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>
65 MODULE_AUTHOR("NTT Corp.");
66 MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
68 MODULE_LICENSE("GPL");
70 static int nilfs_remount(struct super_block *sb, int *flags, char *data);
73 * nilfs_error() - report failure condition on a filesystem
75 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
76 * reporting an error message. It should be called when NILFS detects
77 * incoherences or defects of meta data on disk. As for sustainable
78 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
79 * function should be used instead.
81 * The segment constructor must not call this function because it can
84 void nilfs_error(struct super_block *sb, const char *function,
87 struct nilfs_sb_info *sbi = NILFS_SB(sb);
91 printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function);
96 if (!(sb->s_flags & MS_RDONLY)) {
97 struct the_nilfs *nilfs = sbi->s_nilfs;
99 if (!nilfs_test_opt(sbi, ERRORS_CONT))
100 nilfs_detach_segment_constructor(sbi);
102 down_write(&nilfs->ns_sem);
103 if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
104 nilfs->ns_mount_state |= NILFS_ERROR_FS;
105 nilfs->ns_sbp[0]->s_state |=
106 cpu_to_le16(NILFS_ERROR_FS);
107 nilfs_commit_super(sbi, 1);
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_common(struct the_nilfs *nilfs)
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, nilfs->ns_bdi);
148 return &ii->vfs_inode;
151 struct inode *nilfs_alloc_inode(struct super_block *sb)
153 return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
156 void nilfs_destroy_inode(struct inode *inode)
158 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
161 static void init_once(void *obj)
163 struct nilfs_inode_info *ii = obj;
165 INIT_LIST_HEAD(&ii->i_dirty);
166 #ifdef CONFIG_NILFS_XATTR
167 init_rwsem(&ii->xattr_sem);
169 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
170 ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union;
171 inode_init_once(&ii->vfs_inode);
174 static int nilfs_init_inode_cache(void)
176 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
177 sizeof(struct nilfs_inode_info),
178 0, SLAB_RECLAIM_ACCOUNT,
181 return (nilfs_inode_cachep == NULL) ? -ENOMEM : 0;
184 static inline void nilfs_destroy_inode_cache(void)
186 kmem_cache_destroy(nilfs_inode_cachep);
189 static void nilfs_clear_inode(struct inode *inode)
191 struct nilfs_inode_info *ii = NILFS_I(inode);
194 * Free resources allocated in nilfs_read_inode(), here.
196 BUG_ON(!list_empty(&ii->i_dirty));
200 if (test_bit(NILFS_I_BMAP, &ii->i_state))
201 nilfs_bmap_clear(ii->i_bmap);
203 nilfs_btnode_cache_clear(&ii->i_btnode_cache);
206 static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb)
208 struct the_nilfs *nilfs = sbi->s_nilfs;
210 int barrier_done = 0;
212 if (nilfs_test_opt(sbi, BARRIER)) {
213 set_buffer_ordered(nilfs->ns_sbh[0]);
217 set_buffer_dirty(nilfs->ns_sbh[0]);
218 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
219 if (err == -EOPNOTSUPP && barrier_done) {
220 nilfs_warning(sbi->s_super, __func__,
221 "barrier-based sync failed. "
222 "disabling barriers\n");
223 nilfs_clear_opt(sbi, BARRIER);
225 clear_buffer_ordered(nilfs->ns_sbh[0]);
230 "NILFS: unable to write superblock (err=%d)\n", err);
231 if (err == -EIO && nilfs->ns_sbh[1]) {
232 nilfs_fall_back_super_block(nilfs);
236 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
239 * The latest segment becomes trailable from the position
240 * written in superblock.
242 clear_nilfs_discontinued(nilfs);
244 /* update GC protection for recent segments */
245 if (nilfs->ns_sbh[1]) {
248 set_buffer_dirty(nilfs->ns_sbh[1]);
249 if (!sync_dirty_buffer(nilfs->ns_sbh[1]))
250 sbp = nilfs->ns_sbp[1];
254 spin_lock(&nilfs->ns_last_segment_lock);
255 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
256 spin_unlock(&nilfs->ns_last_segment_lock);
263 int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
265 struct the_nilfs *nilfs = sbi->s_nilfs;
266 struct nilfs_super_block **sbp = nilfs->ns_sbp;
267 sector_t nfreeblocks;
271 /* nilfs->sem must be locked by the caller. */
272 if (sbp[0]->s_magic != NILFS_SUPER_MAGIC) {
273 if (sbp[1] && sbp[1]->s_magic == NILFS_SUPER_MAGIC)
274 nilfs_swap_super_block(nilfs);
276 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
281 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
283 printk(KERN_ERR "NILFS: failed to count free blocks\n");
286 spin_lock(&nilfs->ns_last_segment_lock);
287 sbp[0]->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
288 sbp[0]->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
289 sbp[0]->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
290 spin_unlock(&nilfs->ns_last_segment_lock);
293 nilfs->ns_sbwtime[0] = t;
294 sbp[0]->s_free_blocks_count = cpu_to_le64(nfreeblocks);
295 sbp[0]->s_wtime = cpu_to_le64(t);
297 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
298 (unsigned char *)sbp[0],
300 if (dupsb && sbp[1]) {
301 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
302 nilfs->ns_sbwtime[1] = t;
304 sbi->s_super->s_dirt = 0;
305 return nilfs_sync_super(sbi, dupsb);
308 static void nilfs_put_super(struct super_block *sb)
310 struct nilfs_sb_info *sbi = NILFS_SB(sb);
311 struct the_nilfs *nilfs = sbi->s_nilfs;
315 nilfs_detach_segment_constructor(sbi);
317 if (!(sb->s_flags & MS_RDONLY)) {
318 down_write(&nilfs->ns_sem);
319 nilfs->ns_sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state);
320 nilfs_commit_super(sbi, 1);
321 up_write(&nilfs->ns_sem);
323 down_write(&nilfs->ns_super_sem);
324 if (nilfs->ns_current == sbi)
325 nilfs->ns_current = NULL;
326 up_write(&nilfs->ns_super_sem);
328 nilfs_detach_checkpoint(sbi);
329 put_nilfs(sbi->s_nilfs);
331 sb->s_fs_info = NULL;
332 nilfs_put_sbinfo(sbi);
337 static int nilfs_sync_fs(struct super_block *sb, int wait)
339 struct nilfs_sb_info *sbi = NILFS_SB(sb);
340 struct the_nilfs *nilfs = sbi->s_nilfs;
343 /* This function is called when super block should be written back */
345 err = nilfs_construct_segment(sb);
347 down_write(&nilfs->ns_sem);
349 nilfs_commit_super(sbi, 1);
350 up_write(&nilfs->ns_sem);
355 int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
357 struct the_nilfs *nilfs = sbi->s_nilfs;
358 struct nilfs_checkpoint *raw_cp;
359 struct buffer_head *bh_cp;
362 down_write(&nilfs->ns_super_sem);
363 list_add(&sbi->s_list, &nilfs->ns_supers);
364 up_write(&nilfs->ns_super_sem);
366 sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size);
370 down_read(&nilfs->ns_segctor_sem);
371 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
373 up_read(&nilfs->ns_segctor_sem);
375 if (err == -ENOENT || err == -EINVAL) {
377 "NILFS: Invalid checkpoint "
378 "(checkpoint number=%llu)\n",
379 (unsigned long long)cno);
384 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
387 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
388 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
390 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
394 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
396 nilfs_mdt_destroy(sbi->s_ifile);
399 down_write(&nilfs->ns_super_sem);
400 list_del_init(&sbi->s_list);
401 up_write(&nilfs->ns_super_sem);
406 void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
