Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2: (36 commits)
  nilfs2: eliminate sparse warning - "context imbalance"
  nilfs2: eliminate sparse warnings - "symbol not declared"
  nilfs2: get rid of bdi from nilfs object
  nilfs2: change license of exported header file
  nilfs2: add bdev freeze/thaw support
  nilfs2: accept 64-bit checkpoint numbers in cp mount option
  nilfs2: remove own inode allocator and destructor for metadata files
  nilfs2: get rid of back pointer to writable sb instance
  nilfs2: get rid of mi_nilfs back pointer to nilfs object
  nilfs2: see state of root dentry for mount check of snapshots
  nilfs2: use iget for all metadata files
  nilfs2: get rid of GCDAT inode
  nilfs2: add routines to redirect access to buffers of DAT file
  nilfs2: add routines to roll back state of DAT file
  nilfs2: add routines to save and restore bmap state
  nilfs2: do not allocate nilfs_mdt_info structure to gc-inodes
  nilfs2: allow nilfs_clear_inode to clear metadata file inodes
  nilfs2: get rid of snapshot mount flag
  nilfs2: simplify life cycle management of nilfs object
  nilfs2: do not allocate multiple super block instances for a device
  ...

fs/nilfs2/Makefile

@@ -2,4 +2,4 @@ obj-$(CONFIG_NILFS2_FS) += nilfs2.o
 nilfs2-y := inode.o file.o dir.o super.o namei.o page.o mdt.o \
 	btnode.o bmap.o btree.o direct.o dat.o recovery.o \
 	the_nilfs.o segbuf.o segment.o cpfile.o sufile.o \
-	ifile.o alloc.o gcinode.o ioctl.o gcdat.o
+	ifile.o alloc.o gcinode.o ioctl.o

fs/nilfs2/bmap.c

@@ -533,18 +533,20 @@ void nilfs_bmap_init_gc(struct nilfs_bmap *bmap)
 	nilfs_btree_init_gc(bmap);
 }
 
-void nilfs_bmap_init_gcdat(struct nilfs_bmap *gcbmap, struct nilfs_bmap *bmap)
+void nilfs_bmap_save(const struct nilfs_bmap *bmap,
+		     struct nilfs_bmap_store *store)
 {
-	memcpy(gcbmap, bmap, sizeof(*bmap));
-	init_rwsem(&gcbmap->b_sem);
-	lockdep_set_class(&bmap->b_sem, &nilfs_bmap_dat_lock_key);
-	gcbmap->b_inode = &NILFS_BMAP_I(gcbmap)->vfs_inode;
+	memcpy(store->data, bmap->b_u.u_data, sizeof(store->data));
+	store->last_allocated_key = bmap->b_last_allocated_key;
+	store->last_allocated_ptr = bmap->b_last_allocated_ptr;
+	store->state = bmap->b_state;
 }
 
-void nilfs_bmap_commit_gcdat(struct nilfs_bmap *gcbmap, struct nilfs_bmap *bmap)
+void nilfs_bmap_restore(struct nilfs_bmap *bmap,
+			const struct nilfs_bmap_store *store)
 {
-	memcpy(bmap, gcbmap, sizeof(*bmap));
-	init_rwsem(&bmap->b_sem);
-	lockdep_set_class(&bmap->b_sem, &nilfs_bmap_dat_lock_key);
-	bmap->b_inode = &NILFS_BMAP_I(bmap)->vfs_inode;
+	memcpy(bmap->b_u.u_data, store->data, sizeof(store->data));
+	bmap->b_last_allocated_key = store->last_allocated_key;
+	bmap->b_last_allocated_ptr = store->last_allocated_ptr;
+	bmap->b_state = store->state;
 }

fs/nilfs2/bmap.h

@@ -135,6 +135,12 @@ struct nilfs_bmap {
 /* state */
 #define NILFS_BMAP_DIRTY	0x00000001
 
+struct nilfs_bmap_store {
+	__le64 data[NILFS_BMAP_SIZE / sizeof(__le64)];
+	__u64 last_allocated_key;
+	__u64 last_allocated_ptr;
+	int state;
+};
 
 int nilfs_bmap_test_and_clear_dirty(struct nilfs_bmap *);
 int nilfs_bmap_read(struct nilfs_bmap *, struct nilfs_inode *);
@@ -153,9 +159,9 @@ int nilfs_bmap_lookup_at_level(struct nilfs_bmap *, __u64, int, __u64 *);
 int nilfs_bmap_mark(struct nilfs_bmap *, __u64, int);
 
 void nilfs_bmap_init_gc(struct nilfs_bmap *);
-void nilfs_bmap_init_gcdat(struct nilfs_bmap *, struct nilfs_bmap *);
-void nilfs_bmap_commit_gcdat(struct nilfs_bmap *, struct nilfs_bmap *);
 
+void nilfs_bmap_save(const struct nilfs_bmap *, struct nilfs_bmap_store *);
+void nilfs_bmap_restore(struct nilfs_bmap *, const struct nilfs_bmap_store *);
 
 static inline int nilfs_bmap_lookup(struct nilfs_bmap *bmap, __u64 key,
 				    __u64 *ptr)

fs/nilfs2/btnode.c

@@ -37,15 +37,7 @@
 
 void nilfs_btnode_cache_init_once(struct address_space *btnc)
 {
-	memset(btnc, 0, sizeof(*btnc));
-	INIT_RADIX_TREE(&btnc->page_tree, GFP_ATOMIC);
-	spin_lock_init(&btnc->tree_lock);
-	INIT_LIST_HEAD(&btnc->private_list);
-	spin_lock_init(&btnc->private_lock);
-
-	spin_lock_init(&btnc->i_mmap_lock);
-	INIT_RAW_PRIO_TREE_ROOT(&btnc->i_mmap);
-	INIT_LIST_HEAD(&btnc->i_mmap_nonlinear);
+	nilfs_mapping_init_once(btnc);
 }
 
 static const struct address_space_operations def_btnode_aops = {
@@ -55,12 +47,7 @@ static const struct address_space_operations def_btnode_aops = {
 void nilfs_btnode_cache_init(struct address_space *btnc,
 			     struct backing_dev_info *bdi)
 {
-	btnc->host = NULL;  /* can safely set to host inode ? */
-	btnc->flags = 0;
-	mapping_set_gfp_mask(btnc, GFP_NOFS);
-	btnc->assoc_mapping = NULL;
-	btnc->backing_dev_info = bdi;
-	btnc->a_ops = &def_btnode_aops;
+	nilfs_mapping_init(btnc, bdi, &def_btnode_aops);
 }
 
 void nilfs_btnode_cache_clear(struct address_space *btnc)

fs/nilfs2/cpfile.c

@@ -863,26 +863,19 @@ int nilfs_cpfile_is_snapshot(struct inode *cpfile, __u64 cno)
  */
 int nilfs_cpfile_change_cpmode(struct inode *cpfile, __u64 cno, int mode)
 {
-	struct the_nilfs *nilfs;
 	int ret;
 
-	nilfs = NILFS_MDT(cpfile)->mi_nilfs;
-
 	switch (mode) {
 	case NILFS_CHECKPOINT:
-		/*
-		 * Check for protecting existing snapshot mounts:
-		 * ns_mount_mutex is used to make this operation atomic and
-		 * exclusive with a new mount job.  Though it doesn't cover
-		 * umount, it's enough for the purpose.
-		 */
-		if (nilfs_checkpoint_is_mounted(nilfs, cno, 1)) {
-			/* Current implementation does not have to protect
-			   plain read-only mounts since they are exclusive
-			   with a read/write mount and are protected from the
-			   cleaner. */
+		if (nilfs_checkpoint_is_mounted(cpfile->i_sb, cno))
+			/*
+			 * Current implementation does not have to protect
+			 * plain read-only mounts since they are exclusive
+			 * with a read/write mount and are protected from the
+			 * cleaner.
+			 */
 			ret = -EBUSY;
-		} else
+		else
 			ret = nilfs_cpfile_clear_snapshot(cpfile, cno);
 		return ret;
 	case NILFS_SNAPSHOT:
@@ -933,27 +926,40 @@ int nilfs_cpfile_get_stat(struct inode *cpfile, struct nilfs_cpstat *cpstat)
 }
 
 /**
- * nilfs_cpfile_read - read cpfile inode
- * @cpfile: cpfile inode
- * @raw_inode: on-disk cpfile inode
- */
-int nilfs_cpfile_read(struct inode *cpfile, struct nilfs_inode *raw_inode)
-{
-	return nilfs_read_inode_common(cpfile, raw_inode);
-}
-
-/**
- * nilfs_cpfile_new - create cpfile
- * @nilfs: nilfs object
+ * nilfs_cpfile_read - read or get cpfile inode
+ * @sb: super block instance
  * @cpsize: size of a checkpoint entry
+ * @raw_inode: on-disk cpfile inode
+ * @inodep: buffer to store the inode
  */
-struct inode *nilfs_cpfile_new(struct the_nilfs *nilfs, size_t cpsize)
+int nilfs_cpfile_read(struct super_block *sb, size_t cpsize,
+		      struct nilfs_inode *raw_inode, struct inode **inodep)
 {
 	struct inode *cpfile;
+	int err;
+
+	cpfile = nilfs_iget_locked(sb, NULL, NILFS_CPFILE_INO);
+	if (unlikely(!cpfile))
+		return -ENOMEM;
+	if (!(cpfile->i_state & I_NEW))
+		goto out;
+
+	err = nilfs_mdt_init(cpfile, NILFS_MDT_GFP, 0);
+	if (err)
+		goto failed;
 
-	cpfile = nilfs_mdt_new(nilfs, NULL, NILFS_CPFILE_INO, 0);
-	if (cpfile)
-		nilfs_mdt_set_entry_size(cpfile, cpsize,
-					 sizeof(struct nilfs_cpfile_header));
-	return cpfile;
+	nilfs_mdt_set_entry_size(cpfile, cpsize,
+				 sizeof(struct nilfs_cpfile_header));
+
+	err = nilfs_read_inode_common(cpfile, raw_inode);
+	if (err)
+		goto failed;
+
+	unlock_new_inode(cpfile);
+ out:
+	*inodep = cpfile;
+	return 0;
+ failed:
+	iget_failed(cpfile);
+	return err;
 }

fs/nilfs2/cpfile.h

@@ -40,7 +40,7 @@ int nilfs_cpfile_get_stat(struct inode *, struct nilfs_cpstat *);
 ssize_t nilfs_cpfile_get_cpinfo(struct inode *, __u64 *, int, void *, unsigned,
 				size_t);
 
-int nilfs_cpfile_read(struct inode *cpfile, struct nilfs_inode *raw_inode);
-struct inode *nilfs_cpfile_new(struct the_nilfs *nilfs, size_t cpsize);
+int nilfs_cpfile_read(struct super_block *sb, size_t cpsize,
+		      struct nilfs_inode *raw_inode, struct inode **inodep);
 
 #endif	/* _NILFS_CPFILE_H */

fs/nilfs2/dat.c

@@ -36,6 +36,7 @@
 struct nilfs_dat_info {
 	struct nilfs_mdt_info mi;
 	struct nilfs_palloc_cache palloc_cache;
+	struct nilfs_shadow_map shadow;
 };
 
 static inline struct nilfs_dat_info *NILFS_DAT_I(struct inode *dat)
@@ -102,7 +103,8 @@ void nilfs_dat_abort_alloc(struct inode *dat, struct nilfs_palloc_req *req)
 	nilfs_palloc_abort_alloc_entry(dat, req);
 }
 
-void nilfs_dat_commit_free(struct inode *dat, struct nilfs_palloc_req *req)
+static void nilfs_dat_commit_free(struct inode *dat,
+				  struct nilfs_palloc_req *req)
 {
 	struct nilfs_dat_entry *entry;
 	void *kaddr;
@@ -327,6 +329,23 @@ int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
 	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
 	if (ret < 0)
 		return ret;
+
+	/*
+	 * The given disk block number (blocknr) is not yet written to
+	 * the device at this point.
+	 *
+	 * To prevent nilfs_dat_translate() from returning the
+	 * uncommited block number, this makes a copy of the entry
+	 * buffer and redirects nilfs_dat_translate() to the copy.
+	 */
+	if (!buffer_nilfs_redirected(entry_bh)) {
+		ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
+		if (ret) {
+			brelse(entry_bh);
+			return ret;
+		}
+	}
+
 	kaddr = kmap_atomic(entry_bh->b_page, KM_USER0);
 	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
 	if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
@@ -371,7 +390,7 @@ int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
  */
 int nilfs_dat_translate(struct inode *dat, __u64 vblocknr, sector_t *blocknrp)
 {
-	struct buffer_head *entry_bh;
+	struct buffer_head *entry_bh, *bh;
 	struct nilfs_dat_entry *entry;
 	sector_t blocknr;
 	void *kaddr;
@@ -381,6 +400,15 @@ int nilfs_dat_translate(struct inode *dat, __u64 vblocknr, sector_t *blocknrp)
 	if (ret < 0)
 		return ret;
 
+	if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
+		bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
+		if (bh) {
+			WARN_ON(!buffer_uptodate(bh));
+			brelse(entry_bh);
+			entry_bh = bh;
+		}
+	}
+
 	kaddr = kmap_atomic(entry_bh->b_page, KM_USER0);
 	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
 	blocknr = le64_to_cpu(entry->de_blocknr);
@@ -436,38 +464,48 @@ ssize_t nilfs_dat_get_vinfo(struct inode *dat, void *buf, unsigned visz,
 }
 
 /**
- * nilfs_dat_read - read dat inode
- * @dat: dat inode
- * @raw_inode: on-disk dat inode
- */
-int nilfs_dat_read(struct inode *dat, struct nilfs_inode *raw_inode)
-{
-	return nilfs_read_inode_common(dat, raw_inode);
-}
-
-/**
- * nilfs_dat_new - create dat file
- * @nilfs: nilfs object
+ * nilfs_dat_read - read or get dat inode
+ * @sb: super block instance
  * @entry_size: size of a dat entry
+ * @raw_inode: on-disk dat inode
+ * @inodep: buffer to store the inode
  */
-struct inode *nilfs_dat_new(struct the_nilfs *nilfs, size_t entry_size)
+int nilfs_dat_read(struct super_block *sb, size_t entry_size,
+		   struct nilfs_inode *raw_inode, struct inode **inodep)
 {
 	static struct lock_class_key dat_lock_key;
 	struct inode *dat;
 	struct nilfs_dat_info *di;
 	int err;
 
-	dat = nilfs_mdt_new(nilfs, NULL, NILFS_DAT_INO, sizeof(*di));
-	if (dat) {
-		err = nilfs_palloc_init_blockgroup(dat, entry_size);
-		if (unlikely(err)) {
-			nilfs_mdt_destroy(dat);
-			return NULL;
-		}
+	dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
+	if (unlikely(!dat))
+		return -ENOMEM;
+	if (!(dat->i_state & I_NEW))
+		goto out;
 
-		di = NILFS_DAT_I(dat);
-		lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
-		nilfs_palloc_setup_cache(dat, &di->palloc_cache);
-	}
-	return dat;
+	err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
+	if (err)
+		goto failed;
+
+	err = nilfs_palloc_init_blockgroup(dat, entry_size);
+	if (err)
+		goto failed;
+
+	di = NILFS_DAT_I(dat);
+	lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
+	nilfs_palloc_setup_cache(dat, &di->palloc_cache);
+	nilfs_mdt_setup_shadow_map(dat, &di->shadow);
+
+	err = nilfs_read_inode_common(dat, raw_inode);
+	if (err)
+		goto failed;
+
+	unlock_new_inode(dat);
+ out:
+	*inodep = dat;
+	return 0;
+ failed:
+	iget_failed(dat);
+	return err;
 }

fs/nilfs2/dat.h

@@ -53,7 +53,7 @@ int nilfs_dat_freev(struct inode *, __u64 *, size_t);
 int nilfs_dat_move(struct inode *, __u64, sector_t);
 ssize_t nilfs_dat_get_vinfo(struct inode *, void *, unsigned, size_t);
 
-int nilfs_dat_read(struct inode *dat, struct nilfs_inode *raw_inode);
-struct inode *nilfs_dat_new(struct the_nilfs *nilfs, size_t entry_size);
+int nilfs_dat_read(struct super_block *sb, size_t entry_size,
+		   struct nilfs_inode *raw_inode, struct inode **inodep);
 
 #endif	/* _NILFS_DAT_H */

fs/nilfs2/export.h

@@ -0,0 +1,17 @@
+#ifndef NILFS_EXPORT_H
+#define NILFS_EXPORT_H
+
+#include <linux/exportfs.h>
+
+extern const struct export_operations nilfs_export_ops;
+
+struct nilfs_fid {
+	u64 cno;
+	u64 ino;
+	u32 gen;
+
+	u32 parent_gen;
+	u64 parent_ino;
+} __attribute__ ((packed));
+
+#endif

fs/nilfs2/gcdat.c

@@ -1,87 +0,0 @@
-/*
- * gcdat.c - NILFS shadow DAT inode for GC
- *
- * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
- *
- * Written by Seiji Kihara <kihara@osrg.net>, Amagai Yoshiji <amagai@osrg.net>,
- *            and Ryusuke Konishi <ryusuke@osrg.net>.
- *
- */
-
-#include <linux/buffer_head.h>
-#include "nilfs.h"
-#include "page.h"
-#include "mdt.h"
-
-int nilfs_init_gcdat_inode(struct the_nilfs *nilfs)
-{
-	struct inode *dat = nilfs->ns_dat, *gcdat = nilfs->ns_gc_dat;
-	struct nilfs_inode_info *dii = NILFS_I(dat), *gii = NILFS_I(gcdat);
-	int err;
-
-	gcdat->i_state = 0;
-	gcdat->i_blocks = dat->i_blocks;
-	gii->i_flags = dii->i_flags;
-	gii->i_state = dii->i_state | (1 << NILFS_I_GCDAT);
-	gii->i_cno = 0;
-	nilfs_bmap_init_gcdat(gii->i_bmap, dii->i_bmap);
-	err = nilfs_copy_dirty_pages(gcdat->i_mapping, dat->i_mapping);
-	if (unlikely(err))
-		return err;
-
-	return nilfs_copy_dirty_pages(&gii->i_btnode_cache,
-				      &dii->i_btnode_cache);
-}
-
-void nilfs_commit_gcdat_inode(struct the_nilfs *nilfs)
-{
-	struct inode *dat = nilfs->ns_dat, *gcdat = nilfs->ns_gc_dat;
-	struct nilfs_inode_info *dii = NILFS_I(dat), *gii = NILFS_I(gcdat);
-	struct address_space *mapping = dat->i_mapping;
-	struct address_space *gmapping = gcdat->i_mapping;
-
-	down_write(&NILFS_MDT(dat)->mi_sem);
-	dat->i_blocks = gcdat->i_blocks;
-	dii->i_flags = gii->i_flags;
-	dii->i_state = gii->i_state & ~(1 << NILFS_I_GCDAT);
-
-	nilfs_bmap_commit_gcdat(gii->i_bmap, dii->i_bmap);
-
-	nilfs_palloc_clear_cache(dat);
-	nilfs_palloc_clear_cache(gcdat);
-	nilfs_clear_dirty_pages(mapping);
-	nilfs_copy_back_pages(mapping, gmapping);
-	/* note: mdt dirty flags should be cleared by segctor. */
-
-	nilfs_clear_dirty_pages(&dii->i_btnode_cache);
-	nilfs_copy_back_pages(&dii->i_btnode_cache, &gii->i_btnode_cache);
-
-	up_write(&NILFS_MDT(dat)->mi_sem);
-}
-
-void nilfs_clear_gcdat_inode(struct the_nilfs *nilfs)
-{
-	struct inode *gcdat = nilfs->ns_gc_dat;
-	struct nilfs_inode_info *gii = NILFS_I(gcdat);
-
-	gcdat->i_state = I_FREEING | I_CLEAR;
-	gii->i_flags = 0;
-
-	nilfs_palloc_clear_cache(gcdat);
-	truncate_inode_pages(gcdat->i_mapping, 0);
-	truncate_inode_pages(&gii->i_btnode_cache, 0);
-}

fs/nilfs2/gcinode.c

@@ -28,13 +28,6 @@
  * gcinodes), and this file provides lookup function of the dummy
  * inodes and their buffer read function.
  *
- * Since NILFS2 keeps up multiple checkpoints/snapshots across GC, it
- * has to treat blocks that belong to a same file but have different
- * checkpoint numbers.  To avoid interference among generations, dummy
- * inodes are managed separately from actual inodes, and their lookup
- * function (nilfs_gc_iget) is designed to be specified with a
- * checkpoint number argument as well as an inode number.
- *
  * Buffers and pages held by the dummy inodes will be released each
  * time after they are copied to a new log.  Dirty blocks made on the
  * current generation and the blocks to be moved by GC never overlap
@@ -175,125 +168,46 @@ int nilfs_gccache_wait_and_mark_dirty(struct buffer_head *bh)
 		}
 		nilfs_btnode_mark_dirty(bh);
 	} else {
-		nilfs_mdt_mark_buffer_dirty(bh);
+		nilfs_mark_buffer_dirty(bh);
 	}
 	return 0;
 }
 
-/*
- * nilfs_init_gccache() - allocate and initialize gc_inode hash table
- * @nilfs - the_nilfs
- *
- * Return Value: On success, 0.
- * On error, a negative error code is returned.
- */
-int nilfs_init_gccache(struct the_nilfs *nilfs)
+int nilfs_init_gcinode(struct inode *inode)
 {
-	int loop;
-
-	BUG_ON(nilfs->ns_gc_inodes_h);
-
-	INIT_LIST_HEAD(&nilfs->ns_gc_inodes);
-
-	nilfs->ns_gc_inodes_h =
-		kmalloc(sizeof(struct hlist_head) * NILFS_GCINODE_HASH_SIZE,
-			GFP_NOFS);
-	if (nilfs->ns_gc_inodes_h == NULL)
-		return -ENOMEM;
-
-	for (loop = 0; loop < NILFS_GCINODE_HASH_SIZE; loop++)
-		INIT_HLIST_HEAD(&nilfs->ns_gc_inodes_h[loop]);
-	return 0;
-}
-
-/*
- * nilfs_destroy_gccache() - free gc_inode hash table
- * @nilfs - the nilfs
- */
-void nilfs_destroy_gccache(struct the_nilfs *nilfs)
-{
-	if (nilfs->ns_gc_inodes_h) {
-		nilfs_remove_all_gcinode(nilfs);
-		kfree(nilfs->ns_gc_inodes_h);
-		nilfs->ns_gc_inodes_h = NULL;
-	}
-}
-
-static struct inode *alloc_gcinode(struct the_nilfs *nilfs, ino_t ino,
-				   __u64 cno)
-{
-	struct inode *inode;
-	struct nilfs_inode_info *ii;
-
-	inode = nilfs_mdt_new_common(nilfs, NULL, ino, GFP_NOFS, 0);
-	if (!inode)
-		return NULL;
+	struct nilfs_inode_info *ii = NILFS_I(inode);
+	struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
 
