Re: AIC7XXXX bug in 2.0.34

Richard B. Johnson (root@chaos.analogic.com)
Fri, 5 Jun 1998 15:44:20 -0400 (EDT)


On Fri, 5 Jun 1998, Nomad the Wanderer wrote:

[SNIPPED]
>
> Do what? Is this a problem with the driver? I can reliablly move alot
> of stuff around on my box in windows, but I've had lockups, etc in
> Linux... I just recently went to a 3940UW. I have 5 4Gig UW drives on
> channel A and 4 devices on Channel B. What Impact should I see now? Ive
> been staying in paranoid mode in that, if I have a couple of the drives
> unmounted it's more reliable and haven't tried all devices in use with
> the 3940controller...
>
> Any more details?

Windows doesn't report any errors. It's the design. That is the reason
why the machines seem to work, then stop, then you have to reinstall
Windows (with all user data lost). So "works with Windows" is not
relevant.

The driver tries its best to fix problems. However it can't recover
from the failure of an active device to respond. If it can't talk
to the device, it can't tell it to reset, etc. I did not write the
driver so I am not trying to defend it. However, since I used to use
these controllers on 4 machines I have lots of (bad) experience.

The bus interface chip cannot possibly drive more than 4 devices
correctly. It is a broken design. The output impedence (resistive
component) is 30 ohms. Simple Ohm's Law shows the problem. I tried to work
out this problem with Adaptec about a year ago. This controller was
aparently not designed by anybody who continued to work for Adaptec.

I suggest that you sell the controller to someone who has 4 SCSI devices
or less and buy a BusLogic BT-958. No. I do not work for BusLogic either.
However, the next time you download a new kernel, look at the .config
file before you do `make config`. You will see what Linus uses.

Cheers,
Dick Johnson
***** FILE SYSTEM MODIFIED *****
Penguin : Linux version 2.1.103 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu