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/exportfs.h>
52 #include <linux/seq_file.h>
53 #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);
158 ii->vfs_inode.i_version = 1;
159 nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi);
160 return &ii->vfs_inode;
163 struct inode *nilfs_alloc_inode(struct super_block *sb)
165 return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
168 void nilfs_destroy_inode(struct inode *inode)
170 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
173 static int nilfs_sync_super(struct nilfs_sb_info *sbi, int flag)
175 struct the_nilfs *nilfs = sbi->s_nilfs;
179 set_buffer_dirty(nilfs->ns_sbh[0]);
180 if (nilfs_test_opt(sbi, BARRIER)) {
181 err = __sync_dirty_buffer(nilfs->ns_sbh[0],
182 WRITE_SYNC | WRITE_FLUSH_FUA);
184 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
189 "NILFS: unable to write superblock (err=%d)\n", err);
190 if (err == -EIO && nilfs->ns_sbh[1]) {
192 * sbp[0] points to newer log than sbp[1],
193 * so copy sbp[0] to sbp[1] to take over sbp[0].
195 memcpy(nilfs->ns_sbp[1], nilfs->ns_sbp[0],
197 nilfs_fall_back_super_block(nilfs);
201 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
203 nilfs->ns_sbwcount++;
206 * The latest segment becomes trailable from the position
207 * written in superblock.
209 clear_nilfs_discontinued(nilfs);
211 /* update GC protection for recent segments */
212 if (nilfs->ns_sbh[1]) {
213 if (flag == NILFS_SB_COMMIT_ALL) {
214 set_buffer_dirty(nilfs->ns_sbh[1]);
215 if (sync_dirty_buffer(nilfs->ns_sbh[1]) < 0)
218 if (le64_to_cpu(nilfs->ns_sbp[1]->s_last_cno) <
219 le64_to_cpu(nilfs->ns_sbp[0]->s_last_cno))
220 sbp = nilfs->ns_sbp[1];
223 spin_lock(&nilfs->ns_last_segment_lock);
224 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
225 spin_unlock(&nilfs->ns_last_segment_lock);
231 void nilfs_set_log_cursor(struct nilfs_super_block *sbp,
232 struct the_nilfs *nilfs)
234 sector_t nfreeblocks;
236 /* nilfs->ns_sem must be locked by the caller. */
237 nilfs_count_free_blocks(nilfs, &nfreeblocks);
238 sbp->s_free_blocks_count = cpu_to_le64(nfreeblocks);
240 spin_lock(&nilfs->ns_last_segment_lock);
241 sbp->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
242 sbp->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
243 sbp->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
244 spin_unlock(&nilfs->ns_last_segment_lock);
247 struct nilfs_super_block **nilfs_prepare_super(struct nilfs_sb_info *sbi,
250 struct the_nilfs *nilfs = sbi->s_nilfs;
251 struct nilfs_super_block **sbp = nilfs->ns_sbp;
253 /* nilfs->ns_sem must be locked by the caller. */
254 if (sbp[0]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) {
256 sbp[1]->s_magic == cpu_to_le16(NILFS_SUPER_MAGIC)) {
257 memcpy(sbp[0], sbp[1], nilfs->ns_sbsize);
259 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
264 sbp[1]->s_magic != cpu_to_le16(NILFS_SUPER_MAGIC)) {
265 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
269 nilfs_swap_super_block(nilfs);
274 int nilfs_commit_super(struct nilfs_sb_info *sbi, int flag)
276 struct the_nilfs *nilfs = sbi->s_nilfs;
277 struct nilfs_super_block **sbp = nilfs->ns_sbp;
280 /* nilfs->ns_sem must be locked by the caller. */
282 nilfs->ns_sbwtime = t;
283 sbp[0]->s_wtime = cpu_to_le64(t);
285 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
286 (unsigned char *)sbp[0],
288 if (flag == NILFS_SB_COMMIT_ALL && sbp[1]) {
289 sbp[1]->s_wtime = sbp[0]->s_wtime;
291 sbp[1]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
292 (unsigned char *)sbp[1],
295 clear_nilfs_sb_dirty(nilfs);
296 return nilfs_sync_super(sbi, flag);
300 * nilfs_cleanup_super() - write filesystem state for cleanup
301 * @sbi: nilfs_sb_info to be unmounted or degraded to read-only
303 * This function restores state flags in the on-disk super block.
304 * This will set "clean" flag (i.e. NILFS_VALID_FS) unless the
305 * filesystem was not clean previously.
307 int nilfs_cleanup_super(struct nilfs_sb_info *sbi)
309 struct nilfs_super_block **sbp;
310 int flag = NILFS_SB_COMMIT;
313 sbp = nilfs_prepare_super(sbi, 0);
315 sbp[0]->s_state = cpu_to_le16(sbi->s_nilfs->ns_mount_state);
316 nilfs_set_log_cursor(sbp[0], sbi->s_nilfs);
317 if (sbp[1] && sbp[0]->s_last_cno == sbp[1]->s_last_cno) {
319 * make the "clean" flag also to the opposite
320 * super block if both super blocks point to
321 * the same checkpoint.
323 sbp[1]->s_state = sbp[0]->s_state;
324 flag = NILFS_SB_COMMIT_ALL;
326 ret = nilfs_commit_super(sbi, flag);
331 static void nilfs_put_super(struct super_block *sb)
333 struct nilfs_sb_info *sbi = NILFS_SB(sb);
334 struct the_nilfs *nilfs = sbi->s_nilfs;
336 nilfs_detach_segment_constructor(sbi);
338 if (!(sb->s_flags & MS_RDONLY)) {
339 down_write(&nilfs->ns_sem);
340 nilfs_cleanup_super(sbi);
341 up_write(&nilfs->ns_sem);
343 down_write(&nilfs->ns_super_sem);
344 if (nilfs->ns_current == sbi)
345 nilfs->ns_current = NULL;
346 up_write(&nilfs->ns_super_sem);
348 nilfs_detach_checkpoint(sbi);
349 put_nilfs(sbi->s_nilfs);
351 sb->s_fs_info = NULL;
352 nilfs_put_sbinfo(sbi);
355 static int nilfs_sync_fs(struct super_block *sb, int wait)
357 struct nilfs_sb_info *sbi = NILFS_SB(sb);
358 struct the_nilfs *nilfs = sbi->s_nilfs;
359 struct nilfs_super_block **sbp;
362 /* This function is called when super block should be written back */
364 err = nilfs_construct_segment(sb);
366 down_write(&nilfs->ns_sem);
367 if (nilfs_sb_dirty(nilfs)) {
368 sbp = nilfs_prepare_super(sbi, nilfs_sb_will_flip(nilfs));
370 nilfs_set_log_cursor(sbp[0], nilfs);
371 nilfs_commit_super(sbi, NILFS_SB_COMMIT);
374 up_write(&nilfs->ns_sem);
379 int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
381 struct the_nilfs *nilfs = sbi->s_nilfs;
382 struct nilfs_checkpoint *raw_cp;
383 struct buffer_head *bh_cp;
386 down_write(&nilfs->ns_super_sem);
387 list_add(&sbi->s_list, &nilfs->ns_supers);
388 up_write(&nilfs->ns_super_sem);
391 sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size);
395 down_read(&nilfs->ns_segctor_sem);
396 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
398 up_read(&nilfs->ns_segctor_sem);
400 if (err == -ENOENT || err == -EINVAL) {
402 "NILFS: Invalid checkpoint "
403 "(checkpoint number=%llu)\n",
404 (unsigned long long)cno);
409 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
412 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
413 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
415 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
419 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
421 nilfs_mdt_destroy(sbi->s_ifile);
425 down_write(&nilfs->ns_super_sem);
426 list_del_init(&sbi->s_list);
427 up_write(&nilfs->ns_super_sem);
432 void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
434 struct the_nilfs *nilfs = sbi->s_nilfs;
436 nilfs_mdt_destroy(sbi->s_ifile);
438 down_write(&nilfs->ns_super_sem);
439 list_del_init(&sbi->s_list);
440 up_write(&nilfs->ns_super_sem);
443 static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
445 struct super_block *sb = dentry->d_sb;
446 struct nilfs_sb_info *sbi = NILFS_SB(sb);
447 struct the_nilfs *nilfs = sbi->s_nilfs;
448 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
449 unsigned long long blocks;
450 unsigned long overhead;
451 unsigned long nrsvblocks;
452 sector_t nfreeblocks;
456 * Compute all of the segment blocks
458 * The blocks before first segment and after last segment
461 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
462 - nilfs->ns_first_data_block;
463 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
466 * Compute the overhead
468 * When distributing meta data blocks outside segment structure,
469 * We must count them as the overhead.
