Adaptec 2940UW (summary)
Jason Costomiris (email@example.com) thought it appropriate to state:
>> I found a statement in the december RedHat (4.1?)
>> claiming that some, but not all Adaptec have problems
>> since 2.0.12.
To set the record straight: it´s in RedHat 4.0, in the unsupported
i386 images section. It states that their bootimage kernel (dunno
which version) had problems with some Adaptec aic7xxx cards, and
that they created another image using the aic7xxx driver from 2.0.12.
Fact: I am, right now, running a 2.0.27 patched with the drivers/scsi
from 2.0.11. The system runs w/o problems. It did not run with the
regular 2.0.27, 2.0.25, 2.0.28, 2.1.15, 2.1.16. Dunno about 2.0.29.
>I've got 9 or 10 of these cards deployed. All of the machines are running
>2.0.29 without a single problem. I've never done anything more than plug
>the card in, turn the machine on and install. No special tweaking
So you´re lucky.
Somewhere between 2.0.11 and 2.0.25, a change in the aic7xxx driver
has happened. From the looks of it, the aic7xxx insists on accessing
a wide scsi disk (IBM DORS, in my case) with 20 MHz scsi-2. This
fails, and the system gets stuck in an loop of negotiating, aborting,
reseting. The Adaptec BIOS is v1.23, 1996.
There might be a problem with cable and/or termination. I have read
that Adaptec boards are a bit picky in this regard. The 16bit cable
is from the kit, but the 8bit bus (having been in operation since
1994, with ISA Adaptec 1542, Asus on-board NCR, and PCI NCR SC200)
has seven SCSI devices and might be a bit long. With "Automatic"
termination selected, low byte termination is reported as disabled,
high byte is reported as enabled. Fits page 1-7 of the manual.
Nonetheless I will check termination once more as soon as I have
done all the other stuff that kept piling up during the last two days.
The earlier kernel accesses the disk with 10MHz no problem. So if
there´s a hardware flaw, the aic7xxx driver might as well use the
lower frequency as a fallback. I would have settled for half the
bandwidth any time during the last days. If Adaptec tolerance for
cables is indeed substandard, this fallback might be worth changing
the driver once more.
Hope that those who might encounter a similar problem will find this
summary helpful. You will probably not be able to use the older
aic7xxx with 2.1.x, I suspect.
P.S.: my gratitude to those maintaining the aic7xxx. I have read claims
during the last two days that Adaptec does not support Linux in any
regard. The 2940, 2940U and 2940UW boards I have dealt with during the
last years were okay, however, this particular one will be the first
and last one I bought myself. There are alternatives (see e.g.