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 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_remount(struct super_block
*sb
, int *flags
, char *data
);
78 static void nilfs_set_error(struct nilfs_sb_info
*sbi
)
80 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
81 struct nilfs_super_block
**sbp
;
83 down_write(&nilfs
->ns_sem
);
84 if (!(nilfs
->ns_mount_state
& NILFS_ERROR_FS
)) {
85 nilfs
->ns_mount_state
|= NILFS_ERROR_FS
;
86 sbp
= nilfs_prepare_super(sbi
, 0);
88 sbp
[0]->s_state
|= cpu_to_le16(NILFS_ERROR_FS
);
90 sbp
[1]->s_state
|= cpu_to_le16(NILFS_ERROR_FS
);
91 nilfs_commit_super(sbi
, NILFS_SB_COMMIT_ALL
);
94 up_write(&nilfs
->ns_sem
);
98 * nilfs_error() - report failure condition on a filesystem
100 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
101 * reporting an error message. It should be called when NILFS detects
102 * incoherences or defects of meta data on disk. As for sustainable
103 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
104 * function should be used instead.
106 * The segment constructor must not call this function because it can
109 void nilfs_error(struct super_block
*sb
, const char *function
,
110 const char *fmt
, ...)
112 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
116 printk(KERN_CRIT
"NILFS error (device %s): %s: ", sb
->s_id
, function
);
121 if (!(sb
->s_flags
& MS_RDONLY
)) {
122 nilfs_set_error(sbi
);
124 if (nilfs_test_opt(sbi
, ERRORS_RO
)) {
125 printk(KERN_CRIT
"Remounting filesystem read-only\n");
126 sb
->s_flags
|= MS_RDONLY
;
130 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
131 panic("NILFS (device %s): panic forced after error\n",
135 void nilfs_warning(struct super_block
*sb
, const char *function
,
136 const char *fmt
, ...)
141 printk(KERN_WARNING
"NILFS warning (device %s): %s: ",
149 struct inode
*nilfs_alloc_inode_common(struct the_nilfs
*nilfs
)
151 struct nilfs_inode_info
*ii
;
153 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 struct nilfs_mdt_info
*mdi
= NILFS_MDT(inode
);
174 kfree(mdi
->mi_bgl
); /* kfree(NULL) is safe */
177 kmem_cache_free(nilfs_inode_cachep
, NILFS_I(inode
));
180 static int nilfs_sync_super(struct nilfs_sb_info
*sbi
, int flag
)
182 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
186 set_buffer_dirty(nilfs
->ns_sbh
[0]);
188 if (nilfs_test_opt(sbi
, BARRIER
)) {
189 err
= __sync_dirty_buffer(nilfs
->ns_sbh
[0],
190 WRITE_SYNC
| WRITE_BARRIER
);
191 if (err
== -EOPNOTSUPP
) {
192 nilfs_warning(sbi
->s_super
, __func__
,
193 "barrier-based sync failed. "
194 "disabling barriers\n");
195 nilfs_clear_opt(sbi
, BARRIER
);
199 err
= sync_dirty_buffer(nilfs
->ns_sbh
[0]);
204 "NILFS: unable to write superblock (err=%d)\n", err
);
205 if (err
== -EIO
&& nilfs
->ns_sbh
[1]) {
207 * sbp[0] points to newer log than sbp[1],
208 * so copy sbp[0] to sbp[1] to take over sbp[0].
210 memcpy(nilfs
->ns_sbp
[1], nilfs
->ns_sbp
[0],
212 nilfs_fall_back_super_block(nilfs
);
216 struct nilfs_super_block
*sbp
= nilfs
->ns_sbp
[0];
218 nilfs
->ns_sbwcount
++;
221 * The latest segment becomes trailable from the position
222 * written in superblock.
