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 struct kmem_cache *nilfs_inode_cachep;
71 struct kmem_cache *nilfs_transaction_cachep;
72 struct kmem_cache *nilfs_segbuf_cachep;
73 struct kmem_cache *nilfs_btree_path_cache;
75 static int nilfs_remount(struct super_block *sb, int *flags, char *data);
77 static void nilfs_set_error(struct nilfs_sb_info *sbi)
79 struct the_nilfs *nilfs = sbi->s_nilfs;
81 down_write(&nilfs->ns_sem);
82 if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
83 nilfs->ns_mount_state |= NILFS_ERROR_FS;
84 nilfs->ns_sbp[0]->s_state |= cpu_to_le16(NILFS_ERROR_FS);
85 nilfs_commit_super(sbi, 1);
87 up_write(&nilfs->ns_sem);
91 * nilfs_error() - report failure condition on a filesystem
93 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
94 * reporting an error message. It should be called when NILFS detects
95 * incoherences or defects of meta data on disk. As for sustainable
96 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
97 * function should be used instead.
99 * The segment constructor must not call this function because it can
102 void nilfs_error(struct super_block *sb, const char *function,
103 const char *fmt, ...)
105 struct nilfs_sb_info *sbi = NILFS_SB(sb);
109 printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function);
114 if (!(sb->s_flags & MS_RDONLY)) {
115 nilfs_set_error(sbi);
117 if (nilfs_test_opt(sbi, ERRORS_RO)) {
118 printk(KERN_CRIT "Remounting filesystem read-only\n");
119 sb->s_flags |= MS_RDONLY;
123 if (nilfs_test_opt(sbi, ERRORS_PANIC))
124 panic("NILFS (device %s): panic forced after error\n",
128 void nilfs_warning(struct super_block *sb, const char *function,
129 const char *fmt, ...)
134 printk(KERN_WARNING "NILFS warning (device %s): %s: ",
142 struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs)
144 struct nilfs_inode_info *ii;
146 ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS);
151 ii->vfs_inode.i_version = 1;
152 nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi);
153 return &ii->vfs_inode;
156 struct inode *nilfs_alloc_inode(struct super_block *sb)
158 return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
161 void nilfs_destroy_inode(struct inode *inode)
163 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
166 static void nilfs_clear_inode(struct inode *inode)
168 struct nilfs_inode_info *ii = NILFS_I(inode);
171 * Free resources allocated in nilfs_read_inode(), here.
173 BUG_ON(!list_empty(&ii->i_dirty));
177 if (test_bit(NILFS_I_BMAP, &ii->i_state))
178 nilfs_bmap_clear(ii->i_bmap);
180 nilfs_btnode_cache_clear(&ii->i_btnode_cache);
183 static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb)
185 struct the_nilfs *nilfs = sbi->s_nilfs;
187 int barrier_done = 0;
189 if (nilfs_test_opt(sbi, BARRIER)) {
190 set_buffer_ordered(nilfs->ns_sbh[0]);
194 set_buffer_dirty(nilfs->ns_sbh[0]);
195 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
196 if (err == -EOPNOTSUPP && barrier_done) {
197 nilfs_warning(sbi->s_super, __func__,
198 "barrier-based sync failed. "
199 "disabling barriers\n");
200 nilfs_clear_opt(sbi, BARRIER);
202 clear_buffer_ordered(nilfs->ns_sbh[0]);
207 "NILFS: unable to write superblock (err=%d)\n", err);
208 if (err == -EIO && nilfs->ns_sbh[1]) {
209 nilfs_fall_back_super_block(nilfs);
213 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
216 * The latest segment becomes trailable from the position
217 * written in superblock.
