Re: [PATCH 1/8] dhcp-server: Add "authoritative" mode

[Andrew Zaborowski <andrew.zaborowski@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1500 bytes --]

On Wed, 28 Jul 2021 at 20:45, Denis Kenzior <denkenz@gmail.com> wrote:
> >>> @@ -640,6 +639,9 @@ static void listener_event(const void *data, size_t len, void *user_data)
> >>>                SERVER_DEBUG("Received DISCOVER, requested IP "NIPQUAD_FMT,
> >>>                                        NIPQUAD(requested_ip_opt));
> >>>
> >>> +             if (server_id_opt && server_id_opt != server->address)
> >>> +                     break;
> >>> +
> >>
> >> So strictly speaking, according to DHCPDISCOVER from the client shouldn't even
> >> have this option...
> >
> > Ok, I can make this check stricter here.  I preserved the original
> > logic where we would ignore the whole message if (erver_id_opt !=
> > server->address) but otherwise we weren't checking whether the
> > server_id was correct or absent.
> >
>
> Well, the thing is that you've introduced a small ambiguity.  Maybe it doesn't
> matter... but...
>
> The prior logic looked at the server_id option if present and compared to our
> own server id.  So the only way you get past the pre-check is if the server_id
> matched or was omitted entirely.
>
> But now, you can have server_id option from the client that is equal to '0', and
> we treat both the same.  So ideally this should be tightened up a bit.

True, if the client includes a 0 server_id we'd treat it as if the
option was omitted.  Let me replace uint32_t server_id_opt with a bool
server_id_opt and a bool server_id_match.

Best regards