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/vfs.h>
49 #include <linux/writeback.h>
50 #include <linux/kobject.h>
51 #include <linux/seq_file.h>
52 #include <linux/mount.h>
66 MODULE_AUTHOR("NTT Corp.");
67 MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
69 MODULE_LICENSE("GPL");
71 static struct kmem_cache
*nilfs_inode_cachep
;
72 struct kmem_cache
*nilfs_transaction_cachep
;
73 struct kmem_cache
*nilfs_segbuf_cachep
;
74 struct kmem_cache
*nilfs_btree_path_cache
;
76 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
, int is_mount
);
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(struct super_block
*sb
)
152 struct nilfs_inode_info
*ii
;
154 ii
= kmem_cache_alloc(nilfs_inode_cachep
, GFP_NOFS
);
160 ii
->vfs_inode
.i_version
= 1;
161 nilfs_btnode_cache_init(&ii
->i_btnode_cache
, sb
->s_bdi
);
162 return &ii
->vfs_inode
;
165 static void nilfs_i_callback(struct rcu_head
*head
)
167 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
168 struct nilfs_mdt_info
*mdi
= NILFS_MDT(inode
);
170 INIT_LIST_HEAD(&inode
->i_dentry
);
173 kfree(mdi
->mi_bgl
); /* kfree(NULL) is safe */
176 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
179 void nilfs_destroy_inode(struct inode
*inode
)
181 call_rcu(&inode
->i_rcu
, nilfs_i_callback
);
184 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
, int flag
)
186 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
190 set_buffer_dirty(nilfs
->ns_sbh
[0]);
191 if (nilfs_test_opt(sbi
, BARRIER
)) {
192 err
= __sync_dirty_buffer(nilfs
->ns_sbh
[0],
193 WRITE_SYNC
| WRITE_FLUSH_FUA
);
195 err
= sync_dirty_buffer(nilfs
->ns_sbh
[0]);
200 "NILFS: unable to write superblock (err=%d)\n", err
);
201 if (err
== -EIO
&& nilfs
->ns_sbh
[1]) {
203 * sbp[0] points to newer log than sbp[1],
204 * so copy sbp[0] to sbp[1] to take over sbp[0].
206 memcpy(nilfs
->ns_sbp
[1], nilfs
->ns_sbp
[0],
208 nilfs_fall_back_super_block(nilfs
);
212 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
214 nilfs
->ns_sbwcount
++;
217 * The latest segment becomes trailable from the position
218 * written in superblock.
220 clear_nilfs_discontinued(nilfs
);
222 /* update GC protection for recent segments */
223 if (nilfs
->ns_sbh
[1]) {
224 if (flag
== NILFS_SB_COMMIT_ALL
) {
225 set_buffer_dirty(nilfs
->ns_sbh
[1]);
226 if (sync_dirty_buffer(nilfs
->ns_sbh
[1]) < 0)
229 if (le64_to_cpu(nilfs
->ns_sbp
[1]->s_last_cno
) <
230 le64_to_cpu(nilfs
->ns_sbp
[0]->s_last_cno
))
231 sbp
= nilfs
->ns_sbp
[1];
234 spin_lock(&nilfs
->ns_last_segment_lock
);
235 nilfs
->ns_prot_seq
= le64_to_cpu(sbp
->s_last_seq
);
236 spin_unlock(&nilfs
->ns_last_segment_lock
);
242 void nilfs_set_log_cursor(struct nilfs_super_block
*sbp
,
243 struct the_nilfs
*nilfs
)
245 sector_t nfreeblocks
;
247 /* nilfs->ns_sem must be locked by the caller. */
248 nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
249 sbp
->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
251 spin_lock(&nilfs
->ns_last_segment_lock
);
252 sbp
->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
253 sbp
->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
254 sbp
->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
255 spin_unlock(&nilfs
->ns_last_segment_lock
);
258 struct nilfs_super_block
**nilfs_prepare_super(struct nilfs_sb_info
*sbi
,
261 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
262 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
264 /* nilfs->ns_sem must be locked by the caller. */
265 if (sbp
[0]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
267 sbp
[1]->s_magic
== cpu_to_le16(NILFS_SUPER_MAGIC
)) {
268 memcpy(sbp
[0], sbp
[1], nilfs
->ns_sbsize