-	inode->i_op = NULL;
-	inode->i_fop = NULL;
+	inode->i_mode = S_IFREG;
+	mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
 	inode->i_mapping->a_ops = &def_gcinode_aops;
+	inode->i_mapping->backing_dev_info = inode->i_sb->s_bdi;
 
-	ii = NILFS_I(inode);
-	ii->i_cno = cno;
 	ii->i_flags = 0;
-	ii->i_state = 1 << NILFS_I_GCINODE;
-	ii->i_bh = NULL;
 	nilfs_bmap_init_gc(ii->i_bmap);
 
-	return inode;
-}
-
-static unsigned long ihash(ino_t ino, __u64 cno)
-{
-	return hash_long((unsigned long)((ino << 2) + cno),
-			 NILFS_GCINODE_HASH_BITS);
-}
-
-/*
- * nilfs_gc_iget() - find or create gc inode with specified (ino,cno)
- */
-struct inode *nilfs_gc_iget(struct the_nilfs *nilfs, ino_t ino, __u64 cno)
-{
-	struct hlist_head *head = nilfs->ns_gc_inodes_h + ihash(ino, cno);
-	struct hlist_node *node;
-	struct inode *inode;
-
-	hlist_for_each_entry(inode, node, head, i_hash) {
-		if (inode->i_ino == ino && NILFS_I(inode)->i_cno == cno)
-			return inode;
-	}
+	/*
+	 * Add the inode to GC inode list. Garbage Collection
+	 * is serialized and no two processes manipulate the
+	 * list simultaneously.
+	 */
+	igrab(inode);
+	list_add(&NILFS_I(inode)->i_dirty, &nilfs->ns_gc_inodes);
 
-	inode = alloc_gcinode(nilfs, ino, cno);
-	if (likely(inode)) {
-		hlist_add_head(&inode->i_hash, head);
-		list_add(&NILFS_I(inode)->i_dirty, &nilfs->ns_gc_inodes);
-	}
-	return inode;
-}
-
-/*
- * nilfs_clear_gcinode() - clear and free a gc inode
- */
-void nilfs_clear_gcinode(struct inode *inode)
-{
-	nilfs_mdt_destroy(inode);
+	return 0;
 }
 
-/*
- * nilfs_remove_all_gcinode() - remove all inodes from the_nilfs
+/**
+ * nilfs_remove_all_gcinodes() - remove all unprocessed gc inodes
  */
-void nilfs_remove_all_gcinode(struct the_nilfs *nilfs)
+void nilfs_remove_all_gcinodes(struct the_nilfs *nilfs)
 {
-	struct hlist_head *head = nilfs->ns_gc_inodes_h;
-	struct hlist_node *node, *n;
-	struct inode *inode;
-	int loop;
+	struct list_head *head = &nilfs->ns_gc_inodes;
+	struct nilfs_inode_info *ii;
 
-	for (loop = 0; loop < NILFS_GCINODE_HASH_SIZE; loop++, head++) {
-		hlist_for_each_entry_safe(inode, node, n, head, i_hash) {
-			hlist_del_init(&inode->i_hash);
-			list_del_init(&NILFS_I(inode)->i_dirty);
-			nilfs_clear_gcinode(inode); /* might sleep */
-		}
+	while (!list_empty(head)) {
+		ii = list_first_entry(head, struct nilfs_inode_info, i_dirty);
+		list_del_init(&ii->i_dirty);
+		iput(&ii->vfs_inode);
 	}
 }

fs/nilfs2/ifile.c

@@ -161,25 +161,46 @@ int nilfs_ifile_get_inode_block(struct inode *ifile, ino_t ino,
 }
 
 /**
- * nilfs_ifile_new - create inode file
- * @sbi: nilfs_sb_info struct
+ * nilfs_ifile_read - read or get ifile inode
+ * @sb: super block instance
+ * @root: root object
  * @inode_size: size of an inode
+ * @raw_inode: on-disk ifile inode
+ * @inodep: buffer to store the inode
  */
-struct inode *nilfs_ifile_new(struct nilfs_sb_info *sbi, size_t inode_size)
+int nilfs_ifile_read(struct super_block *sb, struct nilfs_root *root,
+		     size_t inode_size, struct nilfs_inode *raw_inode,
+		     struct inode **inodep)
 {
 	struct inode *ifile;
 	int err;
 
-	ifile = nilfs_mdt_new(sbi->s_nilfs, sbi->s_super, NILFS_IFILE_INO,
-			      sizeof(struct nilfs_ifile_info));
-	if (ifile) {
-		err = nilfs_palloc_init_blockgroup(ifile, inode_size);
-		if (unlikely(err)) {
-			nilfs_mdt_destroy(ifile);
-			return NULL;
-		}
-		nilfs_palloc_setup_cache(ifile,
-					 &NILFS_IFILE_I(ifile)->palloc_cache);
-	}
-	return ifile;
+	ifile = nilfs_iget_locked(sb, root, NILFS_IFILE_INO);
+	if (unlikely(!ifile))
+		return -ENOMEM;
+	if (!(ifile->i_state & I_NEW))
+		goto out;
+
+	err = nilfs_mdt_init(ifile, NILFS_MDT_GFP,
+			     sizeof(struct nilfs_ifile_info));
+	if (err)
+		goto failed;
+
+	err = nilfs_palloc_init_blockgroup(ifile, inode_size);
+	if (err)
+		goto failed;
+
+	nilfs_palloc_setup_cache(ifile, &NILFS_IFILE_I(ifile)->palloc_cache);
+
+	err = nilfs_read_inode_common(ifile, raw_inode);
+	if (err)
+		goto failed;
+
+	unlock_new_inode(ifile);
+ out:
+	*inodep = ifile;
+	return 0;
+ failed:
+	iget_failed(ifile);
+	return err;
 }

fs/nilfs2/ifile.h

@@ -49,6 +49,8 @@ int nilfs_ifile_create_inode(struct inode *, ino_t *, struct buffer_head **);
 int nilfs_ifile_delete_inode(struct inode *, ino_t);
 int nilfs_ifile_get_inode_block(struct inode *, ino_t, struct buffer_head **);
 
-struct inode *nilfs_ifile_new(struct nilfs_sb_info *sbi, size_t inode_size);
+int nilfs_ifile_read(struct super_block *sb, struct nilfs_root *root,
+		     size_t inode_size, struct nilfs_inode *raw_inode,
+		     struct inode **inodep);
 
 #endif	/* _NILFS_IFILE_H */

fs/nilfs2/inode.c

@@ -34,6 +34,12 @@
 #include "cpfile.h"
 #include "ifile.h"
 
+struct nilfs_iget_args {
+	u64 ino;
+	__u64 cno;
+	struct nilfs_root *root;
+	int for_gc;
+};
 
 /**
  * nilfs_get_block() - get a file block on the filesystem (callback function)
@@ -279,6 +285,7 @@ struct inode *nilfs_new_inode(struct inode *dir, int mode)
 	struct nilfs_sb_info *sbi = NILFS_SB(sb);
 	struct inode *inode;
 	struct nilfs_inode_info *ii;
+	struct nilfs_root *root;
 	int err = -ENOMEM;
 	ino_t ino;
 
@@ -289,15 +296,17 @@ struct inode *nilfs_new_inode(struct inode *dir, int mode)
 	mapping_set_gfp_mask(inode->i_mapping,
 			     mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS);
 
+	root = NILFS_I(dir)->i_root;
 	ii = NILFS_I(inode);
 	ii->i_state = 1 << NILFS_I_NEW;
+	ii->i_root = root;
 
-	err = nilfs_ifile_create_inode(sbi->s_ifile, &ino, &ii->i_bh);
+	err = nilfs_ifile_create_inode(root->ifile, &ino, &ii->i_bh);
 	if (unlikely(err))
 		goto failed_ifile_create_inode;
 	/* reference count of i_bh inherits from nilfs_mdt_read_block() */
 
-	atomic_inc(&sbi->s_inodes_count);
+	atomic_inc(&root->inodes_count);
 	inode_init_owner(inode, dir, mode);
 	inode->i_ino = ino;
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
@@ -320,7 +329,6 @@ struct inode *nilfs_new_inode(struct inode *dir, int mode)
 	/* ii->i_file_acl = 0; */
 	/* ii->i_dir_acl = 0; */
 	ii->i_dir_start_lookup = 0;
-	ii->i_cno = 0;
 	nilfs_set_inode_flags(inode);
 	spin_lock(&sbi->s_next_gen_lock);
 	inode->i_generation = sbi->s_next_generation++;
@@ -350,16 +358,6 @@ struct inode *nilfs_new_inode(struct inode *dir, int mode)
 	return ERR_PTR(err);
 }
 
-void nilfs_free_inode(struct inode *inode)
-{
-	struct super_block *sb = inode->i_sb;
-	struct nilfs_sb_info *sbi = NILFS_SB(sb);
-
-	/* XXX: check error code? Is there any thing I can do? */
-	(void) nilfs_ifile_delete_inode(sbi->s_ifile, inode->i_ino);
-	atomic_dec(&sbi->s_inodes_count);
-}
-
 void nilfs_set_inode_flags(struct inode *inode)
 {
 	unsigned int flags = NILFS_I(inode)->i_flags;
@@ -410,7 +408,6 @@ int nilfs_read_inode_common(struct inode *inode,
 		0 : le32_to_cpu(raw_inode->i_dir_acl);
 #endif
 	ii->i_dir_start_lookup = 0;
-	ii->i_cno = 0;
 	inode->i_generation = le32_to_cpu(raw_inode->i_generation);
 
 	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
@@ -424,7 +421,8 @@ int nilfs_read_inode_common(struct inode *inode,
 	return 0;
 }
 
-static int __nilfs_read_inode(struct super_block *sb, unsigned long ino,
+static int __nilfs_read_inode(struct super_block *sb,
+			      struct nilfs_root *root, unsigned long ino,
 			      struct inode *inode)
 {
 	struct nilfs_sb_info *sbi = NILFS_SB(sb);
@@ -434,11 +432,11 @@ static int __nilfs_read_inode(struct super_block *sb, unsigned long ino,
 	int err;
 
 	down_read(&NILFS_MDT(dat)->mi_sem);	/* XXX */
-	err = nilfs_ifile_get_inode_block(sbi->s_ifile, ino, &bh);
+	err = nilfs_ifile_get_inode_block(root->ifile, ino, &bh);
 	if (unlikely(err))
 		goto bad_inode;
 
-	raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, bh);
+	raw_inode = nilfs_ifile_map_inode(root->ifile, ino, bh);
 
 	err = nilfs_read_inode_common(inode, raw_inode);
 	if (err)
@@ -461,14 +459,14 @@ static int __nilfs_read_inode(struct super_block *sb, unsigned long ino,
 			inode, inode->i_mode,
 			huge_decode_dev(le64_to_cpu(raw_inode->i_device_code)));
 	}
-	nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh);
+	nilfs_ifile_unmap_inode(root->ifile, ino, bh);
 	brelse(bh);
 	up_read(&NILFS_MDT(dat)->mi_sem);	/* XXX */
 	nilfs_set_inode_flags(inode);
 	return 0;
 
  failed_unmap:
-	nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh);
+	nilfs_ifile_unmap_inode(root->ifile, ino, bh);
 	brelse(bh);
 
  bad_inode:
@@ -476,18 +474,95 @@ static int __nilfs_read_inode(struct super_block *sb, unsigned long ino,
 	return err;
 }
 
-struct inode *nilfs_iget(struct super_block *sb, unsigned long ino)
+static int nilfs_iget_test(struct inode *inode, void *opaque)
+{
+	struct nilfs_iget_args *args = opaque;
+	struct nilfs_inode_info *ii;
+
+	if (args->ino != inode->i_ino || args->root != NILFS_I(inode)->i_root)
+		return 0;
+
+	ii = NILFS_I(inode);
+	if (!test_bit(NILFS_I_GCINODE, &ii->i_state))
+		return !args->for_gc;
+
+	return args->for_gc && args->cno == ii->i_cno;
+}
+
+static int nilfs_iget_set(struct inode *inode, void *opaque)
+{
+	struct nilfs_iget_args *args = opaque;
+
+	inode->i_ino = args->ino;
+	if (args->for_gc) {
+		NILFS_I(inode)->i_state = 1 << NILFS_I_GCINODE;
+		NILFS_I(inode)->i_cno = args->cno;
+		NILFS_I(inode)->i_root = NULL;
+	} else {
+		if (args->root && args->ino == NILFS_ROOT_INO)
+			nilfs_get_root(args->root);
+		NILFS_I(inode)->i_root = args->root;
+	}
+	return 0;
+}
+
+struct inode *nilfs_ilookup(struct super_block *sb, struct nilfs_root *root,
+			    unsigned long ino)
+{
+	struct nilfs_iget_args args = {
+		.ino = ino, .root = root, .cno = 0, .for_gc = 0
+	};
+
+	return ilookup5(sb, ino, nilfs_iget_test, &args);
+}
+
+struct inode *nilfs_iget_locked(struct super_block *sb, struct nilfs_root *root,
+				unsigned long ino)
+{
+	struct nilfs_iget_args args = {
+		.ino = ino, .root = root, .cno = 0, .for_gc = 0
+	};
+
+	return iget5_locked(sb, ino, nilfs_iget_test, nilfs_iget_set, &args);
+}
+
+struct inode *nilfs_iget(struct super_block *sb, struct nilfs_root *root,
+			 unsigned long ino)
 {
 	struct inode *inode;
 	int err;
 
-	inode = iget_locked(sb, ino);
+	inode = nilfs_iget_locked(sb, root, ino);
 	if (unlikely(!inode))
 		return ERR_PTR(-ENOMEM);
 	if (!(inode->i_state & I_NEW))
 		return inode;
 
-	err = __nilfs_read_inode(sb, ino, inode);
+	err = __nilfs_read_inode(sb, root, ino, inode);
+	if (unlikely(err)) {
+		iget_failed(inode);
+		return ERR_PTR(err);
+	}
+	unlock_new_inode(inode);
+	return inode;
+}
+
+struct inode *nilfs_iget_for_gc(struct super_block *sb, unsigned long ino,
+				__u64 cno)
+{
+	struct nilfs_iget_args args = {
+		.ino = ino, .root = NULL, .cno = cno, .for_gc = 1
+	};
+	struct inode *inode;
+	int err;
+
+	inode = iget5_locked(sb, ino, nilfs_iget_test, nilfs_iget_set, &args);
+	if (unlikely(!inode))
+		return ERR_PTR(-ENOMEM);
+	if (!(inode->i_state & I_NEW))
+		return inode;
+
+	err = nilfs_init_gcinode(inode);
 	if (unlikely(err)) {
 		iget_failed(inode);
 		return ERR_PTR(err);
@@ -528,21 +603,20 @@ void nilfs_update_inode(struct inode *inode, struct buffer_head *ibh)
 {
 	ino_t ino = inode->i_ino;
 	struct nilfs_inode_info *ii = NILFS_I(inode);
-	struct super_block *sb = inode->i_sb;
-	struct nilfs_sb_info *sbi = NILFS_SB(sb);
+	struct inode *ifile = ii->i_root->ifile;
 	struct nilfs_inode *raw_inode;
 
-	raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, ibh);
+	raw_inode = nilfs_ifile_map_inode(ifile, ino, ibh);
 
 	if (test_and_clear_bit(NILFS_I_NEW, &ii->i_state))
-		memset(raw_inode, 0, NILFS_MDT(sbi->s_ifile)->mi_entry_size);
+		memset(raw_inode, 0, NILFS_MDT(ifile)->mi_entry_size);
 	set_bit(NILFS_I_INODE_DIRTY, &ii->i_state);
 
 	nilfs_write_inode_common(inode, raw_inode, 0);
 		/* XXX: call with has_bmap = 0 is a workaround to avoid
 		   deadlock of bmap. This delays update of i_bmap to just
 		   before writing */
-	nilfs_ifile_unmap_inode(sbi->s_ifile, ino, ibh);
+	nilfs_ifile_unmap_inode(ifile, ino, ibh);
 }
 
 #define NILFS_MAX_TRUNCATE_BLOCKS	16384  /* 64MB for 4KB block */
@@ -617,6 +691,7 @@ void nilfs_truncate(struct inode *inode)
 static void nilfs_clear_inode(struct inode *inode)
 {
 	struct nilfs_inode_info *ii = NILFS_I(inode);
+	struct nilfs_mdt_info *mdi = NILFS_MDT(inode);
 
 	/*
 	 * Free resources allocated in nilfs_read_inode(), here.
@@ -625,10 +700,16 @@ static void nilfs_clear_inode(struct inode *inode)
 	brelse(ii->i_bh);
 	ii->i_bh = NULL;
 
+	if (mdi && mdi->mi_palloc_cache)
+		nilfs_palloc_destroy_cache(inode);
+
 	if (test_bit(NILFS_I_BMAP, &ii->i_state))
 		nilfs_bmap_clear(ii->i_bmap);
 
 	nilfs_btnode_cache_clear(&ii->i_btnode_cache);
+
+	if (ii->i_root && inode->i_ino == NILFS_ROOT_INO)
+		nilfs_put_root(ii->i_root);
 }
 
 void nilfs_evict_inode(struct inode *inode)
@@ -637,7 +718,7 @@ void nilfs_evict_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	struct nilfs_inode_info *ii = NILFS_I(inode);
 
-	if (inode->i_nlink || unlikely(is_bad_inode(inode))) {
+	if (inode->i_nlink || !ii->i_root || unlikely(is_bad_inode(inode))) {
 		if (inode->i_data.nrpages)
 			truncate_inode_pages(&inode->i_data, 0);
 		end_writeback(inode);
@@ -649,12 +730,16 @@ void nilfs_evict_inode(struct inode *inode)
 	if (inode->i_data.nrpages)
 		truncate_inode_pages(&inode->i_data, 0);
 
+	/* TODO: some of the following operations may fail.  */
 	nilfs_truncate_bmap(ii, 0);
 	nilfs_mark_inode_dirty(inode);
 	end_writeback(inode);
+
+	nilfs_ifile_delete_inode(ii->i_root->ifile, inode->i_ino);
+	atomic_dec(&ii->i_root->inodes_count);
+
 	nilfs_clear_inode(inode);
-	nilfs_free_inode(inode);
-	/* nilfs_free_inode() marks inode buffer dirty */
+
 	if (IS_SYNC(inode))
 		nilfs_set_transaction_flag(NILFS_TI_SYNC);
 	nilfs_transaction_commit(sb);
@@ -700,6 +785,17 @@ out_err:
 	return err;
 }
 
+int nilfs_permission(struct inode *inode, int mask)
+{
+	struct nilfs_root *root = NILFS_I(inode)->i_root;
+
+	if ((mask & MAY_WRITE) && root &&
+	    root->cno != NILFS_CPTREE_CURRENT_CNO)
+		return -EROFS; /* snapshot is not writable */
+
+	return generic_permission(inode, mask, NULL);
+}
+
 int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode,
 			   struct buffer_head **pbh)
 {
@@ -709,8 +805,8 @@ int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode,
 	spin_lock(&sbi->s_inode_lock);
 	if (ii->i_bh == NULL) {
 		spin_unlock(&sbi->s_inode_lock);
-		err = nilfs_ifile_get_inode_block(sbi->s_ifile, inode->i_ino,
-						  pbh);
+		err = nilfs_ifile_get_inode_block(ii->i_root->ifile,
+						  inode->i_ino, pbh);
 		if (unlikely(err))
 			return err;
 		spin_lock(&sbi->s_inode_lock);
@@ -790,7 +886,7 @@ int nilfs_mark_inode_dirty(struct inode *inode)
 	}
 	nilfs_update_inode(inode, ibh);
 	nilfs_mdt_mark_buffer_dirty(ibh);
-	nilfs_mdt_mark_dirty(sbi->s_ifile);
+	nilfs_mdt_mark_dirty(NILFS_I(inode)->i_root->ifile);
 	brelse(ibh);
 	return 0;
 }
@@ -808,6 +904,7 @@ int nilfs_mark_inode_dirty(struct inode *inode)
 void nilfs_dirty_inode(struct inode *inode)
 {
 	struct nilfs_transaction_info ti;
+	struct nilfs_mdt_info *mdi = NILFS_MDT(inode);
 
 	if (is_bad_inode(inode)) {
 		nilfs_warning(inode->i_sb, __func__,
@@ -815,6 +912,10 @@ void nilfs_dirty_inode(struct inode *inode)
 		dump_stack();
 		return;
 	}
+	if (mdi) {
+		nilfs_mdt_mark_dirty(inode);
+		return;
+	}
 	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 	nilfs_mark_inode_dirty(inode);
 	nilfs_transaction_commit(inode->i_sb); /* never fails */

fs/nilfs2/ioctl.c

@@ -117,7 +117,7 @@ static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
 	if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
 		goto out;
 
-	mutex_lock(&nilfs->ns_mount_mutex);
+	down_read(&inode->i_sb->s_umount);
 
 	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 	ret = nilfs_cpfile_change_cpmode(
@@ -127,7 +127,7 @@ static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
 	else
 		nilfs_transaction_commit(inode->i_sb); /* never fails */
 
-	mutex_unlock(&nilfs->ns_mount_mutex);
+	up_read(&inode->i_sb->s_umount);
 out:
 	mnt_drop_write(filp->f_path.mnt);
 	return ret;
@@ -333,7 +333,7 @@ static int nilfs_ioctl_move_inode_block(struct inode *inode,
 	return 0;
 }
 
