Re: PATCH for NET=n

Zlatko Calusic (Zlatko.Calusic@CARNet.hr)
30 Dec 1997 18:28:06 +0100


alan@lxorguk.ukuu.org.uk (Alan Cox) writes:

> > >>No - if you say N to CONFIG_NET you should get no networking - not even
> > >>support code for AF_UNIX
> >
> > Fine but You should still get a chance to specify CONFIG_UNIX=y
>
> Only if you say Y to CONFIG_NET - the stuff AF_UNXI depends upon is sockets
>
> > See above... If You specify CONFIG_NET=Y and try to disable anything but
> > CONFIG_UNIX in the networking setup portin you will still get a lot of
> > things in which arent really needed for CONFIG_UNIX only (Did You ever
> > have a look at how many things are using the CONFIG_NET switch in random
> > kernel parts before judgement on the patch I sendid?)
>
> Yes I had a brief look at it - it sucks in network device support, thats
> the only chunk that it shouldnt have draw in. However it is that which wants
> cleaning up _not_ the configuration breaking
>
>

Speaking of CONFIG_NET, small change is needed to compile kernel
without net support.

In file net/core/sock.c, #ifdef is needed around one case statement:

case SO_BINDTODEVICE:
...

in line 290 (at least in 2.1.76).

Problems arise from ip_rt_put() and ip_route_output() functions, which
appear as undefined in the last linking stage of kernel building.

I'm leaving to more clever heads to decide what #ifdef to apply. :)

Regards!

-- 
Posted by Zlatko Calusic           E-mail: <Zlatko.Calusic@CARNet.hr>
---------------------------------------------------------------------
	       Is it ok to use my AM radio after NOON?