);
270 printk(KERN_CRIT
"NILFS: superblock broke on dev %s\n",
275 sbp
[1]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
276 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
280 nilfs_swap_super_block(nilfs
);
285 int nilfs_commit_super(struct nilfs_sb_info
*sbi
, int flag
)
287 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
288 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
291 /* nilfs->ns_sem must be locked by the caller. */
293 nilfs
->ns_sbwtime
= t
;
294 sbp
[0]->s_wtime
= cpu_to_le64(t
);
296 sbp
[0]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
297 (unsigned char *)sbp
[0],
299 if (flag
== NILFS_SB_COMMIT_ALL
&& sbp
[1]) {
300 sbp
[1]->s_wtime
= sbp
[0]->s_wtime
;
302 sbp
[1]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
303 (unsigned char *)sbp
[1],
306 clear_nilfs_sb_dirty(nilfs
);
307 return nilfs_sync_super(sbi
, flag
);
311 * nilfs_cleanup_super() - write filesystem state for cleanup
312 * @sbi: nilfs_sb_info to be unmounted or degraded to read-only
314 * This function restores state flags in the on-disk super block.
315 * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the
316 * filesystem was not clean previously.
318 int nilfs_cleanup_super(struct nilfs_sb_info
*sbi
)
320 struct nilfs_super_block
**sbp
;
321 int flag
= NILFS_SB_COMMIT
;
324 sbp
= nilfs_prepare_super(sbi
, 0);
326 sbp
[0]->s_state
= cpu_to_le16(sbi
->s_nilfs
->ns_mount_state
);
327 nilfs_set_log_cursor(sbp
[0], sbi
->s_nilfs
);
328 if (sbp
[1] && sbp
[0]->s_last_cno
== sbp
[1]->s_last_cno
) {
330 * make the "clean" flag also to the opposite
331 * super block if both super blocks point to
332 * the same checkpoint.
334 sbp
[1]->s_state
= sbp
[0]->s_state
;
335 flag
= NILFS_SB_COMMIT_ALL
;
337 ret
= nilfs_commit_super(sbi
, flag
);
342 static void nilfs_put_super(struct super_block
*sb
)
344 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
345 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
347 nilfs_detach_segment_constructor(sbi
);
349 if (!(sb
->s_flags
& MS_RDONLY
)) {
350 down_write(&nilfs
->ns_sem
);
351 nilfs_cleanup_super(sbi
);
352 up_write(&nilfs
->ns_sem
);
355 iput(nilfs
->ns_sufile
);
356 iput(nilfs
->ns_cpfile
);
359 destroy_nilfs(nilfs
);
361 sb
->s_fs_info
= NULL
;
365 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
367 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
368 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
369 struct nilfs_super_block
**sbp
;
372 /* This function is called when super block should be written back */
374 err
= nilfs_construct_segment(sb
);
376 down_write(&nilfs
->ns_sem
);
377 if (nilfs_sb_dirty(nilfs
)) {
378 sbp
= nilfs_prepare_super(sbi
, nilfs_sb_will_flip(nilfs
));
380 nilfs_set_log_cursor(sbp
[0], nilfs
);
381 nilfs_commit_super(sbi
, NILFS_SB_COMMIT
);
384 up_write(&nilfs
->ns_sem
);
389 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
, int curr_mnt
,
390 struct nilfs_root
**rootp
)
392 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
393 struct nilfs_root
*root
;
394 struct nilfs_checkpoint
*raw_cp
;
395 struct buffer_head
*bh_cp
;
398 root
= nilfs_find_or_create_root(
399 nilfs
, curr_mnt
? NILFS_CPTREE_CURRENT_CNO
: cno
);
404 goto reuse
; /* already attached checkpoint */
406 down_read(&nilfs
->ns_segctor_sem
);
407 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
409 up_read(&nilfs
->ns_segctor_sem
);
411 if (err
== -ENOENT
|| err
== -EINVAL
) {
413 "NILFS: Invalid checkpoint "
414 "(checkpoint number=%llu)\n",
415 (unsigned long long)cno
);
421 err
= nilfs_ifile_read(sbi
->s_super
, root
, nilfs
->ns_inode_size
,
422 &raw_cp
->cp_ifile_inode
, &root
->ifile
