Adaptec 2940 controllers

[Mike Bilow]

Jerry Feldman wrote in a message to Mike Bilow:

 JF> I recently tried to upgrade the former Boston Computer Society 
 JF> server from  an old version of Slackware. The system has a 
 JF> 2940w controller which is one  of the ones that the 2.0 kernels 
 JF> don't like. I have tried several flavors of 2.0 kernels mostly 
 JF> 2.0.18 and 2.0.32.  Other than obtaining a different controller 
 JF> (which I am planning to do),  does anyone know of a driver or 
 JF> patch that could solve the problem. 

Yes, I've been participating in driver development for the Adaptec controllers
and chipsets since the very early days of Linux.  I'll dig up the necessary
information for you and post it here.

 JF> The symptom is that during boot, the kernel detects the SCSI 
 JF> controller,  but does not detect any devices. During 
 JF> installation from ATAPI CD ROM, the  controller is properly 
 JF> probed.

This is kind of strange.  Do you have the controller configured in Ctrl-A to
send "START UNIT" to the devices?  Many hard drives will not answer until this
is done, or at least can be jumpered to act this way.  What is the exact model
number (AHA-2940U, AHA-2940UW, etc.) and ROM revision on your controller?  Is
SCAM enabled, if available in your ROM?  Is auto-termination emabled?

 JF> Most of the messages that I have seen posted  to newsgroups say 
 JF> essentially "don't use adaptec 2940" with Linux. 

Don't panic.  I strongly disagree with advice against the 2940 series.  While
Adaptec has not been the most helpful and cooperative of hardware
manufacturers, the drivers are in a fairly mature state by now and should be
expected to work. There are known problems affecting some the very latest
kernels, such as 2.0.32, but there are simple patches available.

 JF> I am not averse to replacing the controller, but since the
 JF> system is  co-located at an ISP, access to it is limited. If
 JF> anyone can point me to a  reasonable patch, I am planning on
 JF> working on the system for an hour or 2  tomorrow.

I will make a point of posting some information by later this evening.  The
mailing list for AHA-2940 development is actually "aic7xxx," named after the
chipset.  (The actual chipset on the AHA-2940 is AIC-7880.)
 
-- Mike