Dumpster diving for PC Cards at Akihabara Junk Shops

When I was in Japan two weeks ago, I took a trip to Akihabara and found a few junk shops. Well, these places are just second hand stores for used computer hardware and software. There is quite the range of shops. Some are nice, neat, well lit and have plenty of room to move about. Others aren't quite so nice and are very cramped. The further you get away from the street level and from the main streets, the less polished these stores are. The harder it is for a non-Japanese speaker like myself finds it to do business. This was my first trip to Akihabara that I didn't have Japanese speakers to help me out, so I didn't get too far away from the beaten path. I managed to find about 20 PC Cards at these shops, and spent only 2500 yen for them.

First, the cards that worked. The CyQ've ELA-010 and NEC PC-9801N-J12 both worked with the ed driver without changes to the drivers. The TDK LAC-CD02x worked with the fe driver, again without changes. In a way, these were the disappointing cards in the lot. While I didn't have examples of them in my collection, I also didn't expand FreeBSD's support for PC Cards at all.

I was able to modify the fe driver to support two additional cards that I found. One was the Hitachi HT-4840-11. I had to update the fe driver to select an alternative CFE. What's a CFE? It is the Configuration Function Entry in the CIS. These nodes in the CIS describe the resources that the card will use, as well as control different ways the hardware interacts with the system. The PC Card infrastructure usually selects the proper CFE based on the available resources. However, there are cases like this where some CFEs don't work for whatever reason and alternatives need to be selected. That's what happened here. The HT-4840-11 CFE entries less than 10 appear to be designed to be used in a PC-9801 and PC-9821 laptop. They don't work in my modern HP laptop at all. Entries 10 and above all work fine.

The second fe card I was able to make work was the CONTEC CNETPC. Its first entry had a similar problem to the first few of the Hitachi HT-4840-11, but the rest were good. A quick modification to my earlier modification for the Hitachi and I was supporting both. Well, with one caveat. I can't get this card to get any link with any of the hubs that I have. I don't know if this is because the dongles got mixed up over time and I didn't have a good dongle, or if other parts of the card are 'blown out' or otherwise messed up. I had several cards like this.

Next up was a Toshiba JT Ethernet card. It worked out of the box with the ed driver. However, there was an issue. The MAC address showed up as 00:20:16:ff:ff:ff which seemed odd. Especially since 002016016b31 was labelled on the card. Digging into this deeper showed that the card has the bogus-looking address in the CIS. None of the other standard ways to get a MAC produced anything other than 0's. There's no aux entries in the CIS that could be mined for this data either. Since that's the case, I'm almost convinced that this is working the best it can given the circumstances. Also, there was no good dongle for this card...

There were a number of other boring cards. These were cards that I was able to add to the drivers more easily. The NextComK.K NC5310B Ver1.0 seems to work with the fe driver, but I don't know if the driver I have works or not. Also the MACNICA ME1-JEIDA worked with the ed driver, but the dongle didn't seem to be the right one.

Now, on to the odd-balls.

The NEC PC-9801N-J02R is a card made especially for PC-9801 and PC-9821 laptops. I've never been able to make this work with the smc driver that's in the tree. The card has a sonic chip in it, and it should just work. However, it is extremely limited to where it can be mapped in the I/O space (only at 0x880!), and none of the laptops I've ever had worked with this address.

The TDK LAC-CD011 I found turns out to be rare. At one time they were expensive, but no longer. Sadly, I didn't get dongles for this card. There's sources on the web that document it not working with the fe driver, but that it may have the MB86960A inside, which we might be able to make work. Plus, the NetBSD mbe driver supports the CD010, so I may fiddle with the fe driver to see if I can make this work...

The Toshiba LANCT00A doesn't work out of the box at all. The only information about this card on the net today is that once there was a web page that Toshiba said it was a custom-built card. Chances are good that there will never be a driver for this card.

I'll take a look at the other cards another time. They are even weirder, I fear...

1 comment:

Anonymous said...

I recently came accross your blog and have been reading along. I thought I would leave my first comment. I dont know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.