1 files changed, 140 insertions, 38 deletions

diff --git a/fs/fhandle.c b/fs/fhandle.c
index 8a7f86c2139a..6e8cea16790e 100644
--- a/fs/fhandle.c
+++ b/fs/fhandle.c
@@ -115,88 +115,188 @@ SYSCALL_DEFINE5(name_to_handle_at, int, dfd, const char __user *, name,
 	return err;
 }
 
-static struct vfsmount *get_vfsmount_from_fd(int fd)
+static int get_path_from_fd(int fd, struct path *root)
 {
-	struct vfsmount *mnt;
-
 	if (fd == AT_FDCWD) {
 		struct fs_struct *fs = current->fs;
 		spin_lock(&fs->lock);
-		mnt = mntget(fs->pwd.mnt);
+		*root = fs->pwd;
+		path_get(root);
 		spin_unlock(&fs->lock);
 	} else {
 		struct fd f = fdget(fd);
 		if (!f.file)
-			return ERR_PTR(-EBADF);
-		mnt = mntget(f.file->f_path.mnt);
+			return -EBADF;
+		*root = f.file->f_path;
+		path_get(root);
 		fdput(f);
 	}
-	return mnt;
+
+	return 0;
 }
 
+enum handle_to_path_flags {
+	HANDLE_CHECK_PERMS   = (1 << 0),
+	HANDLE_CHECK_SUBTREE = (1 << 1),
+};
+
+struct handle_to_path_ctx {
+	struct path root;
+	enum handle_to_path_flags flags;
+	unsigned int fh_flags;
+};
+
 static int vfs_dentry_acceptable(void *context, struct dentry *dentry)
 {
-	return 1;
+	struct handle_to_path_ctx *ctx = context;
+	struct user_namespace *user_ns = current_user_ns();
+	struct dentry *d, *root = ctx->root.dentry;
+	struct mnt_idmap *idmap = mnt_idmap(ctx->root.mnt);
+	int retval = 0;
+
+	if (!root)
+		return 1;
+
+	/* Old permission model with global CAP_DAC_READ_SEARCH. */
+	if (!ctx->flags)
+		return 1;
+
+	/*
+	 * It's racy as we're not taking rename_lock but we're able to ignore
+	 * permissions and we just need an approximation whether we were able
+	 * to follow a path to the file.
+	 *
+	 * It's also potentially expensive on some filesystems especially if
+	 * there is a deep path.
+	 */
+	d = dget(dentry);
+	while (d != root && !IS_ROOT(d)) {
+		struct dentry *parent = dget_parent(d);
+
+		/*
+		 * We know that we have the ability to override DAC permissions
+		 * as we've verified this earlier via CAP_DAC_READ_SEARCH. But
+		 * we also need to make sure that there aren't any unmapped
+		 * inodes in the path that would prevent us from reaching the
+		 * file.
+		 */
+		if (!privileged_wrt_inode_uidgid(user_ns, idmap,
+						 d_inode(parent))) {
+			dput(d);
+			dput(parent);
+			return retval;
+		}
+
+		dput(d);
+		d = parent;
+	}
+
+	if (!(ctx->flags & HANDLE_CHECK_SUBTREE) || d == root)
+		retval = 1;
+	WARN_ON_ONCE(d != root && d != root->d_sb->s_root);
+	dput(d);
+	return retval;
 }
 
-static int do_handle_to_path(int mountdirfd, struct file_handle *handle,
-			     struct path *path)
+static int do_handle_to_path(struct file_handle *handle, struct path *path,
+			     struct handle_to_path_ctx *ctx)
 {
-	int retval = 0;
 	int handle_dwords;
+	struct vfsmount *mnt = ctx->root.mnt;
 
-	path->mnt = get_vfsmount_from_fd(mountdirfd);
-	if (IS_ERR(path->mnt)) {
-		retval = PTR_ERR(path->mnt);
-		goto out_err;
-	}
 	/* change the handle size to multiple of sizeof(u32) */
 	handle_dwords = handle->handle_bytes >> 2;
-	path->dentry = exportfs_decode_fh(path->mnt,
+	path->dentry = exportfs_decode_fh_raw(mnt,
 					  (struct fid *)handle->f_handle,
 					  handle_dwords, handle->handle_type,
-					  vfs_dentry_acceptable, NULL);
-	if (IS_ERR(path->dentry)) {
-		retval = PTR_ERR(path->dentry);
-		goto out_mnt;
+					  ctx->fh_flags,
+					  vfs_dentry_acceptable, ctx);
+	if (IS_ERR_OR_NULL(path->dentry)) {
+		if (path->dentry == ERR_PTR(-ENOMEM))
+			return -ENOMEM;
+		return -ESTALE;
 	}
+	path->mnt = mntget(mnt);
 	return 0;
-out_mnt:
-	mntput(path->mnt);
-out_err:
-	return retval;
+}
+
+/*
+ * Allow relaxed permissions of file handles if the caller has the
+ * ability to mount the filesystem or create a bind-mount of the
+ * provided @mountdirfd.
+ *
+ * In both cases the caller may be able to get an unobstructed way to
+ * the encoded file handle. If the caller is only able to create a
+ * bind-mount we need to verify that there are no locked mounts on top
+ * of it that could prevent us from getting to the encoded file.
+ *
+ * In principle, locked mounts can prevent the caller from mounting the
+ * filesystem but that only applies to procfs and sysfs neither of which
+ * support decoding file handles.
+ */
+static inline bool may_decode_fh(struct handle_to_path_ctx *ctx,
+				 unsigned int o_flags)
+{
+	struct path *root = &ctx->root;
+
+	/*
+	 * Restrict to O_DIRECTORY to provide a deterministic API that avoids a
+	 * confusing api in the face of disconnected non-dir dentries.
+	 *
+	 * There's only one dentry for each directory inode (VFS rule)...
+	 */
+	if (!(o_flags & O_DIRECTORY))
+		return false;
+
+	if (ns_capable(root->mnt->mnt_sb->s_user_ns, CAP_SYS_ADMIN))
+		ctx->flags = HANDLE_CHECK_PERMS;
+	else if (is_mounted(root->mnt) &&
+		 ns_capable(real_mount(root->mnt)->mnt_ns->user_ns,
+			    CAP_SYS_ADMIN) &&
+		 !has_locked_children(real_mount(root->mnt), root->dentry))
+		ctx->flags = HANDLE_CHECK_PERMS | HANDLE_CHECK_SUBTREE;
+	else
+		return false;
+
+	/* Are we able to override DAC permissions? */
+	if (!ns_capable(current_user_ns(), CAP_DAC_READ_SEARCH))
+		return false;
+
+	ctx->fh_flags = EXPORT_FH_DIR_ONLY;
+	return true;
 }
 
 static int handle_to_path(int mountdirfd, struct file_handle __user *ufh,
-		   struct path *path)
+		   struct path *path, unsigned int o_flags)
 {
 	int retval = 0;
 	struct file_handle f_handle;
 	struct file_handle *handle = NULL;
+	struct handle_to_path_ctx ctx = {};
 
-	/*
-	 * With handle we don't look at the execute bit on the
-	 * directory. Ideally we would like CAP_DAC_SEARCH.
-	 * But we don't have that
-	 */
-	if (!capable(CAP_DAC_READ_SEARCH)) {
-		retval = -EPERM;
+	retval = get_path_from_fd(mountdirfd, &ctx.root);
+	if (retval)
 		goto out_err;
+
+	if (!capable(CAP_DAC_READ_SEARCH) && !may_decode_fh(&ctx, o_flags)) {
+		retval = -EPERM;
+		goto out_path;
 	}
+
 	if (copy_from_user(&f_handle, ufh, sizeof(struct file_handle))) {
 		retval = -EFAULT;
-		goto out_err;
+		goto out_path;
 	}
 	if ((f_handle.handle_bytes > MAX_HANDLE_SZ) ||
 	    (f_handle.handle_bytes == 0)) {
 		retval = -EINVAL;
-		goto out_err;
+		goto out_path;
 	}
 	handle = kmalloc(struct_size(handle, f_handle, f_handle.handle_bytes),
 			 GFP_KERNEL);
 	if (!handle) {
 		retval = -ENOMEM;
-		goto out_err;
+		goto out_path;
 	}
 	/* copy the full handle */
 	*handle = f_handle;
@@ -207,10 +307,12 @@ static int handle_to_path(int mountdirfd, struct file_handle __user *ufh,
 		goto out_handle;
 	}
 
-	retval = do_handle_to_path(mountdirfd, handle, path);
+	retval = do_handle_to_path(handle, path, &ctx);
 
 out_handle:
 	kfree(handle);
+out_path:
+	path_put(&ctx.root);
 out_err:
 	return retval;
 }
@@ -223,7 +325,7 @@ static long do_handle_open(int mountdirfd, struct file_handle __user *ufh,
 	struct file *file;
 	int fd;
 
-	retval = handle_to_path(mountdirfd, ufh, &path);
+	retval = handle_to_path(mountdirfd, ufh, &path, open_flag);
 	if (retval)
 		return retval;