Tech by Androda Comm slot ethernet card

[Arbee]

It's a weird limitation because the MAC address is programmed into the chip by the driver every time you use it. So if they wrote their own driver or even disassembled and modified Farallon's they could make it fully configurable without changing the hardware.

 

[Windoze]

Normally the driver reads the MAC from an I2C EEPROM on the card and programms it into the ethernet chip.
At least that is how it works on PC cards of that era.

 

[Daniël]

Normally the driver reads the MAC from an I2C EEPROM on the card and programms it into the ethernet chip.
At least that is how it works on PC cards of that era.

That does appear to be the case here, as the three MACs each correspond to a different "color" EEPROM.
Some other Farallon cards might have had that driver-based MAC addressing, but the Realtek chip is most definitely getting its MAC from the EEPROM.

 

[Reasons.]

From the Tinker Different thread, it appears the problem is that the driver itself won't work with certain MAC addresses. Their solution was to give three known-good options.

 

[Arbee]

It was fairly common to have drivers reject MAC addresses that didn't have the right manufacturer bytes. Even the ROM driver for the Quadras does it (it has 3 different Apple prefixes that it will allow).

 

[croissantking]

So what would be the best solution to improve this product?

 

[Arbee]

The best way would be to disassemble the driver enough to be rebuildable and replace the MAC address setting with loading it from a config file on the disk instead of the ROM. It wouldn't have to be a "showroom quality" disassembly, you could leave most of it machine-generated and just modify the MAC setting.

 

[NJRoadfan]

Does it work with the generic RTL8139 driver put out by RealTek for classic MacOS? CommSlot II devices are PCI, so the driver should work. The driver even comes with instructions for changing the PCI vendor and device IDs to work with a particular card.

 

[croissantking]

I hope that it gets those improvements down the line. The card has divided opinion here but personally, given it’s the only affordable way to get Ethernet on a Comm Slot II Mac, without using up a valuable PCI slot, I’m happy it exists.

 

[Daniël]

In any case, even when deciding on a fixed pool of reused MAC addresses, I'm not sure why such a low number was decided upon.
I can imagine quite a few scenarios where someone would have more than three Macs that could use such a card.

Say, you want to add Ethernet to a PM4400 with the CSII riser, a PM6400, a PM5500 and a TAM, you're already running into a MAC address conflict.
Sure, you could sacrifice a PCI slot in any of them, but this product was made with the aim of eliminating that entirely...

 

[zigzagjoe]

Considering the driver is almost certainly looking at the first 3 octets of the MAC and doesn't care about the tail 3, while it's not properly kosher to issue new MACs using the OEM's range it'd practically be unlikely to cause an issue with other vintage hardware.

Given that these ROMs are being programmed with the MAC it would be simple to have the programmer jig keep count of a serial number and increment it on each ROM programmed. Start at AA:BB:CCD:Ex:xx and increment x each time. Ideally, the MAC programmed into the card should be written on the card, but I'd call that excusable on a small run product.

Done, and with less effort than needing to track 3 type of eeprom, 3 different card SKUs, etc....