Re: [PATCH 2/2] binder: Use receive_fd() to receive file from another process

[Al Viro <viro@xxxxxxxxxxxxxxxxxx>]

On Fri, Apr 16, 2021 at 06:00:38PM +0200, Christian Brauner wrote:

> (dma_buf_fd() seems like another good candidate. But again, I don't have
> any plans to shove this down anyone's throat.)

Sure, there are candidates for such a helper.  Just as there are legitimate
users of anon_inode_getfd().

However, it is *NOT* something we can use as a vehicle for some hooks (pardon
the obscenity); it won't be consistently used in all cases - it simply is not
feasible for many of the fd_install() users.

Incidentally, looking at the user of receive_fd_user(), I would say that it
would be easier to follow in this form:
	case VDUSE_IOTLB_GET_ENTRY: {
		struct vduse_iotlb_entry entry;
		struct vhost_iotlb_map *map;
		struct vduse_iova_domain *domain = dev->domain;
		struct file *f = NULL;

		if (copy_from_user(&entry, argp, sizeof(entry)))
			return -EFAULT;
		entry.fd = get_unused_fd_flags(perm_to_file_flags(entry.perm));
		if (entry.fd < 0)
			return entry.fd;
		spin_lock(&domain->iotlb_lock);
		map = vhost_iotlb_itree_first(domain->iotlb,
					      entry.start, entry.start + 1);
		if (map) {
			struct vdpa_map_file *map_file = map->opaque;

			f = get_file(map_file->file);
			entry.offset = map_file->offset;
			entry.start = map->start;
			entry.last = map->last;
			entry.perm = map->perm;
		}
		spin_unlock(&domain->iotlb_lock);
		if (!f) {
			put_unused_fd(entry.fd);
			return -EINVAL;
		}
		if (copy_to_user(argp, &entry, sizeof(entry))) {
			put_unused_fd(entry.fd);
			fput(f);
			return -EFAULT;
		}
		// point of no return
		fd_install(entry.fd, f);
		return 0;
	}