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>
53 #include <linux/seq_file.h>
54 #include <linux/mount.h>
67 MODULE_AUTHOR("NTT Corp.");
68 MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
70 MODULE_LICENSE("GPL");
72 struct kmem_cache
*nilfs_inode_cachep
;
73 struct kmem_cache
*nilfs_transaction_cachep
;
74 struct kmem_cache
*nilfs_segbuf_cachep
;
75 struct kmem_cache
*nilfs_btree_path_cache
;
77 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
);
79 static void nilfs_set_error(struct nilfs_sb_info
*sbi
)
81 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
82 struct nilfs_super_block
**sbp
;
84 down_write(&nilfs
->ns_sem
);
85 if (!(nilfs
->ns_mount_state
& NILFS_ERROR_FS
)) {
86 nilfs
->ns_mount_state
|= NILFS_ERROR_FS
;
87 sbp
= nilfs_prepare_super(sbi
, 0);
89 sbp
[0]->s_state
|= cpu_to_le16(NILFS_ERROR_FS
);
91 sbp
[1]->s_state
|= cpu_to_le16(NILFS_ERROR_FS
);
92 nilfs_commit_super(sbi
, NILFS_SB_COMMIT_ALL
);
95 up_write(&nilfs
->ns_sem
);
99 * nilfs_error() - report failure condition on a filesystem
101 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
102 * reporting an error message. It should be called when NILFS detects
103 * incoherences or defects of meta data on disk. As for sustainable
104 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
105 * function should be used instead.
107 * The segment constructor must not call this function because it can
110 void nilfs_error(struct super_block
*sb
, const char *function
,
111 const char *fmt
, ...)
113 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
117 printk(KERN_CRIT
"NILFS error (device %s): %s: ", sb
->s_id
, function
);
122 if (!(sb
->s_flags
& MS_RDONLY
)) {
123 nilfs_set_error(sbi
);
125 if (nilfs_test_opt(sbi
, ERRORS_RO
)) {
126 printk(KERN_CRIT
"Remounting filesystem read-only\n");
127 sb
->s_flags
|= MS_RDONLY
;
131 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
132 panic("NILFS (device %s): panic forced after error\n",
136 void nilfs_warning(struct super_block
*sb
, const char *function
,
137 const char *fmt
, ...)
142 printk(KERN_WARNING
"NILFS warning (device %s): %s: ",
150 struct inode
*nilfs_alloc_inode_common(struct the_nilfs
*nilfs
)
152 struct nilfs_inode_info
*ii
;
154 ii
= kmem_cache_alloc(nilfs_inode_cachep
, GFP_NOFS
);
159 ii
->vfs_inode
.i_version
= 1;
160 nilfs_btnode_cache_init(&ii
->i_btnode_cache
, nilfs
->ns_bdi
);
161 return &ii
->vfs_inode
;
164 struct inode
*nilfs_alloc_inode(struct super_block
*sb
)
166 return nilfs_alloc_inode_common(NILFS_SB(sb
)->s_nilfs
);
169 void nilfs_destroy_inode(struct inode
*inode
)
171 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
174 static void nilfs_clear_inode(struct inode
*inode
)
176 struct nilfs_inode_info
*ii
= NILFS_I(inode
);
179 * Free resources allocated in nilfs_read_inode(), here.
181 BUG_ON(!list_empty(&ii
->i_dirty
));
185 if (test_bit(NILFS_I_BMAP
, &ii
->i_state
))
186 nilfs_bmap_clear(ii
->i_bmap
);
188 nilfs_btnode_cache_clear(&ii
->i_btnode_cache
);
191 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
, int flag
)
193 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
195 int barrier_done
= 0;
197 if (nilfs_test_opt(sbi
, BARRIER
)) {
198 set_buffer_ordered(nilfs
->ns_sbh
[0]);
202 set_buffer_dirty(nilfs
->ns_sbh
[0]);
203 err
= sync_dirty_buffer(nilfs
->ns_sbh
[0]);
204 if (err
== -EOPNOTSUPP
&& barrier_done
) {
205 nilfs_warning(sbi
->s_super
, __func__
,
206 "barrier-based sync failed. "
207 "disabling barriers\n");
208 nilfs_clear_opt(sbi
, BARRIER
);
210 clear_buffer_ordered(nilfs
->ns_sbh
[0]);
215 "NILFS: unable to write superblock (err=%d)\n", err
);
216 if (err
== -EIO
&& nilfs
->ns_sbh
[1]) {
218 * sbp[0] points to newer log than sbp[1],
219 * so copy sbp[0] to sbp[1] to take over sbp[0].
