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 void nilfs_write_super(struct super_block *sb);
71 static int nilfs_remount(struct super_block *sb, int *flags, char *data);
74 * nilfs_error() - report failure condition on a filesystem
76 * nilfs_error() sets an ERROR_FS flag on the superblock as well as
77 * reporting an error message. It should be called when NILFS detects
78 * incoherences or defects of meta data on disk. As for sustainable
79 * errors such as a single-shot I/O error, nilfs_warning() or the printk()
80 * function should be used instead.
82 * The segment constructor must not call this function because it can
85 void nilfs_error(struct super_block *sb, const char *function,
88 struct nilfs_sb_info *sbi = NILFS_SB(sb);
92 printk(KERN_CRIT "NILFS error (device %s): %s: ", sb->s_id, function);
97 if (!(sb->s_flags & MS_RDONLY)) {
98 struct the_nilfs *nilfs = sbi->s_nilfs;
100 if (!nilfs_test_opt(sbi, ERRORS_CONT))
101 nilfs_detach_segment_constructor(sbi);
103 down_write(&nilfs->ns_sem);
104 if (!(nilfs->ns_mount_state & NILFS_ERROR_FS)) {
105 nilfs->ns_mount_state |= NILFS_ERROR_FS;
106 nilfs->ns_sbp[0]->s_state |=
107 cpu_to_le16(NILFS_ERROR_FS);
108 nilfs_commit_super(sbi, 1);
110 up_write(&nilfs->ns_sem);
112 if (nilfs_test_opt(sbi, ERRORS_RO)) {
113 printk(KERN_CRIT "Remounting filesystem read-only\n");
114 sb->s_flags |= MS_RDONLY;
118 if (nilfs_test_opt(sbi, ERRORS_PANIC))
119 panic("NILFS (device %s): panic forced after error\n",
123 void nilfs_warning(struct super_block *sb, const char *function,
124 const char *fmt, ...)
129 printk(KERN_WARNING "NILFS warning (device %s): %s: ",
136 static struct kmem_cache *nilfs_inode_cachep;
138 struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs)
140 struct nilfs_inode_info *ii;
142 ii = kmem_cache_alloc(nilfs_inode_cachep, GFP_NOFS);
147 ii->vfs_inode.i_version = 1;
148 nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi);
149 return &ii->vfs_inode;
152 struct inode *nilfs_alloc_inode(struct super_block *sb)
154 return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
157 void nilfs_destroy_inode(struct inode *inode)
159 kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
162 static void init_once(void *obj)
164 struct nilfs_inode_info *ii = obj;
166 INIT_LIST_HEAD(&ii->i_dirty);
167 #ifdef CONFIG_NILFS_XATTR
168 init_rwsem(&ii->xattr_sem);
170 nilfs_btnode_cache_init_once(&ii->i_btnode_cache);
171 ii->i_bmap = (struct nilfs_bmap *)&ii->i_bmap_union;
172 inode_init_once(&ii->vfs_inode);
175 static int nilfs_init_inode_cache(void)
177 nilfs_inode_cachep = kmem_cache_create("nilfs2_inode_cache",
178 sizeof(struct nilfs_inode_info),
179 0, SLAB_RECLAIM_ACCOUNT,
182 return (nilfs_inode_cachep == NULL) ? -ENOMEM : 0;
185 static inline void nilfs_destroy_inode_cache(void)
187 kmem_cache_destroy(nilfs_inode_cachep);
190 static void nilfs_clear_inode(struct inode *inode)
192 struct nilfs_inode_info *ii = NILFS_I(inode);
195 * Free resources allocated in nilfs_read_inode(), here.
197 BUG_ON(!list_empty(&ii->i_dirty));
201 if (test_bit(NILFS_I_BMAP, &ii->i_state))
202 nilfs_bmap_clear(ii->i_bmap);
204 nilfs_btnode_cache_clear(&ii->i_btnode_cache);
207 static int nilfs_sync_super(struct nilfs_sb_info *sbi, int dupsb)
209 struct the_nilfs *nilfs = sbi->s_nilfs;
211 int barrier_done = 0;
213 if (nilfs_test_opt(sbi, BARRIER)) {
214 set_buffer_ordered(nilfs->ns_sbh[0]);
218 set_buffer_dirty(nilfs->ns_sbh[0]);
219 err = sync_dirty_buffer(nilfs->ns_sbh[0]);
220 if (err == -EOPNOTSUPP && barrier_done) {
221 nilfs_warning(sbi->s_super, __func__,
222 "barrier-based sync failed. "
223 "disabling barriers\n");
224 nilfs_clear_opt(sbi, BARRIER);
226 clear_buffer_ordered(nilfs->ns_sbh[0]);
231 "NILFS: unable to write superblock (err=%d)\n", err);
232 if (err == -EIO && nilfs->ns_sbh[1]) {
233 nilfs_fall_back_super_block(nilfs);
237 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
240 * The latest segment becomes trailable from the position
241 * written in superblock.
243 clear_nilfs_discontinued(nilfs);
245 /* update GC protection for recent segments */
246 if (nilfs->ns_sbh[1]) {
249 set_buffer_dirty(nilfs->ns_sbh[1]);
250 if (!sync_dirty_buffer(nilfs->ns_sbh[1]))
251 sbp = nilfs->ns_sbp[1];
255 spin_lock(&nilfs->ns_last_segment_lock);
256 nilfs->ns_prot_seq = le64_to_cpu(sbp->s_last_seq);
257 spin_unlock(&nilfs->ns_last_segment_lock);
264 int nilfs_commit_super(struct nilfs_sb_info *sbi, int dupsb)
266 struct the_nilfs *nilfs = sbi->s_nilfs;
267 struct nilfs_super_block **sbp = nilfs->ns_sbp;
268 sector_t nfreeblocks;
272 /* nilfs->sem must be locked by the caller. */
273 if (sbp[0]->s_magic != NILFS_SUPER_MAGIC) {
274 if (sbp[1] && sbp[1]->s_magic == NILFS_SUPER_MAGIC)
275 nilfs_swap_super_block(nilfs);
277 printk(KERN_CRIT "NILFS: superblock broke on dev %s\n",
282 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
284 printk(KERN_ERR "NILFS: failed to count free blocks\n");
287 spin_lock(&nilfs->ns_last_segment_lock);
288 sbp[0]->s_last_seq = cpu_to_le64(nilfs->ns_last_seq);
289 sbp[0]->s_last_pseg = cpu_to_le64(nilfs->ns_last_pseg);
290 sbp[0]->s_last_cno = cpu_to_le64(nilfs->ns_last_cno);
291 spin_unlock(&nilfs->ns_last_segment_lock);
294 nilfs->ns_sbwtime[0] = t;
295 sbp[0]->s_free_blocks_count = cpu_to_le64(nfreeblocks);
296 sbp[0]->s_wtime = cpu_to_le64(t);
298 sbp[0]->s_sum = cpu_to_le32(crc32_le(nilfs->ns_crc_seed,
299 (unsigned char *)sbp[0],
301 if (dupsb && sbp[1]) {
302 memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
303 nilfs->ns_sbwtime[1] = t;
305 sbi->s_super->s_dirt = 0;
306 return nilfs_sync_super(sbi, dupsb);
309 static void nilfs_put_super(struct super_block *sb)
311 struct nilfs_sb_info *sbi = NILFS_SB(sb);
312 struct the_nilfs *nilfs = sbi->s_nilfs;
317 nilfs_write_super(sb);
319 nilfs_detach_segment_constructor(sbi);
321 if (!(sb->s_flags & MS_RDONLY)) {
322 down_write(&nilfs->ns_sem);
323 nilfs->ns_sbp[0]->s_state = cpu_to_le16(nilfs->ns_mount_state);
324 nilfs_commit_super(sbi, 1);
325 up_write(&nilfs->ns_sem);
327 down_write(&nilfs->ns_super_sem);
328 if (nilfs->ns_current == sbi)
329 nilfs->ns_current = NULL;
330 up_write(&nilfs->ns_super_sem);
332 nilfs_detach_checkpoint(sbi);
333 put_nilfs(sbi->s_nilfs);
335 sb->s_fs_info = NULL;
336 nilfs_put_sbinfo(sbi);
342 * nilfs_write_super - write super block(s) of NILFS
345 * nilfs_write_super() gets a fs-dependent lock, writes super block(s), and
346 * clears s_dirt. This function is called in the section protected by
349 * The s_dirt flag is managed by each filesystem and we protect it by ns_sem
350 * of the struct the_nilfs. Lock order must be as follows:
353 * 2. down_write(&nilfs->ns_sem)
355 * Inside NILFS, locking ns_sem is enough to protect s_dirt and the buffer
356 * of the super block (nilfs->ns_sbp[]).
