[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Dock I's scsi and Linux
> It is my understanding that the built-in scsi adapter in the
> Dock I is a proprietary design.
It uses the 8-bit Future Domain TMC-950 chip.
> Does anyone know if Linux will recognize the built-in scsi of the
> Dock I? I.e.
> (1) Can one build a Linux kernel by selecting an appropriate
> scsi device driver during the configuration part of a
> kernel build? If so, what scsi driver should one use?
Yes, although with possible caveats -- see below.
Here's a summary of my experience with the Dock I SCSI and older
versions (1.2.x and 1.3.x) of Linux that I posted to the list
in June 1996. I don't know how much of this still applies to more
recent 2.0.x kernels.
==================================================
First of all, I did my initial Linux install (a 1.2.13 kernel from a
Slackware-based distribution) from a borrowed Adaptec 1542CF installed in
the Dock I's slot (at the time, I didn't have the appropriate external
cable for the Dock I's SCSI).
Eventually I got a cable, and was able to get the dock's internal SCSI
adapter recognized, using the bootstring of the form "tmc8xx=<addr>,<irq>"
However, the adapter returned junk from the device identification bus probe
at bootup, and none of the attached devices worked.
After hacking around, I found that removing the compile-time "ARBITRATION"
option from the tmc8xx driver allowed things to mostly work OK, and I
was able to access an attached external 3x CD-ROM. So, I attached an
an additional external SCSI hard drive as well, and built a filesystem,
etc. However, that didn't work so well -- in particular, I'd observe
bus hangs and file corruption on the SCSI hard disk when writing large
files (16mb or so).
Next, I investigated some of the development kernels. I noticed the
changes summary for 1.3.50 mentioned some fixes for the seagate driver.
(The Seagate ST0x and Future Domain TMC-950-based boards, which is what's
in the Dock I, are almost identical, and are both implemented in the
kernel source module "drivers/scsi/seagate.c".) So, I repeated my
experiments with 1.3.85. I still had to remove the "ARBITRATION" option,
but I didn't experience any bus hangs or file corruption, after working
the SCSI devices fairly hard. I was also successful in attaching and
using an external DAT drive (I didn't try that with 1.2.13 at all).
Final summary:
- I needed to remove the "ARBITRATION" option in all cases.
- the driver in 1.2.13 is flakey under load, but seems solid
under 1.3.85 (I picked 1.3.85 more-or-less at random, by the way; my
guess is that it's solid in anything above 1.3.50).
- performance is fairly poor; the hardware really isn't very capable.
- you're very unlikely to be able to do an installation from the Dock 1
SCSI.
Hope this helps somebody. BTW, all this was with a ThinkPad 360Cs with
20 megs of memory.
--
Tim Tyhurst tim@excelsior.com