkpoint number=%llu)\n",
392 (unsigned long long)cno
);
397 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
400 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
401 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
403 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
407 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
409 nilfs_mdt_destroy(sbi
->s_ifile
);
412 down_write(&nilfs
->ns_sem
);
413 list_del_init(&sbi
->s_list
);
414 up_write(&nilfs
->ns_sem
);
419 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
421 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
423 nilfs_mdt_clear(sbi
->s_ifile
);
424 nilfs_mdt_destroy(sbi
->s_ifile
);
426 down_write(&nilfs
->ns_sem
);
427 list_del_init(&sbi
->s_list
);
428 up_write(&nilfs
->ns_sem
);
431 static int nilfs_mark_recovery_complete(struct nilfs_sb_info
*sbi
)
433 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
436 down_write(&nilfs
->ns_sem
);
437 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
438 nilfs
->ns_mount_state
|= NILFS_VALID_FS
;
439 err
= nilfs_commit_super(sbi
);
441 printk(KERN_INFO
"NILFS: recovery complete.\n");
443 up_write(&nilfs
->ns_sem
);
447 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
449 struct super_block
*sb
= dentry
->d_sb
;
450 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
451 unsigned long long blocks
;
452 unsigned long overhead
;
453 unsigned long nrsvblocks
;
454 sector_t nfreeblocks
;
455 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
459 * Compute all of the segment blocks
461 * The blocks before first segment and after last segment
464 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
465 - nilfs
->ns_first_data_block
;
466 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
469 * Compute the overhead
471 * When distributing meta data blocks outside semgent structure,
472 * We must count them as the overhead.
476 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
480 buf
->f_type
= NILFS_SUPER_MAGIC
;
481 buf
->f_bsize
= sb
->s_blocksize
;
482 buf
->f_blocks
= blocks
- overhead
;
483 buf
->f_bfree
= nfreeblocks
;
484 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
485 (buf
->f_bfree
- nrsvblocks
) : 0;
486 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
487 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
488 buf
->f_namelen
= NILFS_NAME_LEN
;
492 static struct super_operations nilfs_sops
= {
493 .alloc_inode
= nilfs_alloc_inode
,
494 .destroy_inode
= nilfs_destroy_inode
,
495 .dirty_inode
= nilfs_dirty_inode
,
496 /* .write_inode = nilfs_write_inode, */
497 /* .put_inode = nilfs_put_inode, */
498 /* .drop_inode = nilfs_drop_inode, */
499 .delete_inode
= nilfs_delete_inode
,
500 .put_super
= nilfs_put_super
,
501 .write_super
= nilfs_write_super
,
502 .sync_fs
= nilfs_sync_fs
,
503 /* .write_super_lockfs */
505 .statfs
= nilfs_statfs
,
506 .remount_fs
= nilfs_remount
,
507 .clear_inode
= nilfs_clear_inode
,
512 static struct inode
*
513 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
517 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
518 ino
!= NILFS_SKETCH_INO
)
519 return ERR_PTR(-ESTALE
);
521 inode
= nilfs_iget(sb
, ino
);
523 return ERR_CAST(inode
);
524 if (generation
&& inode
->i_generation
!= generation
) {
526 return ERR_PTR(-ESTALE
);
532 static struct dentry
*
533 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
