Subject: Dual SCSI problems
To: None <current-users@NetBSD.ORG, port-i386@NetBSD.ORG>
From: Mika Nystroem <email@example.com>
Date: 05/09/1996 03:50:42
It's me again. This time I am having SCSI problems on my dual-PCI pentium
server. The system has been working smoothly for a month, so I guess it is
about time for another set of problems.
Today, I took delivery of a nice new drive for the lab server, a
Quantum XP34300W, wide-SCSI 4.3 GB disk. I proceeded to install it
on the file server...
The system has two AIC7870P SCSI controllers on the motherboard.
They are connected to a SCSI hot-docking backplane via ribbon
cables. Since the 80-pin hot-swappable drives are more than twice
as expensive (when we checked they were, at least) per megabyte
as the Quantum, we decided to go with a standard 68-pin wide SCSI
drive. This means that it cannot (easily) be connected to the same
part of the backplane as the other disks (two 2GB hot-swappable
Barracudas) and "miscellaneous" devices (CD-ROM and 4mm tape).
So... I unplugged the second SCSI bus from the backplane and
connected the 68-pin ribbon cable directly to the drive,
which of course has the termination thing enabled.
There does not seem to be a hardware problem. When the system
boots up, the Adaptec SCSI thing that runs detects all the devices
in scsi controller 0's territory just fine, and then it detects
the Quantum on scsi 1. The NetBSD kernel comes along, it detects
the controllers fine, and then...
(I typed the following in from notes on paper, so there may be an error)
(lots of ahc0 stuff)
pci_map_int: interrupting at irq 11
ahc1 at pci1 dev 14 function 0
ahc1: aic 7870 Wide Channel SCSI Id=7, aic7870, 16 SCBs
scsibus1 at ahc1
ahc1: target 4 using 16 Bit transfers
sd2 at scsibus 1 targ 4 lun 0:<Quantum, XP34300W, L912>
sd2: 4101 MB, 3907 cyl, ...
pci_map_int: interrupting at irq 10
biomask c40 netmask e60 ttymask 1efa
(system hangs here)
Now, it doesn't matter if I leave the drive in or not; if I boot a kernel
*without* the following lines, everything is happy:
ahc1 at pci? dev ? function ? # Adaptec 2940 SCSI controllers
scsibus1 at ahc1
(That is, even if the drive is turned on, etc..) If I just put in the ahc1
line, it hangs.
If I boot a working kernel, the last message above turns into:
biomask 840 netmask a60 ttymask 1afa
Anyone have an idea what might be going on here?
I may be able to get around this by buying another cable and setting
up the hot-swapping backplane in an odd way so that all the devices
can coexist on scsibus0, but I'd really rather see it work the way
Department of Computer Science
California Institute of Technology
Pasadena, Calif., U.S.A.