408 struct the_nilfs *nilfs = sbi->s_nilfs;
410 nilfs_mdt_destroy(sbi->s_ifile);
412 down_write(&nilfs->ns_super_sem);
413 list_del_init(&sbi->s_list);
414 up_write(&nilfs->ns_super_sem);
417 static int nilfs_mark_recovery_complete(struct nilfs_sb_info *sbi)
419 struct the_nilfs *nilfs = sbi->s_nilfs;
422 down_write(&nilfs->ns_sem);
423 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
424 nilfs->ns_mount_state |= NILFS_VALID_FS;
425 err = nilfs_commit_super(sbi, 1);
427 printk(KERN_INFO "NILFS: recovery complete.\n");
429 up_write(&nilfs->ns_sem);
433 static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
435 struct super_block *sb = dentry->d_sb;
436 struct nilfs_sb_info *sbi = NILFS_SB(sb);
437 struct the_nilfs *nilfs = sbi->s_nilfs;
438 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
439 unsigned long long blocks;
440 unsigned long overhead;
441 unsigned long nrsvblocks;
442 sector_t nfreeblocks;
446 * Compute all of the segment blocks
448 * The blocks before first segment and after last segment
451 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
452 - nilfs->ns_first_data_block;
453 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
456 * Compute the overhead
458 * When distributing meta data blocks outside semgent structure,
459 * We must count them as the overhead.
463 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
467 buf->f_type = NILFS_SUPER_MAGIC;
468 buf->f_bsize = sb->s_blocksize;
469 buf->f_blocks = blocks - overhead;
470 buf->f_bfree = nfreeblocks;
471 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
472 (buf->f_bfree - nrsvblocks) : 0;
473 buf->f_files = atomic_read(&sbi->s_inodes_count);
474 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
475 buf->f_namelen = NILFS_NAME_LEN;
476 buf->f_fsid.val[0] = (u32)id;
477 buf->f_fsid.val[1] = (u32)(id >> 32);
482 static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
484 struct super_block *sb = vfs->mnt_sb;
485 struct nilfs_sb_info *sbi = NILFS_SB(sb);
487 if (!nilfs_test_opt(sbi, BARRIER))
488 seq_printf(seq, ",nobarrier");
489 if (nilfs_test_opt(sbi, SNAPSHOT))
490 seq_printf(seq, ",cp=%llu",
491 (unsigned long long int)sbi->s_snapshot_cno);
492 if (nilfs_test_opt(sbi, ERRORS_RO))
493 seq_printf(seq, ",errors=remount-ro");
494 if (nilfs_test_opt(sbi, ERRORS_PANIC))
495 seq_printf(seq, ",errors=panic");
496 if (nilfs_test_opt(sbi, STRICT_ORDER))
497 seq_printf(seq, ",order=strict");
502 static const struct super_operations nilfs_sops = {
503 .alloc_inode = nilfs_alloc_inode,
504 .destroy_inode = nilfs_destroy_inode,
505 .dirty_inode = nilfs_dirty_inode,
506 /* .write_inode = nilfs_write_inode, */
507 /* .put_inode = nilfs_put_inode, */
508 /* .drop_inode = nilfs_drop_inode, */
509 .delete_inode = nilfs_delete_inode,
510 .put_super = nilfs_put_super,
511 /* .write_super = nilfs_write_super, */
512 .sync_fs = nilfs_sync_fs,
513 /* .write_super_lockfs */
515 .statfs = nilfs_statfs,
516 .remount_fs = nilfs_remount,
517 .clear_inode = nilfs_clear_inode,
519 .show_options = nilfs_show_options
522 static struct inode *
523 nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
527 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
528 ino != NILFS_SKETCH_INO)
529 return ERR_PTR(-ESTALE);
531 inode = nilfs_iget(sb, ino);
533 return ERR_CAST(inode);
534 if (generation && inode->i_generation != generation) {
536 return ERR_PTR(-ESTALE);
542 static struct dentry *
543 nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
546 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
547 nilfs_nfs_get_inode);
550 static struct dentry *
551 nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
554 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