358 * In most cases, VFS functions call lock_super() before calling these
359 * methods. So we must be careful not to bring on deadlocks when using
360 * lock_super(); see generic_shutdown_super(), write_super(), and so on.
362 * Note that order of lock_kernel() and lock_super() depends on contexts
363 * of VFS. We should also note that lock_kernel() can be used in its
364 * protective section and only the outermost one has an effect.
366 static void nilfs_write_super(struct super_block *sb)
368 struct nilfs_sb_info *sbi = NILFS_SB(sb);
369 struct the_nilfs *nilfs = sbi->s_nilfs;
371 down_write(&nilfs->ns_sem);
372 if (!(sb->s_flags & MS_RDONLY)) {
373 struct nilfs_super_block **sbp = nilfs->ns_sbp;
374 u64 t = get_seconds();
377 if (!nilfs_discontinued(nilfs) && t >= nilfs->ns_sbwtime[0] &&
378 t < nilfs->ns_sbwtime[0] + NILFS_SB_FREQ) {
379 up_write(&nilfs->ns_sem);
382 dupsb = sbp[1] && t > nilfs->ns_sbwtime[1] + NILFS_ALTSB_FREQ;
383 nilfs_commit_super(sbi, dupsb);
386 up_write(&nilfs->ns_sem);
389 static int nilfs_sync_fs(struct super_block *sb, int wait)
393 nilfs_write_super(sb);
395 /* This function is called when super block should be written back */
397 err = nilfs_construct_segment(sb);
401 int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
403 struct the_nilfs *nilfs = sbi->s_nilfs;
404 struct nilfs_checkpoint *raw_cp;
405 struct buffer_head *bh_cp;
408 down_write(&nilfs->ns_super_sem);
409 list_add(&sbi->s_list, &nilfs->ns_supers);
410 up_write(&nilfs->ns_super_sem);
412 sbi->s_ifile = nilfs_mdt_new(
413 nilfs, sbi->s_super, NILFS_IFILE_INO, NILFS_IFILE_GFP);
417 err = nilfs_palloc_init_blockgroup(sbi->s_ifile, nilfs->ns_inode_size);
421 down_read(&nilfs->ns_segctor_sem);
422 err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
424 up_read(&nilfs->ns_segctor_sem);
426 if (err == -ENOENT || err == -EINVAL) {
428 "NILFS: Invalid checkpoint "
429 "(checkpoint number=%llu)\n",
430 (unsigned long long)cno);
435 err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
438 atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
439 atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
441 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
445 nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
447 nilfs_mdt_destroy(sbi->s_ifile);
450 down_write(&nilfs->ns_super_sem);
451 list_del_init(&sbi->s_list);
452 up_write(&nilfs->ns_super_sem);
457 void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
459 struct the_nilfs *nilfs = sbi->s_nilfs;
461 nilfs_mdt_clear(sbi->s_ifile);
462 nilfs_mdt_destroy(sbi->s_ifile);
464 down_write(&nilfs->ns_super_sem);
465 list_del_init(&sbi->s_list);
466 up_write(&nilfs->ns_super_sem);
469 static int nilfs_mark_recovery_complete(struct nilfs_sb_info *sbi)
471 struct the_nilfs *nilfs = sbi->s_nilfs;
474 down_write(&nilfs->ns_sem);
475 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
476 nilfs->ns_mount_state |= NILFS_VALID_FS;
477 err = nilfs_commit_super(sbi, 1);
479 printk(KERN_INFO "NILFS: recovery complete.\n");
481 up_write(&nilfs->ns_sem);
485 static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
487 struct super_block *sb = dentry->d_sb;
488 struct nilfs_sb_info *sbi = NILFS_SB(sb);
489 struct the_nilfs *nilfs = sbi->s_nilfs;
490 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
491 unsigned long long blocks;
492 unsigned long overhead;
493 unsigned long nrsvblocks;
494 sector_t nfreeblocks;
498 * Compute all of the segment blocks
500 * The blocks before first segment and after last segment
503 blocks = nilfs->ns_blocks_per_segment * nilfs->ns_nsegments
504 - nilfs->ns_first_data_block;
505 nrsvblocks = nilfs->ns_nrsvsegs * nilfs->ns_blocks_per_segment;
508 * Compute the overhead
510 * When distributing meta data blocks outside semgent structure,
511 * We must count them as the overhead.