-static int nilfs_ioctl_move_blocks(struct the_nilfs *nilfs,
+static int nilfs_ioctl_move_blocks(struct super_block *sb,
 				   struct nilfs_argv *argv, void *buf)
 {
 	size_t nmembs = argv->v_nmembs;
@@ -348,7 +348,7 @@ static int nilfs_ioctl_move_blocks(struct the_nilfs *nilfs,
 	for (i = 0, vdesc = buf; i < nmembs; ) {
 		ino = vdesc->vd_ino;
 		cno = vdesc->vd_cno;
-		inode = nilfs_gc_iget(nilfs, ino, cno);
+		inode = nilfs_iget_for_gc(sb, ino, cno);
 		if (unlikely(inode == NULL)) {
 			ret = -ENOMEM;
 			goto failed;
@@ -356,11 +356,15 @@ static int nilfs_ioctl_move_blocks(struct the_nilfs *nilfs,
 		do {
 			ret = nilfs_ioctl_move_inode_block(inode, vdesc,
 							   &buffers);
-			if (unlikely(ret < 0))
+			if (unlikely(ret < 0)) {
+				iput(inode);
 				goto failed;
+			}
 			vdesc++;
 		} while (++i < nmembs &&
 			 vdesc->vd_ino == ino && vdesc->vd_cno == cno);
+
+		iput(inode); /* The inode still remains in GC inode list */
 	}
 
 	list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
@@ -566,7 +570,7 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 	}
 
 	/*
-	 * nilfs_ioctl_move_blocks() will call nilfs_gc_iget(),
+	 * nilfs_ioctl_move_blocks() will call nilfs_iget_for_gc(),
 	 * which will operates an inode list without blocking.
 	 * To protect the list from concurrent operations,
 	 * nilfs_ioctl_move_blocks should be atomic operation.
@@ -576,15 +580,16 @@ static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 		goto out_free;
 	}
 
-	ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
+	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
+
+	ret = nilfs_ioctl_move_blocks(inode->i_sb, &argv[0], kbufs[0]);
 	if (ret < 0)
 		printk(KERN_ERR "NILFS: GC failed during preparation: "
 			"cannot read source blocks: err=%d\n", ret);
 	else
 		ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
 
-	if (ret < 0)
-		nilfs_remove_all_gcinode(nilfs);
+	nilfs_remove_all_gcinodes(nilfs);
 	clear_nilfs_gc_running(nilfs);
 
 out_free:

fs/nilfs2/mdt.c

@@ -36,7 +36,6 @@
 
 #define NILFS_MDT_MAX_RA_BLOCKS		(16 - 1)
 
-#define INIT_UNUSED_INODE_FIELDS
 
 static int
 nilfs_mdt_insert_new_block(struct inode *inode, unsigned long block,
@@ -78,25 +77,11 @@ static int nilfs_mdt_create_block(struct inode *inode, unsigned long block,
 						     struct buffer_head *,
 						     void *))
 {
-	struct the_nilfs *nilfs = NILFS_MDT(inode)->mi_nilfs;
 	struct super_block *sb = inode->i_sb;
 	struct nilfs_transaction_info ti;
 	struct buffer_head *bh;
 	int err;
 
-	if (!sb) {
-		/*
-		 * Make sure this function is not called from any
-		 * read-only context.
-		 */
-		if (!nilfs->ns_writer) {
-			WARN_ON(1);
-			err = -EROFS;
-			goto out;
-		}
-		sb = nilfs->ns_writer->s_super;
-	}
-
 	nilfs_transaction_begin(sb, &ti, 0);
 
 	err = -ENOMEM;
@@ -112,7 +97,7 @@ static int nilfs_mdt_create_block(struct inode *inode, unsigned long block,
 	if (buffer_uptodate(bh))
 		goto failed_bh;
 
-	bh->b_bdev = nilfs->ns_bdev;
+	bh->b_bdev = sb->s_bdev;
 	err = nilfs_mdt_insert_new_block(inode, block, bh, init_block);
 	if (likely(!err)) {
 		get_bh(bh);
@@ -129,7 +114,7 @@ static int nilfs_mdt_create_block(struct inode *inode, unsigned long block,
 		err = nilfs_transaction_commit(sb);
 	else
 		nilfs_transaction_abort(sb);
- out:
+
 	return err;
 }
 
@@ -167,9 +152,7 @@ nilfs_mdt_submit_block(struct inode *inode, unsigned long blkoff,
 		unlock_buffer(bh);
 		goto failed_bh;
 	}
-	bh->b_bdev = NILFS_MDT(inode)->mi_nilfs->ns_bdev;
-	bh->b_blocknr = (sector_t)blknum;
-	set_buffer_mapped(bh);
+	map_bh(bh, inode->i_sb, (sector_t)blknum);
 
 	bh->b_end_io = end_buffer_read_sync;
 	get_bh(bh);
@@ -398,35 +381,24 @@ int nilfs_mdt_fetch_dirty(struct inode *inode)
 static int
 nilfs_mdt_write_page(struct page *page, struct writeback_control *wbc)
 {
-	struct inode *inode = container_of(page->mapping,
-					   struct inode, i_data);
-	struct super_block *sb = inode->i_sb;
-	struct the_nilfs *nilfs = NILFS_MDT(inode)->mi_nilfs;
-	struct nilfs_sb_info *writer = NULL;
+	struct inode *inode;
+	struct super_block *sb;
 	int err = 0;
 
 	redirty_page_for_writepage(wbc, page);
 	unlock_page(page);
 
-	if (page->mapping->assoc_mapping)
-		return 0; /* Do not request flush for shadow page cache */
-	if (!sb) {
-		down_read(&nilfs->ns_writer_sem);
-		writer = nilfs->ns_writer;
-		if (!writer) {
-			up_read(&nilfs->ns_writer_sem);
-			return -EROFS;
-		}
-		sb = writer->s_super;
-	}
+	inode = page->mapping->host;
+	if (!inode)
+		return 0;
+
+	sb = inode->i_sb;
 
 	if (wbc->sync_mode == WB_SYNC_ALL)
 		err = nilfs_construct_segment(sb);
 	else if (wbc->for_reclaim)
 		nilfs_flush_segment(sb, inode->i_ino);
 
-	if (writer)
-		up_read(&nilfs->ns_writer_sem);
 	return err;
 }
 
@@ -439,105 +411,27 @@ static const struct address_space_operations def_mdt_aops = {
 static const struct inode_operations def_mdt_iops;
 static const struct file_operations def_mdt_fops;
 
-/*
- * NILFS2 uses pseudo inodes for meta data files such as DAT, cpfile, sufile,
- * ifile, or gcinodes.  This allows the B-tree code and segment constructor
- * to treat them like regular files, and this helps to simplify the
- * implementation.
- *   On the other hand, some of the pseudo inodes have an irregular point:
- * They don't have valid inode->i_sb pointer because their lifetimes are
- * longer than those of the super block structs; they may continue for
- * several consecutive mounts/umounts.  This would need discussions.
- */
-/**
- * nilfs_mdt_new_common - allocate a pseudo inode for metadata file
- * @nilfs: nilfs object
- * @sb: super block instance the metadata file belongs to
- * @ino: inode number
- * @gfp_mask: gfp mask for data pages
- * @objsz: size of the private object attached to inode->i_private
- */
-struct inode *
-nilfs_mdt_new_common(struct the_nilfs *nilfs, struct super_block *sb,
-		     ino_t ino, gfp_t gfp_mask, size_t objsz)
+
+int nilfs_mdt_init(struct inode *inode, gfp_t gfp_mask, size_t objsz)
 {
-	struct inode *inode = nilfs_alloc_inode_common(nilfs);
+	struct nilfs_mdt_info *mi;
 
-	if (!inode)
-		return NULL;
-	else {
-		struct address_space * const mapping = &inode->i_data;
-		struct nilfs_mdt_info *mi;
-
-		mi = kzalloc(max(sizeof(*mi), objsz), GFP_NOFS);
-		if (!mi) {
-			nilfs_destroy_inode(inode);
-			return NULL;
-		}
-		mi->mi_nilfs = nilfs;
-		init_rwsem(&mi->mi_sem);
-
-		inode->i_sb = sb; /* sb may be NULL for some meta data files */
-		inode->i_blkbits = nilfs->ns_blocksize_bits;
-		inode->i_flags = 0;
-		atomic_set(&inode->i_count, 1);
-		inode->i_nlink = 1;
-		inode->i_ino = ino;
-		inode->i_mode = S_IFREG;
-		inode->i_private = mi;
-
-#ifdef INIT_UNUSED_INODE_FIELDS
-		atomic_set(&inode->i_writecount, 0);
-		inode->i_size = 0;
-		inode->i_blocks = 0;
-		inode->i_bytes = 0;
-		inode->i_generation = 0;
-#ifdef CONFIG_QUOTA
-		memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
-#endif
-		inode->i_pipe = NULL;
-		inode->i_bdev = NULL;
-		inode->i_cdev = NULL;
-		inode->i_rdev = 0;
-#ifdef CONFIG_SECURITY
-		inode->i_security = NULL;
-#endif
-		inode->dirtied_when = 0;
-
-		INIT_LIST_HEAD(&inode->i_list);
-		INIT_LIST_HEAD(&inode->i_sb_list);
-		inode->i_state = 0;
-#endif
-
-		spin_lock_init(&inode->i_lock);
-		mutex_init(&inode->i_mutex);
-		init_rwsem(&inode->i_alloc_sem);
-
-		mapping->host = NULL;  /* instead of inode */
-		mapping->flags = 0;
-		mapping_set_gfp_mask(mapping, gfp_mask);
-		mapping->assoc_mapping = NULL;
-		mapping->backing_dev_info = nilfs->ns_bdi;
-
-		inode->i_mapping = mapping;
-	}
+	mi = kzalloc(max(sizeof(*mi), objsz), GFP_NOFS);
+	if (!mi)
+		return -ENOMEM;
 
-	return inode;
-}
+	init_rwsem(&mi->mi_sem);
+	inode->i_private = mi;
 
-struct inode *nilfs_mdt_new(struct the_nilfs *nilfs, struct super_block *sb,
-			    ino_t ino, size_t objsz)
-{
-	struct inode *inode;
-
-	inode = nilfs_mdt_new_common(nilfs, sb, ino, NILFS_MDT_GFP, objsz);
-	if (!inode)
-		return NULL;
+	inode->i_mode = S_IFREG;
+	mapping_set_gfp_mask(inode->i_mapping, gfp_mask);
+	inode->i_mapping->backing_dev_info = inode->i_sb->s_bdi;
 
 	inode->i_op = &def_mdt_iops;
 	inode->i_fop = &def_mdt_fops;
 	inode->i_mapping->a_ops = &def_mdt_aops;
-	return inode;
+
+	return 0;
 }
 
 void nilfs_mdt_set_entry_size(struct inode *inode, unsigned entry_size,
@@ -550,34 +444,159 @@ void nilfs_mdt_set_entry_size(struct inode *inode, unsigned entry_size,
 	mi->mi_first_entry_offset = DIV_ROUND_UP(header_size, entry_size);
 }
 
-void nilfs_mdt_set_shadow(struct inode *orig, struct inode *shadow)
+static const struct address_space_operations shadow_map_aops = {
+	.sync_page		= block_sync_page,
+};
+
+/**
+ * nilfs_mdt_setup_shadow_map - setup shadow map and bind it to metadata file
+ * @inode: inode of the metadata file
+ * @shadow: shadow mapping
+ */
+int nilfs_mdt_setup_shadow_map(struct inode *inode,
+			       struct nilfs_shadow_map *shadow)
 {
-	shadow->i_mapping->assoc_mapping = orig->i_mapping;
-	NILFS_I(shadow)->i_btnode_cache.assoc_mapping =
-		&NILFS_I(orig)->i_btnode_cache;
+	struct nilfs_mdt_info *mi = NILFS_MDT(inode);
+	struct backing_dev_info *bdi = inode->i_sb->s_bdi;
+
+	INIT_LIST_HEAD(&shadow->frozen_buffers);
+	nilfs_mapping_init_once(&shadow->frozen_data);
+	nilfs_mapping_init(&shadow->frozen_data, bdi, &shadow_map_aops);
+	nilfs_mapping_init_once(&shadow->frozen_btnodes);
+	nilfs_mapping_init(&shadow->frozen_btnodes, bdi, &shadow_map_aops);
+	mi->mi_shadow = shadow;
+	return 0;
 }
 
-static void nilfs_mdt_clear(struct inode *inode)
+/**
+ * nilfs_mdt_save_to_shadow_map - copy bmap and dirty pages to shadow map
+ * @inode: inode of the metadata file
+ */
+int nilfs_mdt_save_to_shadow_map(struct inode *inode)
 {
+	struct nilfs_mdt_info *mi = NILFS_MDT(inode);
 	struct nilfs_inode_info *ii = NILFS_I(inode);
+	struct nilfs_shadow_map *shadow = mi->mi_shadow;
+	int ret;
 
-	invalidate_mapping_pages(inode->i_mapping, 0, -1);
-	truncate_inode_pages(inode->i_mapping, 0);
+	ret = nilfs_copy_dirty_pages(&shadow->frozen_data, inode->i_mapping);
+	if (ret)
+		goto out;
+
+	ret = nilfs_copy_dirty_pages(&shadow->frozen_btnodes,
+				     &ii->i_btnode_cache);
+	if (ret)
+		goto out;
 
-	if (test_bit(NILFS_I_BMAP, &ii->i_state))
-		nilfs_bmap_clear(ii->i_bmap);
-	nilfs_btnode_cache_clear(&ii->i_btnode_cache);
+	nilfs_bmap_save(ii->i_bmap, &shadow->bmap_store);
+ out:
+	return ret;
 }
 
-void nilfs_mdt_destroy(struct inode *inode)
+int nilfs_mdt_freeze_buffer(struct inode *inode, struct buffer_head *bh)
 {
-	struct nilfs_mdt_info *mdi = NILFS_MDT(inode);
+	struct nilfs_shadow_map *shadow = NILFS_MDT(inode)->mi_shadow;
+	struct buffer_head *bh_frozen;
+	struct page *page;
+	int blkbits = inode->i_blkbits;
+	int ret = -ENOMEM;
+
+	page = grab_cache_page(&shadow->frozen_data, bh->b_page->index);
+	if (!page)
+		return ret;
+
+	if (!page_has_buffers(page))
+		create_empty_buffers(page, 1 << blkbits, 0);
+
+	bh_frozen = nilfs_page_get_nth_block(page, bh_offset(bh) >> blkbits);
+	if (bh_frozen) {
+		if (!buffer_uptodate(bh_frozen))
+			nilfs_copy_buffer(bh_frozen, bh);
+		if (list_empty(&bh_frozen->b_assoc_buffers)) {
+			list_add_tail(&bh_frozen->b_assoc_buffers,
+				      &shadow->frozen_buffers);
+			set_buffer_nilfs_redirected(bh);
+		} else {
+			brelse(bh_frozen); /* already frozen */
+		}
+		ret = 0;
+	}
+	unlock_page(page);
+	page_cache_release(page);
+	return ret;
+}
+
+struct buffer_head *
+nilfs_mdt_get_frozen_buffer(struct inode *inode, struct buffer_head *bh)
+{
+	struct nilfs_shadow_map *shadow = NILFS_MDT(inode)->mi_shadow;
+	struct buffer_head *bh_frozen = NULL;
+	struct page *page;
+	int n;
+
+	page = find_lock_page(&shadow->frozen_data, bh->b_page->index);
+	if (page) {
+		if (page_has_buffers(page)) {
+			n = bh_offset(bh) >> inode->i_blkbits;
+			bh_frozen = nilfs_page_get_nth_block(page, n);
+		}
+		unlock_page(page);
+		page_cache_release(page);
+	}
+	return bh_frozen;
+}
+
+static void nilfs_release_frozen_buffers(struct nilfs_shadow_map *shadow)
+{
+	struct list_head *head = &shadow->frozen_buffers;
+	struct buffer_head *bh;
+
+	while (!list_empty(head)) {
+		bh = list_first_entry(head, struct buffer_head,
+				      b_assoc_buffers);
+		list_del_init(&bh->b_assoc_buffers);
+		brelse(bh); /* drop ref-count to make it releasable */
+	}
+}
+
+/**
+ * nilfs_mdt_restore_from_shadow_map - restore dirty pages and bmap state
+ * @inode: inode of the metadata file
+ */
+void nilfs_mdt_restore_from_shadow_map(struct inode *inode)
+{
+	struct nilfs_mdt_info *mi = NILFS_MDT(inode);
+	struct nilfs_inode_info *ii = NILFS_I(inode);
+	struct nilfs_shadow_map *shadow = mi->mi_shadow;
+
+	down_write(&mi->mi_sem);
 
-	if (mdi->mi_palloc_cache)
-		nilfs_palloc_destroy_cache(inode);
-	nilfs_mdt_clear(inode);
+	if (mi->mi_palloc_cache)
+		nilfs_palloc_clear_cache(inode);
+
+	nilfs_clear_dirty_pages(inode->i_mapping);
+	nilfs_copy_back_pages(inode->i_mapping, &shadow->frozen_data);
+
+	nilfs_clear_dirty_pages(&ii->i_btnode_cache);
+	nilfs_copy_back_pages(&ii->i_btnode_cache, &shadow->frozen_btnodes);
+
+	nilfs_bmap_restore(ii->i_bmap, &shadow->bmap_store);
+
+	up_write(&mi->mi_sem);
+}
+
+/**
+ * nilfs_mdt_clear_shadow_map - truncate pages in shadow map caches
+ * @inode: inode of the metadata file
+ */
+void nilfs_mdt_clear_shadow_map(struct inode *inode)
+{
+	struct nilfs_mdt_info *mi = NILFS_MDT(inode);
+	struct nilfs_shadow_map *shadow = mi->mi_shadow;
 
-	kfree(mdi->mi_bgl); /* kfree(NULL) is safe */
-	kfree(mdi);
-	nilfs_destroy_inode(inode);
+	down_write(&mi->mi_sem);
+	nilfs_release_frozen_buffers(shadow);
+	truncate_inode_pages(&shadow->frozen_data, 0);
+	truncate_inode_pages(&shadow->frozen_btnodes, 0);
+	up_write(&mi->mi_sem);
 }

fs/nilfs2/mdt.h

@@ -28,26 +28,33 @@
 #include "nilfs.h"
 #include "page.h"
 
+struct nilfs_shadow_map {
+	struct nilfs_bmap_store bmap_store;
+	struct address_space frozen_data;
+	struct address_space frozen_btnodes;
+	struct list_head frozen_buffers;
+};
+
 /**
  * struct nilfs_mdt_info - on-memory private data of meta data files
- * @mi_nilfs: back pointer to the_nilfs struct
  * @mi_sem: reader/writer semaphore for meta data operations
  * @mi_bgl: per-blockgroup locking
  * @mi_entry_size: size of an entry
  * @mi_first_entry_offset: offset to the first entry
  * @mi_entries_per_block: number of entries in a block
  * @mi_palloc_cache: persistent object allocator cache
+ * @mi_shadow: shadow of bmap and page caches
  * @mi_blocks_per_group: number of blocks in a group
  * @mi_blocks_per_desc_block: number of blocks per descriptor block
  */
 struct nilfs_mdt_info {
-	struct the_nilfs       *mi_nilfs;
 	struct rw_semaphore	mi_sem;
 	struct blockgroup_lock *mi_bgl;
 	unsigned		mi_entry_size;
 	unsigned		mi_first_entry_offset;
 	unsigned long		mi_entries_per_block;
 	struct nilfs_palloc_cache *mi_palloc_cache;
+	struct nilfs_shadow_map *mi_shadow;
 	unsigned long		mi_blocks_per_group;
 	unsigned long		mi_blocks_per_desc_block;
 };
@@ -59,9 +66,7 @@ static inline struct nilfs_mdt_info *NILFS_MDT(const struct inode *inode)
 
 static inline struct the_nilfs *NILFS_I_NILFS(struct inode *inode)
 {
-	struct super_block *sb = inode->i_sb;
-
-	return sb ? NILFS_SB(sb)->s_nilfs : NILFS_MDT(inode)->mi_nilfs;
+	return NILFS_SB(inode->i_sb)->s_nilfs;
 }
 
 /* Default GFP flags using highmem */
@@ -76,14 +81,17 @@ int nilfs_mdt_forget_block(struct inode *, unsigned long);
 int nilfs_mdt_mark_block_dirty(struct inode *, unsigned long);
 int nilfs_mdt_fetch_dirty(struct inode *);
 
-struct inode *nilfs_mdt_new(struct the_nilfs *, struct super_block *, ino_t,
-			    size_t);
-struct inode *nilfs_mdt_new_common(struct the_nilfs *, struct super_block *,
-				   ino_t, gfp_t, size_t);
-void nilfs_mdt_destroy(struct inode *);
+int nilfs_mdt_init(struct inode *inode, gfp_t gfp_mask, size_t objsz);
 void nilfs_mdt_set_entry_size(struct inode *, unsigned, unsigned);
-void nilfs_mdt_set_shadow(struct inode *, struct inode *);
 
+int nilfs_mdt_setup_shadow_map(struct inode *inode,
+			       struct nilfs_shadow_map *shadow);
+int nilfs_mdt_save_to_shadow_map(struct inode *inode);
+void nilfs_mdt_restore_from_shadow_map(struct inode *inode);
+void nilfs_mdt_clear_shadow_map(struct inode *inode);
+int nilfs_mdt_freeze_buffer(struct inode *inode, struct buffer_head *bh);
+struct buffer_head *nilfs_mdt_get_frozen_buffer(struct inode *inode,
+						struct buffer_head *bh);
 
 #define nilfs_mdt_mark_buffer_dirty(bh)	nilfs_mark_buffer_dirty(bh)
 
@@ -100,7 +108,7 @@ static inline void nilfs_mdt_clear_dirty(struct inode *inode)
 
 static inline __u64 nilfs_mdt_cno(struct inode *inode)
 {
-	return NILFS_MDT(inode)->mi_nilfs->ns_cno;
+	return NILFS_I_NILFS(inode)->ns_cno;
 }
 
 #define nilfs_mdt_bgl_lock(inode, bg) \

fs/nilfs2/namei.c

@@ -40,7 +40,11 @@
 
 #include <linux/pagemap.h>
 #include "nilfs.h"
+#include "export.h"
 