224 clear_nilfs_discontinued(nilfs
);
226 /* update GC protection for recent segments */
227 if (nilfs
->ns_sbh
[1]) {
228 if (flag
== NILFS_SB_COMMIT_ALL
) {
229 set_buffer_dirty(nilfs
->ns_sbh
[1]);
230 if (sync_dirty_buffer(nilfs
->ns_sbh
[1]) < 0)
233 if (le64_to_cpu(nilfs
->ns_sbp
[1]->s_last_cno
) <
234 le64_to_cpu(nilfs
->ns_sbp
[0]->s_last_cno
))
235 sbp
= nilfs
->ns_sbp
[1];
238 spin_lock(&nilfs
->ns_last_segment_lock
);
239 nilfs
->ns_prot_seq
= le64_to_cpu(sbp
->s_last_seq
);
240 spin_unlock(&nilfs
->ns_last_segment_lock
);
246 void nilfs_set_log_cursor(struct nilfs_super_block
*sbp
,
247 struct the_nilfs
*nilfs
)
249 sector_t nfreeblocks
;
251 /* nilfs->ns_sem must be locked by the caller. */
252 nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
253 sbp
->s_free_blocks_count
= cpu_to_le64(nfreeblocks
);
255 spin_lock(&nilfs
->ns_last_segment_lock
);
256 sbp
->s_last_seq
= cpu_to_le64(nilfs
->ns_last_seq
);
257 sbp
->s_last_pseg
= cpu_to_le64(nilfs
->ns_last_pseg
);
258 sbp
->s_last_cno
= cpu_to_le64(nilfs
->ns_last_cno
);
259 spin_unlock(&nilfs
->ns_last_segment_lock
);
262 struct nilfs_super_block
**nilfs_prepare_super(struct nilfs_sb_info
*sbi
,
265 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
266 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
268 /* nilfs->ns_sem must be locked by the caller. */
269 if (sbp
[0]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
271 sbp
[1]->s_magic
== cpu_to_le16(NILFS_SUPER_MAGIC
)) {
272 memcpy(sbp
[0], sbp
[1], nilfs
->ns_sbsize
);
274 printk(KERN_CRIT
"NILFS: superblock broke on dev %s\n",
279 sbp
[1]->s_magic
!= cpu_to_le16(NILFS_SUPER_MAGIC
)) {
280 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
284 nilfs_swap_super_block(nilfs
);
289 int nilfs_commit_super(struct nilfs_sb_info
*sbi
, int flag
)
291 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
292 struct nilfs_super_block
**sbp
= nilfs
->ns_sbp
;
295 /* nilfs->ns_sem must be locked by the caller. */
297 nilfs
->ns_sbwtime
= t
;
298 sbp
[0]->s_wtime
= cpu_to_le64(t
);
300 sbp
[0]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
301 (unsigned char *)sbp
[0],
303 if (flag
== NILFS_SB_COMMIT_ALL
&& sbp
[1]) {
304 sbp
[1]->s_wtime
= sbp
[0]->s_wtime
;
306 sbp
[1]->s_sum
= cpu_to_le32(crc32_le(nilfs
->ns_crc_seed
,
307 (unsigned char *)sbp
[1],
310 clear_nilfs_sb_dirty(nilfs
);
311 return nilfs_sync_super(sbi
, flag
);
315 * nilfs_cleanup_super() - write filesystem state for cleanup
316 * @sbi: nilfs_sb_info to be unmounted or degraded to read-only
318 * This function restores state flags in the on-disk super block.
319 * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the
320 * filesystem was not clean previously.
322 int nilfs_cleanup_super(struct nilfs_sb_info
*sbi
)
324 struct nilfs_super_block
**sbp
;
325 int flag
= NILFS_SB_COMMIT
;
328 sbp
= nilfs_prepare_super(sbi
, 0);
330 sbp
[0]->s_state
= cpu_to_le16(sbi
->s_nilfs
->ns_mount_state
);
331 nilfs_set_log_cursor(sbp
[0], sbi
->s_nilfs
);
332 if (sbp
[1] && sbp
[0]->s_last_cno
== sbp
[1]->s_last_cno
) {
334 * make the "clean" flag also to the opposite
335 * super block if both super blocks point to
336 * the same checkpoint.
