2001-12-25 16:17:09

by Manfred Spraul

[permalink] [raw]
Subject: Re:

> When I went to build 2.4.17 on a dinky box (486, 16M RAM), the
> config option was missing. The box is a wall mount and is not very
> capable of multiple kernel experimentation alas. Can someone
> supply some background as to what has happened?

It seems that RTNETLINK is now unconditionally enabled, I don't know
why.

--
Manfred




2001-12-25 19:14:58

by Jeff Garzik

[permalink] [raw]
Subject: Re:

On Tue, Dec 25, 2001 at 05:17:01PM +0100, Manfred Spraul wrote:
> > When I went to build 2.4.17 on a dinky box (486, 16M RAM), the
> > config option was missing. The box is a wall mount and is not very
> > capable of multiple kernel experimentation alas. Can someone
> > supply some background as to what has happened?
>
> It seems that RTNETLINK is now unconditionally enabled, I don't know
> why.

It's required by newer RedHat and MDK initscripts, perhaps others.
ip, iproute and similar utilities use it, and so since it's commonly
required DaveM made it unconditional... I think the checkin comment was
something along the lines of "make it unconditional unless Alan
complains about kernel bloat" :)

Jeff


2001-12-25 19:25:09

by J Sloan

[permalink] [raw]
Subject: Re: RTNETLINK

Legacy Fishtank wrote:

> It's required by newer RedHat and MDK initscripts, perhaps others.
> ip, iproute and similar utilities use it, and so since it's commonly
> required DaveM made it unconditional... I think the checkin comment was
> something along the lines of "make it unconditional unless Alan
> complains about kernel bloat" :)

Hmm, perhaps RTNETLINK should be enabled
IFF networking is selected? I see to remember
that was the idea being bandied about...

Anyway, Merry Christmas to all -

jjs

2001-12-25 21:30:20

by Kurt Roeckx

[permalink] [raw]
Subject: Re:

On Tue, Dec 25, 2001 at 02:14:41PM -0500, Legacy Fishtank wrote:
> On Tue, Dec 25, 2001 at 05:17:01PM +0100, Manfred Spraul wrote:
> > It seems that RTNETLINK is now unconditionally enabled, I don't know
> > why.
>
> It's required by newer RedHat and MDK initscripts, perhaps others.
> ip, iproute and similar utilities use it, and so since it's commonly
> required DaveM made it unconditional... I think the checkin comment was
> something along the lines of "make it unconditional unless Alan
> complains about kernel bloat" :)

But ifconfig and route don't use it, and now you can't do certain
things you could before.

One thing that comes to mind is showing the ipv6 routig cache,
because it only made that proc entry when it's not enabled.

Should I mention my kernel got bigger?

(I also was under the impression that this was a stable series.)


Kurt

2001-12-25 21:53:12

by Alan

[permalink] [raw]
Subject: Re: RTNETLINK

>
> Legacy Fishtank wrote:
>
> > It's required by newer RedHat and MDK initscripts, perhaps others.
> > ip, iproute and similar utilities use it, and so since it's commonly
> > required DaveM made it unconditional... I think the checkin comment was
> > something along the lines of "make it unconditional unless Alan
> > complains about kernel bloat" :)
>
> Hmm, perhaps RTNETLINK should be enabled
> IFF networking is selected? I see to remember
> that was the idea being bandied about...
>
> Anyway, Merry Christmas to all -
>
> jjs
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>

2001-12-25 21:54:07

by Alan

[permalink] [raw]
Subject: Re:

> It's required by newer RedHat and MDK initscripts, perhaps others.
> ip, iproute and similar utilities use it, and so since it's commonly

Basically because Dave refused to recognize lots of embedded setups don't
need the netlink crap and couldn't just accept defaulting it to Y we all
get lumbered with it

> required DaveM made it unconditional... I think the checkin comment was
> something along the lines of "make it unconditional unless Alan
> complains about kernel bloat" :)

And I did complain. "Red Hat needs XYZ so we make it mandatory" is not an
appropriate approach to a problem.

Alan

2002-01-03 00:09:57

by David Miller

[permalink] [raw]
Subject: Re:

From: Alan Cox <[email protected]>
Date: Tue, 25 Dec 2001 22:03:59 +0000 (GMT)

> required DaveM made it unconditional... I think the checkin comment was
> something along the lines of "make it unconditional unless Alan
> complains about kernel bloat" :)

And I did complain. "Red Hat needs XYZ so we make it mandatory" is not an
appropriate approach to a problem.

[ Just got back from British Columbia... ]

No you did not complain. I asked you specifically if it was ok, and
your response was that turning netlink/rtnetlink on by default was
fine with you.

It has zilch to do with redhat anything, in fact I had to ask vendors
first if they could still fit the kernel on their boot disks if I
added ~5K of object code to kernels with networking enabled.

It has everything to do with iproute2 and tcp_diag using it.

Franks a lot,
David S. Miller
[email protected]

2002-01-03 00:17:25

by Alan

[permalink] [raw]
Subject: Re:

> No you did not complain. I asked you specifically if it was ok, and
> your response was that turning netlink/rtnetlink on by default was
> fine with you.

But not forcing it always on - thats not what I said.

| From: Alan Cox <[email protected]>
| Date: Wed, 31 Oct 2001 13:09:16 -0500 (EST)
|
| Ask Arjan. I'm certainly of the opinion it should edefault to Y in Linus
| config.in

Alan