+#define NILFS_FID_SIZE_NON_CONNECTABLE \
+	(offsetof(struct nilfs_fid, parent_gen) / 4)
+#define NILFS_FID_SIZE_CONNECTABLE	(sizeof(struct nilfs_fid) / 4)
 
 static inline int nilfs_add_nondir(struct dentry *dentry, struct inode *inode)
 {
@@ -70,29 +74,13 @@ nilfs_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
 	ino = nilfs_inode_by_name(dir, &dentry->d_name);
 	inode = NULL;
 	if (ino) {
-		inode = nilfs_iget(dir->i_sb, ino);
+		inode = nilfs_iget(dir->i_sb, NILFS_I(dir)->i_root, ino);
 		if (IS_ERR(inode))
 			return ERR_CAST(inode);
 	}
 	return d_splice_alias(inode, dentry);
 }
 
-struct dentry *nilfs_get_parent(struct dentry *child)
-{
-	unsigned long ino;
-	struct inode *inode;
-	struct qstr dotdot = {.name = "..", .len = 2};
-
-	ino = nilfs_inode_by_name(child->d_inode, &dotdot);
-	if (!ino)
-		return ERR_PTR(-ENOENT);
-
-	inode = nilfs_iget(child->d_inode->i_sb, ino);
-	if (IS_ERR(inode))
-		return ERR_CAST(inode);
-	return d_obtain_alias(inode);
-}
-
 /*
  * By the time this is called, we already have created
  * the directory cache entry for the new file, but it
@@ -468,6 +456,115 @@ out:
 	return err;
 }
 
+/*
+ * Export operations
+ */
+static struct dentry *nilfs_get_parent(struct dentry *child)
+{
+	unsigned long ino;
+	struct inode *inode;
+	struct qstr dotdot = {.name = "..", .len = 2};
+	struct nilfs_root *root;
+
+	ino = nilfs_inode_by_name(child->d_inode, &dotdot);
+	if (!ino)
+		return ERR_PTR(-ENOENT);
+
+	root = NILFS_I(child->d_inode)->i_root;
+
+	inode = nilfs_iget(child->d_inode->i_sb, root, ino);
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+
+	return d_obtain_alias(inode);
+}
+
+static struct dentry *nilfs_get_dentry(struct super_block *sb, u64 cno,
+				       u64 ino, u32 gen)
+{
+	struct nilfs_root *root;
+	struct inode *inode;
+
+	if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO)
+		return ERR_PTR(-ESTALE);
+
+	root = nilfs_lookup_root(NILFS_SB(sb)->s_nilfs, cno);
+	if (!root)
+		return ERR_PTR(-ESTALE);
+
+	inode = nilfs_iget(sb, root, ino);
+	nilfs_put_root(root);
+
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+	if (gen && inode->i_generation != gen) {
+		iput(inode);
+		return ERR_PTR(-ESTALE);
+	}
+	return d_obtain_alias(inode);
+}
+
+static struct dentry *nilfs_fh_to_dentry(struct super_block *sb, struct fid *fh,
+					 int fh_len, int fh_type)
+{
+	struct nilfs_fid *fid = (struct nilfs_fid *)fh;
+
+	if ((fh_len != NILFS_FID_SIZE_NON_CONNECTABLE &&
+	     fh_len != NILFS_FID_SIZE_CONNECTABLE) ||
+	    (fh_type != FILEID_NILFS_WITH_PARENT &&
+	     fh_type != FILEID_NILFS_WITHOUT_PARENT))
+		return NULL;
+
+	return nilfs_get_dentry(sb, fid->cno, fid->ino, fid->gen);
+}
+
+static struct dentry *nilfs_fh_to_parent(struct super_block *sb, struct fid *fh,
+					 int fh_len, int fh_type)
+{
+	struct nilfs_fid *fid = (struct nilfs_fid *)fh;
+
+	if (fh_len != NILFS_FID_SIZE_CONNECTABLE ||
+	    fh_type != FILEID_NILFS_WITH_PARENT)
+		return NULL;
+
+	return nilfs_get_dentry(sb, fid->cno, fid->parent_ino, fid->parent_gen);
+}
+
+static int nilfs_encode_fh(struct dentry *dentry, __u32 *fh, int *lenp,
+			   int connectable)
+{
+	struct nilfs_fid *fid = (struct nilfs_fid *)fh;
+	struct inode *inode = dentry->d_inode;
+	struct nilfs_root *root = NILFS_I(inode)->i_root;
+	int type;
+
+	if (*lenp < NILFS_FID_SIZE_NON_CONNECTABLE ||
+	    (connectable && *lenp < NILFS_FID_SIZE_CONNECTABLE))
+		return 255;
+
+	fid->cno = root->cno;
+	fid->ino = inode->i_ino;
+	fid->gen = inode->i_generation;
+
+	if (connectable && !S_ISDIR(inode->i_mode)) {
+		struct inode *parent;
+
+		spin_lock(&dentry->d_lock);
+		parent = dentry->d_parent->d_inode;
+		fid->parent_ino = parent->i_ino;
+		fid->parent_gen = parent->i_generation;
+		spin_unlock(&dentry->d_lock);
+
+		type = FILEID_NILFS_WITH_PARENT;
+		*lenp = NILFS_FID_SIZE_CONNECTABLE;
+	} else {
+		type = FILEID_NILFS_WITHOUT_PARENT;
+		*lenp = NILFS_FID_SIZE_NON_CONNECTABLE;
+	}
+
+	return type;
+}
+
 const struct inode_operations nilfs_dir_inode_operations = {
 	.create		= nilfs_create,
 	.lookup		= nilfs_lookup,
@@ -491,4 +588,12 @@ const struct inode_operations nilfs_symlink_inode_operations = {
 	.readlink	= generic_readlink,
 	.follow_link	= page_follow_link_light,
 	.put_link	= page_put_link,
+	.permission     = nilfs_permission,
+};
+
+const struct export_operations nilfs_export_ops = {
+	.encode_fh = nilfs_encode_fh,
+	.fh_to_dentry = nilfs_fh_to_dentry,
+	.fh_to_parent = nilfs_fh_to_parent,
+	.get_parent = nilfs_get_parent,
 };

fs/nilfs2/nilfs.h

@@ -59,6 +59,7 @@ struct nilfs_inode_info {
 #endif
 	struct buffer_head *i_bh;	/* i_bh contains a new or dirty
 					   disk inode */
+	struct nilfs_root *i_root;
 	struct inode vfs_inode;
 };
 
@@ -100,7 +101,6 @@ enum {
 	NILFS_I_INODE_DIRTY,		/* write_inode is requested */
 	NILFS_I_BMAP,			/* has bmap and btnode_cache */
 	NILFS_I_GCINODE,		/* inode for GC, on memory only */
-	NILFS_I_GCDAT,			/* shadow DAT, on memory only */
 };
 
 /*
@@ -192,7 +192,7 @@ static inline int nilfs_doing_construction(void)
 
 static inline struct inode *nilfs_dat_inode(const struct the_nilfs *nilfs)
 {
-	return nilfs_doing_gc() ? nilfs->ns_gc_dat : nilfs->ns_dat;
+	return nilfs->ns_dat;
 }
 
 /*
@@ -200,12 +200,9 @@ static inline struct inode *nilfs_dat_inode(const struct the_nilfs *nilfs)
  */
 #ifdef CONFIG_NILFS_POSIX_ACL
 #error "NILFS: not yet supported POSIX ACL"
-extern int nilfs_permission(struct inode *, int, struct nameidata *);
 extern int nilfs_acl_chmod(struct inode *);
 extern int nilfs_init_acl(struct inode *, struct inode *);
 #else
-#define nilfs_permission   NULL
-
 static inline int nilfs_acl_chmod(struct inode *inode)
 {
 	return 0;
@@ -247,11 +244,19 @@ extern int nilfs_get_block(struct inode *, sector_t, struct buffer_head *, int);
 extern void nilfs_set_inode_flags(struct inode *);
 extern int nilfs_read_inode_common(struct inode *, struct nilfs_inode *);
 extern void nilfs_write_inode_common(struct inode *, struct nilfs_inode *, int);
-extern struct inode *nilfs_iget(struct super_block *, unsigned long);
+struct inode *nilfs_ilookup(struct super_block *sb, struct nilfs_root *root,
+			    unsigned long ino);
+struct inode *nilfs_iget_locked(struct super_block *sb, struct nilfs_root *root,
+				unsigned long ino);
+struct inode *nilfs_iget(struct super_block *sb, struct nilfs_root *root,
+			 unsigned long ino);
+extern struct inode *nilfs_iget_for_gc(struct super_block *sb,
+				       unsigned long ino, __u64 cno);
 extern void nilfs_update_inode(struct inode *, struct buffer_head *);
 extern void nilfs_truncate(struct inode *);
 extern void nilfs_evict_inode(struct inode *);
 extern int nilfs_setattr(struct dentry *, struct iattr *);
+int nilfs_permission(struct inode *inode, int mask);
 extern int nilfs_load_inode_block(struct nilfs_sb_info *, struct inode *,
 				  struct buffer_head **);
 extern int nilfs_inode_dirty(struct inode *);
@@ -260,11 +265,7 @@ extern int nilfs_set_file_dirty(struct nilfs_sb_info *, struct inode *,
 extern int nilfs_mark_inode_dirty(struct inode *);
 extern void nilfs_dirty_inode(struct inode *);
 
-/* namei.c */
-extern struct dentry *nilfs_get_parent(struct dentry *);
-
 /* super.c */
-extern struct inode *nilfs_alloc_inode_common(struct the_nilfs *);
 extern struct inode *nilfs_alloc_inode(struct super_block *);
 extern void nilfs_destroy_inode(struct inode *);
 extern void nilfs_error(struct super_block *, const char *, const char *, ...)
@@ -283,8 +284,9 @@ extern struct nilfs_super_block **nilfs_prepare_super(struct nilfs_sb_info *,
 						      int flip);
 extern int nilfs_commit_super(struct nilfs_sb_info *, int);
 extern int nilfs_cleanup_super(struct nilfs_sb_info *);
-extern int nilfs_attach_checkpoint(struct nilfs_sb_info *, __u64);
-extern void nilfs_detach_checkpoint(struct nilfs_sb_info *);
+int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno, int curr_mnt,
+			    struct nilfs_root **root);
+int nilfs_checkpoint_is_mounted(struct super_block *sb, __u64 cno);
 
 /* gcinode.c */
 int nilfs_gccache_submit_read_data(struct inode *, sector_t, sector_t, __u64,
@@ -292,16 +294,8 @@ int nilfs_gccache_submit_read_data(struct inode *, sector_t, sector_t, __u64,
 int nilfs_gccache_submit_read_node(struct inode *, sector_t, __u64,
 				   struct buffer_head **);
 int nilfs_gccache_wait_and_mark_dirty(struct buffer_head *);
-int nilfs_init_gccache(struct the_nilfs *);
-void nilfs_destroy_gccache(struct the_nilfs *);
-void nilfs_clear_gcinode(struct inode *);
-struct inode *nilfs_gc_iget(struct the_nilfs *, ino_t, __u64);
-void nilfs_remove_all_gcinode(struct the_nilfs *);
-
-/* gcdat.c */
-int nilfs_init_gcdat_inode(struct the_nilfs *);
-void nilfs_commit_gcdat_inode(struct the_nilfs *);
-void nilfs_clear_gcdat_inode(struct the_nilfs *);
+int nilfs_init_gcinode(struct inode *inode);
+void nilfs_remove_all_gcinodes(struct the_nilfs *nilfs);
 
 /*
  * Inodes and files operations

fs/nilfs2/page.c

@@ -79,8 +79,8 @@ struct buffer_head *nilfs_grab_buffer(struct inode *inode,
 {
 	int blkbits = inode->i_blkbits;
 	pgoff_t index = blkoff >> (PAGE_CACHE_SHIFT - blkbits);
-	struct page *page, *opage;
-	struct buffer_head *bh, *obh;
+	struct page *page;
+	struct buffer_head *bh;
 
 	page = grab_cache_page(mapping, index);
 	if (unlikely(!page))
@@ -92,30 +92,6 @@ struct buffer_head *nilfs_grab_buffer(struct inode *inode,
 		page_cache_release(page);
 		return NULL;
 	}
-	if (!buffer_uptodate(bh) && mapping->assoc_mapping != NULL) {
-		/*
-		 * Shadow page cache uses assoc_mapping to point its original
-		 * page cache.  The following code tries the original cache
-		 * if the given cache is a shadow and it didn't hit.
-		 */
-		opage = find_lock_page(mapping->assoc_mapping, index);
-		if (!opage)
-			return bh;
-
-		obh = __nilfs_get_page_block(opage, blkoff, index, blkbits,
-					     b_state);
-		if (buffer_uptodate(obh)) {
-			nilfs_copy_buffer(bh, obh);
-			if (buffer_dirty(obh)) {
-				nilfs_mark_buffer_dirty(bh);
-				if (!buffer_nilfs_node(bh) && NILFS_MDT(inode))
-					nilfs_mdt_mark_dirty(inode);
-			}
-		}
-		brelse(obh);
-		unlock_page(opage);
-		page_cache_release(opage);
-	}
 	return bh;
 }
 
@@ -131,6 +107,7 @@ void nilfs_forget_buffer(struct buffer_head *bh)
 	lock_buffer(bh);
 	clear_buffer_nilfs_volatile(bh);
 	clear_buffer_nilfs_checked(bh);
+	clear_buffer_nilfs_redirected(bh);
 	clear_buffer_dirty(bh);
 	if (nilfs_page_buffers_clean(page))
 		__nilfs_clear_page_dirty(page);
@@ -483,6 +460,7 @@ void nilfs_clear_dirty_pages(struct address_space *mapping)
 				clear_buffer_dirty(bh);
 				clear_buffer_nilfs_volatile(bh);
 				clear_buffer_nilfs_checked(bh);
+				clear_buffer_nilfs_redirected(bh);
 				clear_buffer_uptodate(bh);
 				clear_buffer_mapped(bh);
 				unlock_buffer(bh);
@@ -513,6 +491,31 @@ unsigned nilfs_page_count_clean_buffers(struct page *page,
 	}
 	return nc;
 }
+ 
+void nilfs_mapping_init_once(struct address_space *mapping)
+{
+	memset(mapping, 0, sizeof(*mapping));
+	INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
+	spin_lock_init(&mapping->tree_lock);
+	INIT_LIST_HEAD(&mapping->private_list);
+	spin_lock_init(&mapping->private_lock);
+
+	spin_lock_init(&mapping->i_mmap_lock);
+	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
+	INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
+}
+
+void nilfs_mapping_init(struct address_space *mapping,
+			struct backing_dev_info *bdi,
+			const struct address_space_operations *aops)
+{
+	mapping->host = NULL;
+	mapping->flags = 0;
+	mapping_set_gfp_mask(mapping, GFP_NOFS);
+	mapping->assoc_mapping = NULL;
+	mapping->backing_dev_info = bdi;
+	mapping->a_ops = aops;
+}
 
 /*
  * NILFS2 needs clear_page_dirty() in the following two cases:

fs/nilfs2/page.h

@@ -35,12 +35,14 @@ enum {
 	BH_NILFS_Node,
 	BH_NILFS_Volatile,
 	BH_NILFS_Checked,
+	BH_NILFS_Redirected,
 };
 
 BUFFER_FNS(NILFS_Allocated, nilfs_allocated)	/* nilfs private buffers */
 BUFFER_FNS(NILFS_Node, nilfs_node)		/* nilfs node buffers */
 BUFFER_FNS(NILFS_Volatile, nilfs_volatile)
 BUFFER_FNS(NILFS_Checked, nilfs_checked)	/* buffer is verified */
+BUFFER_FNS(NILFS_Redirected, nilfs_redirected)	/* redirected to a copy */
 
 
 void nilfs_mark_buffer_dirty(struct buffer_head *bh);
@@ -59,6 +61,10 @@ void nilfs_free_private_page(struct page *);
 int nilfs_copy_dirty_pages(struct address_space *, struct address_space *);
 void nilfs_copy_back_pages(struct address_space *, struct address_space *);
 void nilfs_clear_dirty_pages(struct address_space *);
+void nilfs_mapping_init_once(struct address_space *mapping);
+void nilfs_mapping_init(struct address_space *mapping,
+			struct backing_dev_info *bdi,
+			const struct address_space_operations *aops);
 unsigned nilfs_page_count_clean_buffers(struct page *, unsigned, unsigned);
 
 #define NILFS_PAGE_BUG(page, m, a...) \

fs/nilfs2/recovery.c

@@ -440,7 +440,6 @@ static int nilfs_prepare_segment_for_recovery(struct the_nilfs *nilfs,
 	segnum[2] = ri->ri_segnum;
 	segnum[3] = ri->ri_nextnum;
 
-	nilfs_attach_writer(nilfs, sbi);
 	/*
 	 * Releasing the next segment of the latest super root.
 	 * The next segment is invalidated by this recovery.
@@ -480,7 +479,6 @@ static int nilfs_prepare_segment_for_recovery(struct the_nilfs *nilfs,
 
  failed:
 	/* No need to recover sufile because it will be destroyed on error */
-	nilfs_detach_writer(nilfs, sbi);
 	return err;
 }
 
@@ -504,6 +502,7 @@ static int nilfs_recovery_copy_block(struct the_nilfs *nilfs,
 
 static int nilfs_recover_dsync_blocks(struct the_nilfs *nilfs,
 				      struct nilfs_sb_info *sbi,
+				      struct nilfs_root *root,
 				      struct list_head *head,
 				      unsigned long *nr_salvaged_blocks)
 {
@@ -515,7 +514,7 @@ static int nilfs_recover_dsync_blocks(struct the_nilfs *nilfs,
 	int err = 0, err2 = 0;
 
 	list_for_each_entry_safe(rb, n, head, list) {
-		inode = nilfs_iget(sbi->s_super, rb->ino);
+		inode = nilfs_iget(sbi->s_super, root, rb->ino);
 		if (IS_ERR(inode)) {
 			err = PTR_ERR(inode);
 			inode = NULL;
@@ -578,6 +577,7 @@ static int nilfs_recover_dsync_blocks(struct the_nilfs *nilfs,
  */
 static int nilfs_do_roll_forward(struct the_nilfs *nilfs,
 				 struct nilfs_sb_info *sbi,
+				 struct nilfs_root *root,
 				 struct nilfs_recovery_info *ri)
 {
 	struct buffer_head *bh_sum = NULL;
@@ -597,7 +597,6 @@ static int nilfs_do_roll_forward(struct the_nilfs *nilfs,
 	};
 	int state = RF_INIT_ST;
 
-	nilfs_attach_writer(nilfs, sbi);
 	pseg_start = ri->ri_lsegs_start;
 	seg_seq = ri->ri_lsegs_start_seq;
 	segnum = nilfs_get_segnum_of_block(nilfs, pseg_start);
@@ -649,7 +648,7 @@ static int nilfs_do_roll_forward(struct the_nilfs *nilfs,
 				goto failed;
 			if (flags & NILFS_SS_LOGEND) {
 				err = nilfs_recover_dsync_blocks(
-					nilfs, sbi, &dsync_blocks,
+					nilfs, sbi, root, &dsync_blocks,
 					&nsalvaged_blocks);
 				if (unlikely(err))
 					goto failed;
@@ -688,7 +687,6 @@ static int nilfs_do_roll_forward(struct the_nilfs *nilfs,
  out:
 	brelse(bh_sum);
 	dispose_recovery_list(&dsync_blocks);
-	nilfs_detach_writer(nilfs, sbi);
 	return err;
 
  confused:
@@ -746,19 +744,20 @@ int nilfs_salvage_orphan_logs(struct the_nilfs *nilfs,
 			      struct nilfs_sb_info *sbi,
 			      struct nilfs_recovery_info *ri)
 {
+	struct nilfs_root *root;
 	int err;
 
 	if (ri->ri_lsegs_start == 0 || ri->ri_lsegs_end == 0)
 		return 0;
 
-	err = nilfs_attach_checkpoint(sbi, ri->ri_cno);
+	err = nilfs_attach_checkpoint(sbi, ri->ri_cno, true, &root);
 	if (unlikely(err)) {
 		printk(KERN_ERR
 		       "NILFS: error loading the latest checkpoint.\n");
 		return err;
 	}
 
-	err = nilfs_do_roll_forward(nilfs, sbi, ri);
+	err = nilfs_do_roll_forward(nilfs, sbi, root, ri);
 	if (unlikely(err))
 		goto failed;
 
@@ -770,7 +769,7 @@ int nilfs_salvage_orphan_logs(struct the_nilfs *nilfs,
 			goto failed;
 		}
 
-		err = nilfs_attach_segment_constructor(sbi);
+		err = nilfs_attach_segment_constructor(sbi, root);
 		if (unlikely(err))
 			goto failed;
 
@@ -788,7 +787,7 @@ int nilfs_salvage_orphan_logs(struct the_nilfs *nilfs,
 	}
 
  failed:
-	nilfs_detach_checkpoint(sbi);
+	nilfs_put_root(root);
 	return err;
 }
 

fs/nilfs2/sb.h

@@ -42,11 +42,6 @@ struct nilfs_sc_info;
  * NILFS super-block data in memory
  */
 struct nilfs_sb_info {
-	/* Snapshot status */
-	__u64 s_snapshot_cno;		/* Checkpoint number */
-	atomic_t s_inodes_count;
-	atomic_t s_blocks_count;	/* Reserved (might be deleted) */
-
 	/* Mount options */
 	unsigned long s_mount_opt;
 	uid_t s_resuid;
@@ -59,8 +54,6 @@ struct nilfs_sb_info {
 	/* Fundamental members */
 	struct super_block *s_super;	/* reverse pointer to super_block */
 	struct the_nilfs *s_nilfs;
-	struct list_head s_list;	/* list head for nilfs->ns_supers */
-	atomic_t s_count;		/* reference count */
 
 	/* Segment constructor */
 	struct list_head s_dirty_files;	/* dirty files list */
@@ -68,9 +61,6 @@ struct nilfs_sb_info {
 	spinlock_t s_inode_lock;	/* Lock for the nilfs inode.
 					   It covers s_dirty_files list */
 
-	/* Metadata files */
-	struct inode *s_ifile;		/* index file inode */
-
 	/* Inode allocator */
 	spinlock_t s_next_gen_lock;
 	u32 s_next_generation;

fs/nilfs2/segbuf.c

@@ -371,7 +371,8 @@ static int nilfs_segbuf_submit_bio(struct nilfs_segment_buffer *segbuf,
 	struct bio *bio = wi->bio;
 	int err;
 