338 sbp
[1]->s_state
= sbp
[0]->s_state
;
339 flag
= NILFS_SB_COMMIT_ALL
;
341 ret
= nilfs_commit_super(sbi
, flag
);
346 static void nilfs_put_super(struct super_block
*sb
)
348 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
349 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
351 nilfs_detach_segment_constructor(sbi
);
353 if (!(sb
->s_flags
& MS_RDONLY
)) {
354 down_write(&nilfs
->ns_sem
);
355 nilfs_cleanup_super(sbi
);
356 up_write(&nilfs
->ns_sem
);
358 down_write(&nilfs
->ns_super_sem
);
359 if (nilfs
->ns_current
== sbi
)
360 nilfs
->ns_current
= NULL
;
361 list_del_init(&sbi
->s_list
);
362 up_write(&nilfs
->ns_super_sem
);
364 put_nilfs(sbi
->s_nilfs
);
366 sb
->s_fs_info
= NULL
;
367 nilfs_put_sbinfo(sbi
);
370 static int nilfs_sync_fs(struct super_block
*sb
, int wait
)
372 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
373 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
374 struct nilfs_super_block
**sbp
;
377 /* This function is called when super block should be written back */
379 err
= nilfs_construct_segment(sb
);
381 down_write(&nilfs
->ns_sem
);
382 if (nilfs_sb_dirty(nilfs
)) {
383 sbp
= nilfs_prepare_super(sbi
, nilfs_sb_will_flip(nilfs
));
385 nilfs_set_log_cursor(sbp
[0], nilfs
);
386 nilfs_commit_super(sbi
, NILFS_SB_COMMIT
);
389 up_write(&nilfs
->ns_sem
);
394 int nilfs_attach_checkpoint(struct nilfs_sb_info
*sbi
, __u64 cno
, int curr_mnt
,
395 struct nilfs_root
**rootp
)
397 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
398 struct nilfs_root
*root
;
399 struct nilfs_checkpoint
*raw_cp
;
400 struct buffer_head
*bh_cp
;
403 root
= nilfs_find_or_create_root(
404 nilfs
, curr_mnt
? NILFS_CPTREE_CURRENT_CNO
: cno
);
409 goto reuse
; /* already attached checkpoint */
411 root
->ifile
= nilfs_ifile_new(sbi
, nilfs
->ns_inode_size
);
415 down_read(&nilfs
->ns_segctor_sem
);
416 err
= nilfs_cpfile_get_checkpoint(nilfs
->ns_cpfile
, cno
, 0, &raw_cp
,
418 up_read(&nilfs
->ns_segctor_sem
);
420 if (err
== -ENOENT
|| err
== -EINVAL
) {
422 "NILFS: Invalid checkpoint "
423 "(checkpoint number=%llu)\n",
424 (unsigned long long)cno
);
429 err
= nilfs_read_inode_common(root
->ifile
, &raw_cp
->cp_ifile_inode
);
433 atomic_set(&root
->inodes_count
, le64_to_cpu(raw_cp
->cp_inodes_count
));
434 atomic_set(&root
->blocks_count
, le64_to_cpu(raw_cp
->cp_blocks_count
));
436 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
443 nilfs_cpfile_put_checkpoint(nilfs
->ns_cpfile
, cno
, bh_cp
);
445 nilfs_put_root(root
);
450 static int nilfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
452 struct super_block
*sb
= dentry
->d_sb
;
453 struct nilfs_root
*root
= NILFS_I(dentry
->d_inode
)->i_root
;
454 struct the_nilfs
*nilfs
= root
->nilfs
;
455 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
456 unsigned long long blocks
;
457 unsigned long overhead
;
458 unsigned long nrsvblocks
;
459 sector_t nfreeblocks
;
463 * Compute all of the segment blocks
465 * The blocks before first segment and after last segment
468 blocks
= nilfs
->ns_blocks_per_segment
* nilfs
->ns_nsegments
469 - nilfs
->ns_first_data_block
;
470 nrsvblocks
= nilfs
->ns_nrsvsegs
* nilfs
->ns_blocks_per_segment
;
473 * Compute the overhead
475 * When distributing meta data blocks outside segment structure,
476 * We must count them as the overhead.
480 err
= nilfs_count_free_blocks(nilfs
, &nfreeblocks
);
484 buf
->f_type
= NILFS_SUPER_MAGIC
;
485 buf
->f_bsize
= sb
->s_blocksize
;
486 buf
->f_blocks
= blocks
- overhead
;
487 buf
->f_bfree
= nfreeblocks
;
488 buf
->f_bavail
= (buf
->f_bfree
>= nrsvblocks
) ?
489 (buf
->f_bfree
- nrsvblocks
) : 0;
490 buf
->f_files
= atomic_read(&root
->inodes_count
);
491 buf
->f_ffree
= 0; /* nilfs_count_free_inodes(sb); */
492 buf
->f_namelen
= NILFS_NAME_LEN
;
493 buf
->f_fsid
.val
[0] = (u32
)id
;
494 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
499 static int nilfs_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
501 struct super_block
*sb
= vfs
->mnt_sb
;
502 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
504 if (!nilfs_test_opt(sbi
, BARRIER
))
505 seq_puts(seq
, ",nobarrier");
506 if (nilfs_test_opt(sbi
, SNAPSHOT
))
507 seq_printf(seq
, ",cp=%llu",
508 (unsigned long long int)sbi
->s_snapshot_cno
);
509 if (nilfs_test_opt(sbi
, ERRORS_PANIC
))
510 seq_puts(seq
, ",errors=panic");
511 if (nilfs_test_opt(sbi
, ERRORS_CONT
))
512 seq_puts(seq
, ",errors=continue");
513 if (nilfs_test_opt(sbi
, STRICT_ORDER
))
514 seq_puts(seq
, ",order=strict");
515 if (nilfs_test_opt(sbi
, NORECOVERY
))
516 seq_puts(seq
, ",norecovery");
517 if (nilfs_test_opt(sbi
, DISCARD
))
518 seq_puts(seq
, ",discard");
523 static const struct super_operations nilfs_sops
= {
524 .alloc_inode
= nilfs_alloc_inode
,
525 .destroy_inode
= nilfs_destroy_inode
,
526 .dirty_inode
= nilfs_dirty_inode
,
527 /* .write_inode = nilfs_write_inode, */
528 /* .put_inode = nilfs_put_inode, */
529 /* .drop_inode = nilfs_drop_inode, */
530 .evict_inode
= nilfs_evict_inode
,
531 .put_super
= nilfs_put_super
,
532 /* .write_super = nilfs_write_super, */
533 .sync_fs
= nilfs_sync_fs
,
534 /* .write_super_lockfs */
536 .statfs
= nilfs_statfs
,
537 .remount_fs
= nilfs_remount
,
539 .show_options
= nilfs_show_options
543 Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
544 Opt_barrier
, Opt_nobarrier
, Opt_snapshot
, Opt_order
, Opt_norecovery
,
545 Opt_discard
, Opt_nodiscard
, Opt_err
,
548 static match_table_t tokens
= {
549 {Opt_err_cont
, "errors=continue"},
550 {Opt_err_panic
, "errors=panic"},
551 {Opt_err_ro
, "errors=remount-ro"},
552 {Opt_barrier
, "barrier"},
553 {Opt_nobarrier
, "nobarrier"},
554 {Opt_snapshot
, "cp=%u"},
555 {Opt_order
, "order=%s"},
556 {Opt_norecovery
, "norecovery"},
557 {Opt_discard
, "discard"},
558 {Opt_nodiscard
, "nodiscard"},
562 static int parse_options(char *options
, struct super_block
*sb
, int is_remount
)
564 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
566 substring_t args
[MAX_OPT_ARGS
];
572 while ((p
= strsep(&options
, ",")) != NULL
) {
577 token
= match_token(p
, tokens
, args
);
580 nilfs_set_opt(sbi
, BARRIER
);
583 nilfs_clear_opt(sbi
, BARRIER
);
586 if (strcmp(args
[0].from
, "relaxed") == 0)
587 /* Ordered data semantics */
588 nilfs_clear_opt(sbi
, STRICT_ORDER
);
589 else if (strcmp(args
[0].from
, "strict") == 0)
590 /* Strict in-order semantics */
591 nilfs_set_opt(sbi
, STRICT_ORDER
);
596 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_PANIC
);
599 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_RO
);
602 nilfs_write_opt(sbi
, ERROR_MODE
, ERRORS_CONT
);
605 if (match_int(&args
[0], &option
) || option
<= 0)
608 if (!nilfs_test_opt(sbi
, SNAPSHOT
)) {
610 "NILFS: cannot change regular "
611 "mount to snapshot.\n");
613 } else if (option
!= sbi
->s_snapshot_cno
) {
615 "NILFS: cannot remount to a "
616 "different snapshot.\n");
621 if (!(sb
->s_flags
& MS_RDONLY
)) {
622 printk(KERN_ERR
"NILFS: cannot mount snapshot "
623 "read/write. A read-only option is "
627 sbi
->s_snapshot_cno
= option
;
628 nilfs_set_opt(sbi
, SNAPSHOT
);
631 nilfs_set_opt(sbi
, NORECOVERY
);
634 nilfs_set_opt(sbi
, DISCARD
);
637 nilfs_clear_opt(sbi
, DISCARD
);
641 "NILFS: Unrecognized mount option \"%s\"\n", p
);
649 nilfs_set_default_options(struct nilfs_sb_info
*sbi
,
650 struct nilfs_super_block
*sbp
)
653 NILFS_MOUNT_ERRORS_RO
| NILFS_MOUNT_BARRIER
;
656 static int nilfs_setup_super(struct nilfs_sb_info
*sbi
)
658 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
