Re: [PATCH v2 1/2] cfg80211: Userspace may inform kernel of mesh auth method.

[Johannes Berg <johannes@sipsolutions.net>]

On Mon, 2013-05-06 at 16:49 -0700, Colleen Twitty wrote:

> +	if (tb[NL80211_MESH_SETUP_AUTH_PROTOCOL]) {
> +		if (!setup->user_mpm)
> +			return -EINVAL;
> +		switch (nla_get_u8(tb[NL80211_MESH_SETUP_AUTH_PROTOCOL])) {
> +		case NL80211_AUTHTYPE_SAE:
> +			setup->auth_id = IEEE80211_AUTH_PROTO_SAE;
> +			break;
> +		case NL80211_AUTHTYPE_OPEN_SYSTEM:
> +			setup->auth_id = IEEE80211_AUTH_PROTO_NONE;
> +			break;
> +		default:
> +			return -EINVAL;
> +		}
> +	}

Ok one more question. Does it actually make sense to check the auth
protocol?

>From what I see, the entire auth protocol (e.g. SAE) is handled entirely
in userspace, so if somebody invents a new protocol (e.g. number 3), or
wants to use 802.1X (number 2) they should also be able to implement
that completely in userspace. As such, why validate it? It could be a
valid protocol?

Or am I missing something and there's some kernel part involved in
handling the auth protocol, so we need to check that the kernel code
actually supports it?

johannes