515 err = nilfs_count_free_blocks(nilfs, &nfreeblocks);
519 buf->f_type = NILFS_SUPER_MAGIC;
520 buf->f_bsize = sb->s_blocksize;
521 buf->f_blocks = blocks - overhead;
522 buf->f_bfree = nfreeblocks;
523 buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
524 (buf->f_bfree - nrsvblocks) : 0;
525 buf->f_files = atomic_read(&sbi->s_inodes_count);
526 buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
527 buf->f_namelen = NILFS_NAME_LEN;
528 buf->f_fsid.val[0] = (u32)id;
529 buf->f_fsid.val[1] = (u32)(id >> 32);
534 static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
536 struct super_block *sb = vfs->mnt_sb;
537 struct nilfs_sb_info *sbi = NILFS_SB(sb);
539 if (!nilfs_test_opt(sbi, BARRIER))
540 seq_printf(seq, ",barrier=off");
541 if (nilfs_test_opt(sbi, SNAPSHOT))
542 seq_printf(seq, ",cp=%llu",
543 (unsigned long long int)sbi->s_snapshot_cno);
544 if (nilfs_test_opt(sbi, ERRORS_RO))
545 seq_printf(seq, ",errors=remount-ro");
546 if (nilfs_test_opt(sbi, ERRORS_PANIC))
547 seq_printf(seq, ",errors=panic");
548 if (nilfs_test_opt(sbi, STRICT_ORDER))
549 seq_printf(seq, ",order=strict");
554 static struct super_operations nilfs_sops = {
555 .alloc_inode = nilfs_alloc_inode,
556 .destroy_inode = nilfs_destroy_inode,
557 .dirty_inode = nilfs_dirty_inode,
558 /* .write_inode = nilfs_write_inode, */
559 /* .put_inode = nilfs_put_inode, */
560 /* .drop_inode = nilfs_drop_inode, */
561 .delete_inode = nilfs_delete_inode,
562 .put_super = nilfs_put_super,
563 .write_super = nilfs_write_super,
564 .sync_fs = nilfs_sync_fs,
565 /* .write_super_lockfs */
567 .statfs = nilfs_statfs,
568 .remount_fs = nilfs_remount,
569 .clear_inode = nilfs_clear_inode,
571 .show_options = nilfs_show_options
574 static struct inode *
575 nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
579 if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
580 ino != NILFS_SKETCH_INO)
581 return ERR_PTR(-ESTALE);
583 inode = nilfs_iget(sb, ino);
585 return ERR_CAST(inode);
586 if (generation && inode->i_generation != generation) {
588 return ERR_PTR(-ESTALE);
594 static struct dentry *
595 nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
598 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
599 nilfs_nfs_get_inode);
602 static struct dentry *
603 nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
606 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
607 nilfs_nfs_get_inode);
610 static struct export_operations nilfs_export_ops = {
611 .fh_to_dentry = nilfs_fh_to_dentry,
612 .fh_to_parent = nilfs_fh_to_parent,
613 .get_parent = nilfs_get_parent,
617 Opt_err_cont, Opt_err_panic, Opt_err_ro,
618 Opt_barrier, Opt_snapshot, Opt_order,
622 static match_table_t tokens = {
623 {Opt_err_cont, "errors=continue"},
624 {Opt_err_panic, "errors=panic"},
625 {Opt_err_ro, "errors=remount-ro"},
626 {Opt_barrier, "barrier=%s"},
627 {Opt_snapshot, "cp=%u"},
628 {Opt_order, "order=%s"},
632 static int match_bool(substring_t *s, int *result)
634 int len = s->to - s->from;
636 if (strncmp(s->from, "on", len) == 0)
638 else if (strncmp(s->from, "off", len) == 0)
645 static int parse_options(char *options, struct super_block *sb)
