Subject: Re: Minor nit with the wt0 (Wangtek Tape) driver.
To: Curt Sampson <cjs@portal.ca>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: current-users
Date: 11/12/1997 03:21:48
On Tue, 11 Nov 1997 17:27:26 -0800 (PST), 
  Curt Sampson <cjs@portal.ca> writes:

Curt,

Short answer: fix the ie driver. It's the only non-port-accouting
driver left in GENERIC/INSTALL that uses 0x300.  If necessary, do
_just_ port accounting for the 1.3 release, not an entire bus-space
rework.

>I think it is. If you need network access to install, NetBSD has
>not been installable on a machine with an SMC Ethernet card at
>0x300 for a couple of years now, and it looks as if this release
>will not be installable on those machines either. I'm not sure if
>this is due to the Mitsumi or wt, but I'd sure like to see it fixed.

I too think that this is broken.  (The port overlap of the ed driver
was one of the arguments in favour of splitting the ed driver way, way
back when, as I recall.)

But it *can't* be due to the eithe the misumi or the wt driver, since
they're not configured in either the GENERIC or the INSTALL kernels.
(Even though the mitsumi does bus-space and port accounting and can be
made `safe' by a trivial fix to its always-attach probe.)

If there's any problem, it *has* to be with the ie driver which
(to coin a phrase), ``very incorrectly'' doesn't do port allocation.

There have already been reports of the ie1 STARLAN probe breaking the
old ed2 driver. I can't see anywhere this has been specificaly fixed:
Eric S. Hvozda <hvozda@netcom.com>'s initial workaround was to build a
custom kernel without an ie1 device, and now it looks like Eric is
using PCMCIA. (or maybe Eric has a de-650 in one machine and a real
SMC card in another, I can't tell..).

If this has not been fixed, then Curt is right: *exactly* the same
problem will hit the we driver on real non-PCMCIA isa SMC cards at 0x300.
A we1 card will get properly probed and attached, and will then get
trashed by the ie driver probing for ie1.

At this point in the release cycle, maybe we should drop a
port-accounting check into the probe routine for the ISA ie driver?
Unless someone has time to bus-space the whole darn ie.c and test it??
('lo, Jason?)