);
426 atomic_set(&root
->inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
427 atomic_set(&root
->blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
429 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
436 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
438 nilfs_put_root(root
);
443 static int nilfs_freeze(struct super_block
*sb
)
445 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
446 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
449 if (sb
->s_flags
& MS_RDONLY
)
452 /* Mark super block clean */
453 down_write(&nilfs
->ns_sem
);
454 err
= nilfs_cleanup_super(sbi
);
455 up_write(&nilfs
->ns_sem
);
459 static int nilfs_unfreeze(struct super_block
*sb
)
461 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
462 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
464 if (sb
->s_flags
& MS_RDONLY
)
467 down_write(&nilfs
->ns_sem
);
468 nilfs_setup_super(sbi
, false);
469 up_write(&nilfs
->ns_sem
);
473 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
475 struct super_block
*sb
= dentry
->d_sb
;
476 struct nilfs_root
*root
= NILFS_I(dentry
->d_inode
)->i_root
;
477 struct the_nilfs
*nilfs
= root
->nilfs
;
478 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
479 unsigned long long blocks
;
480 unsigned long overhead
;
481 unsigned long nrsvblocks
;
482 sector_t nfreeblocks
;
486 * Compute all of the segment blocks
488 * The blocks before first segment and after last segment
491 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
492 - nilfs
->ns_first_data_block
;
493 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
496 * Compute the overhead
498 * When distributing meta data blocks outside segment structure,
499 * We must count them as the overhead.
503 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
507 buf
->f_type
= NILFS_SUPER_MAGIC
;
508 buf
->f_bsize
= sb
->s_blocksize
;
509 buf
->f_blocks
= blocks
- overhead
;
510 buf
->f_bfree
= nfreeblocks
;
511 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
512 (buf
->f_bfree
- nrsvblocks
) : 0;
513 buf
->f_files
= atomic_read(&root
->inodes_count
);
514 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
515 buf
->f_namelen
= NILFS_NAME_LEN
;
516 buf
->f_fsid
.val
[0] = (u32
)id
;
517 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
522 static int nilfs_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
524 struct super_block
*sb
= vfs
->mnt_sb
;
525 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
526 struct nilfs_root
*root
= NILFS_I(vfs
->mnt_root
->d_inode
)->i_root
;
528 if (!nilfs_test_opt(sbi
, BARRIER
))
529 seq_puts(seq
, ",nobarrier");
530 if (root
->cno
!= NILFS_CPTREE_CURRENT_CNO
)
531 seq_printf(seq
, ",cp=%llu", (unsigned long long)root
->cno
);
532 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
533 seq_puts(seq
, ",errors=panic");
534 if (nilfs_test_opt(sbi
, ERRORS_CONT
))
535 seq_puts(seq
, ",errors=continue");
536 if (nilfs_test_opt(sbi
, STRICT_ORDER
))
537 seq_puts(seq
, ",order=strict");
538 if (nilfs_test_opt(sbi
, NORECOVERY
))
539 seq_puts(seq
, ",norecovery");
540 if (nilfs_test_opt(sbi
, DISCARD
))
541 seq_puts(seq
, ",discard");
546 static const struct super_operations nilfs_sops
= {
547 .alloc_inode
= nilfs_alloc_inode
,
548 .destroy_inode
= nilfs_destroy_inode
,
549 .dirty_inode
= nilfs_dirty_inode
,
550 /* .write_inode = nilfs_write_inode, */
551 /* .put_inode = nilfs_put_inode, */
552 /* .drop_inode = nilfs_drop_inode, */
553 .evict_inode
= nilfs_evict_inode
,
554 .put_super
= nilfs_put_super
,
555 /* .write_super = nilfs_write_super, */
556 .sync_fs
= nilfs_sync_fs
,
557 .freeze_fs
= nilfs_freeze
,
558 .unfreeze_fs