536 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
537 nilfs_nfs_get_inode
);
540 static struct dentry
*
541 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
544 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
545 nilfs_nfs_get_inode
);
548 static struct export_operations nilfs_export_ops
= {
549 .fh_to_dentry
= nilfs_fh_to_dentry
,
550 .fh_to_parent
= nilfs_fh_to_parent
,
551 .get_parent
= nilfs_get_parent
,
555 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
556 Opt_barrier
, Opt_snapshot
, Opt_order
,
560 static match_table_t tokens
= {
561 {Opt_err_cont
, "errors=continue"},
562 {Opt_err_panic
, "errors=panic"},
563 {Opt_err_ro
, "errors=remount-ro"},
564 {Opt_barrier
, "barrier=%s"},
565 {Opt_snapshot
, "cp=%u"},
566 {Opt_order
, "order=%s"},
570 static int match_bool(substring_t
*s
, int *result
)
572 int len
= s
->to
- s
->from
;
574 if (strncmp(s
->from
, "on", len
) == 0)
576 else if (strncmp(s
->from
, "off", len
) == 0)
583 static int parse_options(char *options
, struct super_block
*sb
)
585 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
587 substring_t args
[MAX_OPT_ARGS
];
593 while ((p
= strsep(&options
, ",")) != NULL
) {
598 token
= match_token(p
, tokens
, args
);
601 if (match_bool(&args
[0], &option
))
604 nilfs_set_opt(sbi
, BARRIER
);
606 nilfs_clear_opt(sbi
, BARRIER
);
609 if (strcmp(args
[0].from
, "relaxed") == 0)
610 /* Ordered data semantics */
611 nilfs_clear_opt(sbi
, STRICT_ORDER
);
612 else if (strcmp(args
[0].from
, "strict") == 0)
613 /* Strict in-order semantics */
614 nilfs_set_opt(sbi
, STRICT_ORDER
);
619 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
622 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
625 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
628 if (match_int(&args
[0], &option
) || option
<= 0)
630 if (!(sb
->s_flags
& MS_RDONLY
))
632 sbi
->s_snapshot_cno
= option
;
633 nilfs_set_opt(sbi
, SNAPSHOT
);
637 "NILFS: Unrecognized mount option \"%s\"\n", p
);
645 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
646 struct nilfs_super_block
*sbp
)
649 NILFS_MOUNT_ERRORS_CONT
| NILFS_MOUNT_BARRIER
;
652 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
654 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
655 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
;
656 int max_mnt_count
= le16_to_cpu(sbp
->s_max_mnt_count
);
657 int mnt_count
= le16_to_cpu(sbp
->s_mnt_count
);
659 /* nilfs->sem must be locked by the caller. */
660 if (!(nilfs
->ns_mount_state
& NILFS_VALID_FS
)) {
661 printk(KERN_WARNING
"NILFS warning: mounting unchecked fs\n");
662 } else if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
664 "NILFS warning: mounting fs with errors\n");
666 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
668 "NILFS warning: maximal mount count reached\n");
672 sbp
->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
674 sbp
->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
675 sbp
->s_state
= cpu_to_le16(le16_to_cpu(sbp
->s_state
) & ~NILFS_VALID_FS
);
676 sbp
->s_mtime
= cpu_to_le64(get_seconds());
677 return nilfs_commit_super(sbi
);
680 struct nilfs_super_block
*
681 nilfs_load_super_block(struct super_block
*sb
, struct buffer_head
**pbh
)
684 unsigned long offset
, sb_index
;
687 * Adjusting block size
688 * Blocksize will be enlarged when it is smaller than hardware
690 * Disk format of superblock does not change.
