2003-03-15 15:41:13

by Andre Tomt

[permalink] [raw]
Subject: problems with DFE-580TX (sundance) in 2.4.20

Hi,

I recently got hold of a D-Link DFE-580TX 4 port Server Adapter for
testing, but we're having serious issues with it. It is a card with a
Intel PCI-to-PCI bridge on it, having 4 sundance chips on their "own"
bus.

I actually got it to work pretty well on a AMD k6-2 system with VIA
chipset for the few days it was in that machine, but all other systems
have so far failed. Another VIA worked for a while, but failed after a
cold boot. Even Intel ones, both cheap "home" mainboards, and more
expensive server mainboards have otherwise failed completely. That
means, I've tested in about 5-6 machines, one working, one partly
working, the rest fails.

This is what it prints out when bailing out:
eth2: D-Link DFE-580TX 4 port Server Adapter at 0xf88f1000,
6f:ef:6f:ef:6f:ef, IRQ 12.
eth2: No MII transceiver found, aborting. ASIC status f000ef6f
eth2: D-Link DFE-580TX 4 port Server Adapter at 0xf88f1000,
6f:ef:6f:ef:6f:ef, IRQ 10.
eth2: No MII transceiver found, aborting. ASIC status f000ef6f
eth2: D-Link DFE-580TX 4 port Server Adapter at 0xf88f1000,
6f:ef:6f:ef:6f:ef, IRQ 11.
eth2: No MII transceiver found, aborting. ASIC status f000ef6f
eth2: D-Link DFE-580TX 4 port Server Adapter at 0xf88f1000,
6f:ef:6f:ef:6f:ef, IRQ 11.
eth2: No MII transceiver found, aborting. ASIC status f000ef6f

then the module unloads.

More comprehensive information is attached in a text file, to avoid
linewrapping issues. The kernelversion shown in dmesg is "almost
vanilla" - it has some bugfixes from BK applied, mostly the ext3-stuff.
The card behaves exactly the same on a pristine kernel, so I don't think
it will be a issue.


Attachments:
typescript.txt (26.59 kB)

2003-03-16 17:17:18

by Andre Tomt

[permalink] [raw]
Subject: Re: problems with DFE-580TX (sundance) in 2.4.20

On Sat, 2003-03-15 at 16:51, Andre Tomt wrote:
> Hi,
>
> I recently got hold of a D-Link DFE-580TX 4 port Server Adapter for
> testing, but we're having serious issues with it. It is a card with a
> Intel PCI-to-PCI bridge on it, having 4 sundance chips on their "own"
> bus.

Seems like we solved it, maybe I should read the warnings on MMIO access
to the NIC registers in the kernel configuration next time - I'm a bit
ashamed now :-)

I have no possibility to try other hardware setups and coldboots until
monday, but so far it seems to be working, over about 5 warm boots in
the Intel setup.

--
Mvh,
Andr? Tomt
[email protected]