Subject: Re: lo(4) as a clonable interface
To: None <>
From: Andrew Brown <>
List: tech-net
Date: 07/24/2004 01:24:19
On Fri, Jul 23, 2004 at 10:20:59PM +0200, Peter Postma wrote:
>On Fri, Jul 23, 2004 at 09:58:22PM +0200, Quentin Garnier wrote:
>> I made a similar patch a while ago, but then got discouraged by the uses
>> of loif[0], mostly in the INET6 code.  I think your patch (as mine did)
>> makes those uses even more hacky than they currently are, so I don't see
>> the overall benefit.
>> I'd like to see the loif[0] uses fixed somehow, they seems wrong to me.
>Hmmm, I've been thinking about making the loop interface mandatory and
>remove it from the kernel config. There are many places where the code
>assumes to have a loop interface (like in inet6 code). And the kernel
>doesn't even compile without 'pseudo-device loop'...

if you can clean up the interaction between INET6 and loopback
interfaces in general (i tried this once, but gave up because i
couldn't figure out how to make the second interface not automagically
get ::1 as an address), make the loopback interface non-optional, make
the first instance (ie, lo0) indestructable (you mustn't be able to
remove it), and have it be pre-created when the kernel boots (perhaps
even in a general manner), this would all be cool.  :)

|-----< "CODE WARRIOR" >-----|             * "ah!  i see you have the internet (Andrew Brown)                that goes *ping*!"       * "information is power -- share the wealth."