hostfs: Fix ephemeral inodes

hostfs creates a new inode for each opened or created file, which
created useless inode allocations and forbade identifying a host file
with a kernel inode.

Fix this uncommon filesystem behavior by tying kernel inodes to host
file's inode and device IDs. Even if the host filesystem inodes may be
recycled, this cannot happen while a file referencing it is opened,
which is the case with hostfs. It should be noted that hostfs inode IDs
may not be unique for the same hostfs superblock because multiple host's
(backed) superblocks may be used.

Delete inodes when dropping them to force backed host's file descriptors
closing.

This enables to entirely remove ARCH_EPHEMERAL_INODES, and then makes
Landlock fully supported by UML. This is very useful for testing
changes.

These changes also factor out and simplify some helpers thanks to the
new hostfs_inode_update() and the hostfs_iget() revamp: read_name(),
hostfs_create(), hostfs_lookup(), hostfs_mknod(), and
hostfs_fill_sb_common().

A following commit with new Landlock tests check this new hostfs inode
consistency.

Cc: Anton Ivanov <[e-mail]>
Cc: Johannes Berg <[e-mail]>
Acked-by: Richard Weinberger <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Mickaël Salaün <[e-mail]>

arch/Kconfig

···
 config ARCH_NO_PREEMPT
 	bool
 
-config ARCH_EPHEMERAL_INODES
-	def_bool n
-	help
-	  An arch should select this symbol if it doesn't keep track of inode
-	  instances on its own, but instead relies on something else (e.g. the
-	  host kernel for an UML kernel).
-
 config ARCH_SUPPORTS_RT
 	bool
 

arch/um/Kconfig

···
 config UML
 	bool
 	default y
-	select ARCH_EPHEMERAL_INODES
 	select ARCH_HAS_FORTIFY_SOURCE
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_HAS_KCOV

fs/hostfs/hostfs.h

···
 	unsigned long long blocks;
 	unsigned int maj;
 	unsigned int min;
+	dev_t dev;
 };
 
 extern int stat_file(const char *path, struct hostfs_stat *p, int fd);

fs/hostfs/hostfs_kern.c

···
 	fmode_t mode;
 	struct inode vfs_inode;
 	struct mutex open_mutex;
+	dev_t dev;
 };
 
 static inline struct hostfs_inode_info *HOSTFS_I(struct inode *inode)
···
 	return ERR_PTR(n);
 }
 
-static struct inode *hostfs_iget(struct super_block *sb)
-{
-	struct inode *inode = new_inode(sb);
-	if (!inode)
-		return ERR_PTR(-ENOMEM);
-	return inode;
-}
-
 static int hostfs_statfs(struct dentry *dentry, struct kstatfs *sf)
 {
 	/*
···
 		return NULL;
 	hi->fd = -1;
 	hi->mode = 0;
+	hi->dev = 0;
 	inode_init_once(&hi->vfs_inode);
 	mutex_init(&hi->open_mutex);
 	return &hi->vfs_inode;
···
 	if (HOSTFS_I(inode)->fd != -1) {
 		close_file(&HOSTFS_I(inode)->fd);
 		HOSTFS_I(inode)->fd = -1;
+		HOSTFS_I(inode)->dev = 0;
 	}
 }
 
···
 static const struct super_operations hostfs_sbops = {
 	.alloc_inode	= hostfs_alloc_inode,
 	.free_inode	= hostfs_free_inode,
+	.drop_inode	= generic_delete_inode,
 	.evict_inode	= hostfs_evict_inode,
 	.statfs		= hostfs_statfs,
 	.show_options	= hostfs_show_options,
···
 	.write_end	= hostfs_write_end,
 };
 
-static int read_name(struct inode *ino, char *name)
+static int hostfs_inode_update(struct inode *ino, const struct hostfs_stat *st)
 {
+	set_nlink(ino, st->nlink);
+	i_uid_write(ino, st->uid);
+	i_gid_write(ino, st->gid);
+	ino->i_atime =
+		(struct timespec64){ st->atime.tv_sec, st->atime.tv_nsec };
+	ino->i_mtime =
+		(struct timespec64){ st->mtime.tv_sec, st->mtime.tv_nsec };
+	ino->i_ctime =
+		(struct timespec64){ st->ctime.tv_sec, st->ctime.tv_nsec };
+	ino->i_size = st->size;
+	ino->i_blocks = st->blocks;
+	return 0;
+}
+
+static int hostfs_inode_set(struct inode *ino, void *data)
+{
+	struct hostfs_stat *st = data;
 	dev_t rdev;
-	struct hostfs_stat st;
-	int err = stat_file(name, &st, -1);
-	if (err)
-		return err;
 