555 nilfs_nfs_get_inode);
558 static const struct export_operations nilfs_export_ops = {
559 .fh_to_dentry = nilfs_fh_to_dentry,
560 .fh_to_parent = nilfs_fh_to_parent,
561 .get_parent = nilfs_get_parent,
565 Opt_err_cont, Opt_err_panic, Opt_err_ro,
566 Opt_nobarrier, Opt_snapshot, Opt_order,
570 static match_table_t tokens = {
571 {Opt_err_cont, "errors=continue"},
572 {Opt_err_panic, "errors=panic"},
573 {Opt_err_ro, "errors=remount-ro"},
574 {Opt_nobarrier, "nobarrier"},
575 {Opt_snapshot, "cp=%u"},
576 {Opt_order, "order=%s"},
580 static int parse_options(char *options, struct super_block *sb)
582 struct nilfs_sb_info *sbi = NILFS_SB(sb);
584 substring_t args[MAX_OPT_ARGS];
590 while ((p = strsep(&options, ",")) != NULL) {
595 token = match_token(p, tokens, args);
598 nilfs_clear_opt(sbi, BARRIER);
601 if (strcmp(args[0].from, "relaxed") == 0)
602 /* Ordered data semantics */
603 nilfs_clear_opt(sbi, STRICT_ORDER);
604 else if (strcmp(args[0].from, "strict") == 0)
605 /* Strict in-order semantics */
606 nilfs_set_opt(sbi, STRICT_ORDER);
611 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
614 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
617 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
620 if (match_int(&args[0], &option) || option <= 0)
622 if (!(sb->s_flags & MS_RDONLY))
624 sbi->s_snapshot_cno = option;
625 nilfs_set_opt(sbi, SNAPSHOT);
629 "NILFS: Unrecognized mount option \"%s\"\n", p);
637 nilfs_set_default_options(struct nilfs_sb_info *sbi,
638 struct nilfs_super_block *sbp)
641 NILFS_MOUNT_ERRORS_CONT | NILFS_MOUNT_BARRIER;
644 static int nilfs_setup_super(struct nilfs_sb_info *sbi)
646 struct the_nilfs *nilfs = sbi->s_nilfs;
647 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
648 int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count);
649 int mnt_count = le16_to_cpu(sbp->s_mnt_count);
651 /* nilfs->sem must be locked by the caller. */
652 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
653 printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
654 } else if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
656 "NILFS warning: mounting fs with errors\n");
658 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
660 "NILFS warning: maximal mount count reached\n");
664 sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
666 sbp->s_mnt_count = cpu_to_le16(mnt_count + 1);
667 sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS);
668 sbp->s_mtime = cpu_to_le64(get_seconds());
669 return nilfs_commit_super(sbi, 1);
672 struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
673 u64 pos, int blocksize,
674 struct buffer_head **pbh)
676 unsigned long long sb_index = pos;
677 unsigned long offset;
679 offset = do_div(sb_index, blocksize);
680 *pbh = sb_bread(sb, sb_index);
683 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
686 int nilfs_store_magic_and_option(struct super_block *sb,
687 struct nilfs_super_block *sbp,
690 struct nilfs_sb_info *sbi = NILFS_SB(sb);
692 sb->s_magic = le16_to_cpu(sbp->s_magic);
694 /* FS independent flags */
695 #ifdef NILFS_ATIME_DISABLE
696 sb->s_flags |= MS_NOATIME;
699 nilfs_set_default_options(sbi, sbp);
701 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
702 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
703 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
704 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
706 return !parse_options(data, sb) ? -EINVAL : 0 ;
710 * nilfs_fill_super() - initialize a super block instance
712 * @data: mount options
713 * @silent: silent mode flag
714 * @nilfs: the_nilfs struct
716 * This function is called exclusively by nilfs->ns_mount_mutex.
717 * So, the recovery process is protected from other simultaneous mounts.