219 clear_nilfs_discontinued(nilfs);
221 /* update GC protection for recent segments */
222 if (nilfs->ns_sbh[1]) {
225 set_buffer_dirty(nilfs->ns_sbh[1]);
226 if (!sync_dirty_buffer(nilfs->ns_sbh[1]))
227 sbp = nilfs->ns_sbp[1];
231 spin_lock(&nilfs->ns_last_segment_lock);
232 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
233 spin_unlock(&nilfs->ns_last_segment_lock);
240 void nilfs_set_log_cursor(struct nilfs_super_block *sbp,
241 struct the_nilfs *nilfs)
243 sector_t nfreeblocks;
245 /* nilfs->ns_sem must be locked by the caller. */
246 nilfs_count_free_blocks(nilfs, &nfreeblocks);
247 sbp->s_free_blocks_count = cpu_to_le64(nfreeblocks);
249 spin_lock(&nilfs->ns_last_segment_lock);
250 sbp->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
251 sbp->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
252 sbp->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
253 spin_unlock(&nilfs->ns_last_segment_lock);
256 int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
258 struct the_nilfs *nilfs = sbi->s_nilfs;
259 struct nilfs_super_block **sbp = nilfs->ns_sbp;
262 /* nilfs->sem must be locked by the caller. */
263 if (sbp[0]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) {
264 if (sbp[1] && sbp[1]->s_magic == cpu_to_le16(NILFS_SUPER_MAGIC))
265 nilfs_swap_super_block(nilfs);
267 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
272 nilfs_set_log_cursor(sbp[0], nilfs);
275 nilfs->ns_sbwtime[0] = t;
276 sbp[0]->s_wtime = cpu_to_le64(t);
278 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
279 (unsigned char *)sbp[0],
281 if (dupsb && sbp[1]) {
282 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
283 nilfs->ns_sbwtime[1] = t;
285 clear_nilfs_sb_dirty(nilfs);
286 return nilfs_sync_super(sbi, dupsb);
290 * nilfs_cleanup_super() - write filesystem state for cleanup
291 * @sbi: nilfs_sb_info to be unmounted or degraded to read-only
293 * This function restores state flags in the on-disk super block.
294 * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the
295 * filesystem was not clean previously.
297 int nilfs_cleanup_super(struct nilfs_sb_info *sbi)
299 struct nilfs_super_block **sbp = sbi->s_nilfs->ns_sbp;
302 sbp[0]->s_state = cpu_to_le16(sbi->s_nilfs->ns_mount_state);
303 ret = nilfs_commit_super(sbi, 1);
307 static void nilfs_put_super(struct super_block *sb)
309 struct nilfs_sb_info *sbi = NILFS_SB(sb);
310 struct the_nilfs *nilfs = sbi->s_nilfs;
314 nilfs_detach_segment_constructor(sbi);
316 if (!(sb->s_flags & MS_RDONLY)) {
317 down_write(&nilfs->ns_sem);
318 nilfs_cleanup_super(sbi);
319 up_write(&nilfs->ns_sem);
321 down_write(&nilfs->ns_super_sem);
322 if (nilfs->ns_current == sbi)
323 nilfs->ns_current = NULL;
324 up_write(&nilfs->ns_super_sem);
326 nilfs_detach_checkpoint(sbi);
327 put_nilfs(sbi->s_nilfs);
329 sb->s_fs_info = NULL;
330 nilfs_put_sbinfo(sbi);
335 static int nilfs_sync_fs(struct super_block *sb, int wait)
337 struct nilfs_sb_info *sbi = NILFS_SB(sb);
338 struct the_nilfs *nilfs = sbi->s_nilfs;
341 /* This function is called when super block should be written back */
343 err = nilfs_construct_segment(sb);
345 down_write(&nilfs->ns_sem);
346 if (nilfs_sb_dirty(nilfs))
347 nilfs_commit_super(sbi, 1);
348 up_write(&nilfs->ns_sem);
353 int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
355 struct the_nilfs *nilfs = sbi->s_nilfs;
356 struct nilfs_checkpoint *raw_cp;
357 struct buffer_head *bh_cp;
360 down_write(&nilfs->ns_super_sem);
361 list_add(&sbi->s_list, &nilfs->ns_supers);
362 up_write(&nilfs->ns_super_sem);
364 sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size);
368 down_read(&nilfs->ns_segctor_sem);
369 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
371 up_read(&nilfs->ns_segctor_sem);
373 if (err == -ENOENT || err == -EINVAL) {
375 "NILFS: Invalid checkpoint "
376 "(checkpoint number=%llu)\n",
377 (unsigned long long)cno);
382 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
385 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
386 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
388 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
392 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
394 nilfs_mdt_destroy(sbi->s_ifile);
397 down_write(&nilfs->ns_super_sem);
398 list_del_init(&sbi->s_list);
399 up_write(&nilfs->ns_super_sem);
404 void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
406 struct the_nilfs *nilfs = sbi->s_nilfs;
408 nilfs_mdt_destroy(sbi->s_ifile);
410 down_write(&nilfs->ns_super_sem);
411 list_del_init(&sbi->s_list);
412 up_write(&nilfs->ns_super_sem);
415 static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
417 struct super_block *sb = dentry->d_sb;
418 struct nilfs_sb_info *sbi = NILFS_SB(sb);
419 struct the_nilfs *nilfs = sbi->s_nilfs;
420 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
421 unsigned long long blocks;
422 unsigned long overhead;
423 unsigned long nrsvblocks;
424 sector_t nfreeblocks;
428 * Compute all of the segment blocks
430 * The blocks before first segment and after last segment
433 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
434 - nilfs->ns_first_data_block;
435 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
438 * Compute the overhead
440 * When distributing meta data blocks outside segment structure,
441 * We must count them as the overhead.
445 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
449 buf->f_type = NILFS_SUPER_MAGIC;
450 buf->f_bsize = sb->s_blocksize;
451 buf->f_blocks = blocks - overhead;
452 buf->f_bfree = nfreeblocks;
453 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
454 (buf->f_bfree - nrsvblocks) : 0;
455 buf->f_files = atomic_read(&sbi->s_inodes_count);
456 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
457 buf->f_namelen = NILFS_NAME_LEN;
458 buf->f_fsid.val[0] = (u32)id;
459 buf->f_fsid.val[1] = (u32)(id >> 32);
464 static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
466 struct super_block *sb = vfs->mnt_sb;
467 struct nilfs_sb_info *sbi = NILFS_SB(sb);
469 if (!nilfs_test_opt(sbi, BARRIER))
470 seq_printf(seq, ",nobarrier");
471 if (nilfs_test_opt(sbi, SNAPSHOT))
472 seq_printf(seq, ",cp=%llu",
473 (unsigned long long int)sbi->s_snapshot_cno);
474 if (nilfs_test_opt(sbi, ERRORS_PANIC))
475 seq_printf(seq, ",errors=panic");
476 if (nilfs_test_opt(sbi, ERRORS_CONT))
477 seq_printf(seq, ",errors=continue");
478 if (nilfs_test_opt(sbi, STRICT_ORDER))
479 seq_printf(seq, ",order=strict");
480 if (nilfs_test_opt(sbi, NORECOVERY))
481 seq_printf(seq, ",norecovery");
482 if (nilfs_test_opt(sbi, DISCARD))
483 seq_printf(seq, ",discard");
488 static const struct super_operations nilfs_sops = {
489 .alloc_inode = nilfs_alloc_inode,
490 .destroy_inode = nilfs_destroy_inode,
491 .dirty_inode = nilfs_dirty_inode,
492 /* .write_inode = nilfs_write_inode, */
493 /* .put_inode = nilfs_put_inode, */
494 /* .drop_inode = nilfs_drop_inode, */
495 .delete_inode = nilfs_delete_inode,
496 .put_super = nilfs_put_super,
497 /* .write_super = nilfs_write_super, */
498 .sync_fs = nilfs_sync_fs,
499 /* .write_super_lockfs */
501 .statfs = nilfs_statfs,
502 .remount_fs = nilfs_remount,
503 .clear_inode = nilfs_clear_inode,
505 .show_options = nilfs_show_options
508 static struct inode *
509 nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
513 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
514 ino != NILFS_SKETCH_INO)
515 return ERR_PTR(-ESTALE);
517 inode = nilfs_iget(sb, ino);
519 return ERR_CAST(inode);
520 if (generation && inode->i_generation != generation) {
522 return ERR_PTR(-ESTALE);
528 static struct dentry *
529 nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
532 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
533 nilfs_nfs_get_inode);
536 static struct dentry *
537 nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
540 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
541 nilfs_nfs_get_inode);
544 static const struct export_operations nilfs_export_ops = {
545 .fh_to_dentry = nilfs_fh_to_dentry,
546 .fh_to_parent = nilfs_fh_to_parent,
547 .get_parent = nilfs_get_parent,
551 Opt_err_cont, Opt_err_panic, Opt_err_ro,
552 Opt_nobarrier, Opt_snapshot, Opt_order, Opt_norecovery,
553 Opt_discard, Opt_err,
556 static match_table_t tokens = {
557 {Opt_err_cont, "errors=continue"},
558 {Opt_err_panic, "errors=panic"},
559 {Opt_err_ro, "errors=remount-ro"},
560 {Opt_nobarrier, "nobarrier"},
561 {Opt_snapshot, "cp=%u"},
562 {Opt_order, "order=%s"},
563 {Opt_norecovery, "norecovery"},
564 {Opt_discard, "discard"},
568 static int parse_options(char *options, struct super_block *sb)
570 struct nilfs_sb_info *sbi = NILFS_SB(sb);
572 substring_t args[MAX_OPT_ARGS];
578 while ((p = strsep(&options, ",")) != NULL) {
583 token = match_token(p, tokens, args);
586 nilfs_clear_opt(sbi, BARRIER);
589 if (strcmp(args[0].from, "relaxed") == 0)
590 /* Ordered data semantics */
591 nilfs_clear_opt(sbi, STRICT_ORDER);
592 else if (strcmp(args[0].from, "strict") == 0)
593 /* Strict in-order semantics */
594 nilfs_set_opt(sbi, STRICT_ORDER);
599 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
602 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
605 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
608 if (match_int(&args[0], &option) || option <= 0)
610 if (!(sb->s_flags & MS_RDONLY))
612 sbi->s_snapshot_cno = option;
613 nilfs_set_opt(sbi, SNAPSHOT);
616 nilfs_set_opt(sbi, NORECOVERY);
619 nilfs_set_opt(sbi, DISCARD);
623 "NILFS: Unrecognized mount option \"%s\"\n", p);
631 nilfs_set_default_options(struct nilfs_sb_info *sbi,
632 struct nilfs_super_block *sbp)
635 NILFS_MOUNT_ERRORS_RO | NILFS_MOUNT_BARRIER;
638 static int nilfs_setup_super(struct nilfs_sb_info *sbi)
640 struct the_nilfs *nilfs = sbi->s_nilfs;
641 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
642 int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count);
643 int mnt_count = le16_to_cpu(sbp->s_mnt_count);
645 /* nilfs->sem must be locked by the caller. */
646 if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
648 "NILFS warning: mounting fs with errors\n");
650 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
652 "NILFS warning: maximal mount count reached\n");
656 sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
658 sbp->s_mnt_count = cpu_to_le16(mnt_count + 1);
659 sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS);
660 sbp->s_mtime = cpu_to_le64(get_seconds());
661 return nilfs_commit_super(sbi, 1);
664 struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
665 u64 pos, int blocksize,
666 struct buffer_head **pbh)
668 unsigned long long sb_index = pos;
669 unsigned long offset;
671 offset = do_div(sb_index, blocksize);
672 *pbh = sb_bread(sb, sb_index);
675 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
678 int nilfs_store_magic_and_option(struct super_block *sb,
679 struct nilfs_super_block *sbp,
682 struct nilfs_sb_info *sbi = NILFS_SB(sb);
684 sb->s_magic = le16_to_cpu(sbp->s_magic);
686 /* FS independent flags */
687 #ifdef NILFS_ATIME_DISABLE
688 sb->s_flags |= MS_NOATIME;
691 nilfs_set_default_options(sbi, sbp);
693 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
694 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
695 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
696 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
698 return !parse_options(data, sb) ? -EINVAL : 0 ;
702 * nilfs_fill_super() - initialize a super block instance
704 * @data: mount options
705 * @silent: silent mode flag
706 * @nilfs: the_nilfs struct
708 * This function is called exclusively by nilfs->ns_mount_mutex.
709 * So, the recovery process is protected from other simultaneous mounts.
712 nilfs_fill_super(struct super_block *sb, void *data, int silent,
713 struct the_nilfs *nilfs)
715 struct nilfs_sb_info *sbi;
720 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
727 sbi->s_nilfs = nilfs;
729 atomic_set(&sbi->s_count, 1);
731 err = init_nilfs(nilfs, sbi, (char *)data);
735 spin_lock_init(&sbi->s_inode_lock);
736 INIT_LIST_HEAD(&sbi->s_dirty_files);
737 INIT_LIST_HEAD(&sbi->s_list);
740 * Following initialization is overlapped because
741 * nilfs_sb_info structure has been cleared at the beginning.
742 * But we reserve them to keep our interest and make ready
743 * for the future change.
745 get_random_bytes(&sbi->s_next_generation,
746 sizeof(sbi->s_next_generation));
747 spin_lock_init(&sbi->s_next_gen_lock);
749 sb->s_op = &nilfs_sops;
750 sb->s_export_op = &nilfs_export_ops;
753 sb->s_bdi = nilfs->ns_bdi;
755 err = load_nilfs(nilfs, sbi);
759 cno = nilfs_last_cno(nilfs);
761 if (sb->s_flags & MS_RDONLY) {
762 if (nilfs_test_opt(sbi, SNAPSHOT)) {
763 down_read(&nilfs->ns_segctor_sem);
764 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
765 sbi->s_snapshot_cno);
766 up_read(&nilfs->ns_segctor_sem);
774 "NILFS: The specified checkpoint is "
776 "(checkpoint number=%llu).\n",
777 (unsigned long long)sbi->s_snapshot_cno);
781 cno = sbi->s_snapshot_cno;
785 err = nilfs_attach_checkpoint(sbi, cno);
787 printk(KERN_ERR "NILFS: error loading a checkpoint"
788 " (checkpoint number=%llu).\n", (unsigned long long)cno);
792 if (!(sb->s_flags & MS_RDONLY)) {
793 err = nilfs_attach_segment_constructor(sbi);
795 goto failed_checkpoint;
798 root = nilfs_iget(sb, NILFS_ROOT_INO);
800 printk(KERN_ERR "NILFS: get root inode failed\n");
804 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
806 printk(KERN_ERR "NILFS: corrupt root inode.\n");
810 sb->s_root = d_alloc_root(root);
813 printk(KERN_ERR "NILFS: get root dentry failed\n");
818 if (!(sb->s_flags & MS_RDONLY)) {
819 down_write(&nilfs->ns_sem);
820 nilfs_setup_super(sbi);
821 up_write(&nilfs->ns_sem);
824 down_write(&nilfs->ns_super_sem);
825 if (!nilfs_test_opt(sbi, SNAPSHOT))
826 nilfs->ns_current = sbi;
827 up_write(&nilfs->ns_super_sem);
832 nilfs_detach_segment_constructor(sbi);
835 nilfs_detach_checkpoint(sbi);
839 sb->s_fs_info = NULL;
840 nilfs_put_sbinfo(sbi);
844 static int nilfs_remount(struct super_block *sb, int *flags, char *data)
846 struct nilfs_sb_info *sbi = NILFS_SB(sb);
847 struct the_nilfs *nilfs = sbi->s_nilfs;
848 unsigned long old_sb_flags;
849 struct nilfs_mount_options old_opts;
850 int was_snapshot, err;
854 down_write(&nilfs->ns_super_sem);
855 old_sb_flags = sb->s_flags;
856 old_opts.mount_opt = sbi->s_mount_opt;
857 old_opts.snapshot_cno = sbi->s_snapshot_cno;
858 was_snapshot = nilfs_test_opt(sbi, SNAPSHOT);
860 if (!parse_options(data, sb)) {
864 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
868 if (!(*flags & MS_RDONLY)) {
869 printk(KERN_ERR "NILFS (device %s): cannot remount "
870 "snapshot read/write.\n",
873 } else if (sbi->s_snapshot_cno != old_opts.snapshot_cno) {
874 printk(KERN_ERR "NILFS (device %s): cannot "
875 "remount to a different snapshot.\n",
880 if (nilfs_test_opt(sbi, SNAPSHOT)) {
881 printk(KERN_ERR "NILFS (device %s): cannot change "
882 "a regular mount to a snapshot.\n",
888 if (!nilfs_valid_fs(nilfs)) {
889 printk(KERN_WARNING "NILFS (device %s): couldn't "
890 "remount because the filesystem is in an "
891 "incomplete recovery state.\n", sb->s_id);
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;
903 * Remounting a valid RW partition RDONLY, so set
904 * the RDONLY flag and then mark the partition as valid again.
906 down_write(&nilfs->ns_sem);
907 nilfs_cleanup_super(sbi);
908 up_write(&nilfs->ns_sem);
911 * Mounting a RDONLY partition read-write, so reread and
912 * store the current valid flag. (It may have been changed
913 * by fsck since we originally mounted the partition.)
915 sb->s_flags &= ~MS_RDONLY;
917 err = nilfs_attach_segment_constructor(sbi);
921 down_write(&nilfs->ns_sem);
922 nilfs_setup_super(sbi);
923 up_write(&nilfs->ns_sem);
926 up_write(&nilfs->ns_super_sem);
931 sb->s_flags = old_sb_flags;
932 sbi->s_mount_opt = old_opts.mount_opt;
933 sbi->s_snapshot_cno = old_opts.snapshot_cno;
934 up_write(&nilfs->ns_super_sem);
939 struct nilfs_super_data {
940 struct block_device *bdev;
941 struct nilfs_sb_info *sbi;
947 * nilfs_identify - pre-read mount options needed to identify mount instance
948 * @data: mount options
949 * @sd: nilfs_super_data
951 static int nilfs_identify(char *data, struct nilfs_super_data *sd)
953 char *p, *options = data;
954 substring_t args[MAX_OPT_ARGS];
959 p = strsep(&options, ",");
960 if (p != NULL && *p) {
961 token = match_token(p, tokens, args);
962 if (token == Opt_snapshot) {
963 if (!(sd->flags & MS_RDONLY))
966 ret = match_int(&args[0], &option);
977 "NILFS: invalid mount option: %s\n", p);
981 BUG_ON(options == data);
982 *(options - 1) = ',';
987 static int nilfs_set_bdev_super(struct super_block *s, void *data)
989 struct nilfs_super_data *sd = data;
991 s->s_bdev = sd->bdev;
992 s->s_dev = s->s_bdev->bd_dev;
996 static int nilfs_test_bdev_super(struct super_block *s, void *data)
998 struct nilfs_super_data *sd = data;
1000 return sd->sbi && s->s_fs_info == (void *)sd->sbi;
1004 nilfs_get_sb(struct file_system_type *fs_type, int flags,
1005 const char *dev_name, void *data, struct vfsmount *mnt)
1007 struct nilfs_super_data sd;
1008 struct super_block *s;
1009 fmode_t mode = FMODE_READ;
1010 struct the_nilfs *nilfs;
1011 int err, need_to_close = 1;
1013 if (!(flags & MS_RDONLY))
1014 mode |= FMODE_WRITE;
1016 sd.bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1017 if (IS_ERR(sd.bdev))
1018 return PTR_ERR(sd.bdev);
1021 * To get mount instance using sget() vfs-routine, NILFS needs
1022 * much more information than normal filesystems to identify mount
1023 * instance. For snapshot mounts, not only a mount type (ro-mount
1024 * or rw-mount) but also a checkpoint number is required.
1028 if (nilfs_identify((char *)data, &sd)) {
1033 nilfs = find_or_create_nilfs(sd.bdev);
1039 mutex_lock(&nilfs->ns_mount_mutex);
1043 * Check if an exclusive mount exists or not.
1044 * Snapshot mounts coexist with a current mount
1045 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1046 * ro-mount are mutually exclusive.
1048 down_read(&nilfs->ns_super_sem);
1049 if (nilfs->ns_current &&
1050 ((nilfs->ns_current->s_super->s_flags ^ flags)
1052 up_read(&nilfs->ns_super_sem);
1056 up_read(&nilfs->ns_super_sem);
1060 * Find existing nilfs_sb_info struct
1062 sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
1065 * Get super block instance holding the nilfs_sb_info struct.
1066 * A new instance is allocated if no existing mount is present or
1067 * existing instance has been unmounted.
1069 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
1071 nilfs_put_sbinfo(sd.sbi);
1079 char b[BDEVNAME_SIZE];
1081 /* New superblock instance created */
1084 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1085 sb_set_blocksize(s, block_size(sd.bdev));
1087 err = nilfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0,
1092 s->s_flags |= MS_ACTIVE;
1096 mutex_unlock(&nilfs->ns_mount_mutex);
1099 close_bdev_exclusive(sd.bdev, mode);
1100 simple_set_mnt(mnt, s);
1104 mutex_unlock(&nilfs->ns_mount_mutex);
1107 close_bdev_exclusive(sd.bdev, mode);
1112 /* Abandoning the newly allocated superblock */
1113 mutex_unlock(&nilfs->ns_mount_mutex);
1115 deactivate_locked_super(s);
1117 * deactivate_locked_super() invokes close_bdev_exclusive().
1118 * We must finish all post-cleaning before this call;
1119 * put_nilfs() needs the block device.
1124 struct file_system_type nilfs_fs_type = {
1125 .owner = THIS_MODULE,
1127 .get_sb = nilfs_get_sb,
1128 .kill_sb = kill_block_super,
1129 .fs_flags = FS_REQUIRES_DEV,
1132 static void nilfs_inode_init_once(void *obj)
1134 struct nilfs_inode_info *ii = obj;
1136 INIT_LIST_HEAD(&ii->i_dirty);
1137 #ifdef CONFIG_NILFS_XATTR
1138 init_rwsem(&ii->xattr_sem);
1140 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
1141 ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union;
1142 inode_init_once(&ii->vfs_inode);
1145 static void nilfs_segbuf_init_once(void *obj)
1147 memset(obj, 0, sizeof(struct nilfs_segment_buffer));
1150 static void nilfs_destroy_cachep(void)
1152 if (nilfs_inode_cachep)
1153 kmem_cache_destroy(nilfs_inode_cachep);
1154 if (nilfs_transaction_cachep)
1155 kmem_cache_destroy(nilfs_transaction_cachep);
1156 if (nilfs_segbuf_cachep)
1157 kmem_cache_destroy(nilfs_segbuf_cachep);
1158 if (nilfs_btree_path_cache)
1159 kmem_cache_destroy(nilfs_btree_path_cache);
1162 static int __init nilfs_init_cachep(void)
1164 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
1165 sizeof(struct nilfs_inode_info), 0,
1166 SLAB_RECLAIM_ACCOUNT, nilfs_inode_init_once);
1167 if (!nilfs_inode_cachep)
1170 nilfs_transaction_cachep = kmem_cache_create("nilfs2_transaction_cache",
1171 sizeof(struct nilfs_transaction_info), 0,
1172 SLAB_RECLAIM_ACCOUNT, NULL);
1173 if (!nilfs_transaction_cachep)
1176 nilfs_segbuf_cachep = kmem_cache_create("nilfs2_segbuf_cache",
1177 sizeof(struct nilfs_segment_buffer), 0,
1178 SLAB_RECLAIM_ACCOUNT, nilfs_segbuf_init_once);
1179 if (!nilfs_segbuf_cachep)
1182 nilfs_btree_path_cache = kmem_cache_create("nilfs2_btree_path_cache",
1183 sizeof(struct nilfs_btree_path) * NILFS_BTREE_LEVEL_MAX,
1185 if (!nilfs_btree_path_cache)
1191 nilfs_destroy_cachep();
1195 static int __init init_nilfs_fs(void)
1199 err = nilfs_init_cachep();
1203 err = register_filesystem(&nilfs_fs_type);
1207 printk(KERN_INFO "NILFS version 2 loaded\n");
1211 nilfs_destroy_cachep();
1216 static void __exit exit_nilfs_fs(void)
1218 nilfs_destroy_cachep();
1219 unregister_filesystem(&nilfs_fs_type);
1222 module_init(init_nilfs_fs)
1223 module_exit(exit_nilfs_fs)