-	if (segbuf->sb_nbio > 0 && bdi_write_congested(wi->nilfs->ns_bdi)) {
+	if (segbuf->sb_nbio > 0 &&
+	    bdi_write_congested(segbuf->sb_super->s_bdi)) {
 		wait_for_completion(&segbuf->sb_bio_event);
 		segbuf->sb_nbio--;
 		if (unlikely(atomic_read(&segbuf->sb_err))) {

fs/nilfs2/segment.c

@@ -191,6 +191,8 @@ int nilfs_transaction_begin(struct super_block *sb,
 	if (ret > 0)
 		return 0;
 
+	vfs_check_frozen(sb, SB_FREEZE_WRITE);
+
 	sbi = NILFS_SB(sb);
 	nilfs = sbi->s_nilfs;
 	down_read(&nilfs->ns_segctor_sem);
@@ -366,8 +368,7 @@ static int nilfs_segctor_reset_segment_buffer(struct nilfs_sc_info *sci)
 
 	if (nilfs_doing_gc())
 		flags = NILFS_SS_GC;
-	err = nilfs_segbuf_reset(segbuf, flags, sci->sc_seg_ctime,
-				 sci->sc_sbi->s_nilfs->ns_cno);
+	err = nilfs_segbuf_reset(segbuf, flags, sci->sc_seg_ctime, sci->sc_cno);
 	if (unlikely(err))
 		return err;
 
@@ -440,17 +441,26 @@ static void nilfs_segctor_end_finfo(struct nilfs_sc_info *sci,
 	struct nilfs_finfo *finfo;
 	struct nilfs_inode_info *ii;
 	struct nilfs_segment_buffer *segbuf;
+	__u64 cno;
 
 	if (sci->sc_blk_cnt == 0)
 		return;
 
 	ii = NILFS_I(inode);
+
+	if (test_bit(NILFS_I_GCINODE, &ii->i_state))
+		cno = ii->i_cno;
+	else if (NILFS_ROOT_METADATA_FILE(inode->i_ino))
+		cno = 0;
+	else
+		cno = sci->sc_cno;
+
 	finfo = nilfs_segctor_map_segsum_entry(sci, &sci->sc_finfo_ptr,
 						 sizeof(*finfo));
 	finfo->fi_ino = cpu_to_le64(inode->i_ino);
 	finfo->fi_nblocks = cpu_to_le32(sci->sc_blk_cnt);
 	finfo->fi_ndatablk = cpu_to_le32(sci->sc_datablk_cnt);
-	finfo->fi_cno = cpu_to_le64(ii->i_cno);
+	finfo->fi_cno = cpu_to_le64(cno);
 
 	segbuf = sci->sc_curseg;
 	segbuf->sb_sum.sumbytes = sci->sc_binfo_ptr.offset +
@@ -755,12 +765,12 @@ static void nilfs_dispose_list(struct nilfs_sb_info *sbi,
 	}
 }
 
-static int nilfs_test_metadata_dirty(struct nilfs_sb_info *sbi)
+static int nilfs_test_metadata_dirty(struct the_nilfs *nilfs,
+				     struct nilfs_root *root)
 {
-	struct the_nilfs *nilfs = sbi->s_nilfs;
 	int ret = 0;
 
-	if (nilfs_mdt_fetch_dirty(sbi->s_ifile))
+	if (nilfs_mdt_fetch_dirty(root->ifile))
 		ret++;
 	if (nilfs_mdt_fetch_dirty(nilfs->ns_cpfile))
 		ret++;
@@ -785,7 +795,7 @@ static int nilfs_segctor_confirm(struct nilfs_sc_info *sci)
 	struct nilfs_sb_info *sbi = sci->sc_sbi;
 	int ret = 0;
 
-	if (nilfs_test_metadata_dirty(sbi))
+	if (nilfs_test_metadata_dirty(sbi->s_nilfs, sci->sc_root))
 		set_bit(NILFS_SC_DIRTY, &sci->sc_flags);
 
 	spin_lock(&sbi->s_inode_lock);
@@ -801,7 +811,7 @@ static void nilfs_segctor_clear_metadata_dirty(struct nilfs_sc_info *sci)
 	struct nilfs_sb_info *sbi = sci->sc_sbi;
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 
-	nilfs_mdt_clear_dirty(sbi->s_ifile);
+	nilfs_mdt_clear_dirty(sci->sc_root->ifile);
 	nilfs_mdt_clear_dirty(nilfs->ns_cpfile);
 	nilfs_mdt_clear_dirty(nilfs->ns_sufile);
 	nilfs_mdt_clear_dirty(nilfs_dat_inode(nilfs));
@@ -848,9 +858,9 @@ static int nilfs_segctor_fill_in_checkpoint(struct nilfs_sc_info *sci)
 	raw_cp->cp_snapshot_list.ssl_next = 0;
 	raw_cp->cp_snapshot_list.ssl_prev = 0;
 	raw_cp->cp_inodes_count =
-		cpu_to_le64(atomic_read(&sbi->s_inodes_count));
+		cpu_to_le64(atomic_read(&sci->sc_root->inodes_count));
 	raw_cp->cp_blocks_count =
-		cpu_to_le64(atomic_read(&sbi->s_blocks_count));
+		cpu_to_le64(atomic_read(&sci->sc_root->blocks_count));
 	raw_cp->cp_nblk_inc =
 		cpu_to_le64(sci->sc_nblk_inc + sci->sc_nblk_this_inc);
 	raw_cp->cp_create = cpu_to_le64(sci->sc_seg_ctime);
@@ -861,7 +871,8 @@ static int nilfs_segctor_fill_in_checkpoint(struct nilfs_sc_info *sci)
 	else
 		nilfs_checkpoint_set_minor(raw_cp);
 
-	nilfs_write_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode, 1);
+	nilfs_write_inode_common(sci->sc_root->ifile,
+				 &raw_cp->cp_ifile_inode, 1);
 	nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, nilfs->ns_cno, bh_cp);
 	return 0;
 
@@ -886,13 +897,12 @@ static void nilfs_fill_in_file_bmap(struct inode *ifile,
 	}
 }
 
-static void nilfs_segctor_fill_in_file_bmap(struct nilfs_sc_info *sci,
-					    struct inode *ifile)
+static void nilfs_segctor_fill_in_file_bmap(struct nilfs_sc_info *sci)
 {
 	struct nilfs_inode_info *ii;
 
 	list_for_each_entry(ii, &sci->sc_dirty_files, i_dirty) {
-		nilfs_fill_in_file_bmap(ifile, ii);
+		nilfs_fill_in_file_bmap(sci->sc_root->ifile, ii);
 		set_bit(NILFS_I_COLLECTED, &ii->i_state);
 	}
 }
@@ -1135,7 +1145,7 @@ static int nilfs_segctor_collect_blocks(struct nilfs_sc_info *sci, int mode)
 		sci->sc_stage.flags |= NILFS_CF_IFILE_STARTED;
 		/* Fall through */
 	case NILFS_ST_IFILE:
-		err = nilfs_segctor_scan_file(sci, sbi->s_ifile,
+		err = nilfs_segctor_scan_file(sci, sci->sc_root->ifile,
 					      &nilfs_sc_file_ops);
 		if (unlikely(err))
 			break;
@@ -1900,6 +1910,7 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
 			set_buffer_uptodate(bh);
 			clear_buffer_dirty(bh);
 			clear_buffer_nilfs_volatile(bh);
+			clear_buffer_nilfs_redirected(bh);
 			if (bh == segbuf->sb_super_root) {
 				if (bh->b_page != bd_page) {
 					end_page_writeback(bd_page);
@@ -1936,11 +1947,9 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
 
 	nilfs_drop_collected_inodes(&sci->sc_dirty_files);
 
-	if (nilfs_doing_gc()) {
+	if (nilfs_doing_gc())
 		nilfs_drop_collected_inodes(&sci->sc_gc_inodes);
-		if (update_sr)
-			nilfs_commit_gcdat_inode(nilfs);
-	} else
+	else
 		nilfs->ns_nongc_ctime = sci->sc_seg_ctime;
 
 	sci->sc_nblk_inc += sci->sc_nblk_this_inc;
@@ -1976,7 +1985,7 @@ static int nilfs_segctor_check_in_files(struct nilfs_sc_info *sci,
 					struct nilfs_sb_info *sbi)
 {
 	struct nilfs_inode_info *ii, *n;
-	__u64 cno = sbi->s_nilfs->ns_cno;
+	struct inode *ifile = sci->sc_root->ifile;
 
 	spin_lock(&sbi->s_inode_lock);
  retry:
@@ -1987,14 +1996,14 @@ static int nilfs_segctor_check_in_files(struct nilfs_sc_info *sci,
 
 			spin_unlock(&sbi->s_inode_lock);
 			err = nilfs_ifile_get_inode_block(
-				sbi->s_ifile, ii->vfs_inode.i_ino, &ibh);
+				ifile, ii->vfs_inode.i_ino, &ibh);
 			if (unlikely(err)) {
 				nilfs_warning(sbi->s_super, __func__,
 					      "failed to get inode block.\n");
 				return err;
 			}
 			nilfs_mdt_mark_buffer_dirty(ibh);
-			nilfs_mdt_mark_dirty(sbi->s_ifile);
+			nilfs_mdt_mark_dirty(ifile);
 			spin_lock(&sbi->s_inode_lock);
 			if (likely(!ii->i_bh))
 				ii->i_bh = ibh;
@@ -2002,7 +2011,6 @@ static int nilfs_segctor_check_in_files(struct nilfs_sc_info *sci,
 				brelse(ibh);
 			goto retry;
 		}
-		ii->i_cno = cno;
 
 		clear_bit(NILFS_I_QUEUED, &ii->i_state);
 		set_bit(NILFS_I_BUSY, &ii->i_state);
@@ -2011,8 +2019,6 @@ static int nilfs_segctor_check_in_files(struct nilfs_sc_info *sci,
 	}
 	spin_unlock(&sbi->s_inode_lock);
 
-	NILFS_I(sbi->s_ifile)->i_cno = cno;
-
 	return 0;
 }
 
@@ -2021,19 +2027,13 @@ static void nilfs_segctor_check_out_files(struct nilfs_sc_info *sci,
 {
 	struct nilfs_transaction_info *ti = current->journal_info;
 	struct nilfs_inode_info *ii, *n;
-	__u64 cno = sbi->s_nilfs->ns_cno;
 
 	spin_lock(&sbi->s_inode_lock);
 	list_for_each_entry_safe(ii, n, &sci->sc_dirty_files, i_dirty) {
 		if (!test_and_clear_bit(NILFS_I_UPDATED, &ii->i_state) ||
-		    test_bit(NILFS_I_DIRTY, &ii->i_state)) {
-			/* The current checkpoint number (=nilfs->ns_cno) is
-			   changed between check-in and check-out only if the
-			   super root is written out.  So, we can update i_cno
-			   for the inodes that remain in the dirty list. */
-			ii->i_cno = cno;
+		    test_bit(NILFS_I_DIRTY, &ii->i_state))
 			continue;
-		}
+
 		clear_bit(NILFS_I_BUSY, &ii->i_state);
 		brelse(ii->i_bh);
 		ii->i_bh = NULL;
@@ -2054,12 +2054,13 @@ static int nilfs_segctor_do_construct(struct nilfs_sc_info *sci, int mode)
 	int err;
 
 	sci->sc_stage.scnt = NILFS_ST_INIT;
+	sci->sc_cno = nilfs->ns_cno;
 
 	err = nilfs_segctor_check_in_files(sci, sbi);
 	if (unlikely(err))
 		goto out;
 
-	if (nilfs_test_metadata_dirty(sbi))
+	if (nilfs_test_metadata_dirty(nilfs, sci->sc_root))
 		set_bit(NILFS_SC_DIRTY, &sci->sc_flags);
 
 	if (nilfs_segctor_clean(sci))
@@ -2091,7 +2092,7 @@ static int nilfs_segctor_do_construct(struct nilfs_sc_info *sci, int mode)
 			goto failed;
 
 		if (sci->sc_stage.flags & NILFS_CF_IFILE_STARTED)
-			nilfs_segctor_fill_in_file_bmap(sci, sbi->s_ifile);
+			nilfs_segctor_fill_in_file_bmap(sci);
 
 		if (mode == SC_LSEG_SR &&
 		    sci->sc_stage.scnt >= NILFS_ST_CPFILE) {
@@ -2452,9 +2453,8 @@ nilfs_remove_written_gcinodes(struct the_nilfs *nilfs, struct list_head *head)
 	list_for_each_entry_safe(ii, n, head, i_dirty) {
 		if (!test_bit(NILFS_I_UPDATED, &ii->i_state))
 			continue;
-		hlist_del_init(&ii->vfs_inode.i_hash);
 		list_del_init(&ii->i_dirty);
-		nilfs_clear_gcinode(&ii->vfs_inode);
+		iput(&ii->vfs_inode);
 	}
 }
 
@@ -2472,13 +2472,15 @@ int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
 
 	nilfs_transaction_lock(sbi, &ti, 1);
 
-	err = nilfs_init_gcdat_inode(nilfs);
+	err = nilfs_mdt_save_to_shadow_map(nilfs->ns_dat);
 	if (unlikely(err))
 		goto out_unlock;
 
 	err = nilfs_ioctl_prepare_clean_segments(nilfs, argv, kbufs);
-	if (unlikely(err))
+	if (unlikely(err)) {
+		nilfs_mdt_restore_from_shadow_map(nilfs->ns_dat);
 		goto out_unlock;
+	}
 
 	sci->sc_freesegs = kbufs[4];
 	sci->sc_nfreesegs = argv[4].v_nmembs;
@@ -2510,7 +2512,7 @@ int nilfs_clean_segments(struct super_block *sb, struct nilfs_argv *argv,
  out_unlock:
 	sci->sc_freesegs = NULL;
 	sci->sc_nfreesegs = 0;
-	nilfs_clear_gcdat_inode(nilfs);
+	nilfs_mdt_clear_shadow_map(nilfs->ns_dat);
 	nilfs_transaction_unlock(sbi);
 	return err;
 }
@@ -2672,6 +2674,8 @@ static int nilfs_segctor_start_thread(struct nilfs_sc_info *sci)
 }
 
 static void nilfs_segctor_kill_thread(struct nilfs_sc_info *sci)
+	__acquires(&sci->sc_state_lock)
+	__releases(&sci->sc_state_lock)
 {
 	sci->sc_state |= NILFS_SEGCTOR_QUIT;
 
@@ -2686,7 +2690,8 @@ static void nilfs_segctor_kill_thread(struct nilfs_sc_info *sci)
 /*
  * Setup & clean-up functions
  */
-static struct nilfs_sc_info *nilfs_segctor_new(struct nilfs_sb_info *sbi)
+static struct nilfs_sc_info *nilfs_segctor_new(struct nilfs_sb_info *sbi,
+					       struct nilfs_root *root)
 {
 	struct nilfs_sc_info *sci;
 
@@ -2697,6 +2702,9 @@ static struct nilfs_sc_info *nilfs_segctor_new(struct nilfs_sb_info *sbi)
 	sci->sc_sbi = sbi;
 	sci->sc_super = sbi->s_super;
 
+	nilfs_get_root(root);
+	sci->sc_root = root;
+
 	init_waitqueue_head(&sci->sc_wait_request);
 	init_waitqueue_head(&sci->sc_wait_daemon);
 	init_waitqueue_head(&sci->sc_wait_task);
@@ -2771,6 +2779,8 @@ static void nilfs_segctor_destroy(struct nilfs_sc_info *sci)
 	WARN_ON(!list_empty(&sci->sc_segbufs));
 	WARN_ON(!list_empty(&sci->sc_write_logs));
 
+	nilfs_put_root(sci->sc_root);
+
 	down_write(&sbi->s_nilfs->ns_segctor_sem);
 
 	del_timer_sync(&sci->sc_timer);
@@ -2780,6 +2790,7 @@ static void nilfs_segctor_destroy(struct nilfs_sc_info *sci)
 /**
  * nilfs_attach_segment_constructor - attach a segment constructor
  * @sbi: nilfs_sb_info
+ * @root: root object of the current filesystem tree
  *
  * nilfs_attach_segment_constructor() allocates a struct nilfs_sc_info,
  * initializes it, and starts the segment constructor.
@@ -2789,9 +2800,9 @@ static void nilfs_segctor_destroy(struct nilfs_sc_info *sci)
  *
  * %-ENOMEM - Insufficient memory available.
  */
-int nilfs_attach_segment_constructor(struct nilfs_sb_info *sbi)
+int nilfs_attach_segment_constructor(struct nilfs_sb_info *sbi,
+				     struct nilfs_root *root)
 {
-	struct the_nilfs *nilfs = sbi->s_nilfs;
 	int err;
 
 	if (NILFS_SC(sbi)) {
@@ -2803,14 +2814,12 @@ int nilfs_attach_segment_constructor(struct nilfs_sb_info *sbi)
 		nilfs_detach_segment_constructor(sbi);
 	}
 
-	sbi->s_sc_info = nilfs_segctor_new(sbi);
+	sbi->s_sc_info = nilfs_segctor_new(sbi, root);
 	if (!sbi->s_sc_info)
 		return -ENOMEM;
 
-	nilfs_attach_writer(nilfs, sbi);
 	err = nilfs_segctor_start_thread(NILFS_SC(sbi));
 	if (err) {
-		nilfs_detach_writer(nilfs, sbi);
 		kfree(sbi->s_sc_info);
 		sbi->s_sc_info = NULL;
 	}
@@ -2847,5 +2856,4 @@ void nilfs_detach_segment_constructor(struct nilfs_sb_info *sbi)
 	up_write(&nilfs->ns_segctor_sem);
 
 	nilfs_dispose_list(sbi, &garbage_list, 1);
-	nilfs_detach_writer(nilfs, sbi);
 }

fs/nilfs2/segment.h

@@ -29,6 +29,8 @@
 #include <linux/nilfs2_fs.h>
 #include "sb.h"
 
+struct nilfs_root;
+
 /**
  * struct nilfs_recovery_info - Recovery information
  * @ri_need_recovery: Recovery status
@@ -87,6 +89,7 @@ struct nilfs_segsum_pointer {
  * struct nilfs_sc_info - Segment constructor information
  * @sc_super: Back pointer to super_block struct
  * @sc_sbi: Back pointer to nilfs_sb_info struct
+ * @sc_root: root object of the current filesystem tree
  * @sc_nblk_inc: Block count of current generation
  * @sc_dirty_files: List of files to be written
  * @sc_gc_inodes: List of GC inodes having blocks to be written
@@ -107,6 +110,7 @@ struct nilfs_segsum_pointer {
  * @sc_datablk_cnt: Data block count of a file
  * @sc_nblk_this_inc: Number of blocks included in the current logical segment
  * @sc_seg_ctime: Creation time
+ * @sc_cno: checkpoint number of current log
  * @sc_flags: Internal flags
  * @sc_state_lock: spinlock for sc_state and so on
  * @sc_state: Segctord state flags
@@ -128,6 +132,7 @@ struct nilfs_segsum_pointer {
 struct nilfs_sc_info {
 	struct super_block     *sc_super;
 	struct nilfs_sb_info   *sc_sbi;
+	struct nilfs_root      *sc_root;
 
 	unsigned long		sc_nblk_inc;
 
@@ -156,7 +161,7 @@ struct nilfs_sc_info {
 	unsigned long		sc_datablk_cnt;
 	unsigned long		sc_nblk_this_inc;
 	time_t			sc_seg_ctime;
-
+	__u64			sc_cno;
 	unsigned long		sc_flags;
 
 	spinlock_t		sc_state_lock;
@@ -230,7 +235,8 @@ extern void nilfs_flush_segment(struct super_block *, ino_t);
 extern int nilfs_clean_segments(struct super_block *, struct nilfs_argv *,
 				void **);
 
-extern int nilfs_attach_segment_constructor(struct nilfs_sb_info *);
+int nilfs_attach_segment_constructor(struct nilfs_sb_info *sbi,
+				     struct nilfs_root *root);
 extern void nilfs_detach_segment_constructor(struct nilfs_sb_info *);
 
 /* recovery.c */

fs/nilfs2/sufile.c

@@ -505,7 +505,7 @@ int nilfs_sufile_get_stat(struct inode *sufile, struct nilfs_sustat *sustat)
 {
 	struct buffer_head *header_bh;
 	struct nilfs_sufile_header *header;
-	struct the_nilfs *nilfs = NILFS_MDT(sufile)->mi_nilfs;
+	struct the_nilfs *nilfs = NILFS_I_NILFS(sufile);
 	void *kaddr;
 	int ret;
 
@@ -583,7 +583,7 @@ ssize_t nilfs_sufile_get_suinfo(struct inode *sufile, __u64 segnum, void *buf,
 	struct nilfs_segment_usage *su;
 	struct nilfs_suinfo *si = buf;
 	size_t susz = NILFS_MDT(sufile)->mi_entry_size;
-	struct the_nilfs *nilfs = NILFS_MDT(sufile)->mi_nilfs;
+	struct the_nilfs *nilfs = NILFS_I_NILFS(sufile);
 	void *kaddr;
 	unsigned long nsegs, segusages_per_block;
 	ssize_t n;
@@ -635,46 +635,55 @@ ssize_t nilfs_sufile_get_suinfo(struct inode *sufile, __u64 segnum, void *buf,
 }
 
 /**
- * nilfs_sufile_read - read sufile inode
- * @sufile: sufile inode
+ * nilfs_sufile_read - read or get sufile inode
+ * @sb: super block instance
+ * @susize: size of a segment usage entry
  * @raw_inode: on-disk sufile inode
+ * @inodep: buffer to store the inode
  */
-int nilfs_sufile_read(struct inode *sufile, struct nilfs_inode *raw_inode)
+int nilfs_sufile_read(struct super_block *sb, size_t susize,
+		      struct nilfs_inode *raw_inode, struct inode **inodep)
 {
-	struct nilfs_sufile_info *sui = NILFS_SUI(sufile);
+	struct inode *sufile;
+	struct nilfs_sufile_info *sui;
 	struct buffer_head *header_bh;
 	struct nilfs_sufile_header *header;
 	void *kaddr;
-	int ret;
+	int err;
 