720 nilfs_fill_super(struct super_block *sb, void *data, int silent,
721 struct the_nilfs *nilfs)
723 struct nilfs_sb_info *sbi;
728 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
735 sbi->s_nilfs = nilfs;
737 atomic_set(&sbi->s_count, 1);
739 err = init_nilfs(nilfs, sbi, (char *)data);
743 spin_lock_init(&sbi->s_inode_lock);
744 INIT_LIST_HEAD(&sbi->s_dirty_files);
745 INIT_LIST_HEAD(&sbi->s_list);
748 * Following initialization is overlapped because
749 * nilfs_sb_info structure has been cleared at the beginning.
750 * But we reserve them to keep our interest and make ready
751 * for the future change.
753 get_random_bytes(&sbi->s_next_generation,
754 sizeof(sbi->s_next_generation));
755 spin_lock_init(&sbi->s_next_gen_lock);
757 sb->s_op = &nilfs_sops;
758 sb->s_export_op = &nilfs_export_ops;
762 if (!nilfs_loaded(nilfs)) {
763 err = load_nilfs(nilfs, sbi);
767 cno = nilfs_last_cno(nilfs);
769 if (sb->s_flags & MS_RDONLY) {
770 if (nilfs_test_opt(sbi, SNAPSHOT)) {
771 down_read(&nilfs->ns_segctor_sem);
772 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
773 sbi->s_snapshot_cno);
774 up_read(&nilfs->ns_segctor_sem);
782 "NILFS: The specified checkpoint is "
784 "(checkpoint number=%llu).\n",
785 (unsigned long long)sbi->s_snapshot_cno);
789 cno = sbi->s_snapshot_cno;
791 /* Read-only mount */
792 sbi->s_snapshot_cno = cno;
795 err = nilfs_attach_checkpoint(sbi, cno);
797 printk(KERN_ERR "NILFS: error loading a checkpoint"
798 " (checkpoint number=%llu).\n", (unsigned long long)cno);
802 if (!(sb->s_flags & MS_RDONLY)) {
803 err = nilfs_attach_segment_constructor(sbi);
805 goto failed_checkpoint;
808 root = nilfs_iget(sb, NILFS_ROOT_INO);
810 printk(KERN_ERR "NILFS: get root inode failed\n");
814 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
816 printk(KERN_ERR "NILFS: corrupt root inode.\n");
820 sb->s_root = d_alloc_root(root);
823 printk(KERN_ERR "NILFS: get root dentry failed\n");
828 if (!(sb->s_flags & MS_RDONLY)) {
829 down_write(&nilfs->ns_sem);
830 nilfs_setup_super(sbi);
831 up_write(&nilfs->ns_sem);
834 err = nilfs_mark_recovery_complete(sbi);
836 printk(KERN_ERR "NILFS: recovery failed.\n");
840 down_write(&nilfs->ns_super_sem);
841 if (!nilfs_test_opt(sbi, SNAPSHOT))
842 nilfs->ns_current = sbi;
843 up_write(&nilfs->ns_super_sem);
852 nilfs_detach_segment_constructor(sbi);
855 nilfs_detach_checkpoint(sbi);
859 sb->s_fs_info = NULL;
860 nilfs_put_sbinfo(sbi);
864 static int nilfs_remount(struct super_block *sb, int *flags, char *data)
866 struct nilfs_sb_info *sbi = NILFS_SB(sb);
867 struct nilfs_super_block *sbp;
868 struct the_nilfs *nilfs = sbi->s_nilfs;
869 unsigned long old_sb_flags;
870 struct nilfs_mount_options old_opts;
875 down_write(&nilfs->ns_super_sem);
876 old_sb_flags = sb->s_flags;
877 old_opts.mount_opt = sbi->s_mount_opt;
878 old_opts.snapshot_cno = sbi->s_snapshot_cno;
880 if (!parse_options(data, sb)) {
884 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
886 if ((*flags & MS_RDONLY) &&
887 sbi->s_snapshot_cno != old_opts.snapshot_cno) {
888 printk(KERN_WARNING "NILFS (device %s): couldn't "
889 "remount to a different snapshot. \n",
895 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
897 if (*flags & MS_RDONLY) {
898 /* Shutting down the segment constructor */
899 nilfs_detach_segment_constructor(sbi);
900 sb->s_flags |= MS_RDONLY;
902 sbi->s_snapshot_cno = nilfs_last_cno(nilfs);
903 /* nilfs_set_opt(sbi, SNAPSHOT); */
906 * Remounting a valid RW partition RDONLY, so set
907 * the RDONLY flag and then mark the partition as valid again.