473 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
477 buf->f_type = NILFS_SUPER_MAGIC;
478 buf->f_bsize = sb->s_blocksize;
479 buf->f_blocks = blocks - overhead;
480 buf->f_bfree = nfreeblocks;
481 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
482 (buf->f_bfree - nrsvblocks) : 0;
483 buf->f_files = atomic_read(&sbi->s_inodes_count);
484 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
485 buf->f_namelen = NILFS_NAME_LEN;
486 buf->f_fsid.val[0] = (u32)id;
487 buf->f_fsid.val[1] = (u32)(id >> 32);
492 static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
494 struct super_block *sb = vfs->mnt_sb;
495 struct nilfs_sb_info *sbi = NILFS_SB(sb);
497 if (!nilfs_test_opt(sbi, BARRIER))
498 seq_puts(seq, ",nobarrier");
499 if (nilfs_test_opt(sbi, SNAPSHOT))
500 seq_printf(seq, ",cp=%llu",
501 (unsigned long long int)sbi->s_snapshot_cno);
502 if (nilfs_test_opt(sbi, ERRORS_PANIC))
503 seq_puts(seq, ",errors=panic");
504 if (nilfs_test_opt(sbi, ERRORS_CONT))
505 seq_puts(seq, ",errors=continue");
506 if (nilfs_test_opt(sbi, STRICT_ORDER))
507 seq_puts(seq, ",order=strict");
508 if (nilfs_test_opt(sbi, NORECOVERY))
509 seq_puts(seq, ",norecovery");
510 if (nilfs_test_opt(sbi, DISCARD))
511 seq_puts(seq, ",discard");
516 static const struct super_operations nilfs_sops = {
517 .alloc_inode = nilfs_alloc_inode,
518 .destroy_inode = nilfs_destroy_inode,
519 .dirty_inode = nilfs_dirty_inode,
520 /* .write_inode = nilfs_write_inode, */
521 /* .put_inode = nilfs_put_inode, */
522 /* .drop_inode = nilfs_drop_inode, */
523 .evict_inode = nilfs_evict_inode,
524 .put_super = nilfs_put_super,
525 /* .write_super = nilfs_write_super, */
526 .sync_fs = nilfs_sync_fs,
527 /* .write_super_lockfs */
529 .statfs = nilfs_statfs,
530 .remount_fs = nilfs_remount,
532 .show_options = nilfs_show_options
535 static struct inode *
536 nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
540 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
541 ino != NILFS_SKETCH_INO)
542 return ERR_PTR(-ESTALE);
544 inode = nilfs_iget(sb, ino);
546 return ERR_CAST(inode);
547 if (generation && inode->i_generation != generation) {
549 return ERR_PTR(-ESTALE);
555 static struct dentry *
556 nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
559 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
560 nilfs_nfs_get_inode);
563 static struct dentry *
564 nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
567 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
568 nilfs_nfs_get_inode);
571 static const struct export_operations nilfs_export_ops = {
572 .fh_to_dentry = nilfs_fh_to_dentry,
573 .fh_to_parent = nilfs_fh_to_parent,
574 .get_parent = nilfs_get_parent,
578 Opt_err_cont, Opt_err_panic, Opt_err_ro,
579 Opt_barrier, Opt_nobarrier, Opt_snapshot, Opt_order, Opt_norecovery,
580 Opt_discard, Opt_nodiscard, Opt_err,
583 static match_table_t tokens = {
584 {Opt_err_cont, "errors=continue"},
585 {Opt_err_panic, "errors=panic"},
586 {Opt_err_ro, "errors=remount-ro"},
587 {Opt_barrier, "barrier"},
588 {Opt_nobarrier, "nobarrier"},
589 {Opt_snapshot, "cp=%u"},
590 {Opt_order, "order=%s"},
591 {Opt_norecovery, "norecovery"},
592 {Opt_discard, "discard"},
593 {Opt_nodiscard, "nodiscard"},
597 static int parse_options(char *options, struct super_block *sb, int is_remount)
599 struct nilfs_sb_info *sbi = NILFS_SB(sb);
601 substring_t args[MAX_OPT_ARGS];
607 while ((p = strsep(&options, ",")) != NULL) {
612 token = match_token(p, tokens, args);
615 nilfs_set_opt(sbi, BARRIER);
618 nilfs_clear_opt(sbi, BARRIER);
621 if (strcmp(args[0].from, "relaxed") == 0)
622 /* Ordered data semantics */
623 nilfs_clear_opt(sbi, STRICT_ORDER);
624 else if (strcmp(args[0].from, "strict") == 0)
625 /* Strict in-order semantics */
626 nilfs_set_opt(sbi, STRICT_ORDER);
631 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
634 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
637 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
640 if (match_int(&args[0], &option) || option <= 0)
643 if (!nilfs_test_opt(sbi, SNAPSHOT)) {
645 "NILFS: cannot change regular "
646 "mount to snapshot.\n");
648 } else if (option != sbi->s_snapshot_cno) {
650 "NILFS: cannot remount to a "
651 "different snapshot.\n");
656 if (!(sb->s_flags & MS_RDONLY)) {
657 printk(KERN_ERR "NILFS: cannot mount snapshot "
658 "read/write. A read-only option is "
662 sbi->s_snapshot_cno = option;
663 nilfs_set_opt(sbi, SNAPSHOT);
666 nilfs_set_opt(sbi, NORECOVERY);
669 nilfs_set_opt(sbi, DISCARD);
672 nilfs_clear_opt(sbi, DISCARD);
676 "NILFS: Unrecognized mount option \"%s\"\n", p);
684 nilfs_set_default_options(struct nilfs_sb_info *sbi,
685 struct nilfs_super_block *sbp)
688 NILFS_MOUNT_ERRORS_RO | NILFS_MOUNT_BARRIER;
691 static int nilfs_setup_super(struct nilfs_sb_info *sbi)
693 struct the_nilfs *nilfs = sbi->s_nilfs;
694 struct nilfs_super_block **sbp;
698 /* nilfs->ns_sem must be locked by the caller. */
699 sbp = nilfs_prepare_super(sbi, 0);
703 max_mnt_count = le16_to_cpu(sbp[0]->s_max_mnt_count);
704 mnt_count = le16_to_cpu(sbp[0]->s_mnt_count);
706 if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
708 "NILFS warning: mounting fs with errors\n");
710 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
712 "NILFS warning: maximal mount count reached\n");
716 sbp[0]->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
718 sbp[0]->s_mnt_count = cpu_to_le16(mnt_count + 1);
720 cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS);
721 sbp[0]->s_mtime = cpu_to_le64(get_seconds());
722 /* synchronize sbp[1] with sbp[0] */
723 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
724 return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL);
727 struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
728 u64 pos, int blocksize,
729 struct buffer_head **pbh)
731 unsigned long long sb_index = pos;
732 unsigned long offset;
734 offset = do_div(sb_index, blocksize);
735 *pbh = sb_bread(sb, sb_index);
738 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
741 int nilfs_store_magic_and_option(struct super_block *sb,
742 struct nilfs_super_block *sbp,
745 struct nilfs_sb_info *sbi = NILFS_SB(sb);
747 sb->s_magic = le16_to_cpu(sbp->s_magic);
749 /* FS independent flags */
750 #ifdef NILFS_ATIME_DISABLE
751 sb->s_flags |= MS_NOATIME;
754 nilfs_set_default_options(sbi, sbp);
756 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
757 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
758 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
759 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
761 return !parse_options(data, sb, 0) ? -EINVAL : 0 ;
764 int nilfs_check_feature_compatibility(struct super_block *sb,
765 struct nilfs_super_block *sbp)
769 features = le64_to_cpu(sbp->s_feature_incompat) &
770 ~NILFS_FEATURE_INCOMPAT_SUPP;
772 printk(KERN_ERR "NILFS: couldn't mount because of unsupported "
773 "optional features (%llx)\n",
774 (unsigned long long)features);
777 features = le64_to_cpu(sbp->s_feature_compat_ro) &
778 ~NILFS_FEATURE_COMPAT_RO_SUPP;
779 if (!(sb->s_flags & MS_RDONLY) && features) {
780 printk(KERN_ERR "NILFS: couldn't mount RDWR because of "
781 "unsupported optional features (%llx)\n",
782 (unsigned long long)features);
789 * nilfs_fill_super() - initialize a super block instance
791 * @data: mount options
792 * @silent: silent mode flag
793 * @nilfs: the_nilfs struct
795 * This function is called exclusively by nilfs->ns_mount_mutex.
796 * So, the recovery process is protected from other simultaneous mounts.
799 nilfs_fill_super(struct super_block *sb, void *data, int silent,
800 struct the_nilfs *nilfs)
802 struct nilfs_sb_info *sbi;
807 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
814 sbi->s_nilfs = nilfs;
816 atomic_set(&sbi->s_count, 1);
818 err = init_nilfs(nilfs, sbi, (char *)data);
822 spin_lock_init(&sbi->s_inode_lock);
823 INIT_LIST_HEAD(&sbi->s_dirty_files);
824 INIT_LIST_HEAD(&sbi->s_list);
827 * Following initialization is overlapped because
828 * nilfs_sb_info structure has been cleared at the beginning.
829 * But we reserve them to keep our interest and make ready
830 * for the future change.
832 get_random_bytes(&sbi->s_next_generation,
833 sizeof(sbi->s_next_generation));
834 spin_lock_init(&sbi->s_next_gen_lock);
836 sb->s_op = &nilfs_sops;
837 sb->s_export_op = &nilfs_export_ops;
840 sb->s_bdi = nilfs->ns_bdi;
842 err = load_nilfs(nilfs, sbi);
846 cno = nilfs_last_cno(nilfs);
848 if (sb->s_flags & MS_RDONLY) {
849 if (nilfs_test_opt(sbi, SNAPSHOT)) {
850 down_read(&nilfs->ns_segctor_sem);
851 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
852 sbi->s_snapshot_cno);
853 up_read(&nilfs->ns_segctor_sem);
861 "NILFS: The specified checkpoint is "
863 "(checkpoint number=%llu).\n",
864 (unsigned long long)sbi->s_snapshot_cno);
868 cno = sbi->s_snapshot_cno;
872 err = nilfs_attach_checkpoint(sbi, cno);
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);
882 goto failed_checkpoint;
885 root = nilfs_iget(sb, NILFS_ROOT_INO);
887 printk(KERN_ERR "NILFS: get root inode failed\n");
891 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
893 printk(KERN_ERR "NILFS: corrupt root inode.\n");
897 sb->s_root = d_alloc_root(root);
900 printk(KERN_ERR "NILFS: get root dentry failed\n");
905 if (!(sb->s_flags & MS_RDONLY)) {
906 down_write(&nilfs->ns_sem);
907 nilfs_setup_super(sbi);
908 up_write(&nilfs->ns_sem);
911 down_write(&nilfs->ns_super_sem);
912 if (!nilfs_test_opt(sbi, SNAPSHOT))
913 nilfs->ns_current = sbi;
914 up_write(&nilfs->ns_super_sem);
919 nilfs_detach_segment_constructor(sbi);
922 nilfs_detach_checkpoint(sbi);
926 sb->s_fs_info = NULL;
927 nilfs_put_sbinfo(sbi);
931 static int nilfs_remount(struct super_block *sb, int *flags, char *data)
933 struct nilfs_sb_info *sbi = NILFS_SB(sb);
934 struct the_nilfs *nilfs = sbi->s_nilfs;
935 unsigned long old_sb_flags;
936 struct nilfs_mount_options old_opts;
937 int was_snapshot, err;
939 down_write(&nilfs->ns_super_sem);
940 old_sb_flags = sb->s_flags;
941 old_opts.mount_opt = sbi->s_mount_opt;
942 old_opts.snapshot_cno = sbi->s_snapshot_cno;
943 was_snapshot = nilfs_test_opt(sbi, SNAPSHOT);
945 if (!parse_options(data, sb, 1)) {
949 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
952 if (was_snapshot && !(*flags & MS_RDONLY)) {
953 printk(KERN_ERR "NILFS (device %s): cannot remount snapshot "
954 "read/write.\n", sb->s_id);
958 if (!nilfs_valid_fs(nilfs)) {
959 printk(KERN_WARNING "NILFS (device %s): couldn't "
960 "remount because the filesystem is in an "
961 "incomplete recovery state.\n", sb->s_id);
965 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
967 if (*flags & MS_RDONLY) {
968 /* Shutting down the segment constructor */
969 nilfs_detach_segment_constructor(sbi);
970 sb->s_flags |= MS_RDONLY;
973 * Remounting a valid RW partition RDONLY, so set
974 * the RDONLY flag and then mark the partition as valid again.
976 down_write(&nilfs->ns_sem);
977 nilfs_cleanup_super(sbi);
978 up_write(&nilfs->ns_sem);
983 * Mounting a RDONLY partition read-write, so reread and
984 * store the current valid flag. (It may have been changed
985 * by fsck since we originally mounted the partition.)
987 down_read(&nilfs->ns_sem);
988 features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) &
989 ~NILFS_FEATURE_COMPAT_RO_SUPP;
990 up_read(&nilfs->ns_sem);
992 printk(KERN_WARNING "NILFS (device %s): couldn't "
993 "remount RDWR because of unsupported optional "
995 sb->s_id, (unsigned long long)features);
1000 sb->s_flags &= ~MS_RDONLY;
1002 err = nilfs_attach_segment_constructor(sbi);
1006 down_write(&nilfs->ns_sem);
1007 nilfs_setup_super(sbi);
1008 up_write(&nilfs->ns_sem);
1011 up_write(&nilfs->ns_super_sem);
1015 sb->s_flags = old_sb_flags;
1016 sbi->s_mount_opt = old_opts.mount_opt;
1017 sbi->s_snapshot_cno = old_opts.snapshot_cno;
1018 up_write(&nilfs->ns_super_sem);
1022 struct nilfs_super_data {
1023 struct block_device *bdev;
1024 struct nilfs_sb_info *sbi;
1030 * nilfs_identify - pre-read mount options needed to identify mount instance
1031 * @data: mount options
1032 * @sd: nilfs_super_data
1034 static int nilfs_identify(char *data, struct nilfs_super_data *sd)
1036 char *p, *options = data;
1037 substring_t args[MAX_OPT_ARGS];
1042 p = strsep(&options, ",");
1043 if (p != NULL && *p) {
1044 token = match_token(p, tokens, args);
1045 if (token == Opt_snapshot) {
1046 if (!(sd->flags & MS_RDONLY))
1049 ret = match_int(&args[0], &option);
1060 "NILFS: invalid mount option: %s\n", p);
1064 BUG_ON(options == data);
1065 *(options - 1) = ',';
1070 static int nilfs_set_bdev_super(struct super_block *s, void *data)
1072 struct nilfs_super_data *sd = data;
1074 s->s_bdev = sd->bdev;
1075 s->s_dev = s->s_bdev->bd_dev;
1079 static int nilfs_test_bdev_super(struct super_block *s, void *data)
1081 struct nilfs_super_data *sd = data;
1083 return sd->sbi && s->s_fs_info == (void *)sd->sbi;
1087 nilfs_get_sb(struct file_system_type *fs_type, int flags,
1088 const char *dev_name, void *data, struct vfsmount *mnt)
1090 struct nilfs_super_data sd;
1091 struct super_block *s;
1092 fmode_t mode = FMODE_READ;
1093 struct the_nilfs *nilfs;
1094 int err, need_to_close = 1;
1096 if (!(flags & MS_RDONLY))
1097 mode |= FMODE_WRITE;
1099 sd.bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1100 if (IS_ERR(sd.bdev))
1101 return PTR_ERR(sd.bdev);
1104 * To get mount instance using sget() vfs-routine, NILFS needs
1105 * much more information than normal filesystems to identify mount
1106 * instance. For snapshot mounts, not only a mount type (ro-mount
1107 * or rw-mount) but also a checkpoint number is required.
1111 if (nilfs_identify((char *)data, &sd)) {
1116 nilfs = find_or_create_nilfs(sd.bdev);
1122 mutex_lock(&nilfs->ns_mount_mutex);
1126 * Check if an exclusive mount exists or not.
1127 * Snapshot mounts coexist with a current mount
1128 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1129 * ro-mount are mutually exclusive.
1131 down_read(&nilfs->ns_super_sem);
1132 if (nilfs->ns_current &&
1133 ((nilfs->ns_current->s_super->s_flags ^ flags)
1135 up_read(&nilfs->ns_super_sem);
1139 up_read(&nilfs->ns_super_sem);
1143 * Find existing nilfs_sb_info struct
1145 sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
1148 * Get super block instance holding the nilfs_sb_info struct.
1149 * A new instance is allocated if no existing mount is present or
1150 * existing instance has been unmounted.
1152 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
1154 nilfs_put_sbinfo(sd.sbi);
1162 char b[BDEVNAME_SIZE];
1164 /* New superblock instance created */
1167 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1168 sb_set_blocksize(s, block_size(sd.bdev));
1170 err = nilfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0,
1175 s->s_flags |= MS_ACTIVE;
1179 mutex_unlock(&nilfs->ns_mount_mutex);
1182 close_bdev_exclusive(sd.bdev, mode);
1183 simple_set_mnt(mnt, s);
1187 mutex_unlock(&nilfs->ns_mount_mutex);
1190 close_bdev_exclusive(sd.bdev, mode);
1194 /* Abandoning the newly allocated superblock */
1195 mutex_unlock(&nilfs->ns_mount_mutex);
1197 deactivate_locked_super(s);
1199 * deactivate_locked_super() invokes close_bdev_exclusive().
1200 * We must finish all post-cleaning before this call;
1201 * put_nilfs() needs the block device.
1206 struct file_system_type nilfs_fs_type = {
1207 .owner = THIS_MODULE,
1209 .get_sb = nilfs_get_sb,
1210 .kill_sb = kill_block_super,
1211 .fs_flags = FS_REQUIRES_DEV,
1214 static void nilfs_inode_init_once(void *obj)
1216 struct nilfs_inode_info *ii = obj;
1218 INIT_LIST_HEAD(&ii->i_dirty);
1219 #ifdef CONFIG_NILFS_XATTR
1220 init_rwsem(&ii->xattr_sem);
1222 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
1223 ii->i_bmap = &ii->i_bmap_data;
1224 inode_init_once(&ii->vfs_inode);
1227 static void nilfs_segbuf_init_once(void *obj)
1229 memset(obj, 0, sizeof(struct nilfs_segment_buffer));
1232 static void nilfs_destroy_cachep(void)
1234 if (nilfs_inode_cachep)
1235 kmem_cache_destroy(nilfs_inode_cachep);
1236 if (nilfs_transaction_cachep)
1237 kmem_cache_destroy(nilfs_transaction_cachep);
1238 if (nilfs_segbuf_cachep)
1239 kmem_cache_destroy(nilfs_segbuf_cachep);
1240 if (nilfs_btree_path_cache)
1241 kmem_cache_destroy(nilfs_btree_path_cache);
1244 static int __init nilfs_init_cachep(void)
1246 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
1247 sizeof(struct nilfs_inode_info), 0,
1248 SLAB_RECLAIM_ACCOUNT, nilfs_inode_init_once);
1249 if (!nilfs_inode_cachep)
1252 nilfs_transaction_cachep = kmem_cache_create("nilfs2_transaction_cache",
1253 sizeof(struct nilfs_transaction_info), 0,
1254 SLAB_RECLAIM_ACCOUNT, NULL);
1255 if (!nilfs_transaction_cachep)
1258 nilfs_segbuf_cachep = kmem_cache_create("nilfs2_segbuf_cache",
1259 sizeof(struct nilfs_segment_buffer), 0,
1260 SLAB_RECLAIM_ACCOUNT, nilfs_segbuf_init_once);
1261 if (!nilfs_segbuf_cachep)
1264 nilfs_btree_path_cache = kmem_cache_create("nilfs2_btree_path_cache",
1265 sizeof(struct nilfs_btree_path) * NILFS_BTREE_LEVEL_MAX,
1267 if (!nilfs_btree_path_cache)
1273 nilfs_destroy_cachep();
1277 static int __init init_nilfs_fs(void)
1281 err = nilfs_init_cachep();
1285 err = register_filesystem(&nilfs_fs_type);
1289 printk(KERN_INFO "NILFS version 2 loaded\n");
1293 nilfs_destroy_cachep();
1298 static void __exit exit_nilfs_fs(void)
1300 nilfs_destroy_cachep();
1301 unregister_filesystem(&nilfs_fs_type);
1304 module_init(init_nilfs_fs)
1305 module_exit(exit_nilfs_fs)