-	ret = nilfs_read_inode_common(sufile, raw_inode);
-	if (ret < 0)
-		return ret;
+	sufile = nilfs_iget_locked(sb, NULL, NILFS_SUFILE_INO);
+	if (unlikely(!sufile))
+		return -ENOMEM;
+	if (!(sufile->i_state & I_NEW))
+		goto out;
 
-	ret = nilfs_sufile_get_header_block(sufile, &header_bh);
-	if (!ret) {
-		kaddr = kmap_atomic(header_bh->b_page, KM_USER0);
-		header = kaddr + bh_offset(header_bh);
-		sui->ncleansegs = le64_to_cpu(header->sh_ncleansegs);
-		kunmap_atomic(kaddr, KM_USER0);
-		brelse(header_bh);
-	}
-	return ret;
-}
+	err = nilfs_mdt_init(sufile, NILFS_MDT_GFP, sizeof(*sui));
+	if (err)
+		goto failed;
 
-/**
- * nilfs_sufile_new - create sufile
- * @nilfs: nilfs object
- * @susize: size of a segment usage entry
- */
-struct inode *nilfs_sufile_new(struct the_nilfs *nilfs, size_t susize)
-{
-	struct inode *sufile;
+	nilfs_mdt_set_entry_size(sufile, susize,
+				 sizeof(struct nilfs_sufile_header));
+
+	err = nilfs_read_inode_common(sufile, raw_inode);
+	if (err)
+		goto failed;
+
+	err = nilfs_sufile_get_header_block(sufile, &header_bh);
+	if (err)
+		goto failed;
 
-	sufile = nilfs_mdt_new(nilfs, NULL, NILFS_SUFILE_INO,
-			       sizeof(struct nilfs_sufile_info));
-	if (sufile)
-		nilfs_mdt_set_entry_size(sufile, susize,
-					 sizeof(struct nilfs_sufile_header));
-	return sufile;
+	sui = NILFS_SUI(sufile);
+	kaddr = kmap_atomic(header_bh->b_page, KM_USER0);
+	header = kaddr + bh_offset(header_bh);
+	sui->ncleansegs = le64_to_cpu(header->sh_ncleansegs);
+	kunmap_atomic(kaddr, KM_USER0);
+	brelse(header_bh);
+
+	unlock_new_inode(sufile);
+ out:
+	*inodep = sufile;
+	return 0;
+ failed:
+	iget_failed(sufile);
+	return err;
 }

fs/nilfs2/sufile.h

@@ -31,7 +31,7 @@
 
 static inline unsigned long nilfs_sufile_get_nsegments(struct inode *sufile)
 {
-	return NILFS_MDT(sufile)->mi_nilfs->ns_nsegments;
+	return NILFS_I_NILFS(sufile)->ns_nsegments;
 }
 
 unsigned long nilfs_sufile_get_ncleansegs(struct inode *sufile);
@@ -61,8 +61,8 @@ void nilfs_sufile_do_cancel_free(struct inode *, __u64, struct buffer_head *,
 void nilfs_sufile_do_set_error(struct inode *, __u64, struct buffer_head *,
 			       struct buffer_head *);
 
-int nilfs_sufile_read(struct inode *sufile, struct nilfs_inode *raw_inode);
-struct inode *nilfs_sufile_new(struct the_nilfs *nilfs, size_t susize);
+int nilfs_sufile_read(struct super_block *sb, size_t susize,
+		      struct nilfs_inode *raw_inode, struct inode **inodep);
 
 /**
  * nilfs_sufile_scrap - make a segment garbage

fs/nilfs2/super.c

@@ -48,10 +48,10 @@
 #include <linux/vfs.h>
 #include <linux/writeback.h>
 #include <linux/kobject.h>
-#include <linux/exportfs.h>
 #include <linux/seq_file.h>
 #include <linux/mount.h>
 #include "nilfs.h"
+#include "export.h"
 #include "mdt.h"
 #include "alloc.h"
 #include "btree.h"
@@ -68,11 +68,12 @@ MODULE_DESCRIPTION("A New Implementation of the Log-structured Filesystem "
 		   "(NILFS)");
 MODULE_LICENSE("GPL");
 
-struct kmem_cache *nilfs_inode_cachep;
+static struct kmem_cache *nilfs_inode_cachep;
 struct kmem_cache *nilfs_transaction_cachep;
 struct kmem_cache *nilfs_segbuf_cachep;
 struct kmem_cache *nilfs_btree_path_cache;
 
+static int nilfs_setup_super(struct nilfs_sb_info *sbi, int is_mount);
 static int nilfs_remount(struct super_block *sb, int *flags, char *data);
 
 static void nilfs_set_error(struct nilfs_sb_info *sbi)
@@ -146,7 +147,7 @@ void nilfs_warning(struct super_block *sb, const char *function,
 }
 
 
-struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs)
+struct inode *nilfs_alloc_inode(struct super_block *sb)
 {
 	struct nilfs_inode_info *ii;
 
@@ -155,18 +156,20 @@ struct inode *nilfs_alloc_inode_common(struct the_nilfs *nilfs)
 		return NULL;
 	ii->i_bh = NULL;
 	ii->i_state = 0;
+	ii->i_cno = 0;
 	ii->vfs_inode.i_version = 1;
-	nilfs_btnode_cache_init(&ii->i_btnode_cache, nilfs->ns_bdi);
+	nilfs_btnode_cache_init(&ii->i_btnode_cache, sb->s_bdi);
 	return &ii->vfs_inode;
 }
 
-struct inode *nilfs_alloc_inode(struct super_block *sb)
-{
-	return nilfs_alloc_inode_common(NILFS_SB(sb)->s_nilfs);
-}
-
 void nilfs_destroy_inode(struct inode *inode)
 {
+	struct nilfs_mdt_info *mdi = NILFS_MDT(inode);
+
+	if (mdi) {
+		kfree(mdi->mi_bgl); /* kfree(NULL) is safe */
+		kfree(mdi);
+	}
 	kmem_cache_free(nilfs_inode_cachep, NILFS_I(inode));
 }
 
@@ -340,16 +343,15 @@ static void nilfs_put_super(struct super_block *sb)
 		nilfs_cleanup_super(sbi);
 		up_write(&nilfs->ns_sem);
 	}
-	down_write(&nilfs->ns_super_sem);
-	if (nilfs->ns_current == sbi)
-		nilfs->ns_current = NULL;
-	up_write(&nilfs->ns_super_sem);
 
-	nilfs_detach_checkpoint(sbi);
-	put_nilfs(sbi->s_nilfs);
+	iput(nilfs->ns_sufile);
+	iput(nilfs->ns_cpfile);
+	iput(nilfs->ns_dat);
+
+	destroy_nilfs(nilfs);
 	sbi->s_super = NULL;
 	sb->s_fs_info = NULL;
-	nilfs_put_sbinfo(sbi);
+	kfree(sbi);
 }
 
 static int nilfs_sync_fs(struct super_block *sb, int wait)
@@ -376,21 +378,22 @@ static int nilfs_sync_fs(struct super_block *sb, int wait)
 	return err;
 }
 
-int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
+int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno, int curr_mnt,
+			    struct nilfs_root **rootp)
 {
 	struct the_nilfs *nilfs = sbi->s_nilfs;
+	struct nilfs_root *root;
 	struct nilfs_checkpoint *raw_cp;
 	struct buffer_head *bh_cp;
-	int err;
+	int err = -ENOMEM;
 
-	down_write(&nilfs->ns_super_sem);
-	list_add(&sbi->s_list, &nilfs->ns_supers);
-	up_write(&nilfs->ns_super_sem);
+	root = nilfs_find_or_create_root(
+		nilfs, curr_mnt ? NILFS_CPTREE_CURRENT_CNO : cno);
+	if (!root)
+		return err;
 
-	err = -ENOMEM;
-	sbi->s_ifile = nilfs_ifile_new(sbi, nilfs->ns_inode_size);
-	if (!sbi->s_ifile)
-		goto delist;
+	if (root->ifile)
+		goto reuse; /* already attached checkpoint */
 
 	down_read(&nilfs->ns_segctor_sem);
 	err = nilfs_cpfile_get_checkpoint(nilfs->ns_cpfile, cno, 0, &raw_cp,
@@ -406,45 +409,64 @@ int nilfs_attach_checkpoint(struct nilfs_sb_info *sbi, __u64 cno)
 		}
 		goto failed;
 	}
-	err = nilfs_read_inode_common(sbi->s_ifile, &raw_cp->cp_ifile_inode);
-	if (unlikely(err))
+
+	err = nilfs_ifile_read(sbi->s_super, root, nilfs->ns_inode_size,
+			       &raw_cp->cp_ifile_inode, &root->ifile);
+	if (err)
 		goto failed_bh;
-	atomic_set(&sbi->s_inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
-	atomic_set(&sbi->s_blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
+
+	atomic_set(&root->inodes_count, le64_to_cpu(raw_cp->cp_inodes_count));
+	atomic_set(&root->blocks_count, le64_to_cpu(raw_cp->cp_blocks_count));
 
 	nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
+
+ reuse:
+	*rootp = root;
 	return 0;
 
  failed_bh:
 	nilfs_cpfile_put_checkpoint(nilfs->ns_cpfile, cno, bh_cp);
  failed:
-	nilfs_mdt_destroy(sbi->s_ifile);
-	sbi->s_ifile = NULL;
+	nilfs_put_root(root);
+
+	return err;
+}
 
- delist:
-	down_write(&nilfs->ns_super_sem);
-	list_del_init(&sbi->s_list);
-	up_write(&nilfs->ns_super_sem);
+static int nilfs_freeze(struct super_block *sb)
+{
+	struct nilfs_sb_info *sbi = NILFS_SB(sb);
+	struct the_nilfs *nilfs = sbi->s_nilfs;
+	int err;
+
+	if (sb->s_flags & MS_RDONLY)
+		return 0;
 
+	/* Mark super block clean */
+	down_write(&nilfs->ns_sem);
+	err = nilfs_cleanup_super(sbi);
+	up_write(&nilfs->ns_sem);
 	return err;
 }
 
-void nilfs_detach_checkpoint(struct nilfs_sb_info *sbi)
+static int nilfs_unfreeze(struct super_block *sb)
 {
+	struct nilfs_sb_info *sbi = NILFS_SB(sb);
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 
-	nilfs_mdt_destroy(sbi->s_ifile);
-	sbi->s_ifile = NULL;
-	down_write(&nilfs->ns_super_sem);
-	list_del_init(&sbi->s_list);
-	up_write(&nilfs->ns_super_sem);
+	if (sb->s_flags & MS_RDONLY)
+		return 0;
+
+	down_write(&nilfs->ns_sem);
+	nilfs_setup_super(sbi, false);
+	up_write(&nilfs->ns_sem);
+	return 0;
 }
 
 static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct super_block *sb = dentry->d_sb;
-	struct nilfs_sb_info *sbi = NILFS_SB(sb);
-	struct the_nilfs *nilfs = sbi->s_nilfs;
+	struct nilfs_root *root = NILFS_I(dentry->d_inode)->i_root;
+	struct the_nilfs *nilfs = root->nilfs;
 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 	unsigned long long blocks;
 	unsigned long overhead;
@@ -480,7 +502,7 @@ static int nilfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	buf->f_bfree = nfreeblocks;
 	buf->f_bavail = (buf->f_bfree >= nrsvblocks) ?
 		(buf->f_bfree - nrsvblocks) : 0;
-	buf->f_files = atomic_read(&sbi->s_inodes_count);
+	buf->f_files = atomic_read(&root->inodes_count);
 	buf->f_ffree = 0; /* nilfs_count_free_inodes(sb); */
 	buf->f_namelen = NILFS_NAME_LEN;
 	buf->f_fsid.val[0] = (u32)id;
@@ -493,12 +515,12 @@ static int nilfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 {
 	struct super_block *sb = vfs->mnt_sb;
 	struct nilfs_sb_info *sbi = NILFS_SB(sb);
+	struct nilfs_root *root = NILFS_I(vfs->mnt_root->d_inode)->i_root;
 
 	if (!nilfs_test_opt(sbi, BARRIER))
 		seq_puts(seq, ",nobarrier");
-	if (nilfs_test_opt(sbi, SNAPSHOT))
-		seq_printf(seq, ",cp=%llu",
-			   (unsigned long long int)sbi->s_snapshot_cno);
+	if (root->cno != NILFS_CPTREE_CURRENT_CNO)
+		seq_printf(seq, ",cp=%llu", (unsigned long long)root->cno);
 	if (nilfs_test_opt(sbi, ERRORS_PANIC))
 		seq_puts(seq, ",errors=panic");
 	if (nilfs_test_opt(sbi, ERRORS_CONT))
@@ -524,6 +546,8 @@ static const struct super_operations nilfs_sops = {
 	.put_super      = nilfs_put_super,
 	/* .write_super    = nilfs_write_super, */
 	.sync_fs        = nilfs_sync_fs,
+	.freeze_fs	= nilfs_freeze,
+	.unfreeze_fs	= nilfs_unfreeze,
 	/* .write_super_lockfs */
 	/* .unlockfs */
 	.statfs         = nilfs_statfs,
@@ -532,48 +556,6 @@ static const struct super_operations nilfs_sops = {
 	.show_options = nilfs_show_options
 };
 
-static struct inode *
-nilfs_nfs_get_inode(struct super_block *sb, u64 ino, u32 generation)
-{
-	struct inode *inode;
-
-	if (ino < NILFS_FIRST_INO(sb) && ino != NILFS_ROOT_INO &&
-	    ino != NILFS_SKETCH_INO)
-		return ERR_PTR(-ESTALE);
-
-	inode = nilfs_iget(sb, ino);
-	if (IS_ERR(inode))
-		return ERR_CAST(inode);
-	if (generation && inode->i_generation != generation) {
-		iput(inode);
-		return ERR_PTR(-ESTALE);
-	}
-
-	return inode;
-}
-
-static struct dentry *
-nilfs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
-		   int fh_type)
-{
-	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
-				    nilfs_nfs_get_inode);
-}
-
-static struct dentry *
-nilfs_fh_to_parent(struct super_block *sb, struct fid *fid, int fh_len,
-		   int fh_type)
-{
-	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
-				    nilfs_nfs_get_inode);
-}
-
-static const struct export_operations nilfs_export_ops = {
-	.fh_to_dentry = nilfs_fh_to_dentry,
-	.fh_to_parent = nilfs_fh_to_parent,
-	.get_parent = nilfs_get_parent,
-};
-
 enum {
 	Opt_err_cont, Opt_err_panic, Opt_err_ro,
 	Opt_barrier, Opt_nobarrier, Opt_snapshot, Opt_order, Opt_norecovery,
@@ -599,7 +581,6 @@ static int parse_options(char *options, struct super_block *sb, int is_remount)
 	struct nilfs_sb_info *sbi = NILFS_SB(sb);
 	char *p;
 	substring_t args[MAX_OPT_ARGS];
-	int option;
 
 	if (!options)
 		return 1;
@@ -637,30 +618,12 @@ static int parse_options(char *options, struct super_block *sb, int is_remount)
 			nilfs_write_opt(sbi, ERROR_MODE, ERRORS_CONT);
 			break;
 		case Opt_snapshot:
-			if (match_int(&args[0], &option) || option <= 0)
-				return 0;
 			if (is_remount) {
-				if (!nilfs_test_opt(sbi, SNAPSHOT)) {
-					printk(KERN_ERR
-					       "NILFS: cannot change regular "
-					       "mount to snapshot.\n");
-					return 0;
-				} else if (option != sbi->s_snapshot_cno) {
-					printk(KERN_ERR
-					       "NILFS: cannot remount to a "
-					       "different snapshot.\n");
-					return 0;
-				}
-				break;
-			}
-			if (!(sb->s_flags & MS_RDONLY)) {
-				printk(KERN_ERR "NILFS: cannot mount snapshot "
-				       "read/write.  A read-only option is "
-				       "required.\n");
+				printk(KERN_ERR
+				       "NILFS: \"%s\" option is invalid "
+				       "for remount.\n", p);
 				return 0;
 			}
-			sbi->s_snapshot_cno = option;
-			nilfs_set_opt(sbi, SNAPSHOT);
 			break;
 		case Opt_norecovery:
 			nilfs_set_opt(sbi, NORECOVERY);
@@ -688,7 +651,7 @@ nilfs_set_default_options(struct nilfs_sb_info *sbi,
 		NILFS_MOUNT_ERRORS_RO | NILFS_MOUNT_BARRIER;
 }
 
-static int nilfs_setup_super(struct nilfs_sb_info *sbi)
+static int nilfs_setup_super(struct nilfs_sb_info *sbi, int is_mount)
 {
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 	struct nilfs_super_block **sbp;
@@ -700,6 +663,9 @@ static int nilfs_setup_super(struct nilfs_sb_info *sbi)
 	if (!sbp)
 		return -EIO;
 
+	if (!is_mount)
+		goto skip_mount_setup;
+
 	max_mnt_count = le16_to_cpu(sbp[0]->s_max_mnt_count);
 	mnt_count = le16_to_cpu(sbp[0]->s_mnt_count);
 
@@ -716,9 +682,11 @@ static int nilfs_setup_super(struct nilfs_sb_info *sbi)
 		sbp[0]->s_max_mnt_count = cpu_to_le16(NILFS_DFL_MAX_MNT_COUNT);
 
 	sbp[0]->s_mnt_count = cpu_to_le16(mnt_count + 1);
+	sbp[0]->s_mtime = cpu_to_le64(get_seconds());
+
+skip_mount_setup:
 	sbp[0]->s_state =
 		cpu_to_le16(le16_to_cpu(sbp[0]->s_state) & ~NILFS_VALID_FS);
-	sbp[0]->s_mtime = cpu_to_le64(get_seconds());
 	/* synchronize sbp[1] with sbp[0] */
 	memcpy(sbp[1], sbp[0], nilfs->ns_sbsize);
 	return nilfs_commit_super(sbi, NILFS_SB_COMMIT_ALL);
@@ -785,22 +753,156 @@ int nilfs_check_feature_compatibility(struct super_block *sb,
 	return 0;
 }
 
+static int nilfs_get_root_dentry(struct super_block *sb,
+				 struct nilfs_root *root,
+				 struct dentry **root_dentry)
+{
+	struct inode *inode;
+	struct dentry *dentry;
+	int ret = 0;
+
+	inode = nilfs_iget(sb, root, NILFS_ROOT_INO);
+	if (IS_ERR(inode)) {
+		printk(KERN_ERR "NILFS: get root inode failed\n");
+		ret = PTR_ERR(inode);
+		goto out;
+	}
+	if (!S_ISDIR(inode->i_mode) || !inode->i_blocks || !inode->i_size) {
+		iput(inode);
+		printk(KERN_ERR "NILFS: corrupt root inode.\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (root->cno == NILFS_CPTREE_CURRENT_CNO) {
+		dentry = d_find_alias(inode);
+		if (!dentry) {
+			dentry = d_alloc_root(inode);
+			if (!dentry) {
+				iput(inode);
+				ret = -ENOMEM;
+				goto failed_dentry;
+			}
+		} else {
+			iput(inode);
+		}
+	} else {
+		dentry = d_obtain_alias(inode);
+		if (IS_ERR(dentry)) {
+			ret = PTR_ERR(dentry);
+			goto failed_dentry;
+		}
+	}
+	*root_dentry = dentry;
+ out:
+	return ret;
+
+ failed_dentry:
+	printk(KERN_ERR "NILFS: get root dentry failed\n");
+	goto out;
+}
+
+static int nilfs_attach_snapshot(struct super_block *s, __u64 cno,
+				 struct dentry **root_dentry)
+{
+	struct the_nilfs *nilfs = NILFS_SB(s)->s_nilfs;
+	struct nilfs_root *root;
+	int ret;
+
+	down_read(&nilfs->ns_segctor_sem);
+	ret = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile, cno);
+	up_read(&nilfs->ns_segctor_sem);
+	if (ret < 0) {
+		ret = (ret == -ENOENT) ? -EINVAL : ret;
+		goto out;
+	} else if (!ret) {
+		printk(KERN_ERR "NILFS: The specified checkpoint is "
+		       "not a snapshot (checkpoint number=%llu).\n",
+		       (unsigned long long)cno);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = nilfs_attach_checkpoint(NILFS_SB(s), cno, false, &root);
+	if (ret) {
+		printk(KERN_ERR "NILFS: error loading snapshot "
+		       "(checkpoint number=%llu).\n",
+	       (unsigned long long)cno);
+		goto out;
+	}
+	ret = nilfs_get_root_dentry(s, root, root_dentry);
+	nilfs_put_root(root);
+ out:
+	return ret;
+}
+
+static int nilfs_tree_was_touched(struct dentry *root_dentry)
+{
+	return atomic_read(&root_dentry->d_count) > 1;
+}
+
+/**
+ * nilfs_try_to_shrink_tree() - try to shrink dentries of a checkpoint
+ * @root_dentry: root dentry of the tree to be shrunk
+ *
+ * This function returns true if the tree was in-use.
+ */
+static int nilfs_try_to_shrink_tree(struct dentry *root_dentry)
+{
+	if (have_submounts(root_dentry))
+		return true;
+	shrink_dcache_parent(root_dentry);
+	return nilfs_tree_was_touched(root_dentry);
+}
+
+int nilfs_checkpoint_is_mounted(struct super_block *sb, __u64 cno)
+{
+	struct the_nilfs *nilfs = NILFS_SB(sb)->s_nilfs;
+	struct nilfs_root *root;
+	struct inode *inode;
+	struct dentry *dentry;
+	int ret;
+
+	if (cno < 0 || cno > nilfs->ns_cno)
+		return false;
+
+	if (cno >= nilfs_last_cno(nilfs))
+		return true;	/* protect recent checkpoints */
+
+	ret = false;
+	root = nilfs_lookup_root(NILFS_SB(sb)->s_nilfs, cno);
+	if (root) {
+		inode = nilfs_ilookup(sb, root, NILFS_ROOT_INO);
+		if (inode) {
+			dentry = d_find_alias(inode);
+			if (dentry) {
+				if (nilfs_tree_was_touched(dentry))
+					ret = nilfs_try_to_shrink_tree(dentry);
+				dput(dentry);
+			}
+			iput(inode);
+		}
+		nilfs_put_root(root);
+	}
+	return ret;
+}
+
 /**
  * nilfs_fill_super() - initialize a super block instance
  * @sb: super_block
  * @data: mount options
  * @silent: silent mode flag
- * @nilfs: the_nilfs struct
  *
  * This function is called exclusively by nilfs->ns_mount_mutex.
  * So, the recovery process is protected from other simultaneous mounts.
  */
 static int
-nilfs_fill_super(struct super_block *sb, void *data, int silent,
-		 struct the_nilfs *nilfs)
+nilfs_fill_super(struct super_block *sb, void *data, int silent)
 {
+	struct the_nilfs *nilfs;
 	struct nilfs_sb_info *sbi;
-	struct inode *root;
+	struct nilfs_root *fsroot;
+	struct backing_dev_info *bdi;
 	__u64 cno;
 	int err;
 
