Re: kernel 2.1.131 booting on Miata... but other problems...

["B. James Phillippe" <bryan terran org>]

On Fri, 4 Dec 1998, Designer 7 wrote:

> But first, just a list of some observations I've made....
...
> However, that's about where all the good things stop.
> 
> With 2.1.131 I am now having problems with my tulip driver/ethernet 
> card. First I noticed the 131 kernel comes with a very old version of 
> tulip.c ... I replaced it with the newer version that was with 2.0.36 
> and recompiled... yet, same problem. 

The version of the tulip driver that comes in the devel kernels is not
stock; there are two code paths on tulip right now (bleah).  One that
Donald does (ie. the latest) and the one in the devel kernels, which is
based on an older driver, but has fast-routing and some other tweaks.

> The symptoms of the problem are: 
> when I try to use ipfwadm, I get the error message I got when I used the 
> default ipfwadm that came with RH 5.2. Which is as if it wasn't compiled 

ipfwadm is obsolete in the new kernels; you need ipchains.  There is a good
deal of documentation in the kernel; see Documentation/Changes.  In case
you want to dual boot 2.0 and 2.1, there is an ipchains patch for 2.0.  See
the ipchains website.

> also, my routing table is awkward. Every entry is duplicated. But if I 
> shutdown all the interfaces (3 of them, eth0,eth1, eth2) and restart 
> them manually, the routing table is fine.

Two things to check here; make sure you have the latest init scripts and
the latest net-tools.  Network interfaces were changed in many ways in the
2.1 kernel.  I think there is documentation on this in the kernel, as well.
I think the specific issue you're dealing with is that routes are
automatically assigned on interface configuration now.

> the third problem I think I just figured out... for some odd reason eth0 
> and eth1 got swapped... they both use the tulip driver. So thats 
> probably why the I couldn't get out to the net using eth0.... as my net 
> connection was switched to eth1.

That's odd... not one I've heard of.  Which console do you use?  The tulip
driver works by iterating through the PCI BIOS in search of successive
devices matching the class/devid.  If it's a problem for you, there is a
#define in the tulip.c driver that can be set for reverse detection.  This
would at least allow you to keep the same interface names across 2.0/2.1
boots.  BTW, you might want to try the de4x5 driver as well.

-bp
--
B. James Phillippe	. bryan@terran.org
Linux Engineer/Admin	. http://www.terran.org/~bryan
Member since 1.1.59	. finger:bryan@earth.terran.org