221 memcpy(nilfs
->ns_sbp
[1], nilfs
->ns_sbp
[0],
223 nilfs_fall_back_super_block(nilfs
);
227 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
229 nilfs
->ns_sbwcount
++;
232 * The latest segment becomes trailable from the position
233 * written in superblock.
235 clear_nilfs_discontinued(nilfs
);
237 /* update GC protection for recent segments */
238 if (nilfs
->ns_sbh
[1]) {
239 if (flag
== NILFS_SB_COMMIT_ALL
) {
240 set_buffer_dirty(nilfs
->ns_sbh
[1]);
241 if (sync_dirty_buffer(nilfs
->ns_sbh
[1]) < 0)
244 if (le64_to_cpu(nilfs
->ns_sbp
[1]->s_last_cno
) <
245 le64_to_cpu(nilfs
->ns_sbp
[0]->s_last_cno
))
246 sbp
= nilfs
->ns_sbp
[1];
249 spin_lock(&nilfs
->ns_last_segment_lock
);
250 nilfs
->ns_prot_seq
= le64_to_cpu(sbp
->s_last_seq
);
251 spin_unlock(&nilfs
->ns_last_segment_lock
);
257 void nilfs_set_log_cursor(struct nilfs_super_block
*sbp
,
258 struct the_nilfs
*nilfs
)
260 sector_t nfreeblocks
;
262 /* nilfs->ns_sem must be locked by the caller. */
263 nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
264 sbp
->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
266 spin_lock(&nilfs
->ns_last_segment_lock
);
267 sbp
->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
268 sbp
->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
269 sbp
->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
270 spin_unlock(&nilfs
->ns_last_segment_lock
);
273 struct nilfs_super_block
**nilfs_prepare_super(struct nilfs_sb_info
*sbi
,
276 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
277 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
279 /* nilfs->ns_sem must be locked by the caller. */
280 if (sbp
[0]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
282 sbp
[1]->s_magic
== cpu_to_le16(NILFS_SUPER_MAGIC
)) {
283 memcpy(sbp
[0], sbp
[1], nilfs
->ns_sbsize
);
285 printk(KERN_CRIT
"NILFS: superblock broke on dev %s\n",
290 sbp
[1]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
291 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
295 nilfs_swap_super_block(nilfs
);
300 int nilfs_commit_super(struct nilfs_sb_info
*sbi
, int flag
)
302 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
303 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
306 /* nilfs->ns_sem must be locked by the caller. */
308 nilfs
->ns_sbwtime
= t
;
309 sbp
[0]->s_wtime
= cpu_to_le64(t
);
311 sbp
[0]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
312 (unsigned char *)sbp
[0],
314 if (flag
== NILFS_SB_COMMIT_ALL
&& sbp
[1]) {
315 sbp
[1]->s_wtime
= sbp
[0]->s_wtime
;
317 sbp
[1]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
318 (unsigned char *)sbp
[1],
321 clear_nilfs_sb_dirty(nilfs
);
322 return nilfs_sync_super(sbi
, flag
);
326 * nilfs_cleanup_super() - write filesystem state for cleanup
327 * @sbi: nilfs_sb_info to be unmounted or degraded to read-only
329 * This function restores state flags in the on-disk super block.
330 * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the
331 * filesystem was not clean previously.
333 int nilfs_cleanup_super(struct nilfs_sb_info
*sbi
)
335 struct nilfs_super_block
**sbp
;
336 int flag
= NILFS_SB_COMMIT
;
339 sbp
= nilfs_prepare_super(sbi
, 0);
341 sbp
[0]->s_state
= cpu_to_le16(sbi
->s_nilfs
->ns_mount_state
);
342 nilfs_set_log_cursor(sbp
[0], sbi
->s_nilfs
);
343 if (sbp
[1] && sbp
[0]->s_last_cno
== sbp
[1]->s_last_cno
) {
345 * make the "clean" flag also to the opposite
346 * super block if both super blocks point to
347 * the same checkpoint.
