Re: tulip module no longer works on a500
On 10/9/05, Grant Grundler <grundler@parisc-linux.org> wrote:
> On Sun, Oct 09, 2005 at 05:29:19PM -0500, David H. Barr wrote:
> > Short summary: onboard DEC 21142/43 can't / won't even come up
> I read your bug report (good notes, btw)
Thanks! My first real go round.
> (a) why do install kernel and resulting installed kernel
> enumerate NICs in different order?
> AFAIK, NICs are only discovered in "PCI Bus" order.
> Only way to change the ethX names is with nameif.
I'm going to entitle this one "It was late and I was tired" (thanks
Gary Larson!) I have not been able to reproduce this behaviour on
subsequent attempts.
It -is- worth noting that the addin NIC won't come up off a hard
boot. I have to hardboot, login over ttyS0, and issue a soft
reboot. I suspect this behaviour coupled with my cable swapping
led me to mis-diagnose as an eth0/eth1 switch.
> (b) fixes to support 21142 Phy (your add-on card)
> only got posted a few monthes ago.
Aren't BOTH of my net interfaces 21142 Phy? The only difference
is "rev 21" vs "rev 30" AFAICS.
> I don't expect your add-on NIC to work with 2.6.8-2.
Quite bizarre. I've just finished a netinstall of debian-base and
apt-get dist-upgraded my way to testing via that very add-in NIC.
Still using 2.6.8-2-32-smp; want to try a newer uni-proc kernel later.
> (c) built-in NIC should work fine. No idea what's going on here.
> I'm not aware of any code changes that might break this.
Tried a fresh net install with -no- additional cards in (just the naked
c200 + drives) and I can of course get a basic system, but none of
the things mentioned so far in this thread can bring up the onboard tulip.
> > http://dhbarr.freeshell.org/c200/lspci-vv.txt
> > http://dhbarr.freeshell.org/c200/dmesg.txt
>
> eth0: Digital DS21143 Tulip rev 33 at 0xf2802000, 00:60:B0:B2:F8:65, IRQ 99.
> eth1: Digital DS21143 Tulip rev 48 at 0xf2803000, 00:60:B0:FD:B4:59, IRQ 96.
>
> d) why is this kernel reporting MMIO space?
> Other kernels used IO Port space. At least I thought it did...maybe
> the 0x0 IO Port space address is just a red herring.
> My preference is the kernel use MMIO space at this point.
> Older kernels might still need to use IO Port pace to support
> card-mode Dino Tulips (aka GSC 100BT cards).
I'm going to have to chew / google on this one, as it's a bit over my head.
We'll see if I make any headway tomorrow with the new kernel.
Reply to: