braun@m10ux.UUCP (MHx7079 mh) (10/21/87)
I've been looking at all these postings about hard disk controllers that won't work under Minix, and I am starting to wonder. Is the hardware interface to the PC disk controllers screwed up, or what? I assume that by now, the Minix driver code has gotten the general bugs out of it. Is the problem due to different controllers trying to remain backwards (bug-for-bug) compatible with the original PC controller? Does the wierdness of the DMA controller have anything to do with it? The reason that I wonder is that I have put together SCSI disk and tape systems for my Z-80 CP/M computer, and I found that once my SCSI host adapter and driver code worked correctly, I had none of the types of problems I see in the PC world. In fact, it took about an hour to plug a SCSI tape drive into the system and get it working. Is this another example of IBM's "technical excellence"? -- Doug Braun AT+T Bell Labs, Murray Hill, NJ m10ux!braun 201 582-7039
jcmorris@mitre-bedford.ARPA (Joseph C. Morris) (10/22/87)
References: A recent posting to comp.os.minix observed that there is a large number of special cases which have to be handled when using some of the brand "X" hard disk controllers. While IBM can be faulted to some degree for not having put the best gee-whiz controller in their machines, it's not really fair to blame it for the failure of the clone vendors to provide products which meet the documented interface specs. (Of course, there *are* problems with undocumented specs, but that's another question.) In any case, here's another gotcha: in a PC or XT with both a hard disk and a Hardcard (the hard-disk-on-a-card) it turns out that the microcode in the Hardcard is using interrupt vectors 7C-7D-7E-7F to store some pointers. The vendor's documentation doesn't indicate that this is being done. (The PC documentation identifies these vectors - at addresses 1F0-1FF - as "unused".) These data areas are used ONLY if the PC has both a normal hard disk controller AND a Hardcard. There is no problem if the Hardcard is the only disk in the machine. DOS itself isn't bothered by the use of the memory locations, but if you have some other program using those addresses you've got a problem. In our case it was YTERM, which documents its use of interupts 7D-7E-7F. I don't know if MINIX will be affected, but with more people trying to make Minix and DOS coexist and needing more disk space to do it in, someone is likely to be zapped by this. Joe Morris
jcmorris@mitre-bedford.ARPA (Joseph C. Morris) (04/15/88)
A while back I posted a warning about the Plus International HardCard and its nasty habit of using some vectors without telling you about it. In case anyone's held off from a HardCard because of that, I finally was able to talk about the problem with the product manager, who informed me that the HardCard 20 meg devices manufactured since November, and all 40 meg units, have relocated the data from main memory to a scratchpad within the HardCard itself. Owners of older 20 meg HardCards can get a no-cost upgrade kit. I haven't had a chance to try this, so I offer no warranties, and your milage may vary... Joe Morris