349 sbp
[1]->s_state
= sbp
[0]->s_state
;
350 flag
= NILFS_SB_COMMIT_ALL
;
352 ret
= nilfs_commit_super(sbi
, flag
);
357 static void nilfs_put_super(struct super_block
*sb
)
359 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
360 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
364 nilfs_detach_segment_constructor(sbi
);
366 if (!(sb
->s_flags
& MS_RDONLY
)) {
367 down_write(&nilfs
->ns_sem
);
368 nilfs_cleanup_super(sbi
);
369 up_write(&nilfs
->ns_sem
);
371 down_write(&nilfs
->ns_super_sem
);
372 if (nilfs
->ns_current
== sbi
)
373 nilfs
->ns_current
= NULL
;
374 up_write(&nilfs
->ns_super_sem
);
376 nilfs_detach_checkpoint(sbi
);
377 put_nilfs(sbi
->s_nilfs
);
379 sb
->s_fs_info
= NULL
;
380 nilfs_put_sbinfo(sbi
);
385 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
387 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
388 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
389 struct nilfs_super_block
**sbp
;
392 /* This function is called when super block should be written back */
394 err
= nilfs_construct_segment(sb
);
396 down_write(&nilfs
->ns_sem
);
397 if (nilfs_sb_dirty(nilfs
)) {
398 sbp
= nilfs_prepare_super(sbi
, nilfs_sb_will_flip(nilfs
));
400 nilfs_set_log_cursor(sbp
[0], nilfs
);
401 nilfs_commit_super(sbi
, NILFS_SB_COMMIT
);
404 up_write(&nilfs
->ns_sem
);
409 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
)
411 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
412 struct nilfs_checkpoint
*raw_cp
;
413 struct buffer_head
*bh_cp
;
416 down_write(&nilfs
->ns_super_sem
);
417 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
418 up_write(&nilfs
->ns_super_sem
);
420 sbi
->s_ifile
= nilfs_ifile_new(sbi
, nilfs
->ns_inode_size
);
424 down_read(&nilfs
->ns_segctor_sem
);
425 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
427 up_read(&nilfs
->ns_segctor_sem
);
429 if (err
== -ENOENT
|| err
== -EINVAL
) {
431 "NILFS: Invalid checkpoint "
432 "(checkpoint number=%llu)\n",
433 (unsigned long long)cno
);
438 err
= nilfs_read_inode_common(sbi
->s_ifile
, &raw_cp
->cp_ifile_inode
);
441 atomic_set(&sbi
->s_inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
442 atomic_set(&sbi
->s_blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
444 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
448 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
450 nilfs_mdt_destroy(sbi
->s_ifile
);
453 down_write(&nilfs
->ns_super_sem
);
454 list_del_init(&sbi
->s_list
);
455 up_write(&nilfs
->ns_super_sem
);
460 void nilfs_detach_checkpoint(struct nilfs_sb_info
*sbi
)
462 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
464 nilfs_mdt_destroy(sbi
->s_ifile
);
466 down_write(&nilfs
->ns_super_sem
);
467 list_del_init(&sbi
->s_list
);
468 up_write(&nilfs
->ns_super_sem
);
471 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
473 struct super_block
*sb
= dentry
->d_sb
;
474 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
475 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
476 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
477 unsigned long long blocks
;
478 unsigned long overhead
;
479 unsigned long nrsvblocks
;
480 sector_t nfreeblocks
;
484 * Compute all of the segment blocks
486 * The blocks before first segment and after last segment
489 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
490 - nilfs
->ns_first_data_block
;
491 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
494 * Compute the overhead
496 * When distributing meta data blocks outside segment structure,
497 * We must count them as the overhead.