692 blocksize
= sb_min_blocksize(sb
, BLOCK_SIZE
);
695 "NILFS: unable to set blocksize of superblock\n");
698 sb_index
= NILFS_SB_OFFSET_BYTES
/ blocksize
;
699 offset
= NILFS_SB_OFFSET_BYTES
% blocksize
;
701 *pbh
= sb_bread(sb
, sb_index
);
703 printk(KERN_ERR
"NILFS: unable to read superblock\n");
706 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
709 struct nilfs_super_block
*
710 nilfs_reload_super_block(struct super_block
*sb
, struct buffer_head
**pbh
,
713 struct nilfs_super_block
*sbp
;
714 unsigned long offset
, sb_index
;
715 int hw_blocksize
= bdev_hardsect_size(sb
->s_bdev
);
717 if (blocksize
< hw_blocksize
) {
719 "NILFS: blocksize %d too small for device "
720 "(sector-size = %d).\n",
721 blocksize
, hw_blocksize
);
725 sb_set_blocksize(sb
, blocksize
);
727 sb_index
= NILFS_SB_OFFSET_BYTES
/ blocksize
;
728 offset
= NILFS_SB_OFFSET_BYTES
% blocksize
;
730 *pbh
= sb_bread(sb
, sb_index
);
733 "NILFS: cannot read superblock on 2nd try.\n");
737 sbp
= (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
738 if (sbp
->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
740 "NILFS: !? Magic mismatch on 2nd try.\n");
752 int nilfs_store_magic_and_option(struct super_block
*sb
,
753 struct nilfs_super_block
*sbp
,
756 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
758 /* trying to fill super (1st stage) */
759 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
761 /* FS independent flags */
762 #ifdef NILFS_ATIME_DISABLE
763 sb
->s_flags
|= MS_NOATIME
;
766 if (sb
->s_magic
!= NILFS_SUPER_MAGIC
) {
767 printk("NILFS: Can't find nilfs on dev %s.\n", sb
->s_id
);
771 nilfs_set_default_options(sbi
, sbp
);
773 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
774 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
775 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
776 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
778 if (!parse_options(data
, sb
))
785 * nilfs_fill_super() - initialize a super block instance
787 * @data: mount options
788 * @silent: silent mode flag
789 * @nilfs: the_nilfs struct
791 * This function is called exclusively by bd_mount_mutex.
792 * So, the recovery process is protected from other simultaneous mounts.
795 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
796 struct the_nilfs
*nilfs
)
798 struct nilfs_sb_info
*sbi
;
803 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
810 sbi
->s_nilfs
= nilfs
;
813 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
817 spin_lock_init(&sbi
->s_inode_lock
);
818 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
819 INIT_LIST_HEAD(&sbi
->s_list
);
822 * Following initialization is overlapped because
823 * nilfs_sb_info structure has been cleared at the beginning.
824 * But we reserve them to keep our interest and make ready
825 * for the future change.
827 get_random_bytes(&sbi
->s_next_generation
,
828 sizeof(sbi
->s_next_generation
));
829 spin_lock_init(&sbi
->s_next_gen_lock
);
831 sb
->s_op
= &nilfs_sops
;
832 sb
->s_export_op
= &nilfs_export_ops
;
835 if (!nilfs_loaded(nilfs
)) {
836 err
= load_nilfs(nilfs
, sbi
);
840 cno
= nilfs_last_cno(nilfs
);
842 if (sb
->s_flags
& MS_RDONLY
) {
843 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
844 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
845 sbi
->s_snapshot_cno
);
850 "NILFS: The specified checkpoint is "
852 "(checkpoint number=%llu).\n",
853 (unsigned long long)sbi
->s_snapshot_cno
);
857 cno
= sbi
->s_snapshot_cno
;
859 /* Read-only mount */
860 sbi
->s_snapshot_cno
= cno
;
863 err
= nilfs_attach_checkpoint(sbi
, cno
);
865 printk(KERN_ERR
"NILFS: error loading a checkpoint"
866 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
870 if (!(sb
->s_flags
& MS_RDONLY
)) {
871 err
= nilfs_attach_segment_constructor(sbi
, NULL
);
873 goto failed_checkpoint
;
876 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
878 printk(KERN_ERR
"NILFS: get root inode failed\n");
882 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
884 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
888 sb
->s_root
= d_alloc_root(root
);
891 printk(KERN_ERR
"NILFS: get root dentry failed\n");
896 if (!(sb
->s_flags
& MS_RDONLY
)) {
897 down_write(&nilfs
->ns_sem
);
898 nilfs_setup_super(sbi
);
899 up_write(&nilfs
->ns_sem
);
902 err
= nilfs_mark_recovery_complete(sbi
);
904 printk(KERN_ERR
"NILFS: recovery failed.\n");
915 nilfs_detach_segment_constructor(sbi
);
918 nilfs_detach_checkpoint(sbi
);
922 sb
->s_fs_info
= NULL
;
927 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
929 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
930 struct nilfs_super_block
*sbp
;
931 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
932 unsigned long old_sb_flags
;
933 struct nilfs_mount_options old_opts
;
936 old_sb_flags
= sb
->s_flags
;
937 old_opts
.mount_opt
= sbi
->s_mount_opt
;
938 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
940 if (!parse_options(data
, sb
)) {
944 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
946 if ((*flags
& MS_RDONLY
) &&
947 sbi
->s_snapshot_cno
!= old_opts
.snapshot_cno
) {
948 printk(KERN_WARNING
"NILFS (device %s): couldn't "
949 "remount to a different snapshot. \n",
955 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
957 if (*flags
& MS_RDONLY
) {
958 /* Shutting down the segment constructor */
959 nilfs_detach_segment_constructor(sbi
);
960 sb
->s_flags
|= MS_RDONLY
;
962 sbi
->s_snapshot_cno
= nilfs_last_cno(nilfs
);
963 /* nilfs_set_opt(sbi, SNAPSHOT); */
966 * Remounting a valid RW partition RDONLY, so set
967 * the RDONLY flag and then mark the partition as valid again.
