hanifl@ultima.socs.uts.edu.au (Greyham Stoney) (10/29/90)
I have a strange and annoying problem with an Amstrad PC1512 640k and its newly installed Kyocera 20MB MFM Hard Disk and SUNIX [That's a company, not an operating system! :-] UMC-chipset Controller. What happens is that the system will run for a while, and then suddenly start generating "Sector Not Found" errors on reads and writes. A quick test program I wrote using the Turbo-C absread() function (ie: DOS device-independant direct access) which repeatedly read sectors from the hard disk and flagged any errors, showed that the chance of it hanging seemed to be related to the length of the seek required by the drive: Reading the same sector over and over again never errored Reading two sectors on adjacent tracks never errored Reading a random sector anywhere on the disk errored after 130000 tries or so Reading the first and last sector on the disk in alternation errored after 5000 tries. Once the latchup occurs, attempting to read ANY sector from the Hard Disk will yeild a Sector Not Found error; and the condition can only be cleared by a reboot (CTRL-ALT-DEL). After rebooting, those sectors it couldn't find before are alive and well (and living on the hard disk).... until the next latchup. You can still park the hard drive (by loading DEBUG from floppy and jumping to c800:9), but that doesn't help any. It doesn't seem to affect the floppy drive at all: after it's latched up, you can still read the floppy; After booting from the Hard Disk however, the floppy drive steps very noisily, suggesting that its step rate has been lowered (from what it was before I installed the HD controlled). It soon resumes its faster, quiter operation after its first access though (this may be a red herring). Weird; Ok, but it gets better: Running the Amstrad mouse driver seems to exacerbate the problem NO END. With the mouse driver installed [makes no difference whether the mouse is actually plugged in or not though] the test that alternately reads the first and last sectors on the disk errors almost immediately. Even more curiously, it seems to recover occasionally (unlike without the mouse, where once it errors, it doesn't recover) and makes griding noises like the drive is trying to recal every few errors. Ready for more?: Loading the mouse driver from autoexec.bat will hand the machine on bootup, UNLESS you also put a DRIVPARM statement in config.sys. With the DRIVPARM statement the machine will boot to the dos prompt - and then probably hang if you try anything complex. I've tried totally removing autoexec.bat and config.sys, and the error persists. Perhaps the HD bios is incompatible with something?. Like I said, not using MOUSE alleviates the problem, but doesn't eliminate it. For the record, the controller uses a UMC83C003L & UM83C001, and the bios ROM has a label on it saying "Name: 3100 Version: 9/15". I don't suppose anyone could tell me what's going on here?. A hint perhaps?. How about a vague assertion?. Would you belive a handwavy generalization? (Ah... It's the old Sector Not Found latchup trick!) Please please mail me a copy if you post anything on this. thanks! Greyham.