659 struct nilfs_super_block
**sbp
;
663 /* nilfs->ns_sem must be locked by the caller. */
664 sbp
= nilfs_prepare_super(sbi
, 0);
668 max_mnt_count
= le16_to_cpu(sbp
[0]->s_max_mnt_count
);
669 mnt_count
= le16_to_cpu(sbp
[0]->s_mnt_count
);
671 if (nilfs
->ns_mount_state
& NILFS_ERROR_FS
) {
673 "NILFS warning: mounting fs with errors\n");
675 } else if (max_mnt_count
>= 0 && mnt_count
>= max_mnt_count
) {
677 "NILFS warning: maximal mount count reached\n");
681 sbp
[0]->s_max_mnt_count
= cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT
);
683 sbp
[0]->s_mnt_count
= cpu_to_le16(mnt_count
+ 1);
685 cpu_to_le16(le16_to_cpu(sbp
[0]->s_state
) & ~NILFS_VALID_FS
);
686 sbp
[0]->s_mtime
= cpu_to_le64(get_seconds());
687 /* synchronize sbp[1] with sbp[0] */
688 memcpy(sbp
[1], sbp
[0], nilfs
->ns_sbsize
);
689 return nilfs_commit_super(sbi
, NILFS_SB_COMMIT_ALL
);
692 struct nilfs_super_block
*nilfs_read_super_block(struct super_block
*sb
,
693 u64 pos
, int blocksize
,
694 struct buffer_head
**pbh
)
696 unsigned long long sb_index
= pos
;
697 unsigned long offset
;
699 offset
= do_div(sb_index
, blocksize
);
700 *pbh
= sb_bread(sb
, sb_index
);
703 return (struct nilfs_super_block
*)((char *)(*pbh
)->b_data
+ offset
);
706 int nilfs_store_magic_and_option(struct super_block
*sb
,
707 struct nilfs_super_block
*sbp
,
710 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
712 sb
->s_magic
= le16_to_cpu(sbp
->s_magic
);
714 /* FS independent flags */
715 #ifdef NILFS_ATIME_DISABLE
716 sb
->s_flags
|= MS_NOATIME
;
719 nilfs_set_default_options(sbi
, sbp
);
721 sbi
->s_resuid
= le16_to_cpu(sbp
->s_def_resuid
);
722 sbi
->s_resgid
= le16_to_cpu(sbp
->s_def_resgid
);
723 sbi
->s_interval
= le32_to_cpu(sbp
->s_c_interval
);
724 sbi
->s_watermark
= le32_to_cpu(sbp
->s_c_block_max
);
726 return !parse_options(data
, sb
, 0) ? -EINVAL
: 0 ;
729 int nilfs_check_feature_compatibility(struct super_block
*sb
,
730 struct nilfs_super_block
*sbp
)
734 features
= le64_to_cpu(sbp
->s_feature_incompat
) &
735 ~NILFS_FEATURE_INCOMPAT_SUPP
;
737 printk(KERN_ERR
"NILFS: couldn't mount because of unsupported "
738 "optional features (%llx)\n",
739 (unsigned long long)features
);
742 features
= le64_to_cpu(sbp
->s_feature_compat_ro
) &
743 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
744 if (!(sb
->s_flags
& MS_RDONLY
) && features
) {
745 printk(KERN_ERR
"NILFS: couldn't mount RDWR because of "
746 "unsupported optional features (%llx)\n",
747 (unsigned long long)features
);
753 static int nilfs_get_root_dentry(struct super_block
*sb
,
754 struct nilfs_root
*root
,
755 struct dentry
**root_dentry
)
758 struct dentry
*dentry
;
761 inode
= nilfs_iget(sb
, root
, NILFS_ROOT_INO
);
763 printk(KERN_ERR
"NILFS: get root inode failed\n");
764 ret
= PTR_ERR(inode
);
767 if (!S_ISDIR(inode
->i_mode
) || !inode
->i_blocks
|| !inode
->i_size
) {
769 printk(KERN_ERR
"NILFS: corrupt root inode.\n");
774 dentry
= d_alloc_root(inode
);
777 printk(KERN_ERR
"NILFS: get root dentry failed\n");
781 *root_dentry
= dentry
;
787 * nilfs_fill_super() - initialize a super block instance
789 * @data: mount options
790 * @silent: silent mode flag
791 * @nilfs: the_nilfs struct
793 * This function is called exclusively by nilfs->ns_mount_mutex.
794 * So, the recovery process is protected from other simultaneous mounts.
797 nilfs_fill_super(struct super_block
*sb
, void *data
, int silent
,
798 struct the_nilfs
*nilfs
)
800 struct nilfs_sb_info
*sbi
;
801 struct nilfs_root
*fsroot
;
805 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
812 sbi
->s_nilfs
= nilfs
;
814 atomic_set(&sbi
->s_count
, 1);
816 err
= init_nilfs(nilfs
, sbi
, (char *)data
);
820 spin_lock_init(&sbi
->s_inode_lock
);
821 INIT_LIST_HEAD(&sbi
->s_dirty_files
);
822 INIT_LIST_HEAD(&sbi
->s_list
);
825 * Following initialization is overlapped because
826 * nilfs_sb_info structure has been cleared at the beginning.
827 * But we reserve them to keep our interest and make ready
828 * for the future change.
830 get_random_bytes(&sbi
->s_next_generation
,
831 sizeof(sbi
->s_next_generation
));
832 spin_lock_init(&sbi
->s_next_gen_lock
);
834 sb
->s_op
= &nilfs_sops
;
835 sb
->s_export_op
= &nilfs_export_ops
;
838 sb
->s_bdi
= nilfs
->ns_bdi
;
840 err
= load_nilfs(nilfs
, sbi
);
844 cno
= nilfs_last_cno(nilfs
);
847 if (sb
->s_flags
& MS_RDONLY
) {
848 if (nilfs_test_opt(sbi
, SNAPSHOT
)) {
849 down_read(&nilfs
->ns_segctor_sem
);
850 err
= nilfs_cpfile_is_snapshot(nilfs
->ns_cpfile
,
851 sbi
->s_snapshot_cno
);
852 up_read(&nilfs
->ns_segctor_sem
);
860 "NILFS: The specified checkpoint is "
862 "(checkpoint number=%llu).\n",
863 (unsigned long long)sbi
->s_snapshot_cno
);
867 cno
= sbi
->s_snapshot_cno
;
872 err
= nilfs_attach_checkpoint(sbi
, cno
, curr_mnt
, &fsroot
);
874 printk(KERN_ERR
"NILFS: error loading a checkpoint"
875 " (checkpoint number=%llu).\n", (unsigned long long)cno
);
879 if (!(sb
->s_flags
& MS_RDONLY
)) {
880 err
= nilfs_attach_segment_constructor(sbi
, fsroot
);
882 goto failed_checkpoint
;
885 err
= nilfs_get_root_dentry(sb
, fsroot
, &sb
->s_root
);
889 nilfs_put_root(fsroot
);
891 if (!(sb
->s_flags
& MS_RDONLY
)) {
892 down_write(&nilfs
->ns_sem
);
893 nilfs_setup_super(sbi
);
894 up_write(&nilfs
->ns_sem
);
897 down_write(&nilfs
->ns_super_sem
);
898 list_add(&sbi
->s_list
, &nilfs
->ns_supers
);
899 if (!nilfs_test_opt(sbi
, SNAPSHOT
))
900 nilfs
->ns_current
= sbi
;
901 up_write(&nilfs
->ns_super_sem
);
906 nilfs_detach_segment_constructor(sbi
);
909 nilfs_put_root(fsroot
);
913 sb
->s_fs_info
= NULL
;
914 nilfs_put_sbinfo(sbi
);
918 static int nilfs_remount(struct super_block
*sb
, int *flags
, char *data
)
920 struct nilfs_sb_info
*sbi
= NILFS_SB(sb
);
921 struct the_nilfs
*nilfs
= sbi
->s_nilfs
;
922 unsigned long old_sb_flags
;
923 struct nilfs_mount_options old_opts
;
924 int was_snapshot
, err
;
926 down_write(&nilfs
->ns_super_sem
);
927 old_sb_flags
= sb
->s_flags
;
928 old_opts
.mount_opt
= sbi
->s_mount_opt
;
929 old_opts
.snapshot_cno
= sbi
->s_snapshot_cno
;
930 was_snapshot
= nilfs_test_opt(sbi
, SNAPSHOT
);
932 if (!parse_options(data
, sb
, 1)) {
936 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
);
939 if (was_snapshot
&& !(*flags
& MS_RDONLY
)) {
940 printk(KERN_ERR
"NILFS (device %s): cannot remount snapshot "
941 "read/write.\n", sb
->s_id
);
945 if (!nilfs_valid_fs(nilfs
)) {
946 printk(KERN_WARNING
"NILFS (device %s): couldn't "
947 "remount because the filesystem is in an "
948 "incomplete recovery state.\n", sb
->s_id
);
952 if ((*flags
& MS_RDONLY
) == (sb
->s_flags
& MS_RDONLY
))
954 if (*flags
& MS_RDONLY
) {
955 /* Shutting down the segment constructor */
956 nilfs_detach_segment_constructor(sbi
);
957 sb
->s_flags
|= MS_RDONLY
;
960 * Remounting a valid RW partition RDONLY, so set
961 * the RDONLY flag and then mark the partition as valid again.
963 down_write(&nilfs
->ns_sem
);
964 nilfs_cleanup_super(sbi
);
965 up_write(&nilfs
->ns_sem
);
968 struct nilfs_root
*root
;
971 * Mounting a RDONLY partition read-write, so reread and
972 * store the current valid flag. (It may have been changed
973 * by fsck since we originally mounted the partition.)
975 down_read(&nilfs
->ns_sem
);
976 features
= le64_to_cpu(nilfs
->ns_sbp
[0]->s_feature_compat_ro
) &
977 ~NILFS_FEATURE_COMPAT_RO_SUPP
;
978 up_read(&nilfs
->ns_sem
);
980 printk(KERN_WARNING
"NILFS (device %s): couldn't "
981 "remount RDWR because of unsupported optional "
983 sb
->s_id
, (unsigned long long)features
);
988 sb
->s_flags
&= ~MS_RDONLY
;
990 root
= NILFS_I(sb
->s_root
->d_inode
)->i_root
;
991 err
= nilfs_attach_segment_constructor(sbi
, root
);
995 down_write(&nilfs
->ns_sem
);
996 nilfs_setup_super(sbi
);
997 up_write(&nilfs
->ns_sem
);
1000 up_write(&nilfs
->ns_super_sem
);
1004 sb
->s_flags
= old_sb_flags
;
1005 sbi
->s_mount_opt
= old_opts
.mount_opt
;
1006 sbi
->s_snapshot_cno
= old_opts
.snapshot_cno
;
1007 up_write(&nilfs
->ns_super_sem
);
1011 struct nilfs_super_data
{
1012 struct block_device
*bdev
;
1013 struct nilfs_sb_info
*sbi
;
1019 * nilfs_identify - pre-read mount options needed to identify mount instance
1020 * @data: mount options
1021 * @sd: nilfs_super_data
1023 static int nilfs_identify(char *data
, struct nilfs_super_data
*sd
)
1025 char *p
, *options
= data
;
1026 substring_t args
[MAX_OPT_ARGS
];
1031 p
= strsep(&options
, ",");
1032 if (p
!= NULL
&& *p
) {
1033 token
= match_token(p
, tokens
, args
);
1034 if (token
== Opt_snapshot
) {
1035 if (!(sd
->flags
& MS_RDONLY
))
1038 ret
= match_int(&args
[0], &option
);
1049 "NILFS: invalid mount option: %s\n", p
);
1053 BUG_ON(options
== data
);
1054 *(options
- 1) = ',';
1059 static int nilfs_set_bdev_super(struct super_block
*s
, void *data
)
1061 struct nilfs_super_data
*sd
= data
;
1063 s
->s_bdev
= sd
->bdev
;
1064 s
->s_dev
= s
->s_bdev
->bd_dev
;
1068 static int nilfs_test_bdev_super(struct super_block
*s
, void *data
)
1070 struct nilfs_super_data
*sd
= data
;
1072 return sd
->sbi
&& s
->s_fs_info
== (void *)sd
->sbi
;
1076 nilfs_get_sb(struct file_system_type
*fs_type
, int flags
,
1077 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1079 struct nilfs_super_data sd
;
1080 struct super_block
*s
;
1081 fmode_t mode
= FMODE_READ
;
1082 struct the_nilfs
*nilfs
;
1083 int err
, need_to_close
= 1;
1085 if (!(flags
& MS_RDONLY
))
1086 mode
|= FMODE_WRITE
;
1088 sd
.bdev
= open_bdev_exclusive(dev_name
, mode
, fs_type
);
1089 if (IS_ERR(sd
.bdev
))
1090 return PTR_ERR(sd
.bdev
);
1093 * To get mount instance using sget() vfs-routine, NILFS needs
1094 * much more information than normal filesystems to identify mount
1095 * instance. For snapshot mounts, not only a mount type (ro-mount
1096 * or rw-mount) but also a checkpoint number is required.
1100 if (nilfs_identify((char *)data
, &sd
)) {
1105 nilfs
= find_or_create_nilfs(sd
.bdev
);
1111 mutex_lock(&nilfs
->ns_mount_mutex
);
1115 * Check if an exclusive mount exists or not.
1116 * Snapshot mounts coexist with a current mount
1117 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1118 * ro-mount are mutually exclusive.
1120 down_read(&nilfs
->ns_super_sem
);
1121 if (nilfs
->ns_current
&&
1122 ((nilfs
->ns_current
->s_super
->s_flags
^ flags
)
1124 up_read(&nilfs
->ns_super_sem
);
1128 up_read(&nilfs
->ns_super_sem
);
1132 * Find existing nilfs_sb_info struct
1134 sd
.sbi
= nilfs_find_sbinfo(nilfs
, !(flags
& MS_RDONLY
), sd
.cno
);
1137 * Get super block instance holding the nilfs_sb_info struct.
1138 * A new instance is allocated if no existing mount is present or
1139 * existing instance has been unmounted.
1141 s
= sget(fs_type
, nilfs_test_bdev_super
, nilfs_set_bdev_super
, &sd
);
1143 nilfs_put_sbinfo(sd
.sbi
);
1151 char b
[BDEVNAME_SIZE
];
1153 /* New superblock instance created */
1156 strlcpy(s
->s_id
, bdevname(sd
.bdev
, b
), sizeof(s
->s_id
));
1157 sb_set_blocksize(s
, block_size(sd
.bdev
));
1159 err
= nilfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0,
1164 s
->s_flags
|= MS_ACTIVE
;
1168 mutex_unlock(&nilfs
->ns_mount_mutex
);
1171 close_bdev_exclusive(sd
.bdev
, mode
);
1172 simple_set_mnt(mnt
, s
);
1176 mutex_unlock(&nilfs
->ns_mount_mutex
);
1179 close_bdev_exclusive(sd
.bdev
, mode
);
1183 /* Abandoning the newly allocated superblock */
1184 mutex_unlock(&nilfs
->ns_mount_mutex
);
1186 deactivate_locked_super(s
);
1188 * deactivate_locked_super() invokes close_bdev_exclusive().
1189 * We must finish all post-cleaning before this call;
1190 * put_nilfs() needs the block device.
1195 struct file_system_type nilfs_fs_type
= {
1196 .owner
= THIS_MODULE
,
1198 .get_sb
= nilfs_get_sb
,
1199 .kill_sb
= kill_block_super
,
1200 .fs_flags
= FS_REQUIRES_DEV
,
1203 static void nilfs_inode_init_once(void *obj
)
1205 struct nilfs_inode_info
*ii
= obj
;
1207 INIT_LIST_HEAD(&ii
->i_dirty
);
1208 #ifdef CONFIG_NILFS_XATTR
1209 init_rwsem(&ii
->xattr_sem
);
1211 nilfs_btnode_cache_init_once(&ii
->i_btnode_cache
);
1212 ii
->i_bmap
= &ii
->i_bmap_data
;
1213 inode_init_once(&ii
->vfs_inode
);
1216 static void nilfs_segbuf_init_once(void *obj
)
1218 memset(obj
, 0, sizeof(struct nilfs_segment_buffer
));
1221 static void nilfs_destroy_cachep(void)
1223 if (nilfs_inode_cachep
)
1224 kmem_cache_destroy(nilfs_inode_cachep
);
1225 if (nilfs_transaction_cachep
)
1226 kmem_cache_destroy(nilfs_transaction_cachep
);
1227 if (nilfs_segbuf_cachep
)
1228 kmem_cache_destroy(nilfs_segbuf_cachep
);
1229 if (nilfs_btree_path_cache
)
1230 kmem_cache_destroy(nilfs_btree_path_cache
);
1233 static int __init
nilfs_init_cachep(void)
1235 nilfs_inode_cachep
= kmem_cache_create("nilfs2_inode_cache",
1236 sizeof(struct nilfs_inode_info
), 0,
1237 SLAB_RECLAIM_ACCOUNT
, nilfs_inode_init_once
);
1238 if (!nilfs_inode_cachep
)
1241 nilfs_transaction_cachep
= kmem_cache_create("nilfs2_transaction_cache",
1242 sizeof(struct nilfs_transaction_info
), 0,
1243 SLAB_RECLAIM_ACCOUNT
, NULL
);
1244 if (!nilfs_transaction_cachep
)
1247 nilfs_segbuf_cachep
= kmem_cache_create("nilfs2_segbuf_cache",
1248 sizeof(struct nilfs_segment_buffer
), 0,
1249 SLAB_RECLAIM_ACCOUNT
, nilfs_segbuf_init_once
);
1250 if (!nilfs_segbuf_cachep
)
1253 nilfs_btree_path_cache
= kmem_cache_create("nilfs2_btree_path_cache",
1254 sizeof(struct nilfs_btree_path
) * NILFS_BTREE_LEVEL_MAX
,
1256 if (!nilfs_btree_path_cache
)
1262 nilfs_destroy_cachep();
1266 static int __init
init_nilfs_fs(void)
1270 err
= nilfs_init_cachep();
1274 err
= register_filesystem(&nilfs_fs_type
);
1278 printk(KERN_INFO
"NILFS version 2 loaded\n");
1282 nilfs_destroy_cachep();
1287 static void __exit
exit_nilfs_fs(void)
1289 nilfs_destroy_cachep();
1290 unregister_filesystem(&nilfs_fs_type
);
1293 module_init(init_nilfs_fs
)
1294 module_exit(exit_nilfs_fs
)