647 struct nilfs_sb_info *sbi = NILFS_SB(sb);
649 substring_t args[MAX_OPT_ARGS];
655 while ((p = strsep(&options, ",")) != NULL) {
660 token = match_token(p, tokens, args);
663 if (match_bool(&args[0], &option))
666 nilfs_set_opt(sbi, BARRIER);
668 nilfs_clear_opt(sbi, BARRIER);
671 if (strcmp(args[0].from, "relaxed") == 0)
672 /* Ordered data semantics */
673 nilfs_clear_opt(sbi, STRICT_ORDER);
674 else if (strcmp(args[0].from, "strict") == 0)
675 /* Strict in-order semantics */
676 nilfs_set_opt(sbi, STRICT_ORDER);
681 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_PANIC);
684 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_RO);
687 nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
690 if (match_int(&args[0], &option) || option <= 0)
692 if (!(sb->s_flags & MS_RDONLY))
694 sbi->s_snapshot_cno = option;
695 nilfs_set_opt(sbi, SNAPSHOT);
699 "NILFS: Unrecognized mount option \"%s\"\n", p);
707 nilfs_set_default_options(struct nilfs_sb_info *sbi,
708 struct nilfs_super_block *sbp)
711 NILFS_MOUNT_ERRORS_CONT | NILFS_MOUNT_BARRIER;
714 static int nilfs_setup_super(struct nilfs_sb_info *sbi)
716 struct the_nilfs *nilfs = sbi->s_nilfs;
717 struct nilfs_super_block *sbp = nilfs->ns_sbp[0];
718 int max_mnt_count = le16_to_cpu(sbp->s_max_mnt_count);
719 int mnt_count = le16_to_cpu(sbp->s_mnt_count);
721 /* nilfs->sem must be locked by the caller. */
722 if (!(nilfs->ns_mount_state & NILFS_VALID_FS)) {
723 printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
724 } else if (nilfs->ns_mount_state & NILFS_ERROR_FS) {
726 "NILFS warning: mounting fs with errors\n");
728 } else if (max_mnt_count >= 0 && mnt_count >= max_mnt_count) {
730 "NILFS warning: maximal mount count reached\n");
734 sbp->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
736 sbp->s_mnt_count = cpu_to_le16(mnt_count + 1);
737 sbp->s_state = cpu_to_le16(le16_to_cpu(sbp->s_state) & ~NILFS_VALID_FS);
738 sbp->s_mtime = cpu_to_le64(get_seconds());
739 return nilfs_commit_super(sbi, 1);
742 struct nilfs_super_block *nilfs_read_super_block(struct super_block *sb,
743 u64 pos, int blocksize,
744 struct buffer_head **pbh)
746 unsigned long long sb_index = pos;
747 unsigned long offset;
749 offset = do_div(sb_index, blocksize);
750 *pbh = sb_bread(sb, sb_index);
753 return (struct nilfs_super_block *)((char *)(*pbh)->b_data + offset);
756 int nilfs_store_magic_and_option(struct super_block *sb,
757 struct nilfs_super_block *sbp,
760 struct nilfs_sb_info *sbi = NILFS_SB(sb);
762 sb->s_magic = le16_to_cpu(sbp->s_magic);
764 /* FS independent flags */
765 #ifdef NILFS_ATIME_DISABLE
766 sb->s_flags |= MS_NOATIME;
769 nilfs_set_default_options(sbi, sbp);
771 sbi->s_resuid = le16_to_cpu(sbp->s_def_resuid);
772 sbi->s_resgid = le16_to_cpu(sbp->s_def_resgid);
773 sbi->s_interval = le32_to_cpu(sbp->s_c_interval);
774 sbi->s_watermark = le32_to_cpu(sbp->s_c_block_max);
776 return !parse_options(data, sb) ? -EINVAL : 0 ;
780 * nilfs_fill_super() - initialize a super block instance
782 * @data: mount options
783 * @silent: silent mode flag
784 * @nilfs: the_nilfs struct
786 * This function is called exclusively by nilfs->ns_mount_mutex.
787 * So, the recovery process is protected from other simultaneous mounts.
790 nilfs_fill_super(struct super_block *sb, void *data, int silent,
791 struct the_nilfs *nilfs)
793 struct nilfs_sb_info *sbi;
798 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
805 sbi->s_nilfs = nilfs;
807 atomic_set(&sbi->s_count, 1);
809 err = init_nilfs(nilfs, sbi, (char *)data);
813 spin_lock_init(&sbi->s_inode_lock);
814 INIT_LIST_HEAD(&sbi->s_dirty_files);
815 INIT_LIST_HEAD(&sbi->s_list);
818 * Following initialization is overlapped because
819 * nilfs_sb_info structure has been cleared at the beginning.
820 * But we reserve them to keep our interest and make ready
821 * for the future change.
823 get_random_bytes(&sbi->s_next_generation,
824 sizeof(sbi->s_next_generation));
825 spin_lock_init(&sbi->s_next_gen_lock);
827 sb->s_op = &nilfs_sops;
828 sb->s_export_op = &nilfs_export_ops;
832 if (!nilfs_loaded(nilfs)) {
833 err = load_nilfs(nilfs, sbi);
837 cno = nilfs_last_cno(nilfs);
839 if (sb->s_flags & MS_RDONLY) {
840 if (nilfs_test_opt(sbi, SNAPSHOT)) {
841 err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
842 sbi->s_snapshot_cno);
847 "NILFS: The specified checkpoint is "
849 "(checkpoint number=%llu).\n",
850 (unsigned long long)sbi->s_snapshot_cno);
854 cno = sbi->s_snapshot_cno;
856 /* Read-only mount */
857 sbi->s_snapshot_cno = cno;
860 err = nilfs_attach_checkpoint(sbi, cno);
862 printk(KERN_ERR "NILFS: error loading a checkpoint"
863 " (checkpoint number=%llu).\n", (unsigned long long)cno);
867 if (!(sb->s_flags & MS_RDONLY)) {
868 err = nilfs_attach_segment_constructor(sbi);
870 goto failed_checkpoint;
873 root = nilfs_iget(sb, NILFS_ROOT_INO);
875 printk(KERN_ERR "NILFS: get root inode failed\n");
879 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
881 printk(KERN_ERR "NILFS: corrupt root inode.\n");
885 sb->s_root = d_alloc_root(root);
888 printk(KERN_ERR "NILFS: get root dentry failed\n");
893 if (!(sb->s_flags & MS_RDONLY)) {
894 down_write(&nilfs->ns_sem);
895 nilfs_setup_super(sbi);
896 up_write(&nilfs->ns_sem);
899 err = nilfs_mark_recovery_complete(sbi);
901 printk(KERN_ERR "NILFS: recovery failed.\n");
905 down_write(&nilfs->ns_super_sem);
906 if (!nilfs_test_opt(sbi, SNAPSHOT))
907 nilfs->ns_current = sbi;
908 up_write(&nilfs->ns_super_sem);
917 nilfs_detach_segment_constructor(sbi);
920 nilfs_detach_checkpoint(sbi);
924 sb->s_fs_info = NULL;
925 nilfs_put_sbinfo(sbi);
929 static int nilfs_remount(struct super_block *sb, int *flags, char *data)
931 struct nilfs_sb_info *sbi = NILFS_SB(sb);
932 struct nilfs_super_block *sbp;
933 struct the_nilfs *nilfs = sbi->s_nilfs;
934 unsigned long old_sb_flags;
935 struct nilfs_mount_options old_opts;
940 down_write(&nilfs->ns_super_sem);
941 old_sb_flags = sb->s_flags;
942 old_opts.mount_opt = sbi->s_mount_opt;
943 old_opts.snapshot_cno = sbi->s_snapshot_cno;
945 if (!parse_options(data, sb)) {
949 sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
951 if ((*flags & MS_RDONLY) &&
952 sbi->s_snapshot_cno != old_opts.snapshot_cno) {
953 printk(KERN_WARNING "NILFS (device %s): couldn't "
954 "remount to a different snapshot. \n",
960 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
962 if (*flags & MS_RDONLY) {
963 /* Shutting down the segment constructor */
964 nilfs_detach_segment_constructor(sbi);
965 sb->s_flags |= MS_RDONLY;
967 sbi->s_snapshot_cno = nilfs_last_cno(nilfs);
968 /* nilfs_set_opt(sbi, SNAPSHOT); */
971 * Remounting a valid RW partition RDONLY, so set
972 * the RDONLY flag and then mark the partition as valid again.
