Why I had a pissy weekend A short story by Chunky. I got me a nice new toy. Three, in fact. Two 933 PIIIs, and a VP6 motherboard from ABIT. I took out the old BP6, and "just stuck everything in" to the new one, in exactly the same configuration. Including drives and PCI cards. Would it work? Hell, no. That would be easy. This is the story of moving to the new motherboard... And not necessarily 100% technically correct all the way through... If it's wrong, feel free to e-mail me about it. This is based on my own empirical observations. Before I start, I'm going to say that my hardware is as follows: Primary IDE BUS, Master : Western Digital 10G [containing essentially the system] Primary IDE BUS, Slave : Iomega 100M zip Secondary IDE BUS, Master : Creative Encore 6x DVD player Secondary IDE BUS, Slave : HP9500 CD-burner Tertiary IDE BUS, Master : IBM Titan II 15G [containing essentially the games/apps] And an nVidia AGP card. And a whole fistful of PCI cards. [enough to fill my new vp6] I _had_ a BP6, with two 433 celerons. Which I'd highly recommend to anyone, by the way. Partition arrangement on the Western is: parititon 1: /boot parititon 2: / parititon 3: /home parititon 4: /usr Yes, I know it's not ideal. But I set this up a while ago, and [for example], /var is part of /. Which, in retrospect, was really dumb. I basically have a heavily modified slackware 7.1 installation, that I didn't want to have to re-build for the new machine. Contrary to what I read in a lot of forums, it's NOT necessary to do a complete re-install of linux. I think I should say that this was a purely linux exercise. I don't use windows, and I imagine that for windows, a complete reinstall would be easier. And that if you can't work out how to set up the bios settings before you start on a task as pissy as this, then mebbe you shouldn't bother starting. OK, so what did I do? I stuck everyhting in the new motherboard, with the hard drives in exactly the same configuration, and it wouldn't boot. Linux seemed to be unable to see the fourth partition of the first hard drive, and was having trouble running e2fsck on any of the partitions. I'd heard that the VP6 was not so good on the first IDE controller, so I moved the two hard drives onto the second controller and tried booting. I'd also taken out ALL the pci cards, and all the drives except the two hard drives. TWO hard drives left in the machine? This was mostly becasue I was too lazy to pull the hard drives out, as the Western won't work if it's set to master on a bus with only one drive on it, and the jumper settings were hidden behind the other hard drive. It was easier just to set the IBM to be slave and stick them on the same piece of wire than it was to make the western work on it's own. Linux, at this point, needed to be told where the root partition was. For some reason, it thought that the first drive on the tertiary bus [which should have been hde] was hdg. Which I thought was strange. I did this by handing lilo the root=/dev/hdg2 option. And it kept coming up with comments such as "hdg: lost interrupt". When I took a closer look at the loading messages, it seemed that the secondary controller was being assigned interrupt 18, which IIRC is a bad thing on an IBM-architecture block of hardware. Note: This is the point at which I realised I had a chance in hell of fixing it... I consider this the turning point in my investigation... So I disabled the primary controller in the BIOS, and the next time linux came up, the secondary controller had been assigned interrupt 16. And had been recognised as hde. Legal, IBM-friendly stuff. Once linux comes up, it mostly ignores the BIOS settings, so the primary controller began to work after the secondary one. What I'd read was true. The best config for the drives in a VP6 puts the removable media drives on the primary controller, and the fixed media drives on the primary. So, now linux can talk to the primary hard drive [the one with the system on, the Western one], without it all going tits-up. I booted with LILO: linux-242 root=/dev/hde2 nosmp single And after lots of errors about missing hard drives, partitions, etc, I hit it with the following: # /sbin/e2fsck -v -n -f /dev/hde2 # /sbin/e2fsck -v -n -f /dev/hde1 # /sbin/e2fsck -v -n -f /dev/hde3 # /sbin/e2fsck -v -n -f /dev/hde4 To verify that the important system bits still worked. The games and stuff was yet to come, and until I knew it was fairly safe, I didn't fancy mounting them. # mount -o remount,rw /dev/hde2 / # mount /dev/pts # mount /proc # mount /dev/hde1 /boot # mount /dev/hde3 /home # mount /dev/hde4 /usr Trust me, those second two are important. Without at least the second, there, when you go into runlevel 3, you won't be able to login properly. At least, that's what happened with me. Your kernel may not be compiled with PTYs support. In which case it may not be necessary. At this point, most of the system bit was working. I could tell what drives were mounted, I could read lots of important stuff. And then the bit that required a touch of a leap of faith: # init 3 And I had a working system. If you, like me, have done that "remove EVERYTHING that's not entirely necessary", you won't have a mouse plugged in. Yes, I went to that extreme. And I kept losing kb input. It's GPM. Either be creative and get rid of it, or plug in the mouse. [plugging in the mouse did fix it for me, BTW] Few more random comments: don't run vi to edit /etc/lilo.conf and /etc/fstab until your system is in runlevel 3, with / mounted rw [as I had it]. If you do, you will shaft your terminal, and I was having about a 50/50 chance of bringing it back from the dead. I'm going to statically compile up a version of vi for /sbin when I next get the chance. Hmmm. I then edited /etc/fstab and changed all instanced of /dev/hda to /dev/hde, and edited /etc/lilo.conf to make sure they pointed at the right place. Don't forget that last one. It will probably save your bacon. Gary (-; PS These were purely my experiences. My conclusions may be wrong, etc, etc, and feel free to e-mail me about it. PPS Quick rant: which muppet at ABIT made that bloody USB header-to-blanking-plate so short that I can't actually get those two USB ports on my machine? I don't use USB, but that's not the point... AfterWord [added 2001-05-11] Since I wrote that, I've been noticing some strange behaviour. The whole machine was acting strangely, lots of random SIG11s, and out-of-memory errors. Thanks to a piece of software I found called memtest86, which can be found at http://www.memtest86.com/, I found that huge tracts of my memory was broken. The software is really good, I recommend it; these guys have thought properly about what they're doing, before doing it. I know for a fact that the RAM I put in was good, because I had tested it just a few days before, on the old motherboard. The replacement has arrived, and I shall be testing again tonight/tomorrow. I seriously hope it doesn't turn out that my VP6 is smokin' 256M sticks of RAM. Gary (-; Updated 2001-05-14 Well, I got home on Friday night, and put the replacement stick, and another for luck, in the machine. 2*256M DIMMS in slots 0&1 [out of 4 possibles]. After a few of memtest86's tests, and no failures, I did some hard play^H^H^H^H testing, and it didn't bomb on me once. The next night [Saturday], I left memtest86 running overnight, and no failures... I let creatures [a VERY memory intestive app, according to it's designers] run on it's own all day, and then left memtest86 running overnight again... And as far as I know, I have half a gig of working memory in my machine. Gary (-;