@@ -809,19 +911,21 @@ nilfs_fill_super(struct super_block *sb, void *data, int silent,
 		return -ENOMEM;
 
 	sb->s_fs_info = sbi;
+	sbi->s_super = sb;
 
-	get_nilfs(nilfs);
+	nilfs = alloc_nilfs(sb->s_bdev);
+	if (!nilfs) {
+		err = -ENOMEM;
+		goto failed_sbi;
+	}
 	sbi->s_nilfs = nilfs;
-	sbi->s_super = sb;
-	atomic_set(&sbi->s_count, 1);
 
 	err = init_nilfs(nilfs, sbi, (char *)data);
 	if (err)
-		goto failed_sbi;
+		goto failed_nilfs;
 
 	spin_lock_init(&sbi->s_inode_lock);
 	INIT_LIST_HEAD(&sbi->s_dirty_files);
-	INIT_LIST_HEAD(&sbi->s_list);
 
 	/*
 	 * Following initialization is overlapped because
@@ -837,94 +941,59 @@ nilfs_fill_super(struct super_block *sb, void *data, int silent,
 	sb->s_export_op = &nilfs_export_ops;
 	sb->s_root = NULL;
 	sb->s_time_gran = 1;
-	sb->s_bdi = nilfs->ns_bdi;
+
+	bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
+	sb->s_bdi = bdi ? : &default_backing_dev_info;
 
 	err = load_nilfs(nilfs, sbi);
 	if (err)
-		goto failed_sbi;
+		goto failed_nilfs;
 
 	cno = nilfs_last_cno(nilfs);
-
-	if (sb->s_flags & MS_RDONLY) {
-		if (nilfs_test_opt(sbi, SNAPSHOT)) {
-			down_read(&nilfs->ns_segctor_sem);
-			err = nilfs_cpfile_is_snapshot(nilfs->ns_cpfile,
-						       sbi->s_snapshot_cno);
-			up_read(&nilfs->ns_segctor_sem);
-			if (err < 0) {
-				if (err == -ENOENT)
-					err = -EINVAL;
-				goto failed_sbi;
-			}
-			if (!err) {
-				printk(KERN_ERR
-				       "NILFS: The specified checkpoint is "
-				       "not a snapshot "
-				       "(checkpoint number=%llu).\n",
-				       (unsigned long long)sbi->s_snapshot_cno);
-				err = -EINVAL;
-				goto failed_sbi;
-			}
-			cno = sbi->s_snapshot_cno;
-		}
-	}
-
-	err = nilfs_attach_checkpoint(sbi, cno);
+	err = nilfs_attach_checkpoint(sbi, cno, true, &fsroot);
 	if (err) {
-		printk(KERN_ERR "NILFS: error loading a checkpoint"
-		       " (checkpoint number=%llu).\n", (unsigned long long)cno);
-		goto failed_sbi;
+		printk(KERN_ERR "NILFS: error loading last checkpoint "
+		       "(checkpoint number=%llu).\n", (unsigned long long)cno);
+		goto failed_unload;
 	}
 
 	if (!(sb->s_flags & MS_RDONLY)) {
-		err = nilfs_attach_segment_constructor(sbi);
+		err = nilfs_attach_segment_constructor(sbi, fsroot);
 		if (err)
 			goto failed_checkpoint;
 	}
 
-	root = nilfs_iget(sb, NILFS_ROOT_INO);
-	if (IS_ERR(root)) {
-		printk(KERN_ERR "NILFS: get root inode failed\n");
-		err = PTR_ERR(root);
-		goto failed_segctor;
-	}
-	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
-		iput(root);
-		printk(KERN_ERR "NILFS: corrupt root inode.\n");
-		err = -EINVAL;
-		goto failed_segctor;
-	}
-	sb->s_root = d_alloc_root(root);
-	if (!sb->s_root) {
-		iput(root);
-		printk(KERN_ERR "NILFS: get root dentry failed\n");
-		err = -ENOMEM;
+	err = nilfs_get_root_dentry(sb, fsroot, &sb->s_root);
+	if (err)
 		goto failed_segctor;
-	}
+
+	nilfs_put_root(fsroot);
 
 	if (!(sb->s_flags & MS_RDONLY)) {
 		down_write(&nilfs->ns_sem);
-		nilfs_setup_super(sbi);
+		nilfs_setup_super(sbi, true);
 		up_write(&nilfs->ns_sem);
 	}
 
-	down_write(&nilfs->ns_super_sem);
-	if (!nilfs_test_opt(sbi, SNAPSHOT))
-		nilfs->ns_current = sbi;
-	up_write(&nilfs->ns_super_sem);
-
 	return 0;
 
  failed_segctor:
 	nilfs_detach_segment_constructor(sbi);
 
  failed_checkpoint:
-	nilfs_detach_checkpoint(sbi);
+	nilfs_put_root(fsroot);
+
+ failed_unload:
+	iput(nilfs->ns_sufile);
+	iput(nilfs->ns_cpfile);
+	iput(nilfs->ns_dat);
+
+ failed_nilfs:
+	destroy_nilfs(nilfs);
 
  failed_sbi:
-	put_nilfs(nilfs);
 	sb->s_fs_info = NULL;
-	nilfs_put_sbinfo(sbi);
+	kfree(sbi);
 	return err;
 }
 
@@ -934,13 +1003,10 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 	struct the_nilfs *nilfs = sbi->s_nilfs;
 	unsigned long old_sb_flags;
 	struct nilfs_mount_options old_opts;
-	int was_snapshot, err;
+	int err;
 
-	down_write(&nilfs->ns_super_sem);
 	old_sb_flags = sb->s_flags;
 	old_opts.mount_opt = sbi->s_mount_opt;
-	old_opts.snapshot_cno = sbi->s_snapshot_cno;
-	was_snapshot = nilfs_test_opt(sbi, SNAPSHOT);
 
 	if (!parse_options(data, sb, 1)) {
 		err = -EINVAL;
@@ -949,11 +1015,6 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL);
 
 	err = -EINVAL;
-	if (was_snapshot && !(*flags & MS_RDONLY)) {
-		printk(KERN_ERR "NILFS (device %s): cannot remount snapshot "
-		       "read/write.\n", sb->s_id);
-		goto restore_opts;
-	}
 
 	if (!nilfs_valid_fs(nilfs)) {
 		printk(KERN_WARNING "NILFS (device %s): couldn't "
@@ -978,6 +1039,7 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 		up_write(&nilfs->ns_sem);
 	} else {
 		__u64 features;
+		struct nilfs_root *root;
 
 		/*
 		 * Mounting a RDONLY partition read-write, so reread and
@@ -999,23 +1061,21 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 
 		sb->s_flags &= ~MS_RDONLY;
 
-		err = nilfs_attach_segment_constructor(sbi);
+		root = NILFS_I(sb->s_root->d_inode)->i_root;
+		err = nilfs_attach_segment_constructor(sbi, root);
 		if (err)
 			goto restore_opts;
 
 		down_write(&nilfs->ns_sem);
-		nilfs_setup_super(sbi);
+		nilfs_setup_super(sbi, true);
 		up_write(&nilfs->ns_sem);
 	}
  out:
-	up_write(&nilfs->ns_super_sem);
 	return 0;
 
  restore_opts:
 	sb->s_flags = old_sb_flags;
 	sbi->s_mount_opt = old_opts.mount_opt;
-	sbi->s_snapshot_cno = old_opts.snapshot_cno;
-	up_write(&nilfs->ns_super_sem);
 	return err;
 }
 
@@ -1035,7 +1095,7 @@ static int nilfs_identify(char *data, struct nilfs_super_data *sd)
 {
 	char *p, *options = data;
 	substring_t args[MAX_OPT_ARGS];
-	int option, token;
+	int token;
 	int ret = 0;
 
 	do {
@@ -1043,16 +1103,18 @@ static int nilfs_identify(char *data, struct nilfs_super_data *sd)
 		if (p != NULL && *p) {
 			token = match_token(p, tokens, args);
 			if (token == Opt_snapshot) {
-				if (!(sd->flags & MS_RDONLY))
+				if (!(sd->flags & MS_RDONLY)) {
 					ret++;
-				else {
-					ret = match_int(&args[0], &option);
-					if (!ret) {
-						if (option > 0)
-							sd->cno = option;
-						else
-							ret++;
-					}
+				} else {
+					sd->cno = simple_strtoull(args[0].from,
+								  NULL, 0);
+					/*
+					 * No need to see the end pointer;
+					 * match_token() has done syntax
+					 * checking.
+					 */
+					if (sd->cno == 0)
+						ret++;
 				}
 			}
 			if (ret)
@@ -1069,18 +1131,14 @@ static int nilfs_identify(char *data, struct nilfs_super_data *sd)
 
 static int nilfs_set_bdev_super(struct super_block *s, void *data)
 {
-	struct nilfs_super_data *sd = data;
-
-	s->s_bdev = sd->bdev;
+	s->s_bdev = data;
 	s->s_dev = s->s_bdev->bd_dev;
 	return 0;
 }
 
 static int nilfs_test_bdev_super(struct super_block *s, void *data)
 {
-	struct nilfs_super_data *sd = data;
-
-	return sd->sbi && s->s_fs_info == (void *)sd->sbi;
+	return (void *)s->s_bdev == data;
 }
 
 static int
@@ -1090,8 +1148,8 @@ nilfs_get_sb(struct file_system_type *fs_type, int flags,
 	struct nilfs_super_data sd;
 	struct super_block *s;
 	fmode_t mode = FMODE_READ;
-	struct the_nilfs *nilfs;
-	int err, need_to_close = 1;
+	struct dentry *root_dentry;
+	int err, s_new = false;
 
 	if (!(flags & MS_RDONLY))
 		mode |= FMODE_WRITE;
@@ -1100,12 +1158,6 @@ nilfs_get_sb(struct file_system_type *fs_type, int flags,
 	if (IS_ERR(sd.bdev))
 		return PTR_ERR(sd.bdev);
 
-	/*
-	 * To get mount instance using sget() vfs-routine, NILFS needs
-	 * much more information than normal filesystems to identify mount
-	 * instance.  For snapshot mounts, not only a mount type (ro-mount
-	 * or rw-mount) but also a checkpoint number is required.
-	 */
 	sd.cno = 0;
 	sd.flags = flags;
 	if (nilfs_identify((char *)data, &sd)) {
@@ -1113,93 +1165,86 @@ nilfs_get_sb(struct file_system_type *fs_type, int flags,
 		goto failed;
 	}
 
-	nilfs = find_or_create_nilfs(sd.bdev);
-	if (!nilfs) {
-		err = -ENOMEM;
-		goto failed;
-	}
-
-	mutex_lock(&nilfs->ns_mount_mutex);
-
-	if (!sd.cno) {
-		/*
-		 * Check if an exclusive mount exists or not.
-		 * Snapshot mounts coexist with a current mount
-		 * (i.e. rw-mount or ro-mount), whereas rw-mount and
-		 * ro-mount are mutually exclusive.
-		 */
-		down_read(&nilfs->ns_super_sem);
-		if (nilfs->ns_current &&
-		    ((nilfs->ns_current->s_super->s_flags ^ flags)
-		     & MS_RDONLY)) {
-			up_read(&nilfs->ns_super_sem);
-			err = -EBUSY;
-			goto failed_unlock;
-		}
-		up_read(&nilfs->ns_super_sem);
-	}
-
-	/*
-	 * Find existing nilfs_sb_info struct
-	 */
-	sd.sbi = nilfs_find_sbinfo(nilfs, !(flags & MS_RDONLY), sd.cno);
-
 	/*
-	 * Get super block instance holding the nilfs_sb_info struct.
-	 * A new instance is allocated if no existing mount is present or
-	 * existing instance has been unmounted.
+	 * once the super is inserted into the list by sget, s_umount
+	 * will protect the lockfs code from trying to start a snapshot
+	 * while we are mounting
 	 */
-	s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, &sd);
-	if (sd.sbi)
-		nilfs_put_sbinfo(sd.sbi);
-
+	mutex_lock(&sd.bdev->bd_fsfreeze_mutex);
+	if (sd.bdev->bd_fsfreeze_count > 0) {
+		mutex_unlock(&sd.bdev->bd_fsfreeze_mutex);
+		err = -EBUSY;
+		goto failed;
+	}
+	s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, sd.bdev);
+	mutex_unlock(&sd.bdev->bd_fsfreeze_mutex);
 	if (IS_ERR(s)) {
 		err = PTR_ERR(s);
-		goto failed_unlock;
+		goto failed;
 	}
 
 	if (!s->s_root) {
 		char b[BDEVNAME_SIZE];
 
+		s_new = true;
+
 		/* New superblock instance created */
 		s->s_flags = flags;
 		s->s_mode = mode;
 		strlcpy(s->s_id, bdevname(sd.bdev, b), sizeof(s->s_id));
 		sb_set_blocksize(s, block_size(sd.bdev));
 
-		err = nilfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0,
-				       nilfs);
+		err = nilfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 		if (err)
-			goto cancel_new;
+			goto failed_super;
 
 		s->s_flags |= MS_ACTIVE;
-		need_to_close = 0;
+	} else if (!sd.cno) {
+		int busy = false;
+
+		if (nilfs_tree_was_touched(s->s_root)) {
+			busy = nilfs_try_to_shrink_tree(s->s_root);
+			if (busy && (flags ^ s->s_flags) & MS_RDONLY) {
+				printk(KERN_ERR "NILFS: the device already "
+				       "has a %s mount.\n",
+				       (s->s_flags & MS_RDONLY) ?
+				       "read-only" : "read/write");
+				err = -EBUSY;
+				goto failed_super;
+			}
+		}
+		if (!busy) {
+			/*
+			 * Try remount to setup mount states if the current
+			 * tree is not mounted and only snapshots use this sb.
+			 */
+			err = nilfs_remount(s, &flags, data);
+			if (err)
+				goto failed_super;
+		}
 	}
 
-	mutex_unlock(&nilfs->ns_mount_mutex);
-	put_nilfs(nilfs);
-	if (need_to_close)
+	if (sd.cno) {
+		err = nilfs_attach_snapshot(s, sd.cno, &root_dentry);
+		if (err)
+			goto failed_super;
+	} else {
+		root_dentry = dget(s->s_root);
+	}
+
+	if (!s_new)
 		close_bdev_exclusive(sd.bdev, mode);
-	simple_set_mnt(mnt, s);
-	return 0;
 
- failed_unlock:
-	mutex_unlock(&nilfs->ns_mount_mutex);
-	put_nilfs(nilfs);
- failed:
-	close_bdev_exclusive(sd.bdev, mode);
-	return err;
+	mnt->mnt_sb = s;
+	mnt->mnt_root = root_dentry;
+	return 0;
 
- cancel_new:
-	/* Abandoning the newly allocated superblock */
-	mutex_unlock(&nilfs->ns_mount_mutex);
-	put_nilfs(nilfs);
+ failed_super:
 	deactivate_locked_super(s);
-	/*
-	 * deactivate_locked_super() invokes close_bdev_exclusive().
-	 * We must finish all post-cleaning before this call;
-	 * put_nilfs() needs the block device.
-	 */
+
+ failed:
+	if (!s_new)
+		close_bdev_exclusive(sd.bdev, mode);
 	return err;
 }
 

fs/nilfs2/the_nilfs.c

@@ -35,9 +35,6 @@
 #include "segbuf.h"
 
 
-static LIST_HEAD(nilfs_objects);
-static DEFINE_SPINLOCK(nilfs_lock);
-
 static int nilfs_valid_sb(struct nilfs_super_block *sbp);
 
 void nilfs_set_last_segment(struct the_nilfs *nilfs,
@@ -61,16 +58,13 @@ void nilfs_set_last_segment(struct the_nilfs *nilfs,
 }
 
 /**
- * alloc_nilfs - allocate the_nilfs structure
+ * alloc_nilfs - allocate a nilfs object
  * @bdev: block device to which the_nilfs is related
  *
- * alloc_nilfs() allocates memory for the_nilfs and
- * initializes its reference count and locks.
- *
  * Return Value: On success, pointer to the_nilfs is returned.
  * On error, NULL is returned.
  */
-static struct the_nilfs *alloc_nilfs(struct block_device *bdev)
+struct the_nilfs *alloc_nilfs(struct block_device *bdev)
 {
 	struct the_nilfs *nilfs;
 
@@ -79,103 +73,38 @@ static struct the_nilfs *alloc_nilfs(struct block_device *bdev)
 		return NULL;
 
 	nilfs->ns_bdev = bdev;
-	atomic_set(&nilfs->ns_count, 1);
 	atomic_set(&nilfs->ns_ndirtyblks, 0);
 	init_rwsem(&nilfs->ns_sem);
-	init_rwsem(&nilfs->ns_super_sem);
-	mutex_init(&nilfs->ns_mount_mutex);
-	init_rwsem(&nilfs->ns_writer_sem);
-	INIT_LIST_HEAD(&nilfs->ns_list);
-	INIT_LIST_HEAD(&nilfs->ns_supers);
+	INIT_LIST_HEAD(&nilfs->ns_gc_inodes);
 	spin_lock_init(&nilfs->ns_last_segment_lock);
-	nilfs->ns_gc_inodes_h = NULL;
+	nilfs->ns_cptree = RB_ROOT;
+	spin_lock_init(&nilfs->ns_cptree_lock);
 	init_rwsem(&nilfs->ns_segctor_sem);
 
 	return nilfs;
 }
 
 /**
- * find_or_create_nilfs - find or create nilfs object
- * @bdev: block device to which the_nilfs is related
- *
- * find_nilfs() looks up an existent nilfs object created on the
- * device and gets the reference count of the object.  If no nilfs object
- * is found on the device, a new nilfs object is allocated.
- *
- * Return Value: On success, pointer to the nilfs object is returned.
- * On error, NULL is returned.
- */
-struct the_nilfs *find_or_create_nilfs(struct block_device *bdev)
-{
-	struct the_nilfs *nilfs, *new = NULL;
-
- retry:
-	spin_lock(&nilfs_lock);
-	list_for_each_entry(nilfs, &nilfs_objects, ns_list) {
-		if (nilfs->ns_bdev == bdev) {
-			get_nilfs(nilfs);
-			spin_unlock(&nilfs_lock);
-			if (new)
-				put_nilfs(new);
-			return nilfs; /* existing object */
-		}
-	}
-	if (new) {
-		list_add_tail(&new->ns_list, &nilfs_objects);
-		spin_unlock(&nilfs_lock);
-		return new; /* new object */
-	}
-	spin_unlock(&nilfs_lock);
-
-	new = alloc_nilfs(bdev);
-	if (new)
-		goto retry;
-	return NULL; /* insufficient memory */
-}
-
-/**
- * put_nilfs - release a reference to the_nilfs
- * @nilfs: the_nilfs structure to be released
- *
- * put_nilfs() decrements a reference counter of the_nilfs.
- * If the reference count reaches zero, the_nilfs is freed.
+ * destroy_nilfs - destroy nilfs object
+ * @nilfs: nilfs object to be released
  */
-void put_nilfs(struct the_nilfs *nilfs)
+void destroy_nilfs(struct the_nilfs *nilfs)
 {
-	spin_lock(&nilfs_lock);
-	if (!atomic_dec_and_test(&nilfs->ns_count)) {
-		spin_unlock(&nilfs_lock);
-		return;
-	}
-	list_del_init(&nilfs->ns_list);
-	spin_unlock(&nilfs_lock);
-
-	/*
-	 * Increment of ns_count never occurs below because the caller
-	 * of get_nilfs() holds at least one reference to the_nilfs.
-	 * Thus its exclusion control is not required here.
-	 */
-
 	might_sleep();
-	if (nilfs_loaded(nilfs)) {
-		nilfs_mdt_destroy(nilfs->ns_sufile);
-		nilfs_mdt_destroy(nilfs->ns_cpfile);
-		nilfs_mdt_destroy(nilfs->ns_dat);
-		nilfs_mdt_destroy(nilfs->ns_gc_dat);
-	}
 	if (nilfs_init(nilfs)) {
-		nilfs_destroy_gccache(nilfs);
 		brelse(nilfs->ns_sbh[0]);
 		brelse(nilfs->ns_sbh[1]);
 	}
 	kfree(nilfs);
 }
 
-static int nilfs_load_super_root(struct the_nilfs *nilfs, sector_t sr_block)
+static int nilfs_load_super_root(struct the_nilfs *nilfs,
+				 struct super_block *sb, sector_t sr_block)
 {
 	struct buffer_head *bh_sr;
 	struct nilfs_super_root *raw_sr;
 	struct nilfs_super_block **sbp = nilfs->ns_sbp;
+	struct nilfs_inode *rawi;
 	unsigned dat_entry_size, segment_usage_size, checkpoint_size;
 	unsigned inode_size;
 	int err;
@@ -192,40 +121,22 @@ static int nilfs_load_super_root(struct the_nilfs *nilfs, sector_t sr_block)
 
 	inode_size = nilfs->ns_inode_size;
 
-	err = -ENOMEM;
-	nilfs->ns_dat = nilfs_dat_new(nilfs, dat_entry_size);
-	if (unlikely(!nilfs->ns_dat))
+	rawi = (void *)bh_sr->b_data + NILFS_SR_DAT_OFFSET(inode_size);
+	err = nilfs_dat_read(sb, dat_entry_size, rawi, &nilfs->ns_dat);
+	if (err)
 		goto failed;
 
-	nilfs->ns_gc_dat = nilfs_dat_new(nilfs, dat_entry_size);
-	if (unlikely(!nilfs->ns_gc_dat))
+	rawi = (void *)bh_sr->b_data + NILFS_SR_CPFILE_OFFSET(inode_size);
+	err = nilfs_cpfile_read(sb, checkpoint_size, rawi, &nilfs->ns_cpfile);
+	if (err)
 		goto failed_dat;
 
