Subject: Re: mount bug?
To: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
From: Thor Lancelot Simon <>
List: current-users
Date: 08/29/1994 19:37:15
(the answer to the above question is "yes".)

Since I've been pretty happy with the performance of the two BT445S
controllers I use in my VESA machines, when we bought a pile of new PCI
machines for Panix I spec'ed BusLogic PCI controllers for them.

I've had approximately zero luck making them work.

For starters, the BT946C just seems to not like the Dell that sits on my desk.
Not A Big Deal; the Dell's a pretty early PCI machine and I never really
expected to use the PCI bus on it for much.  The symptoms of this are a plain
old failure to boot either the IDE or the SCSI disk, even though the BusLogic
is recognized and can go into auto-test mode.

Not a very serious problem.  But the failure of the four 946C cards I bought
to run in the brand-new PCI 486/66 machines I bought for them to run in is
much more serious.  The symptoms are as follows:

[I'm using a kernel built from GENERICBT of approximately one-week-old

I boot from my IDE disk.  The BusLogic correctly recognizes whatever SCSI disk
I have hooked up to it and installs it as the second hard disk.  The kernel
goes through the motions of autoconfiguring all the relevant devices.  When it
gets to bt0, I get the following:

bt0 at isa0 port 0x330-0x333 irq 11: version 4.2, sync, parity, 32 mbxs

[whatever disks I have hooked up to the scsibus]

pci 0 at isa 0 port 0x0-0x665: configuration mode 1
pci 0 bus 0 device 0: identifier 05051106 class 00000000 not configured
pci 0 bus 0 device 8: identifier 0140104b class 01000000 not configured

[! uh oh !]

The machine comes up, and everything works fine *except* the SCSI disks.
Any attempt to access them leaves a process hanging in disk wait, and
eventually messages about

sd0: timed out
sd0: timed out AGAIN

start to scroll down the screen.  After quite a long time, the machine decides
sd0 isn't a configured device anymore.

This happens whether I have my PCI caches turned on or off.  It happens no
matter what IRQ I set the PCI slot for the BT946C to and no matter what IRQ I
set the card itself to.  It happens whether or not parity, fast transfers, and
sync mode negotiation are turned on in the card settings.  It happens if I
select ISA DMA channel emulation, and it happens if I don't.

I just don't get it.  Didn't people say this card worked with the BT driver?

I'm going to go give it a spin with the GENERICAHA kernel, but I'm not exactly
thrilled by that prospect, either.

I know Rob Kolstad said at Usenix that the original BT946 cards had a bug in
the mailbox protocol that made them extraordinarily slow, if workable at all,
under BSDI, but that BusLogic had fixed the problem.  This was months and
months ago; I doubt I've got four such broken cards, from two different

My motherboards are PCI/VL/ISA, using the VIA VT82C505 chipset.  Does anyone
know of bugs in this chipset?

It's possible that if nobody has any idea what's going on, I could arrange to
lend one of these machines to an appropriate core team member -- Charles, I
guess?  I'm pretty confused right about now, and we're about to roll out a
whole new set of news, mail, uucp, web, ftp, gopher, etc. servers here on NetBSD
/i386 machines; I really want to make this work.  

Anyone who can give me a clue would be very, very welcome to.  I feel pretty
foolish right now.