974 down_write(&nilfs->ns_sem);
975 sbp = nilfs->ns_sbp[0];
976 if (!(sbp->s_state & le16_to_cpu(NILFS_VALID_FS)) &&
977 (nilfs->ns_mount_state & NILFS_VALID_FS))
978 sbp->s_state = cpu_to_le16(nilfs->ns_mount_state);
979 sbp->s_mtime = cpu_to_le64(get_seconds());
980 nilfs_commit_super(sbi, 1);
981 up_write(&nilfs->ns_sem);
984 * Mounting a RDONLY partition read-write, so reread and
985 * store the current valid flag. (It may have been changed
986 * by fsck since we originally mounted the partition.)
988 if (nilfs->ns_current && nilfs->ns_current != sbi) {
989 printk(KERN_WARNING "NILFS (device %s): couldn't "
990 "remount because an RW-mount exists.\n",
995 if (sbi->s_snapshot_cno != nilfs_last_cno(nilfs)) {
996 printk(KERN_WARNING "NILFS (device %s): couldn't "
997 "remount because the current RO-mount is not "
1003 sb->s_flags &= ~MS_RDONLY;
1004 nilfs_clear_opt(sbi, SNAPSHOT);
1005 sbi->s_snapshot_cno = 0;
1007 err = nilfs_attach_segment_constructor(sbi);
1011 down_write(&nilfs->ns_sem);
1012 nilfs_setup_super(sbi);
1013 up_write(&nilfs->ns_sem);
1015 nilfs->ns_current = sbi;
1018 up_write(&nilfs->ns_super_sem);
1023 sb->s_flags = old_sb_flags;
1024 sbi->s_mount_opt = old_opts.mount_opt;
1025 sbi->s_snapshot_cno = old_opts.snapshot_cno;
1026 up_write(&nilfs->ns_super_sem);
1031 struct nilfs_super_data {
1032 struct block_device *bdev;
1033 struct nilfs_sb_info *sbi;
1039 * nilfs_identify - pre-read mount options needed to identify mount instance
1040 * @data: mount options
1041 * @sd: nilfs_super_data
1043 static int nilfs_identify(char *data, struct nilfs_super_data *sd)
1045 char *p, *options = data;
1046 substring_t args[MAX_OPT_ARGS];
1051 p = strsep(&options, ",");
1052 if (p != NULL && *p) {
1053 token = match_token(p, tokens, args);
1054 if (token == Opt_snapshot) {
1055 if (!(sd->flags & MS_RDONLY))
1058 ret = match_int(&args[0], &option);
1069 "NILFS: invalid mount option: %s\n", p);
1073 BUG_ON(options == data);
1074 *(options - 1) = ',';
1079 static int nilfs_set_bdev_super(struct super_block *s, void *data)
1081 struct nilfs_super_data *sd = data;
1083 s->s_bdev = sd->bdev;
1084 s->s_dev = s->s_bdev->bd_dev;
1088 static int nilfs_test_bdev_super(struct super_block *s, void *data)
1090 struct nilfs_super_data *sd = data;
1092 return sd->sbi && s->s_fs_info == (void *)sd->sbi;
1096 nilfs_get_sb(struct file_system_type *fs_type, int flags,
1097 const char *dev_name, void *data, struct vfsmount *mnt)
1099 struct nilfs_super_data sd;
1100 struct super_block *s;
1101 struct the_nilfs *nilfs;
1102 int err, need_to_close = 1;
1104 sd.bdev = open_bdev_exclusive(dev_name, flags, fs_type);
1105 if (IS_ERR(sd.bdev))
1106 return PTR_ERR(sd.bdev);
1109 * To get mount instance using sget() vfs-routine, NILFS needs
1110 * much more information than normal filesystems to identify mount
1111 * instance. For snapshot mounts, not only a mount type (ro-mount
1112 * or rw-mount) but also a checkpoint number is required.
1116 if (nilfs_identify((char *)data, &sd)) {
1121 nilfs = find_or_create_nilfs(sd.bdev);
1127 mutex_lock(&nilfs->ns_mount_mutex);
1131 * Check if an exclusive mount exists or not.
1132 * Snapshot mounts coexist with a current mount
1133 * (i.e. rw-mount or ro-mount), whereas rw-mount and
1134 * ro-mount are mutually exclusive.
1136 down_read(&nilfs->ns_super_sem);
1137 if (nilfs->ns_current &&
1138 ((nilfs->ns_current->s_super->s_flags ^ flags)
1140 up_read(&nilfs->ns_super_sem);
1144 up_read(&nilfs->ns_super_sem);
1148 * Find existing nilfs_sb_info struct
1150 sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
1153 /* trying to get the latest checkpoint. */
1154 sd.cno = nilfs_last_cno(nilfs);
1157 * Get super block instance holding the nilfs_sb_info struct.
1158 * A new instance is allocated if no existing mount is present or
1159 * existing instance has been unmounted.
1161 s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
1163 nilfs_put_sbinfo(sd.sbi);
1171 char b[BDEVNAME_SIZE];
1173 /* New superblock instance created */
1175 strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
1176 sb_set_blocksize(s, block_size(sd.bdev));
1178 err = nilfs_fill_super(s, data, flags & MS_VERBOSE, nilfs);
1182 s->s_flags |= MS_ACTIVE;
1186 mutex_unlock(&nilfs->ns_mount_mutex);
1189 close_bdev_exclusive(sd.bdev, flags);
1190 simple_set_mnt(mnt, s);
1194 mutex_unlock(&nilfs->ns_mount_mutex);
1197 close_bdev_exclusive(sd.bdev, flags);
1202 /* Abandoning the newly allocated superblock */
1203 mutex_unlock(&nilfs->ns_mount_mutex);
1205 up_write(&s->s_umount);
1206 deactivate_super(s);
1208 * deactivate_super() invokes close_bdev_exclusive().
1209 * We must finish all post-cleaning before this call;
1210 * put_nilfs() needs the block device.
1215 struct file_system_type nilfs_fs_type = {
1216 .owner = THIS_MODULE,
1218 .get_sb = nilfs_get_sb,
1219 .kill_sb = kill_block_super,
1220 .fs_flags = FS_REQUIRES_DEV,
1223 static int __init init_nilfs_fs(void)
1227 err = nilfs_init_inode_cache();
1231 err = nilfs_init_transaction_cache();
1233 goto failed_inode_cache;
1235 err = nilfs_init_segbuf_cache();
1237 goto failed_transaction_cache;
1239 err = nilfs_btree_path_cache_init();
1241 goto failed_segbuf_cache;
1243 err = register_filesystem(&nilfs_fs_type);
1245 goto failed_btree_path_cache;
1249 failed_btree_path_cache:
1250 nilfs_btree_path_cache_destroy();
1252 failed_segbuf_cache:
1253 nilfs_destroy_segbuf_cache();
1255 failed_transaction_cache:
1256 nilfs_destroy_transaction_cache();
1259 nilfs_destroy_inode_cache();
1265 static void __exit exit_nilfs_fs(void)
1267 nilfs_destroy_segbuf_cache();
1268 nilfs_destroy_transaction_cache();
1269 nilfs_destroy_inode_cache();
1270 nilfs_btree_path_cache_destroy();
1271 unregister_filesystem(&nilfs_fs_type);
1274 module_init(init_nilfs_fs)
1275 module_exit(exit_nilfs_fs)