909 down_write(&nilfs->ns_sem);
910 sbp = nilfs->ns_sbp[0];
911 if (!(sbp->s_state & le16_to_cpu(NILFS_VALID_FS)) &&
912 (nilfs->ns_mount_state & NILFS_VALID_FS))
913 sbp->s_state = cpu_to_le16(nilfs->ns_mount_state);
914 sbp->s_mtime = cpu_to_le64(get_seconds());
915 nilfs_commit_super(sbi, 1);
916 up_write(&nilfs->ns_sem);
919 * Mounting a RDONLY partition read-write, so reread and
920 * store the current valid flag. (It may have been changed
921 * by fsck since we originally mounted the partition.)
923 if (nilfs->ns_current && nilfs->ns_current != sbi) {
924 printk(KERN_WARNING "NILFS (device %s): couldn't "
925 "remount because an RW-mount exists.\n",
930 if (sbi->s_snapshot_cno != nilfs_last_cno(nilfs)) {
931 printk(KERN_WARNING "NILFS (device %s): couldn't "
932 "remount because the current RO-mount is not "
938 sb->s_flags &= ~MS_RDONLY;
939 nilfs_clear_opt(sbi, SNAPSHOT);
940 sbi->s_snapshot_cno = 0;
942 err = nilfs_attach_segment_constructor(sbi);
946 down_write(&nilfs->ns_sem);
947 nilfs_setup_super(sbi);
948 up_write(&nilfs->ns_sem);
950 nilfs->ns_current = sbi;
953 up_write(&nilfs->ns_super_sem);
958 sb->s_flags = old_sb_flags;
959 sbi->s_mount_opt = old_opts.mount_opt;
960 sbi->s_snapshot_cno = old_opts.snapshot_cno;
961 up_write(&nilfs->ns_super_sem);
966 struct nilfs_super_data {
967 struct block_device *bdev;
968 struct nilfs_sb_info *sbi;
974 * nilfs_identify - pre-read mount options needed to identify mount instance
975 * @data: mount options
976 * @sd: nilfs_super_data
978 static int nilfs_identify(char *data, struct nilfs_super_data *sd)
980 char *p, *options = data;
981 substring_t args[MAX_OPT_ARGS];
986 p = strsep(&options, ",");
987 if (p != NULL && *p) {
988 token = match_token(p, tokens, args);
989 if (token == Opt_snapshot) {
990 if (!(sd->flags & MS_RDONLY))
993 ret = match_int(&args[0], &option);
1004 "NILFS: invalid mount option: %s\n", p);
1008 BUG_ON(options == data);
1009 *(options - 1) = ',';
1014 static int nilfs_set_bdev_super(struct super_block *s, void *data)
1016 struct nilfs_super_data *sd = data;
1018 s->s_bdev = sd->bdev;
1019 s->s_dev = s->s_bdev->bd_dev;
1023 static int nilfs_test_bdev_super(struct super_block *s, void *data)
1025 struct nilfs_super_data *sd = data;
1027 return sd->sbi && s->s_fs_info == (void *)sd->sbi;
1031 nilfs_get_sb(struct file_system_type *fs_type, int flags,
1032 const char *dev_name, void *data, struct vfsmount *mnt)
1034 struct nilfs_super_data sd;
1035 struct super_block *s;
1036 struct the_nilfs *nilfs;
1037 int err, need_to_close = 1;
1039 sd.bdev = open_bdev_exclusive(dev_name, flags, fs_type);
1040 if (IS_ERR(sd.bdev))
1041 return PTR_ERR(sd.bdev);
1044 * To get mount instance using sget() vfs-routine, NILFS needs
1045 * much more information than normal filesystems to identify mount
1046 * instance. For snapshot mounts, not only a mount type (ro-mount
1047 * or rw-mount) but also a checkpoint number is required.
1051 if (nilfs_identify((char *)data, &sd)) {
1056 nilfs = find_or_create_nilfs(sd.bdev);
1062 mutex_lock(&nilfs->ns_mount_mutex);
1066 * Check if an exclusive mount exists or not.
1067 * Snapshot mounts coexist with a current mount
1068 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1069 * ro-mount are mutually exclusive.
1071 down_read(&nilfs->ns_super_sem);
1072 if (nilfs->ns_current &&
1073 ((nilfs->ns_current->s_super->s_flags ^ flags)
1075 up_read(&nilfs->ns_super_sem);
1079 up_read(&nilfs->ns_super_sem);
1083 * Find existing nilfs_sb_info struct
1085 sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
1088 * Get super block instance holding the nilfs_sb_info struct.
1089 * A new instance is allocated if no existing mount is present or
1090 * existing instance has been unmounted.
1092 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
1094 nilfs_put_sbinfo(sd.sbi);
1102 char b[BDEVNAME_SIZE];
1104 /* New superblock instance created */
1106 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1107 sb_set_blocksize(s, block_size(sd.bdev));
1109 err = nilfs_fill_super(s, data, flags & MS_VERBOSE, nilfs);
1113 s->s_flags |= MS_ACTIVE;
1117 mutex_unlock(&nilfs->ns_mount_mutex);
1120 close_bdev_exclusive(sd.bdev, flags);
1121 simple_set_mnt(mnt, s);
1125 mutex_unlock(&nilfs->ns_mount_mutex);
1128 close_bdev_exclusive(sd.bdev, flags);
1133 /* Abandoning the newly allocated superblock */
1134 mutex_unlock(&nilfs->ns_mount_mutex);
1136 up_write(&s->s_umount);
1137 deactivate_super(s);
1139 * deactivate_super() invokes close_bdev_exclusive().
1140 * We must finish all post-cleaning before this call;
1141 * put_nilfs() needs the block device.
1146 struct file_system_type nilfs_fs_type = {
1147 .owner = THIS_MODULE,
1149 .get_sb = nilfs_get_sb,
1150 .kill_sb = kill_block_super,
1151 .fs_flags = FS_REQUIRES_DEV,
1154 static int __init init_nilfs_fs(void)
1158 err = nilfs_init_inode_cache();
1162 err = nilfs_init_transaction_cache();
1164 goto failed_inode_cache;
1166 err = nilfs_init_segbuf_cache();
1168 goto failed_transaction_cache;
1170 err = nilfs_btree_path_cache_init();
1172 goto failed_segbuf_cache;
1174 err = register_filesystem(&nilfs_fs_type);
1176 goto failed_btree_path_cache;
1180 failed_btree_path_cache:
1181 nilfs_btree_path_cache_destroy();
1183 failed_segbuf_cache:
1184 nilfs_destroy_segbuf_cache();
1186 failed_transaction_cache:
1187 nilfs_destroy_transaction_cache();
1190 nilfs_destroy_inode_cache();
1196 static void __exit exit_nilfs_fs(void)
1198 nilfs_destroy_segbuf_cache();
1199 nilfs_destroy_transaction_cache();
1200 nilfs_destroy_inode_cache();
1201 nilfs_btree_path_cache_destroy();
1202 unregister_filesystem(&nilfs_fs_type);
1205 module_init(init_nilfs_fs)
1206 module_exit(exit_nilfs_fs)