501 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
505 buf
->f_type
= NILFS_SUPER_MAGIC
;
506 buf
->f_bsize
= sb
->s_blocksize
;
507 buf
->f_blocks
= blocks
- overhead
;
508 buf
->f_bfree
= nfreeblocks
;
509 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
510 (buf
->f_bfree
- nrsvblocks
) : 0;
511 buf
->f_files
= atomic_read(&sbi
->s_inodes_count
);
512 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
513 buf
->f_namelen
= NILFS_NAME_LEN
;
514 buf
->f_fsid
.val
[0] = (u32
)id
;
515 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
520 static int nilfs_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
522 struct super_block
*sb
= vfs
->mnt_sb
;
523 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
525 if (!nilfs_test_opt(sbi
, BARRIER
))
526 seq_puts(seq
, ",nobarrier");
527 if (nilfs_test_opt(sbi
, SNAPSHOT
))
528 seq_printf(seq
, ",cp=%llu",
529 (unsigned long long int)sbi
->s_snapshot_cno
);
530 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
531 seq_puts(seq
, ",errors=panic");
532 if (nilfs_test_opt(sbi
, ERRORS_CONT
))
533 seq_puts(seq
, ",errors=continue");
534 if (nilfs_test_opt(sbi
, STRICT_ORDER
))
535 seq_puts(seq
, ",order=strict");
536 if (nilfs_test_opt(sbi
, NORECOVERY
))
537 seq_puts(seq
, ",norecovery");
538 if (nilfs_test_opt(sbi
, DISCARD
))
539 seq_puts(seq
, ",discard");
544 static const struct super_operations nilfs_sops
= {
545 .alloc_inode
= nilfs_alloc_inode
,
546 .destroy_inode
= nilfs_destroy_inode
,
547 .dirty_inode
= nilfs_dirty_inode
,
548 /* .write_inode = nilfs_write_inode, */
549 /* .put_inode = nilfs_put_inode, */
550 /* .drop_inode = nilfs_drop_inode, */
551 .delete_inode
= nilfs_delete_inode
,
552 .put_super
= nilfs_put_super
,
553 /* .write_super = nilfs_write_super, */
554 .sync_fs
= nilfs_sync_fs
,
555 /* .write_super_lockfs */
557 .statfs
= nilfs_statfs
,
558 .remount_fs
= nilfs_remount
,
559 .clear_inode
= nilfs_clear_inode
,
561 .show_options
= nilfs_show_options
564 static struct inode
*
565 nilfs_nfs_get_inode(struct super_block
*sb
, u64 ino
, u32 generation
)
569 if (ino
< NILFS_FIRST_INO(sb
) && ino
!= NILFS_ROOT_INO
&&
570 ino
!= NILFS_SKETCH_INO
)
571 return ERR_PTR(-ESTALE
);
573 inode
= nilfs_iget(sb
, ino
);
575 return ERR_CAST(inode
);
576 if (generation
&& inode
->i_generation
!= generation
) {
578 return ERR_PTR(-ESTALE
);
584 static struct dentry
*
585 nilfs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
588 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
589 nilfs_nfs_get_inode
);
592 static struct dentry
*
593 nilfs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
, int fh_len
,
596 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
597 nilfs_nfs_get_inode
);
600 static const struct export_operations nilfs_export_ops
= {
601 .fh_to_dentry
= nilfs_fh_to_dentry
,
602 .fh_to_parent
= nilfs_fh_to_parent
,
603 .get_parent
= nilfs_get_parent
,
607 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
608 Opt_barrier
, Opt_nobarrier
, Opt_snapshot
, Opt_order
, Opt_norecovery
,
609 Opt_discard
, Opt_nodiscard
, Opt_err
,
612 static match_table_t tokens
= {
613 {Opt_err_cont
, "errors=continue"},
614 {Opt_err_panic
, "errors=panic"},
615 {Opt_err_ro
, "errors=remount-ro"},
616 {Opt_barrier
, "barrier"},
617 {Opt_nobarrier
, "nobarrier"},
618 {Opt_snapshot
, "cp=%u"},
619 {Opt_order
, "order=%s"},
620 {Opt_norecovery
, "norecovery"},
621 {Opt_discard
, "discard"},
622 {Opt_nodiscard
, "nodiscard"},
626 static int parse_options(char *options
, struct super_block
*sb
, int is_remount
)
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 nilfs_set_opt(sbi
, BARRIER
);
647 nilfs_clear_opt(sbi
, BARRIER
);
650 if (strcmp(args
[0].from
, "relaxed") == 0)
651 /* Ordered data semantics */
652 nilfs_clear_opt(sbi
, STRICT_ORDER
);
653 else if (strcmp(args
[0].from
, "strict") == 0)
654 /* Strict in-order semantics */
655 nilfs_set_opt(sbi
, STRICT_ORDER
);
660 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
663 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
666 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
669 if (match_int(&args
[0], &option
) || option
<= 0)
672 if (!nilfs_test_opt(sbi
, SNAPSHOT
)) {
674 "NILFS: cannot change regular "
675 "mount to snapshot.\n");
677 } else if (option
!= sbi
->s_snapshot_cno
) {
679 "NILFS: cannot remount to a "
680 "different snapshot.\n");
685 if (!(sb
->s_flags
& MS_RDONLY
)) {
686 printk(KERN_ERR
"NILFS: cannot mount snapshot "
687 "read/write. A read-only option is "
691 sbi
->s_snapshot_cno
= option
;
692 nilfs_set_opt(sbi
, SNAPSHOT
);
695 nilfs_set_opt(sbi
, NORECOVERY
);
698 nilfs_set_opt(sbi
, DISCARD
);
701 nilfs_clear_opt(sbi
, DISCARD
);
705 "NILFS: Unrecognized mount option \"%s\"\n", p
);
713 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
714 struct nilfs_super_block
*sbp
)
717 NILFS_MOUNT_ERRORS_RO
| NILFS_MOUNT_BARRIER
;
720 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
722 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
723 struct nilfs_super_block
**sbp
;
727 /* nilfs->ns_sem must be locked by the caller. */
728 sbp
= nilfs_prepare_super(sbi
, 0);
732 max_mnt_count
= le16_to_cpu(sbp
[0]->s_max_mnt_count
);
733 mnt_count
= le16_to_cpu(sbp
[0]->s_mnt_count
);
735 if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
737 "NILFS warning: mounting fs with errors\n");
739 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
741 "NILFS warning: maximal mount count reached\n");
745 sbp
[0]->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
747 sbp
[0]->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
749 cpu_to_le16(le16_to_cpu(sbp
[0]->s_state
) & ~NILFS_VALID_FS
);
750 sbp
[0]->s_mtime
= cpu_to_le64(get_seconds());
751 /* synchronize sbp[1] with sbp[0] */
752 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
753 return nilfs_commit_super(sbi
, NILFS_SB_COMMIT_ALL
);
756 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
757 u64 pos
, int blocksize
,
758 struct buffer_head
**pbh
)
760 unsigned long long sb_index
= pos
;
761 unsigned long offset
;
763 offset
= do_div(sb_index
, blocksize
);
764 *pbh
= sb_bread(sb
, sb_index
);
767 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
770 int nilfs_store_magic_and_option(struct super_block
*sb
,
771 struct nilfs_super_block
*sbp
,
774 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
776 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
778 /* FS independent flags */
779 #ifdef NILFS_ATIME_DISABLE
780 sb
->s_flags
|= MS_NOATIME
;
783 nilfs_set_default_options(sbi
, sbp
);
785 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
786 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
787 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
788 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
790 return !parse_options(data
, sb
, 0) ? -EINVAL
: 0 ;
793 int nilfs_check_feature_compatibility(struct super_block
*sb
,
794 struct nilfs_super_block
*sbp
)
798 features
= le64_to_cpu(sbp
->s_feature_incompat
) &
799 ~NILFS_FEATURE_INCOMPAT_SUPP
;
801 printk(KERN_ERR
"NILFS: couldn't mount because of unsupported "
802 "optional features (%llx)\n",
803 (unsigned long long)features
);
806 features
= le64_to_cpu(sbp
->s_feature_compat_ro
) &
807 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
808 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
809 printk(KERN_ERR
"NILFS: couldn't mount RDWR because of "
810 "unsupported optional features (%llx)\n",
811 (unsigned long long)features
);
818 * nilfs_fill_super() - initialize a super block instance
820 * @data: mount options
821 * @silent: silent mode flag
822 * @nilfs: the_nilfs struct
824 * This function is called exclusively by nilfs->ns_mount_mutex.
825 * So, the recovery process is protected from other simultaneous mounts.