-	nilfs->ns_cpfile = nilfs_cpfile_new(nilfs, checkpoint_size);
-	if (unlikely(!nilfs->ns_cpfile))
-		goto failed_gc_dat;
-
-	nilfs->ns_sufile = nilfs_sufile_new(nilfs, segment_usage_size);
-	if (unlikely(!nilfs->ns_sufile))
+	rawi = (void *)bh_sr->b_data + NILFS_SR_SUFILE_OFFSET(inode_size);
+	err = nilfs_sufile_read(sb, segment_usage_size, rawi,
+				&nilfs->ns_sufile);
+	if (err)
 		goto failed_cpfile;
 
-	nilfs_mdt_set_shadow(nilfs->ns_dat, nilfs->ns_gc_dat);
-
-	err = nilfs_dat_read(nilfs->ns_dat, (void *)bh_sr->b_data +
-			     NILFS_SR_DAT_OFFSET(inode_size));
-	if (unlikely(err))
-		goto failed_sufile;
-
-	err = nilfs_cpfile_read(nilfs->ns_cpfile, (void *)bh_sr->b_data +
-				NILFS_SR_CPFILE_OFFSET(inode_size));
-	if (unlikely(err))
-		goto failed_sufile;
-
-	err = nilfs_sufile_read(nilfs->ns_sufile, (void *)bh_sr->b_data +
-				NILFS_SR_SUFILE_OFFSET(inode_size));
-	if (unlikely(err))
-		goto failed_sufile;
-
 	raw_sr = (struct nilfs_super_root *)bh_sr->b_data;
 	nilfs->ns_nongc_ctime = le64_to_cpu(raw_sr->sr_nongc_ctime);
 
@@ -233,17 +144,11 @@ static int nilfs_load_super_root(struct the_nilfs *nilfs, sector_t sr_block)
 	brelse(bh_sr);
 	return err;
 
- failed_sufile:
-	nilfs_mdt_destroy(nilfs->ns_sufile);
-
  failed_cpfile:
-	nilfs_mdt_destroy(nilfs->ns_cpfile);
-
- failed_gc_dat:
-	nilfs_mdt_destroy(nilfs->ns_gc_dat);
+	iput(nilfs->ns_cpfile);
 
  failed_dat:
-	nilfs_mdt_destroy(nilfs->ns_dat);
+	iput(nilfs->ns_dat);
 	goto failed;
 }
 
@@ -306,15 +211,6 @@ int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
 	int valid_fs = nilfs_valid_fs(nilfs);
 	int err;
 
-	if (nilfs_loaded(nilfs)) {
-		if (valid_fs ||
-		    ((s_flags & MS_RDONLY) && nilfs_test_opt(sbi, NORECOVERY)))
-			return 0;
-		printk(KERN_ERR "NILFS: the filesystem is in an incomplete "
-		       "recovery state.\n");
-		return -EINVAL;
-	}
-
 	if (!valid_fs) {
 		printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
 		if (s_flags & MS_RDONLY) {
@@ -375,7 +271,7 @@ int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
 			goto scan_error;
 	}
 
-	err = nilfs_load_super_root(nilfs, ri.ri_super_root);
+	err = nilfs_load_super_root(nilfs, sbi->s_super, ri.ri_super_root);
 	if (unlikely(err)) {
 		printk(KERN_ERR "NILFS: error loading super root.\n");
 		goto failed;
@@ -443,10 +339,9 @@ int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
 	goto failed;
 
  failed_unload:
-	nilfs_mdt_destroy(nilfs->ns_cpfile);
-	nilfs_mdt_destroy(nilfs->ns_sufile);
-	nilfs_mdt_destroy(nilfs->ns_dat);
-	nilfs_mdt_destroy(nilfs->ns_gc_dat);
+	iput(nilfs->ns_cpfile);
+	iput(nilfs->ns_sufile);
+	iput(nilfs->ns_dat);
 
  failed:
 	nilfs_clear_recovery_info(&ri);
@@ -468,8 +363,8 @@ static unsigned long long nilfs_max_size(unsigned int blkbits)
 static int nilfs_store_disk_layout(struct the_nilfs *nilfs,
 				   struct nilfs_super_block *sbp)
 {
-	if (le32_to_cpu(sbp->s_rev_level) != NILFS_CURRENT_REV) {
-		printk(KERN_ERR "NILFS: revision mismatch "
+	if (le32_to_cpu(sbp->s_rev_level) < NILFS_MIN_SUPP_REV) {
+		printk(KERN_ERR "NILFS: unsupported revision "
 		       "(superblock rev.=%d.%d, current rev.=%d.%d). "
 		       "Please check the version of mkfs.nilfs.\n",
 		       le32_to_cpu(sbp->s_rev_level),
@@ -631,12 +526,7 @@ static int nilfs_load_super_block(struct the_nilfs *nilfs,
  *
  * init_nilfs() performs common initialization per block device (e.g.
  * reading the super block, getting disk layout information, initializing
- * shared fields in the_nilfs). It takes on some portion of the jobs
- * typically done by a fill_super() routine. This division arises from
- * the nature that multiple NILFS instances may be simultaneously
- * mounted on a device.
- * For multiple mounts on the same device, only the first mount
- * invokes these tasks.
+ * shared fields in the_nilfs).
  *
  * Return Value: On success, 0 is returned. On error, a negative error
  * code is returned.
@@ -645,32 +535,10 @@ int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
 {
 	struct super_block *sb = sbi->s_super;
 	struct nilfs_super_block *sbp;
-	struct backing_dev_info *bdi;
 	int blocksize;
 	int err;
 
 	down_write(&nilfs->ns_sem);
-	if (nilfs_init(nilfs)) {
-		/* Load values from existing the_nilfs */
-		sbp = nilfs->ns_sbp[0];
-		err = nilfs_store_magic_and_option(sb, sbp, data);
-		if (err)
-			goto out;
-
-		err = nilfs_check_feature_compatibility(sb, sbp);
-		if (err)
-			goto out;
-
-		blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size);
-		if (sb->s_blocksize != blocksize &&
-		    !sb_set_blocksize(sb, blocksize)) {
-			printk(KERN_ERR "NILFS: blocksize %d unfit to device\n",
-			       blocksize);
-			err = -EINVAL;
-		}
-		sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits);
-		goto out;
-	}
 
 	blocksize = sb_min_blocksize(sb, NILFS_MIN_BLOCK_SIZE);
 	if (!blocksize) {
@@ -729,18 +597,10 @@ int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
 
 	nilfs->ns_mount_state = le16_to_cpu(sbp->s_state);
 
-	bdi = nilfs->ns_bdev->bd_inode->i_mapping->backing_dev_info;
-	nilfs->ns_bdi = bdi ? : &default_backing_dev_info;
-
 	err = nilfs_store_log_cursor(nilfs, sbp);
 	if (err)
 		goto failed_sbh;
 
-	/* Initialize gcinode cache */
-	err = nilfs_init_gccache(nilfs);
-	if (err)
-		goto failed_sbh;
-
 	set_nilfs_init(nilfs);
 	err = 0;
  out:
@@ -812,79 +672,92 @@ int nilfs_near_disk_full(struct the_nilfs *nilfs)
 	return ncleansegs <= nilfs->ns_nrsvsegs + nincsegs;
 }
 
-/**
- * nilfs_find_sbinfo - find existing nilfs_sb_info structure
- * @nilfs: nilfs object
- * @rw_mount: mount type (non-zero value for read/write mount)
- * @cno: checkpoint number (zero for read-only mount)
- *
- * nilfs_find_sbinfo() returns the nilfs_sb_info structure which
- * @rw_mount and @cno (in case of snapshots) matched.  If no instance
- * was found, NULL is returned.  Although the super block instance can
- * be unmounted after this function returns, the nilfs_sb_info struct
- * is kept on memory until nilfs_put_sbinfo() is called.
- */
-struct nilfs_sb_info *nilfs_find_sbinfo(struct the_nilfs *nilfs,
-					int rw_mount, __u64 cno)
+struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno)
 {
-	struct nilfs_sb_info *sbi;
-
-	down_read(&nilfs->ns_super_sem);
-	/*
-	 * The SNAPSHOT flag and sb->s_flags are supposed to be
-	 * protected with nilfs->ns_super_sem.
-	 */
-	sbi = nilfs->ns_current;
-	if (rw_mount) {
-		if (sbi && !(sbi->s_super->s_flags & MS_RDONLY))
-			goto found; /* read/write mount */
-		else
-			goto out;
-	} else if (cno == 0) {
-		if (sbi && (sbi->s_super->s_flags & MS_RDONLY))
-			goto found; /* read-only mount */
-		else
-			goto out;
+	struct rb_node *n;
+	struct nilfs_root *root;
+
+	spin_lock(&nilfs->ns_cptree_lock);
+	n = nilfs->ns_cptree.rb_node;
+	while (n) {
+		root = rb_entry(n, struct nilfs_root, rb_node);
+
+		if (cno < root->cno) {
+			n = n->rb_left;
+		} else if (cno > root->cno) {
+			n = n->rb_right;
+		} else {
+			atomic_inc(&root->count);
+			spin_unlock(&nilfs->ns_cptree_lock);
+			return root;
+		}
 	}
+	spin_unlock(&nilfs->ns_cptree_lock);
 
-	list_for_each_entry(sbi, &nilfs->ns_supers, s_list) {
-		if (nilfs_test_opt(sbi, SNAPSHOT) &&
-		    sbi->s_snapshot_cno == cno)
-			goto found; /* snapshot mount */
-	}
- out:
-	up_read(&nilfs->ns_super_sem);
 	return NULL;
-
- found:
-	atomic_inc(&sbi->s_count);
-	up_read(&nilfs->ns_super_sem);
-	return sbi;
 }
 
-int nilfs_checkpoint_is_mounted(struct the_nilfs *nilfs, __u64 cno,
-				int snapshot_mount)
+struct nilfs_root *
+nilfs_find_or_create_root(struct the_nilfs *nilfs, __u64 cno)
 {
-	struct nilfs_sb_info *sbi;
-	int ret = 0;
+	struct rb_node **p, *parent;
+	struct nilfs_root *root, *new;
 
-	down_read(&nilfs->ns_super_sem);
-	if (cno == 0 || cno > nilfs->ns_cno)
-		goto out_unlock;
+	root = nilfs_lookup_root(nilfs, cno);
+	if (root)
+		return root;
 
-	list_for_each_entry(sbi, &nilfs->ns_supers, s_list) {
-		if (sbi->s_snapshot_cno == cno &&
-		    (!snapshot_mount || nilfs_test_opt(sbi, SNAPSHOT))) {
-					/* exclude read-only mounts */
-			ret++;
-			break;
+	new = kmalloc(sizeof(*root), GFP_KERNEL);
+	if (!new)
+		return NULL;
+
+	spin_lock(&nilfs->ns_cptree_lock);
+
+	p = &nilfs->ns_cptree.rb_node;
+	parent = NULL;
+
+	while (*p) {
+		parent = *p;
+		root = rb_entry(parent, struct nilfs_root, rb_node);
+
+		if (cno < root->cno) {
+			p = &(*p)->rb_left;
+		} else if (cno > root->cno) {
+			p = &(*p)->rb_right;
+		} else {
+			atomic_inc(&root->count);
+			spin_unlock(&nilfs->ns_cptree_lock);
+			kfree(new);
+			return root;
 		}
 	}
-	/* for protecting recent checkpoints */
-	if (cno >= nilfs_last_cno(nilfs))
-		ret++;
 
- out_unlock:
-	up_read(&nilfs->ns_super_sem);
-	return ret;
+	new->cno = cno;
+	new->ifile = NULL;
+	new->nilfs = nilfs;
+	atomic_set(&new->count, 1);
+	atomic_set(&new->inodes_count, 0);
+	atomic_set(&new->blocks_count, 0);
+
+	rb_link_node(&new->rb_node, parent, p);
+	rb_insert_color(&new->rb_node, &nilfs->ns_cptree);
+
+	spin_unlock(&nilfs->ns_cptree_lock);
+
+	return new;
+}
+
+void nilfs_put_root(struct nilfs_root *root)
+{
+	if (atomic_dec_and_test(&root->count)) {
+		struct the_nilfs *nilfs = root->nilfs;
+
+		spin_lock(&nilfs->ns_cptree_lock);
+		rb_erase(&root->rb_node, &nilfs->ns_cptree);
+		spin_unlock(&nilfs->ns_cptree_lock);
+		if (root->ifile)
+			iput(root->ifile);
+
+		kfree(root);
+	}
 }

fs/nilfs2/the_nilfs.h

@@ -26,6 +26,7 @@
 
 #include <linux/types.h>
 #include <linux/buffer_head.h>
+#include <linux/rbtree.h>
 #include <linux/fs.h>
 #include <linux/blkdev.h>
 #include <linux/backing-dev.h>
@@ -45,22 +46,13 @@ enum {
 /**
  * struct the_nilfs - struct to supervise multiple nilfs mount points
  * @ns_flags: flags
- * @ns_count: reference count
- * @ns_list: list head for nilfs_list
  * @ns_bdev: block device
- * @ns_bdi: backing dev info
- * @ns_writer: back pointer to writable nilfs_sb_info
  * @ns_sem: semaphore for shared states
- * @ns_super_sem: semaphore for global operations across super block instances
- * @ns_mount_mutex: mutex protecting mount process of nilfs
- * @ns_writer_sem: semaphore protecting ns_writer attach/detach
- * @ns_current: back pointer to current mount
  * @ns_sbh: buffer heads of on-disk super blocks
  * @ns_sbp: pointers to super block data
  * @ns_sbwtime: previous write time of super block
  * @ns_sbwcount: write count of super block
  * @ns_sbsize: size of valid data in super block
- * @ns_supers: list of nilfs super block structs
  * @ns_seg_seq: segment sequence counter
  * @ns_segnum: index number of the latest full segment.
  * @ns_nextnum: index number of the full segment index to be used next
@@ -79,9 +71,9 @@ enum {
  * @ns_dat: DAT file inode
  * @ns_cpfile: checkpoint file inode
  * @ns_sufile: segusage file inode
- * @ns_gc_dat: shadow inode of the DAT file inode for GC
+ * @ns_cptree: rb-tree of all mounted checkpoints (nilfs_root)
+ * @ns_cptree_lock: lock protecting @ns_cptree
  * @ns_gc_inodes: dummy inodes to keep live blocks
- * @ns_gc_inodes_h: hash list to keep dummy inode holding live blocks
  * @ns_blocksize_bits: bit length of block size
  * @ns_blocksize: block size
  * @ns_nsegments: number of segments in filesystem
@@ -95,22 +87,9 @@ enum {
  */
 struct the_nilfs {
 	unsigned long		ns_flags;
-	atomic_t		ns_count;
-	struct list_head	ns_list;
 
 	struct block_device    *ns_bdev;
-	struct backing_dev_info *ns_bdi;
-	struct nilfs_sb_info   *ns_writer;
 	struct rw_semaphore	ns_sem;
-	struct rw_semaphore	ns_super_sem;
-	struct mutex		ns_mount_mutex;
-	struct rw_semaphore	ns_writer_sem;
-
-	/*
-	 * components protected by ns_super_sem
-	 */
-	struct nilfs_sb_info   *ns_current;
-	struct list_head	ns_supers;
 
 	/*
 	 * used for
@@ -163,11 +142,13 @@ struct the_nilfs {
 	struct inode	       *ns_dat;
 	struct inode	       *ns_cpfile;
 	struct inode	       *ns_sufile;
-	struct inode	       *ns_gc_dat;
 
-	/* GC inode list and hash table head */
+	/* Checkpoint tree */
+	struct rb_root		ns_cptree;
+	spinlock_t		ns_cptree_lock;
+
+	/* GC inode list */
 	struct list_head	ns_gc_inodes;
-	struct hlist_head      *ns_gc_inodes_h;
 
 	/* Disk layout information (static) */
 	unsigned int		ns_blocksize_bits;
@@ -182,9 +163,6 @@ struct the_nilfs {
 	u32			ns_crc_seed;
 };
 
-#define NILFS_GCINODE_HASH_BITS		8
-#define NILFS_GCINODE_HASH_SIZE		(1<<NILFS_GCINODE_HASH_BITS)
-
 #define THE_NILFS_FNS(bit, name)					\
 static inline void set_nilfs_##name(struct the_nilfs *nilfs)		\
 {									\
@@ -205,6 +183,32 @@ THE_NILFS_FNS(DISCONTINUED, discontinued)
 THE_NILFS_FNS(GC_RUNNING, gc_running)
 THE_NILFS_FNS(SB_DIRTY, sb_dirty)
 
+/**
+ * struct nilfs_root - nilfs root object
+ * @cno: checkpoint number
+ * @rb_node: red-black tree node
+ * @count: refcount of this structure
+ * @nilfs: nilfs object
+ * @ifile: inode file
+ * @root: root inode
+ * @inodes_count: number of inodes
+ * @blocks_count: number of blocks (Reserved)
+ */
+struct nilfs_root {
+	__u64 cno;
+	struct rb_node rb_node;
+
+	atomic_t count;
+	struct the_nilfs *nilfs;
+	struct inode *ifile;
+
+	atomic_t inodes_count;
+	atomic_t blocks_count;
+};
+
+/* Special checkpoint number */
+#define NILFS_CPTREE_CURRENT_CNO	0
+
 /* Minimum interval of periodical update of superblocks (in seconds) */
 #define NILFS_SB_FREQ		10
 
@@ -221,46 +225,25 @@ static inline int nilfs_sb_will_flip(struct the_nilfs *nilfs)
 }
 
 void nilfs_set_last_segment(struct the_nilfs *, sector_t, u64, __u64);
-struct the_nilfs *find_or_create_nilfs(struct block_device *);
-void put_nilfs(struct the_nilfs *);
+struct the_nilfs *alloc_nilfs(struct block_device *bdev);
+void destroy_nilfs(struct the_nilfs *nilfs);
 int init_nilfs(struct the_nilfs *, struct nilfs_sb_info *, char *);
 int load_nilfs(struct the_nilfs *, struct nilfs_sb_info *);
 int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
+struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno);
+struct nilfs_root *nilfs_find_or_create_root(struct the_nilfs *nilfs,
+					     __u64 cno);
+void nilfs_put_root(struct nilfs_root *root);
 struct nilfs_sb_info *nilfs_find_sbinfo(struct the_nilfs *, int, __u64);
-int nilfs_checkpoint_is_mounted(struct the_nilfs *, __u64, int);
 int nilfs_near_disk_full(struct the_nilfs *);
 void nilfs_fall_back_super_block(struct the_nilfs *);
 void nilfs_swap_super_block(struct the_nilfs *);
 
 
-static inline void get_nilfs(struct the_nilfs *nilfs)
-{
-	/* Caller must have at least one reference of the_nilfs. */
-	atomic_inc(&nilfs->ns_count);
-}
-
-static inline void
-nilfs_attach_writer(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
-{
-	down_write(&nilfs->ns_writer_sem);
-	nilfs->ns_writer = sbi;
-	up_write(&nilfs->ns_writer_sem);
-}
-
-static inline void
-nilfs_detach_writer(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
-{
-	down_write(&nilfs->ns_writer_sem);
-	if (sbi == nilfs->ns_writer)
-		nilfs->ns_writer = NULL;
-	up_write(&nilfs->ns_writer_sem);
-}
-
-static inline void nilfs_put_sbinfo(struct nilfs_sb_info *sbi)
+static inline void nilfs_get_root(struct nilfs_root *root)
 {
-	if (atomic_dec_and_test(&sbi->s_count))
-		kfree(sbi);
+	atomic_inc(&root->count);
 }
 
 static inline int nilfs_valid_fs(struct the_nilfs *nilfs)

include/linux/exportfs.h

@@ -67,6 +67,19 @@ enum fid_type {
 	 * 32 bit parent block number, 32 bit parent generation number
 	 */
 	FILEID_UDF_WITH_PARENT = 0x52,
+
+	/*
+	 * 64 bit checkpoint number, 64 bit inode number,
+	 * 32 bit generation number.
+	 */
+	FILEID_NILFS_WITHOUT_PARENT = 0x61,
+
+	/*
+	 * 64 bit checkpoint number, 64 bit inode number,
+	 * 32 bit generation number, 32 bit parent generation.
+	 * 64 bit parent inode number.
+	 */
+	FILEID_NILFS_WITH_PARENT = 0x62,
 };
 
 struct fid {

include/linux/nilfs2_fs.h

@@ -4,16 +4,16 @@
  * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
  *
  * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
+ * it under the terms of the GNU Lesser General Public License as published
+ * by the Free Software Foundation; either version 2.1 of the License, or
  * (at your option) any later version.
  *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
+ * GNU Lesser General Public License for more details.
  *
- * You should have received a copy of the GNU General Public License
+ * You should have received a copy of the GNU Lesser General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  *
@@ -147,7 +147,6 @@ struct nilfs_super_root {
 #define NILFS_MOUNT_ERRORS_CONT		0x0010  /* Continue on errors */
 #define NILFS_MOUNT_ERRORS_RO		0x0020  /* Remount fs ro on errors */
 #define NILFS_MOUNT_ERRORS_PANIC	0x0040  /* Panic on errors */
-#define NILFS_MOUNT_SNAPSHOT		0x0080  /* Snapshot flag */
 #define NILFS_MOUNT_BARRIER		0x1000  /* Use block barriers */
 #define NILFS_MOUNT_STRICT_ORDER	0x2000  /* Apply strict in-order
 						   semantics also for data */
@@ -229,6 +228,7 @@ struct nilfs_super_block {
  */
 #define NILFS_CURRENT_REV	2	/* current major revision */
 #define NILFS_MINOR_REV		0	/* minor revision */
+#define NILFS_MIN_SUPP_REV	2	/* minimum supported revision */
 
 /*
  * Feature set definitions
@@ -269,6 +269,14 @@ struct nilfs_super_block {
 #define NILFS_MIN_NRSVSEGS	8	/* Minimum number of reserved
 					   segments */
 
+/*
+ * We call DAT, cpfile, and sufile root metadata files.  Inodes of
+ * these files are written in super root block instead of ifile, and
+ * garbage collector doesn't keep any past versions of these files.
+ */
+#define NILFS_ROOT_METADATA_FILE(ino) \
+	((ino) >= NILFS_DAT_INO && (ino) <= NILFS_SUFILE_INO)
+
 /*
  * bytes offset of secondary super block
  */