969 down_write(&nilfs
->ns_sem
);
971 if (!(sbp
->s_state
& le16_to_cpu(NILFS_VALID_FS
)) &&
972 (nilfs
->ns_mount_state
& NILFS_VALID_FS
))
973 sbp
->s_state
= cpu_to_le16(nilfs
->ns_mount_state
);
974 sbp
->s_mtime
= cpu_to_le64(get_seconds());
975 nilfs_commit_super(sbi
);
976 up_write(&nilfs
->ns_sem
);
979 * Mounting a RDONLY partition read-write, so reread and
980 * store the current valid flag. (It may have been changed
981 * by fsck since we originally mounted the partition.)
983 down(&sb
->s_bdev
->bd_mount_sem
);
984 /* Check existing RW-mount */
985 if (test_exclusive_mount(sb
->s_type
, sb
->s_bdev
, 0)) {
986 printk(KERN_WARNING
"NILFS (device %s): couldn't "
987 "remount because a RW-mount exists.\n",
990 goto rw_remount_failed
;
992 if (sbi
->s_snapshot_cno
!= nilfs_last_cno(nilfs
)) {
993 printk(KERN_WARNING
"NILFS (device %s): couldn't "
994 "remount because the current RO-mount is not "
998 goto rw_remount_failed
;
1000 sb
->s_flags
&= ~MS_RDONLY
;
1001 nilfs_clear_opt(sbi
, SNAPSHOT
);
1002 sbi
->s_snapshot_cno
= 0;
1004 err
= nilfs_attach_segment_constructor(sbi
, NULL
);
1006 goto rw_remount_failed
;
1008 down_write(&nilfs
->ns_sem
);
1009 nilfs_setup_super(sbi
);
1010 up_write(&nilfs
->ns_sem
);
1012 up(&sb
->s_bdev
->bd_mount_sem
);
1018 up(&sb
->s_bdev
->bd_mount_sem
);
1020 sb
->s_flags
= old_sb_flags
;
1021 sbi
->s_mount_opt
= old_opts
.mount_opt
;
1022 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
1026 struct nilfs_super_data
{
1027 struct block_device
*bdev
;
1033 * nilfs_identify - pre-read mount options needed to identify mount instance
1034 * @data: mount options
1035 * @sd: nilfs_super_data
1037 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1039 char *p
, *options
= data
;
1040 substring_t args
[MAX_OPT_ARGS
];
1045 p
= strsep(&options
, ",");
1046 if (p
!= NULL
&& *p
) {
1047 token
= match_token(p
, tokens
, args
);
1048 if (token
== Opt_snapshot
) {
1049 if (!(sd
->flags
& MS_RDONLY
))
1052 ret
= match_int(&args
[0], &option
);
1063 "NILFS: invalid mount option: %s\n", p
);
1067 BUG_ON(options
== data
);
1068 *(options
- 1) = ',';
1073 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1075 struct nilfs_super_data
*sd
= data
;
1077 s
->s_bdev
= sd
->bdev
;
1078 s
->s_dev
= s
->s_bdev
->bd_dev
;
1082 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1084 struct nilfs_super_data
*sd
= data
;
1086 return s
->s_bdev
== sd
->bdev
;
1089 static int nilfs_test_bdev_super2(struct super_block
*s
, void *data
)
1091 struct nilfs_super_data
*sd
= data
;
1094 if (s
->s_bdev
!= sd
->bdev
)
1097 if (!((s
->s_flags
| sd
->flags
) & MS_RDONLY
))
1098 return 1; /* Reuse an old R/W-mode super_block */
1100 if (s
->s_flags
& sd
->flags
& MS_RDONLY
) {
1101 if (down_read_trylock(&s
->s_umount
)) {
1103 (sd
->cno
== NILFS_SB(s
)->s_snapshot_cno
);
1104 up_read(&s
->s_umount
);
1106 * This path is locked with sb_lock by sget().