828 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
829 struct the_nilfs
*nilfs
)
831 struct nilfs_sb_info
*sbi
;
836 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
843 sbi
->s_nilfs
= nilfs
;
845 atomic_set(&sbi
->s_count
, 1);
847 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
851 spin_lock_init(&sbi
->s_inode_lock
);
852 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
853 INIT_LIST_HEAD(&sbi
->s_list
);
856 * Following initialization is overlapped because
857 * nilfs_sb_info structure has been cleared at the beginning.
858 * But we reserve them to keep our interest and make ready
859 * for the future change.
861 get_random_bytes(&sbi
->s_next_generation
,
862 sizeof(sbi
->s_next_generation
));
863 spin_lock_init(&sbi
->s_next_gen_lock
);
865 sb
->s_op
= &nilfs_sops
;
866 sb
->s_export_op
= &nilfs_export_ops
;
869 sb
->s_bdi
= nilfs
->ns_bdi
;
871 err
= load_nilfs(nilfs
, sbi
);
875 cno
= nilfs_last_cno(nilfs
);
877 if (sb
->s_flags
& MS_RDONLY
) {
878 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
879 down_read(&nilfs
->ns_segctor_sem
);
880 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
881 sbi
->s_snapshot_cno
);
882 up_read(&nilfs
->ns_segctor_sem
);
890 "NILFS: The specified checkpoint is "
892 "(checkpoint number=%llu).\n",
893 (unsigned long long)sbi
->s_snapshot_cno
);
897 cno
= sbi
->s_snapshot_cno
;
901 err
= nilfs_attach_checkpoint(sbi
, cno
);
903 printk(KERN_ERR
"NILFS: error loading a checkpoint"
904 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
908 if (!(sb
->s_flags
& MS_RDONLY
)) {
909 err
= nilfs_attach_segment_constructor(sbi
);
911 goto failed_checkpoint
;
914 root
= nilfs_iget(sb
, NILFS_ROOT_INO
);
916 printk(KERN_ERR
"NILFS: get root inode failed\n");
920 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
922 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
926 sb
->s_root
= d_alloc_root(root
);
929 printk(KERN_ERR
"NILFS: get root dentry failed\n");
934 if (!(sb
->s_flags
& MS_RDONLY
)) {
935 down_write(&nilfs
->ns_sem
);
936 nilfs_setup_super(sbi
);
937 up_write(&nilfs
->ns_sem
);
940 down_write(&nilfs
->ns_super_sem
);
941 if (!nilfs_test_opt(sbi
, SNAPSHOT
))
942 nilfs
->ns_current
= sbi
;
943 up_write(&nilfs
->ns_super_sem
);
948 nilfs_detach_segment_constructor(sbi
);
951 nilfs_detach_checkpoint(sbi
);
955 sb
->s_fs_info
= NULL
;
956 nilfs_put_sbinfo(sbi
);
960 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
962 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
963 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
964 unsigned long old_sb_flags
;
965 struct nilfs_mount_options old_opts
;
966 int was_snapshot
, err
;
970 down_write(&nilfs
->ns_super_sem
);
971 old_sb_flags
= sb
->s_flags
;
972 old_opts
.mount_opt
= sbi
->s_mount_opt
;
973 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
974 was_snapshot
= nilfs_test_opt(sbi
, SNAPSHOT
);
976 if (!parse_options(data
, sb
, 1)) {
980 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
983 if (was_snapshot
&& !(*flags
& MS_RDONLY
)) {
984 printk(KERN_ERR
"NILFS (device %s): cannot remount snapshot "
985 "read/write.\n", sb
->s_id
);
989 if (!nilfs_valid_fs(nilfs
)) {
990 printk(KERN_WARNING
"NILFS (device %s): couldn't "
991 "remount because the filesystem is in an "
992 "incomplete recovery state.\n", sb
->s_id
);
996 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
998 if (*flags
& MS_RDONLY
) {
999 /* Shutting down the segment constructor */
1000 nilfs_detach_segment_constructor(sbi
);
1001 sb
->s_flags
|= MS_RDONLY
;
1004 * Remounting a valid RW partition RDONLY, so set
1005 * the RDONLY flag and then mark the partition as valid again.
1007 down_write(&nilfs
->ns_sem
);
1008 nilfs_cleanup_super(sbi
);
1009 up_write(&nilfs
->ns_sem
);
1014 * Mounting a RDONLY partition read-write, so reread and
1015 * store the current valid flag. (It may have been changed
1016 * by fsck since we originally mounted the partition.)
1018 down_read(&nilfs
->ns_sem
);
1019 features
= le64_to_cpu(nilfs
->ns_sbp
[0]->s_feature_compat_ro
) &
1020 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
1021 up_read(&nilfs
->ns_sem
);
1023 printk(KERN_WARNING
"NILFS (device %s): couldn't "
1024 "remount RDWR because of unsupported optional "
1025 "features (%llx)\n",
1026 sb
->s_id
, (unsigned long long)features
);
1031 sb
->s_flags
&= ~MS_RDONLY
;
1033 err
= nilfs_attach_segment_constructor(sbi
);
1037 down_write(&nilfs
->ns_sem
);
1038 nilfs_setup_super(sbi
);
1039 up_write(&nilfs
->ns_sem
);
1042 up_write(&nilfs
->ns_super_sem
);
1047 sb
->s_flags
= old_sb_flags
;
1048 sbi
->s_mount_opt
= old_opts
.mount_opt
;
1049 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
1050 up_write(&nilfs
->ns_super_sem
);
1055 struct nilfs_super_data
{
1056 struct block_device
*bdev
;
1057 struct nilfs_sb_info
*sbi
;
1063 * nilfs_identify - pre-read mount options needed to identify mount instance
1064 * @data: mount options
1065 * @sd: nilfs_super_data
1067 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1069 char *p
, *options
= data
;
1070 substring_t args
[MAX_OPT_ARGS
];
1075 p
= strsep(&options
, ",");
1076 if (p
!= NULL
&& *p
) {
1077 token
= match_token(p
, tokens
, args
);
1078 if (token
== Opt_snapshot
) {
1079 if (!(sd
->flags
& MS_RDONLY
))
1082 ret
= match_int(&args
[0], &option
);
1093 "NILFS: invalid mount option: %s\n", p
);
1097 BUG_ON(options
== data
);
1098 *(options
- 1) = ',';
1103 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1105 struct nilfs_super_data
*sd
= data
;
1107 s
->s_bdev
= sd
->bdev
;
1108 s
->s_dev
= s
->s_bdev
->bd_dev
;
1112 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1114 struct nilfs_super_data
*sd
= data
;
1116 return sd
->sbi
&& s
->s_fs_info
== (void *)sd
->sbi
;
1120 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1121 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1123 struct nilfs_super_data sd
;
1124 struct super_block
*s
;
1125 fmode_t mode
= FMODE_READ
;
1126 struct the_nilfs
*nilfs
;
1127 int err
, need_to_close
= 1;
1129 if (!(flags
& MS_RDONLY
))
1130 mode
|= FMODE_WRITE
;
1132 sd
.bdev
= open_bdev_exclusive(dev_name
, mode
, fs_type
);
1133 if (IS_ERR(sd
.bdev
))
1134 return PTR_ERR(sd
.bdev
);
1137 * To get mount instance using sget() vfs-routine, NILFS needs
1138 * much more information than normal filesystems to identify mount
1139 * instance. For snapshot mounts, not only a mount type (ro-mount
1140 * or rw-mount) but also a checkpoint number is required.
1144 if (nilfs_identify((char *)data
, &sd
)) {
1149 nilfs
= find_or_create_nilfs(sd
.bdev
);
1155 mutex_lock(&nilfs
->ns_mount_mutex
);
1159 * Check if an exclusive mount exists or not.
1160 * Snapshot mounts coexist with a current mount
1161 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1162 * ro-mount are mutually exclusive.
1164 down_read(&nilfs
->ns_super_sem
);
1165 if (nilfs
->ns_current
&&
1166 ((nilfs
->ns_current
->s_super
->s_flags
^ flags
)
1168 up_read(&nilfs
->ns_super_sem
);
1172 up_read(&nilfs
->ns_super_sem
);
1176 * Find existing nilfs_sb_info struct
1178 sd
.sbi
= nilfs_find_sbinfo(nilfs
, !(flags
& MS_RDONLY
), sd
.cno
);
1181 * Get super block instance holding the nilfs_sb_info struct.
1182 * A new instance is allocated if no existing mount is present or
1183 * existing instance has been unmounted.
