Re: Integrating klibc's ipconfig into netcfg
On Sun, May 14, 2006 at 11:00:59PM +0200, Geert Stappers wrote:
> On Mon, May 15, 2006 at 03:33:08AM +1000, Andrew Pollock wrote:
> > Hello,
> >
> > I've been meaning to do this for probably a year, but I've finally gotten
> > around to actually looking into what is involved in using ipconfig in netcfg
> > as a replacement to dhcp-client, so we can get rid of DHCPv2 altogether.
> >
> > The first observation I've made is that it doesn't appear to pass the
> > hostname in the request.
>
> In dhcp.c is this code snippet:
>
> | case DHCLIENT:
> | /* First, set up dhclient.conf */
> |
> | if ((dc = file_open(DHCLIENT_CONF, "w")))
> | {
> | fprintf(dc, "send dhcp-class-identifier \"d-i\";\n" );
> | if (dhostname)
> | {
> | fprintf(dc, "send host-name \"%s\";\n", dhostname);
> | }
> | fclose(dc);
> | }
>
> So only when dhostname is set, it sends the hostname in the request.
Yup, I figured it wasn't something that happened all the time.
> > I'm attempting to clarify this with the author at the moment,
> > but if someone with more C-fu than myself wants to take a look,
> > I'd appreciate it.
>
> I hope the above text did help.
>
>
> > If indeed this is the case, is this a show stopper? I recall some bugs filed
> > in the past from people trying to install on a cable connection where the
> > cable provider required the hostname in order to issue a lease, so I'm
> > assuming it is.
>
> Beside sending the hostname for cable providers,
> there is also the need to send dhcp-class-identifier "d-i",
> to allow DHCP preseeding.
I think that getting it to set random options is going to be a big ask, it's
too lightweight for that. Might be back to the drawing board...
Does that mean the that cases there are largely irrelevant? I mean pump
doesn't appear to support setting that vendor-class-identifier option...
regards
Andrew
Reply to: