Another 3Com 3C19504 Rescued
An Experience Report From Jeff Deacon
C A U T I O N
Most of the actions Jeff describes herein result in the voiding of the devices warrantee. Jeff and I take no responsibility for any damage which might result from anyone attempting to duplicate our work. In other words, you bear all responsibility for your own actions. If you break your 3C19504 then its your own fault. I do not recommend you reproduce my work! This information is offered for entertainment purposes only, no warrantee is offered or implied.
Jeff now hosts this information, and his further 3Com 3C19504 experiences on his personal website: http://inkt.co.cc/
Obviously a few years have elapsed since Christopher made his 3C19504 box work again for its living. So mine was someone else's cast off. In fact it was literally passed on to me by someone who had retrieved it from a skip (dumpster). The only cost would therefore be my time and any parts that were added or replaced. However, it must be noted that in early April 2008, dealtime.com were announcing that a couple of North American dealers had them in stock for just over $US200 each. If that information is correct, they have held their value remarkably well. But as I was to find out, they are very sturdy, reliable boxes.
My first thing to do was to determine whether there was any hope for the box and its constituent parts. Removing the cover showed that the boards and components were all clean, and in apparently good physical shape. So I took a deep breath, just in case, and plugged a monitor and a mains lead in, just to see what happened. After a bit of whirring and clattering, up came Red Hat on the screen, busy looking for all its bits and pieces. Right, we were onto a winner, turn it off in case I start to do any damage. That was the first conundrum. It responded to the off switch, but then promptly rebooted itself in just the same way that a puppy enjoying the attention of a visitor will never leave you alone! OK. Shut it down, wait for the BIOS screen to appear, and remove the mains lead. That dampened its ardour a bit!
This was a puzzle. I went back to re-read Christopher's pages, and as many of the references that remain live. And the first snag to hit me was disk size. Today, a 10GB disk is of precious little use, except perhaps solely as a POP3 mail server. And I didn't need one of them. At the back of my mind was the idea of using this box as a back up file server. As my main machine is slowly stretching itself into a 160GB disk, the backup program would have to employ a very efficient compression algorithm if this disk were to be of any use. So the project went on hold for a while.
Then increasing the disk size on another machine left me with a 20GB disk with Windows 98 installed on it. I had also put all the install files onto the disk, for those occasions when a new piece of software say "insert Windows CD" or something like that. Nothing ventured, nothing gained, I swapped the original disk for this one, plugged in the necessary keyboard, mouse and monitor, applied the power, and waited. I was quite surprised when Windows 98 actually tried to boot up. I was expecting it to require re-installation. Instead it realised that someone had moved the furniture while it was asleep, and tried to come to terms with it. "Windows has found new hardware" was its constant refrain, but it was getting somewhere. Eventually there was a working Windows desktop in 256 glorious colours at 640x480 resolution! But Windows 98 always was lousy at providing its own graphics drivers, so I was not in the least surprised. Had to do something about it though.
Also on the disk, quite by design, was a copy of Aida32. A rather nice weapon to explore strange machines with. This told me that the chipset was a SIS620, and the onboard graphics were SIS 86C306. So I could go hunting. SIS still exist, and hoorah, they still had the relevant Windows 98 drivers on their web site. A quick download, and they were mine. But this bit of exploration had been done on another PC. Given that there is no floppy drive on the 3Com box, I had to hope that the network card was working. Well. I had a choice of two, so that improved the odds a bit, but I was a bit dejected when plugging into the separate card marked LAN produced no response. But then connection on the motherboard marked WAN on the case did the business. Setting appropriate shares in Windows, and holes in the PC firewalls enabled a transfer of the drivers, swiftly followed by installation.
What a relief it was to get back into 16-bit colour, with 1024x768 resolution. But from here it was mostly plain sailing. I still haven't managed to get the other network card to show any useful signs of activity, but I currently have no need of it. I put it down to needing the same drivers as the on board device, and expect that Windows 98 can't cope with complexity like that. All the other parts work. The USB ports work, well the upper one does so reliably, and the lower one when it fancies it. The BIOS settings needed to be adjusted to allow for using PS2 mouse, instead of the serial mouse that it expected. All in all it was proving to be a nice little box to experiment with.
But if it was to earn its keep as a server, principally for back up copies of data that my business could ill afford to lose, I had to do something about the disk size. For a failed attempt to resolve the backup issue, I had previously purchased a 320GB disk. That is what I wanted inside this box, and working. It was about this time that I chanced upon an ATA133 PCI host adapter card at a reasonable price at Maplin. Best of all, it came with Windows 98 drivers. Would it work? It would certainly overcome the BIOS limitation on disk size. But one more thing was needed, and that was to overcome the 137GB disk size limit built into WIndows 98. I had been keeping the installation up to date with various updates and patches from the good folks at MDGx and MSFN. From them I found another source of reliable information at 48bitLBA.com. Amongst their excellent work was a Windows 98 patch that provides for 48 bit LBA (Logical Block Addressing) of the disk, thus providing for disks up to some phenomenal size. Well it is exceedingly large at the time of writing, by the time you read this it may well seem trivial! (As an aside, I well remember my first laptop. It had a 60 MB (yes that is right Mega-Bytes) hard drive that was going to be "Big enough for all your storage needs"!)
So the software issues seemed to be solved. But where was I going to put the new drive. When fitting the metalwork for the case, the constructor had designed obstructions and supports into all the spare space inside the box. Now I am not especially good at metalwork, and I wasn't relishing the chance to improve my skills. So there had to be a better way. Sticking tongue in cheek I put a copy of the working disk onto a partition of this new 320GB drive. The easy way to do that bit of work is to use xxcopy (see technical note 10) and gdisk, a command line partitioning and formatting utility that came packaged with Symantec/Norton Ghost. It still seems to be around, though there are several other programs calling themselves gdisk these days. I used gdisk instead of the OS supplied fdisk for one simple reason. Fdisk will only make active a partition on the first hard drive. That means it must be on a boot floppy if the hard disk has not been activated. But as we know, this 3Com box has no floppy to boot from. So gdisk, which does not have that limitation was an excellent tool. You might have something else that does the trick. Throughout the copying task, of course the box had been open, and I had inserted a Y power cable splitter to power up both disks.
Now the time had come to test my hope. I determined that if it failed, I would be off to the pub to drown my sorrows! Power down, and disconnect the 20GB disk. Power up. Adjust BIOS to confirm that there were no IDE drives on the system. In return, BIOS tells me that there is no operating system. But I knew it was there, where can it have mislaid it? Then the penny dropped. Either the BIOS needed adjusting some more, or it just wasn't going to work. So power down and restart. Hit DEL again for BIOS Setup, this is getting to be a habit! Found the boot sequence section, and here it is: a phenomenal choice of places to boot from. More than I have ever seen on any other PC I've played with. Knowing that I was intending not to connect an IDE disk, I set all four of the main choices to anything that could be right, and then let it go look anywhere else it wanted for a boot record. I wasn't hanging about. It either worked, or I was going for a beer. I didn't get my beer :-(. I don't for certain know which is the specific setting needed (though I suspect it is SCSI), but it worked and has done so reliably ever since. All the excess hardware came back out of the box, and the lid went back on it. Mission almost accomplished. But I'm getting greedy.
I now had a 500MHz system with a 320GB disk. A quick check of its capabilities showed that it would do USB sound quite nicely, but that asking it to do talking movies from highly compressed files such as .wmv was a bit beyond it. Things started to stutter unacceptably. USB pen drives work nicely in the upper connection. I have not yet investigated why the lower one does not work reliably, but my guess is rainwater while it was lying in the skip that it was rescued from.
Having located the last spot in my working area that it could be located, it was clear that connecting keyboard, mouse and monitor were out of the question. The machine had to resort to its former habits of working "headless". TweakUI Powertoy was pressed into service to make it log on to my little local network automatically. Unlike other PCs that I have tried this with, the 3Com box does it right every time; important if it is a major exercise to connect keyboard to enter the password. TightVNC server installed and set to start when Windows is booted up. Main PC has TightVNC viewer installed, and I can now work on the 3Com box from my normal screen and keyboard. In fact just to prove it works, that is how I started preparing this page! To be fair though, after the first draft, I transferred it to another PC for further improvements because I was getting annoyed with the delays between key press or mouse move, and the outcome appearing on screen.
It was far too early to be throwing this workhorse away. There is plenty of life left in it yet. The cost of the ATA 133 card was about £20. The extra large drive was just shy of £70, but it could have been much less if I had shopped around. I don't have a record of my time spent on the project, but that was because it was fun.
In the future I want to convert it to run on *nix, probably Openfiler or FreeNAS. But not yet. I am not particularly fluent in Unix type systems, and need to learn on a machine that does not have a potential driver or booting problem. Offers of help while I do my first driver compile will probably be appreciated. Or some beer to cry into when it doesn't go smoothly!
Now I had better get on with the work that I intend it to back up.