1107 * So, drop_super() causes deadlock.
1116 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1117 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1119 struct nilfs_super_data sd
;
1120 struct super_block
*s
, *s2
;
1121 struct the_nilfs
*nilfs
= NULL
;
1122 int err
, need_to_close
= 1;
1124 sd
.bdev
= open_bdev_exclusive(dev_name
, flags
, fs_type
);
1125 if (IS_ERR(sd
.bdev
))
1126 return PTR_ERR(sd
.bdev
);
1129 * To get mount instance using sget() vfs-routine, NILFS needs
1130 * much more information than normal filesystems to identify mount
1131 * instance. For snapshot mounts, not only a mount type (ro-mount
1132 * or rw-mount) but also a checkpoint number is required.
1133 * The results are passed in sget() using nilfs_super_data.
1137 if (nilfs_identify((char *)data
, &sd
)) {
1143 * once the super is inserted into the list by sget, s_umount
1144 * will protect the lockfs code from trying to start a snapshot
1145 * while we are mounting
1147 down(&sd
.bdev
->bd_mount_sem
);
1149 (err
= test_exclusive_mount(fs_type
, sd
.bdev
, flags
^ MS_RDONLY
))) {
1150 err
= (err
< 0) ? : -EBUSY
;
1155 * Phase-1: search any existent instance and get the_nilfs
1157 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1163 nilfs
= alloc_nilfs(sd
.bdev
);
1167 struct nilfs_sb_info
*sbi
= NILFS_SB(s
);
1170 * s_umount protects super_block from unmount process;
1171 * It covers pointers of nilfs_sb_info and the_nilfs.
1173 nilfs
= sbi
->s_nilfs
;
1175 up_write(&s
->s_umount
);
1178 * Phase-2: search specified snapshot or R/W mode super_block
1181 /* trying to get the latest checkpoint. */
1182 sd
.cno
= nilfs_last_cno(nilfs
);
1184 s2
= sget(fs_type
, nilfs_test_bdev_super2
,
1185 nilfs_set_bdev_super
, &sd
);
1186 deactivate_super(s
);
1188 * Although deactivate_super() invokes close_bdev_exclusive() at
1189 * kill_block_super(). Here, s is an existent mount; we need
1190 * one more close_bdev_exclusive() call.
1198 char b
[BDEVNAME_SIZE
];
1201 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1202 sb_set_blocksize(s
, block_size(sd
.bdev
));
1204 err
= nilfs_fill_super(s
, data
, flags
& MS_VERBOSE
, nilfs
);
1208 s
->s_flags
|= MS_ACTIVE
;
1210 } else if (!(s
->s_flags
& MS_RDONLY
)) {
1214 up(&sd
.bdev
->bd_mount_sem
);
1217 close_bdev_exclusive(sd
.bdev
, flags
);
1218 simple_set_mnt(mnt
, s
);
1222 up(&sd
.bdev
->bd_mount_sem
);
1225 close_bdev_exclusive(sd
.bdev
, flags
);
1229 up(&sd
.bdev
->bd_mount_sem
);
1231 close_bdev_exclusive(sd
.bdev
, flags
);
1236 /* Abandoning the newly allocated superblock */
1237 up(&sd
.bdev
->bd_mount_sem
);
1240 up_write(&s
->s_umount
);
1241 deactivate_super(s
);
1243 * deactivate_super() invokes close_bdev_exclusive().