= nilfs_unfreeze
,
559 /* .write_super_lockfs */
561 .statfs
= nilfs_statfs
,
562 .remount_fs
= nilfs_remount
,
564 .show_options
= nilfs_show_options
568 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
569 Opt_barrier
, Opt_nobarrier
, Opt_snapshot
, Opt_order
, Opt_norecovery
,
570 Opt_discard
, Opt_nodiscard
, Opt_err
,
573 static match_table_t tokens
= {
574 {Opt_err_cont
, "errors=continue"},
575 {Opt_err_panic
, "errors=panic"},
576 {Opt_err_ro
, "errors=remount-ro"},
577 {Opt_barrier
, "barrier"},
578 {Opt_nobarrier
, "nobarrier"},
579 {Opt_snapshot
, "cp=%u"},
580 {Opt_order
, "order=%s"},
581 {Opt_norecovery
, "norecovery"},
582 {Opt_discard
, "discard"},
583 {Opt_nodiscard
, "nodiscard"},
587 static int parse_options(char *options
, struct super_block
*sb
, int is_remount
)
589 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
591 substring_t args
[MAX_OPT_ARGS
];
596 while ((p
= strsep(&options
, ",")) != NULL
) {
601 token
= match_token(p
, tokens
, args
);
604 nilfs_set_opt(sbi
, BARRIER
);
607 nilfs_clear_opt(sbi
, BARRIER
);
610 if (strcmp(args
[0].from
, "relaxed") == 0)
611 /* Ordered data semantics */
612 nilfs_clear_opt(sbi
, STRICT_ORDER
);
613 else if (strcmp(args
[0].from
, "strict") == 0)
614 /* Strict in-order semantics */
615 nilfs_set_opt(sbi
, STRICT_ORDER
);
620 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
623 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
626 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
631 "NILFS: \"%s\" option is invalid "
632 "for remount.\n", p
);
637 nilfs_set_opt(sbi
, NORECOVERY
);
640 nilfs_set_opt(sbi
, DISCARD
);
643 nilfs_clear_opt(sbi
, DISCARD
);
647 "NILFS: Unrecognized mount option \"%s\"\n", p
);
655 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
656 struct nilfs_super_block
*sbp
)
659 NILFS_MOUNT_ERRORS_RO
| NILFS_MOUNT_BARRIER
;
662 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
, int is_mount
)
664 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
665 struct nilfs_super_block
**sbp
;
669 /* nilfs->ns_sem must be locked by the caller. */
670 sbp
= nilfs_prepare_super(sbi
, 0);
675 goto skip_mount_setup
;
677 max_mnt_count
= le16_to_cpu(sbp
[0]->s_max_mnt_count
);
678 mnt_count
= le16_to_cpu(sbp
[0]->s_mnt_count
);
680 if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
682 "NILFS warning: mounting fs with errors\n");
684 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
686 "NILFS warning: maximal mount count reached\n");
690 sbp
[0]->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
692 sbp
[0]->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
693 sbp
[0]->s_mtime
= cpu_to_le64(get_seconds());
697 cpu_to_le16(le16_to_cpu(sbp
[0]->s_state
) & ~NILFS_VALID_FS
);
698 /* synchronize sbp[1] with sbp[0] */
699 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
700 return nilfs_commit_super(sbi
, NILFS_SB_COMMIT_ALL
);
703 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
704 u64 pos
, int blocksize
,
705 struct buffer_head
**pbh
)
707 unsigned long long sb_index
= pos
;
708 unsigned long offset
;
710 offset
= do_div(sb_index
, blocksize
);
711 *pbh
= sb_bread(sb
, sb_index
);
714 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
717 int nilfs_store_magic_and_option(struct super_block
*sb
,
718 struct nilfs_super_block
*sbp
,
721 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
723 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
725 /* FS independent flags */
726 #ifdef NILFS_ATIME_DISABLE
727 sb
->s_flags
|= MS_NOATIME
;
730 nilfs_set_default_options(sbi
, sbp
);
732 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
733 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
734 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
735 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
737 return !parse_options(data
, sb
, 0) ? -EINVAL
: 0 ;
740 int nilfs_check_feature_compatibility(struct super_block
*sb
,
741 struct nilfs_super_block
*sbp
)
745 features
= le64_to_cpu(sbp
->s_feature_incompat
) &
746 ~NILFS_FEATURE_INCOMPAT_SUPP
;
748 printk(KERN_ERR
"NILFS: couldn't mount because of unsupported "
749 "optional features (%llx)\n",
750 (unsigned long long)features
);
753 features
= le64_to_cpu(sbp
->s_feature_compat_ro
) &
754 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
755 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
756 printk(KERN_ERR
"NILFS: couldn't mount RDWR because of "
757 "unsupported optional features (%llx)\n",
758 (unsigned long long)features
);
764 static int nilfs_get_root_dentry(struct super_block
*sb
,
765 struct nilfs_root
*root
,
766 struct dentry
**root_dentry
)
769 struct dentry
*dentry
;
772 inode
= nilfs_iget(sb
, root
, NILFS_ROOT_INO
);
774 printk(KERN_ERR
"NILFS: get root inode failed\n");
775 ret
= PTR_ERR(inode
);
778 if (!S_ISDIR(inode
->i_mode
) || !inode
->i_blocks
|| !inode
->i_size
) {
780 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
785 if (root
->cno
== NILFS_CPTREE_CURRENT_CNO
) {
786 dentry
= d_find_alias(inode
);
788 dentry
= d_alloc_root(inode
);
798 dentry
= d_obtain_alias(inode
);
799 if (IS_ERR(dentry
)) {
800 ret
= PTR_ERR(dentry
);
804 *root_dentry
= dentry
;
809 printk(KERN_ERR
"NILFS: get root dentry failed\n");
813 static int nilfs_attach_snapshot(struct super_block
*s
, __u64 cno
,
814 struct dentry
**root_dentry
)
816 struct the_nilfs
*nilfs
= NILFS_SB(s
)->s_nilfs
;
817 struct nilfs_root
*root
;
820 down_read(&nilfs
->ns_segctor_sem
);
821 ret
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
, cno
);
822 up_read(&nilfs
->ns_segctor_sem
);
824 ret
= (ret
== -ENOENT
) ? -EINVAL
: ret
;
827 printk(KERN_ERR
"NILFS: The specified checkpoint is "
828 "not a snapshot (checkpoint number=%llu).\n",
829 (unsigned long long)cno
);
834 ret
= nilfs_attach_checkpoint(NILFS_SB(s
), cno
, false, &root
);
836 printk(KERN_ERR
"NILFS: error loading snapshot "
837 "(checkpoint number=%llu).\n",
838 (unsigned long long)cno
);
841 ret
= nilfs_get_root_dentry(s
, root
, root_dentry
);
842 nilfs_put_root(root
);
847 static int nilfs_tree_was_touched(struct dentry
*root_dentry
)
849 return root_dentry
->d_count
> 1;
853 * nilfs_try_to_shrink_tree() - try to shrink dentries of a checkpoint
854 * @root_dentry: root dentry of the tree to be shrunk
856 * This function returns true if the tree was in-use.
858 static int nilfs_try_to_shrink_tree(struct dentry
*root_dentry
)
860 if (have_submounts(root_dentry
))
862 shrink_dcache_parent(root_dentry
);
863 return nilfs_tree_was_touched(root_dentry
);
866 int nilfs_checkpoint_is_mounted(struct super_block
*sb
, __u64 cno
)
868 struct the_nilfs
*nilfs
= NILFS_SB(sb
)->s_nilfs
;
869 struct nilfs_root
*root
;
871 struct dentry
*dentry
;
874 if (cno
< 0 || cno
> nilfs
->ns_cno
)
877 if (cno
>= nilfs_last_cno(nilfs
))
878 return true; /* protect recent checkpoints */
881 root
= nilfs_lookup_root(NILFS_SB(sb
)->s_nilfs
, cno
);
883 inode
= nilfs_ilookup(sb
, root
, NILFS_ROOT_INO
);
885 dentry
= d_find_alias(inode
);
887 if (nilfs_tree_was_touched(dentry
))
888 ret
= nilfs_try_to_shrink_tree(dentry
);
893 nilfs_put_root(root
);
899 * nilfs_fill_super() - initialize a super block instance
901 * @data: mount options
902 * @silent: silent mode flag
904 * This function is called exclusively by nilfs->ns_mount_mutex.
905 * So, the recovery process is protected from other simultaneous mounts.
908 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
)
910 struct the_nilfs
*nilfs
;
911 struct nilfs_sb_info
*sbi
;
912 struct nilfs_root
*fsroot
;
913 struct backing_dev_info
*bdi
;
917 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
924 nilfs
= alloc_nilfs(sb
->s_bdev
);
929 sbi
->s_nilfs
= nilfs
;
931 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
935 spin_lock_init(&sbi
->s_inode_lock
);
936 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
939 * Following initialization is overlapped because
940 * nilfs_sb_info structure has been cleared at the beginning.
941 * But we reserve them to keep our interest and make ready
942 * for the future change.
944 get_random_bytes(&sbi
->s_next_generation
,
945 sizeof(sbi
->s_next_generation
));
946 spin_lock_init(&sbi
->s_next_gen_lock
);
948 sb
->s_op
= &nilfs_sops
;
949 sb
->s_export_op
= &nilfs_export_ops
;
953 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
954 sb
->s_bdi
= bdi
? : &default_backing_dev_info
;
956 err
= load_nilfs(nilfs
, sbi
);
960 cno
= nilfs_last_cno(nilfs
);
961 err
= nilfs_attach_checkpoint(sbi
, cno
, true, &fsroot
);
963 printk(KERN_ERR
"NILFS: error loading last checkpoint "
964 "(checkpoint number=%llu).\n", (unsigned long long)cno
);
968 if (!(sb
->s_flags
& MS_RDONLY
)) {
969 err
= nilfs_attach_segment_constructor(sbi
, fsroot
);
971 goto failed_checkpoint
;
974 err
= nilfs_get_root_dentry(sb
, fsroot
, &sb
->s_root
);
978 nilfs_put_root(fsroot
);
980 if (!(sb
->s_flags
& MS_RDONLY
)) {
981 down_write(&nilfs
->ns_sem
);
982 nilfs_setup_super(sbi
, true);
983 up_write(&nilfs
->ns_sem
);
989 nilfs_detach_segment_constructor(sbi
);
992 nilfs_put_root(fsroot
);
995 iput(nilfs
->ns_sufile
);
996 iput(nilfs
->ns_cpfile
);
1000 destroy_nilfs(nilfs
);
1003 sb
->s_fs_info
= NULL
;
1008 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
1010 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
1011 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
1012 unsigned long old_sb_flags
;
1013 struct nilfs_mount_options old_opts
;
1016 old_sb_flags
= sb
->s_flags
;
1017 old_opts
.mount_opt
= sbi
->s_mount_opt
;
1019 if (!parse_options(data
, sb
, 1)) {
1023 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
1027 if (!nilfs_valid_fs(nilfs
)) {
1028 printk(KERN_WARNING
"NILFS (device %s): couldn't "
1029 "remount because the filesystem is in an "
1030 "incomplete recovery state.\n", sb
->s_id
);
1034 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
1036 if (*flags
& MS_RDONLY
) {
1037 /* Shutting down the segment constructor */
1038 nilfs_detach_segment_constructor(sbi
);
1039 sb
->s_flags
|= MS_RDONLY
;
1042 * Remounting a valid RW partition RDONLY, so set
1043 * the RDONLY flag and then mark the partition as valid again.
1045 down_write(&nilfs
->ns_sem
);
1046 nilfs_cleanup_super(sbi
);
1047 up_write(&nilfs
->ns_sem
);
1050 struct nilfs_root
*root
;
1053 * Mounting a RDONLY partition read-write, so reread and
1054 * store the current valid flag. (It may have been changed
1055 * by fsck since we originally mounted the partition.)
1057 down_read(&nilfs
->ns_sem
);
1058 features
= le64_to_cpu(nilfs
->ns_sbp
[0]->s_feature_compat_ro
) &
1059 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
1060 up_read(&nilfs
->ns_sem
);
1062 printk(KERN_WARNING
"NILFS (device %s): couldn't "
1063 "remount RDWR because of unsupported optional "
1064 "features (%llx)\n",
1065 sb
->s_id
, (unsigned long long)features
);
1070 sb
->s_flags
&= ~MS_RDONLY
;
1072 root
= NILFS_I(sb
->s_root
->d_inode
)->i_root
;
1073 err
= nilfs_attach_segment_constructor(sbi
, root
);
1077 down_write(&nilfs
->ns_sem
);
1078 nilfs_setup_super(sbi
, true);
1079 up_write(&nilfs
->ns_sem
);
1085 sb
->s_flags
= old_sb_flags
;
1086 sbi
->s_mount_opt
= old_opts
.mount_opt
;
1090 struct nilfs_super_data
{
1091 struct block_device
*bdev
;
1092 struct nilfs_sb_info
*sbi
;
1098 * nilfs_identify - pre-read mount options needed to identify mount instance
1099 * @data: mount options
1100 * @sd: nilfs_super_data
1102 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1104 char *p
, *options
= data
;
1105 substring_t args
[MAX_OPT_ARGS
];
1110 p
= strsep(&options
, ",");
1111 if (p
!= NULL
&& *p
) {
1112 token
= match_token(p
, tokens
, args
);
1113 if (token
== Opt_snapshot
) {
1114 if (!(sd
->flags
& MS_RDONLY
)) {
1117 sd
->cno
= simple_strtoull(args
[0].from
,
1120 * No need to see the end pointer;
1121 * match_token() has done syntax
1130 "NILFS: invalid mount option: %s\n", p
);
1134 BUG_ON(options
== data
);
1135 *(options
- 1) = ',';
1140 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1143 s
->s_dev
= s
->s_bdev
->bd_dev
;
1147 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1149 return (void *)s
->s_bdev
== data
;
1152 static struct dentry
*
1153 nilfs_mount(struct file_system_type
*fs_type
, int flags
,
1154 const char *dev_name
, void *data
)
1156 struct nilfs_super_data sd
;
1157 struct super_block
*s
;
1158 fmode_t mode
= FMODE_READ
;
1159 struct dentry
*root_dentry
;
1160 int err
, s_new
= false;
1162 if (!(flags
& MS_RDONLY
))
1163 mode
|= FMODE_WRITE
;
1165 sd
.bdev
= open_bdev_exclusive(dev_name
, mode
, fs_type
);
1166 if (IS_ERR(sd
.bdev
))
1167 return ERR_CAST(sd
.bdev
);
1171 if (nilfs_identify((char *)data
, &sd
)) {
1177 * once the super is inserted into the list by sget, s_umount
1178 * will protect the lockfs code from trying to start a snapshot
1179 * while we are mounting
1181 mutex_lock(&sd
.bdev
->bd_fsfreeze_mutex
);
1182 if (sd
.bdev
->bd_fsfreeze_count
> 0) {
1183 mutex_unlock(&sd
.bdev
->bd_fsfreeze_mutex
);
1187 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, sd
.bdev
);
1188 mutex_unlock(&sd
.bdev
->bd_fsfreeze_mutex
);
1195 char b
[BDEVNAME_SIZE
];
1199 /* New superblock instance created */
1202 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1203 sb_set_blocksize(s
, block_size(sd
.bdev
));
1205 err
= nilfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1209 s
->s_flags
|= MS_ACTIVE
;
1210 } else if (!sd
.cno
) {
1213 if (nilfs_tree_was_touched(s
->s_root
)) {
1214 busy
= nilfs_try_to_shrink_tree(s
->s_root
);
1215 if (busy
&& (flags
^ s
->s_flags
) & MS_RDONLY
) {
1216 printk(KERN_ERR
"NILFS: the device already "
1217 "has a %s mount.\n",
1218 (s
->s_flags
& MS_RDONLY
) ?
1219 "read-only" : "read/write");
1226 * Try remount to setup mount states if the current
1227 * tree is not mounted and only snapshots use this sb.
1229 err
= nilfs_remount(s
, &flags
, data
);
1236 err
= nilfs_attach_snapshot(s
, sd
.cno
, &root_dentry
);
1240 root_dentry
= dget(s
->s_root
);
1244 close_bdev_exclusive(sd
.bdev
, mode
);
1249 deactivate_locked_super(s
);
1253 close_bdev_exclusive(sd
.bdev
, mode
);
1254 return ERR_PTR(err
);
1257 struct file_system_type nilfs_fs_type
= {
1258 .owner
= THIS_MODULE
,
1260 .mount
= nilfs_mount
,
1261 .kill_sb
= kill_block_super
,
1262 .fs_flags
= FS_REQUIRES_DEV
,
1265 static void nilfs_inode_init_once(void *obj
)
1267 struct nilfs_inode_info
*ii
= obj
;
1269 INIT_LIST_HEAD(&ii
->i_dirty
);
1270 #ifdef CONFIG_NILFS_XATTR
1271 init_rwsem(&ii
->xattr_sem
);
1273 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
1274 ii
->i_bmap
= &ii
->i_bmap_data
;
1275 inode_init_once(&ii
->vfs_inode
);
1278 static void nilfs_segbuf_init_once(void *obj
)
1280 memset(obj
, 0, sizeof(struct nilfs_segment_buffer
));
1283 static void nilfs_destroy_cachep(void)
1285 if (nilfs_inode_cachep
)
1286 kmem_cache_destroy(nilfs_inode_cachep
);
1287 if (nilfs_transaction_cachep
)
1288 kmem_cache_destroy(nilfs_transaction_cachep
);
1289 if (nilfs_segbuf_cachep
)
1290 kmem_cache_destroy(nilfs_segbuf_cachep
);
1291 if (nilfs_btree_path_cache
)
1292 kmem_cache_destroy(nilfs_btree_path_cache
);
1295 static int __init
nilfs_init_cachep(void)
1297 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
1298 sizeof(struct nilfs_inode_info
), 0,
1299 SLAB_RECLAIM_ACCOUNT
, nilfs_inode_init_once
);
1300 if (!nilfs_inode_cachep
)
1303 nilfs_transaction_cachep
= kmem_cache_create("nilfs2_transaction_cache",
1304 sizeof(struct nilfs_transaction_info
), 0,
1305 SLAB_RECLAIM_ACCOUNT
, NULL
);
1306 if (!nilfs_transaction_cachep
)
1309 nilfs_segbuf_cachep
= kmem_cache_create("nilfs2_segbuf_cache",
1310 sizeof(struct nilfs_segment_buffer
), 0,
1311 SLAB_RECLAIM_ACCOUNT
, nilfs_segbuf_init_once
);
1312 if (!nilfs_segbuf_cachep
)
1315 nilfs_btree_path_cache
= kmem_cache_create("nilfs2_btree_path_cache",
1316 sizeof(struct nilfs_btree_path
) * NILFS_BTREE_LEVEL_MAX
,
1318 if (!nilfs_btree_path_cache
)
1324 nilfs_destroy_cachep();
1328 static int __init
init_nilfs_fs(void)
1332 err
= nilfs_init_cachep();
1336 err
= register_filesystem(&nilfs_fs_type
);
1340 printk(KERN_INFO
"NILFS version 2 loaded\n");
1344 nilfs_destroy_cachep();
1349 static void __exit
exit_nilfs_fs(void)
1351 nilfs_destroy_cachep();
1352 unregister_filesystem(&nilfs_fs_type
);
1355 module_init(init_nilfs_fs
)
1356 module_exit(exit_nilfs_fs
)