 	/* Reencode maj and min with the kernel encoding.*/
-	rdev = MKDEV(st.maj, st.min);
+	rdev = MKDEV(st->maj, st->min);
 
-	switch (st.mode & S_IFMT) {
+	switch (st->mode & S_IFMT) {
 	case S_IFLNK:
 		ino->i_op = &hostfs_link_iops;
 		break;
···
 	case S_IFBLK:
 	case S_IFIFO:
 	case S_IFSOCK:
-		init_special_inode(ino, st.mode & S_IFMT, rdev);
+		init_special_inode(ino, st->mode & S_IFMT, rdev);
 		ino->i_op = &hostfs_iops;
 		break;
 	case S_IFREG:
···
 		return -EIO;
 	}
 
-	ino->i_ino = st.ino;
-	ino->i_mode = st.mode;
-	set_nlink(ino, st.nlink);
-	i_uid_write(ino, st.uid);
-	i_gid_write(ino, st.gid);
-	ino->i_atime = (struct timespec64){ st.atime.tv_sec, st.atime.tv_nsec };
-	ino->i_mtime = (struct timespec64){ st.mtime.tv_sec, st.mtime.tv_nsec };
-	ino->i_ctime = (struct timespec64){ st.ctime.tv_sec, st.ctime.tv_nsec };
-	ino->i_size = st.size;
-	ino->i_blocks = st.blocks;
-	return 0;
+	HOSTFS_I(ino)->dev = st->dev;
+	ino->i_ino = st->ino;
+	ino->i_mode = st->mode;
+	return hostfs_inode_update(ino, st);
+}
+
+static int hostfs_inode_test(struct inode *inode, void *data)
+{
+	const struct hostfs_stat *st = data;
+
+	return inode->i_ino == st->ino && HOSTFS_I(inode)->dev == st->dev;
+}
+
+static struct inode *hostfs_iget(struct super_block *sb, char *name)
+{
+	struct inode *inode;
+	struct hostfs_stat st;
+	int err = stat_file(name, &st, -1);
+
+	if (err)
+		return ERR_PTR(err);
+
+	inode = iget5_locked(sb, st.ino, hostfs_inode_test, hostfs_inode_set,
+			     &st);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+
+	if (inode->i_state & I_NEW) {
+		unlock_new_inode(inode);
+	} else {
+		spin_lock(&inode->i_lock);
+		hostfs_inode_update(inode, &st);
+		spin_unlock(&inode->i_lock);
+	}
+
+	return inode;
 }
 
 static int hostfs_create(struct mnt_idmap *idmap, struct inode *dir,
···
 {
 	struct inode *inode;
 	char *name;
-	int error, fd;
+	int fd;
 
-	inode = hostfs_iget(dir->i_sb);
-	if (IS_ERR(inode)) {
-		error = PTR_ERR(inode);
-		goto out;
-	}
-
-	error = -ENOMEM;
 	name = dentry_name(dentry);
 	if (name == NULL)
-		goto out_put;
+		return -ENOMEM;
 
 	fd = file_create(name, mode & 0777);
-	if (fd < 0)
-		error = fd;
-	else
-		error = read_name(inode, name);
+	if (fd < 0) {
+		__putname(name);
+		return fd;
+	}
 
+	inode = hostfs_iget(dir->i_sb, name);
 	__putname(name);
-	if (error)
-		goto out_put;
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
 
 	HOSTFS_I(inode)->fd = fd;
 	HOSTFS_I(inode)->mode = FMODE_READ | FMODE_WRITE;
 	d_instantiate(dentry, inode);
 	return 0;
-
- out_put:
-	iput(inode);
- out:
-	return error;
 }
 
 static struct dentry *hostfs_lookup(struct inode *ino, struct dentry *dentry,
 				    unsigned int flags)
 {
-	struct inode *inode;
+	struct inode *inode = NULL;
 	char *name;
-	int err;
 