1244 * We must finish all post-cleaning before this call;
1245 * put_nilfs() and unlocking bd_mount_sem need the block device.
1250 static int nilfs_test_bdev_super3(struct super_block
*s
, void *data
)
1252 struct nilfs_super_data
*sd
= data
;
1255 if (s
->s_bdev
!= sd
->bdev
)
1257 if (down_read_trylock(&s
->s_umount
)) {
1258 ret
= (s
->s_flags
& MS_RDONLY
) && s
->s_root
&&
1259 nilfs_test_opt(NILFS_SB(s
), SNAPSHOT
);
1260 up_read(&s
->s_umount
);
1262 return 0; /* ignore snapshot mounts */
1264 return !((sd
->flags
^ s
->s_flags
) & MS_RDONLY
);
1267 static int __false_bdev_super(struct super_block
*s
, void *data
)
1269 #if 0 /* XXX: workaround for lock debug. This is not good idea */
1270 up_write(&s
->s_umount
);
1276 * test_exclusive_mount - check whether an exclusive RW/RO mount exists or not.
1277 * fs_type: filesystem type
1278 * bdev: block device
1279 * flag: 0 (check rw-mount) or MS_RDONLY (check ro-mount)
1280 * res: pointer to an integer to store result
1282 * This function must be called within a section protected by bd_mount_mutex.
1284 static int test_exclusive_mount(struct file_system_type
*fs_type
,
1285 struct block_device
*bdev
, int flags
)
1287 struct super_block
*s
;
1288 struct nilfs_super_data sd
= { .flags
= flags
, .bdev
= bdev
};
1290 s
= sget(fs_type
, nilfs_test_bdev_super3
, __false_bdev_super
, &sd
);
1292 if (PTR_ERR(s
) != -EFAULT
)
1294 return 0; /* Not found */
1296 up_write(&s
->s_umount
);
1297 deactivate_super(s
);
1298 return 1; /* Found */
1301 struct file_system_type nilfs_fs_type
= {
1302 .owner
= THIS_MODULE
,
1304 .get_sb
= nilfs_get_sb
,
1305 .kill_sb
= kill_block_super
,
1306 .fs_flags
= FS_REQUIRES_DEV
,
1309 static int __init
init_nilfs_fs(void)
1313 err
= nilfs_init_inode_cache();
1317 err
= nilfs_init_transaction_cache();
1319 goto failed_inode_cache
;
1321 err
= nilfs_init_segbuf_cache();
1323 goto failed_transaction_cache
;
1325 err
= nilfs_btree_path_cache_init();
1327 goto failed_segbuf_cache
;
1329 err
= register_filesystem(&nilfs_fs_type
);
1331 goto failed_btree_path_cache
;
1335 failed_btree_path_cache
:
1336 nilfs_btree_path_cache_destroy();
1338 failed_segbuf_cache
:
1339 nilfs_destroy_segbuf_cache();
1341 failed_transaction_cache
:
1342 nilfs_destroy_transaction_cache();
1345 nilfs_destroy_inode_cache();
1351 static void __exit
exit_nilfs_fs(void)
1353 nilfs_destroy_segbuf_cache();
1354 nilfs_destroy_transaction_cache();
1355 nilfs_destroy_inode_cache();
1356 nilfs_btree_path_cache_destroy();
1357 unregister_filesystem(&nilfs_fs_type
);
1360 module_init(init_nilfs_fs
)
1361 module_exit(exit_nilfs_fs
)