1185 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1187 nilfs_put_sbinfo(sd
.sbi
);
1195 char b
[BDEVNAME_SIZE
];
1197 /* New superblock instance created */
1200 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1201 sb_set_blocksize(s
, block_size(sd
.bdev
));
1203 err
= nilfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0,
1208 s
->s_flags
|= MS_ACTIVE
;
1212 mutex_unlock(&nilfs
->ns_mount_mutex
);
1215 close_bdev_exclusive(sd
.bdev
, mode
);
1216 simple_set_mnt(mnt
, s
);
1220 mutex_unlock(&nilfs
->ns_mount_mutex
);
1223 close_bdev_exclusive(sd
.bdev
, mode
);
1228 /* Abandoning the newly allocated superblock */
1229 mutex_unlock(&nilfs
->ns_mount_mutex
);
1231 deactivate_locked_super(s
);
1233 * deactivate_locked_super() invokes close_bdev_exclusive().
1234 * We must finish all post-cleaning before this call;
1235 * put_nilfs() needs the block device.
1240 struct file_system_type nilfs_fs_type
= {
1241 .owner
= THIS_MODULE
,
1243 .get_sb
= nilfs_get_sb
,
1244 .kill_sb
= kill_block_super
,
1245 .fs_flags
= FS_REQUIRES_DEV
,
1248 static void nilfs_inode_init_once(void *obj
)
1250 struct nilfs_inode_info
*ii
= obj
;
1252 INIT_LIST_HEAD(&ii
->i_dirty
);
1253 #ifdef CONFIG_NILFS_XATTR
1254 init_rwsem(&ii
->xattr_sem
);
1256 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
1257 ii
->i_bmap
= &ii
->i_bmap_data
;
1258 inode_init_once(&ii
->vfs_inode
);
1261 static void nilfs_segbuf_init_once(void *obj
)
1263 memset(obj
, 0, sizeof(struct nilfs_segment_buffer
));
1266 static void nilfs_destroy_cachep(void)
1268 if (nilfs_inode_cachep
)
1269 kmem_cache_destroy(nilfs_inode_cachep
);
1270 if (nilfs_transaction_cachep
)
1271 kmem_cache_destroy(nilfs_transaction_cachep
);
1272 if (nilfs_segbuf_cachep
)
1273 kmem_cache_destroy(nilfs_segbuf_cachep
);
1274 if (nilfs_btree_path_cache
)
1275 kmem_cache_destroy(nilfs_btree_path_cache
);
1278 static int __init
nilfs_init_cachep(void)
1280 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
1281 sizeof(struct nilfs_inode_info
), 0,
1282 SLAB_RECLAIM_ACCOUNT
, nilfs_inode_init_once
);
1283 if (!nilfs_inode_cachep
)
1286 nilfs_transaction_cachep
= kmem_cache_create("nilfs2_transaction_cache",
1287 sizeof(struct nilfs_transaction_info
), 0,
1288 SLAB_RECLAIM_ACCOUNT
, NULL
);
1289 if (!nilfs_transaction_cachep
)
1292 nilfs_segbuf_cachep
= kmem_cache_create("nilfs2_segbuf_cache",
1293 sizeof(struct nilfs_segment_buffer
), 0,
1294 SLAB_RECLAIM_ACCOUNT
, nilfs_segbuf_init_once
);
1295 if (!nilfs_segbuf_cachep
)
1298 nilfs_btree_path_cache
= kmem_cache_create("nilfs2_btree_path_cache",
1299 sizeof(struct nilfs_btree_path
) * NILFS_BTREE_LEVEL_MAX
,
1301 if (!nilfs_btree_path_cache
)
1307 nilfs_destroy_cachep();
1311 static int __init
init_nilfs_fs(void)
1315 err
= nilfs_init_cachep();
1319 err
= register_filesystem(&nilfs_fs_type
);
1323 printk(KERN_INFO
"NILFS version 2 loaded\n");
1327 nilfs_destroy_cachep();
1332 static void __exit
exit_nilfs_fs(void)
1334 nilfs_destroy_cachep();
1335 unregister_filesystem(&nilfs_fs_type
);
1338 module_init(init_nilfs_fs
)
1339 module_exit(exit_nilfs_fs
)