-	inode = hostfs_iget(ino->i_sb);
-	if (IS_ERR(inode))
-		goto out;
-
-	err = -ENOMEM;
 	name = dentry_name(dentry);
-	if (name) {
-		err = read_name(inode, name);
-		__putname(name);
+	if (name == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	inode = hostfs_iget(ino->i_sb, name);
+	__putname(name);
+	if (IS_ERR(inode)) {
+		if (PTR_ERR(inode) == -ENOENT)
+			inode = NULL;
+		else
+			return ERR_CAST(inode);
 	}
-	if (err) {
-		iput(inode);
-		inode = (err == -ENOENT) ? NULL : ERR_PTR(err);
-	}
- out:
+
 	return d_splice_alias(inode, dentry);
 }
 
···
 	char *name;
 	int err;
 
-	inode = hostfs_iget(dir->i_sb);
-	if (IS_ERR(inode)) {
-		err = PTR_ERR(inode);
-		goto out;
-	}
-
-	err = -ENOMEM;
 	name = dentry_name(dentry);
 	if (name == NULL)
-		goto out_put;
+		return -ENOMEM;
 
 	err = do_mknod(name, mode, MAJOR(dev), MINOR(dev));
-	if (err)
-		goto out_free;
+	if (err) {
+		__putname(name);
+		return err;
+	}
 
-	err = read_name(inode, name);
+	inode = hostfs_iget(dir->i_sb, name);
 	__putname(name);
-	if (err)
-		goto out_put;
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
 
 	d_instantiate(dentry, inode);
 	return 0;
-
- out_free:
-	__putname(name);
- out_put:
-	iput(inode);
- out:
-	return err;
 }
 
 static int hostfs_rename2(struct mnt_idmap *idmap,
···
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
 	err = super_setup_bdi(sb);
 	if (err)
-		goto out;
+		return err;
 
 	/* NULL is printed as '(null)' by printf(): avoid that. */
 	if (req_root == NULL)
 		req_root = "";
 
-	err = -ENOMEM;
 	sb->s_fs_info = host_root_path =
 		kasprintf(GFP_KERNEL, "%s/%s", root_ino, req_root);
 	if (host_root_path == NULL)
-		goto out;
+		return -ENOMEM;
 
-	root_inode = new_inode(sb);
-	if (!root_inode)
-		goto out;
-
-	err = read_name(root_inode, host_root_path);
-	if (err)
-		goto out_put;
+	root_inode = hostfs_iget(sb, host_root_path);
+	if (IS_ERR(root_inode))
+		return PTR_ERR(root_inode);
 
 	if (S_ISLNK(root_inode->i_mode)) {
-		char *name = follow_link(host_root_path);
-		if (IS_ERR(name)) {
-			err = PTR_ERR(name);
-			goto out_put;
-		}
-		err = read_name(root_inode, name);
+		char *name;
+
+		iput(root_inode);
+		name = follow_link(host_root_path);
+		if (IS_ERR(name))
+			return PTR_ERR(name);
+
+		root_inode = hostfs_iget(sb, name);
 		kfree(name);
-		if (err)
-			goto out_put;
+		if (IS_ERR(root_inode))
+			return PTR_ERR(root_inode);
 	}
 
-	err = -ENOMEM;
 	sb->s_root = d_make_root(root_inode);
 	if (sb->s_root == NULL)
-		goto out;
+		return -ENOMEM;
 
 	return 0;
-
-out_put:
-	iput(root_inode);
-out:
-	return err;
 }
 
 static struct dentry *hostfs_read_sb(struct file_system_type *type,

fs/hostfs/hostfs_user.c

···
 	p->blocks = buf->st_blocks;
 	p->maj = os_major(buf->st_rdev);
 	p->min = os_minor(buf->st_rdev);
+	p->dev = buf->st_dev;
 }
 
 int stat_file(const char *path, struct hostfs_stat *p, int fd)

security/landlock/Kconfig

···
 
 config SECURITY_LANDLOCK
 	bool "Landlock support"
-	depends on SECURITY && !ARCH_EPHEMERAL_INODES
+	depends on SECURITY
 	select SECURITY_PATH
 	help
 	  Landlock is a sandboxing mechanism that enables processes to restrict