From owner-freebsd-bugs  Sun Jul 30 22:20:01 1995
Return-Path: bugs-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.11/8.6.6) id WAA07130
          for bugs-outgoing; Sun, 30 Jul 1995 22:20:01 -0700
Received: from gatekeeper.alpharel.com (gatekeeper.ALPHAREL.COM [204.118.5.2])
          by freefall.cdrom.com (8.6.11/8.6.6) with ESMTP id WAA07118
          ; Sun, 30 Jul 1995 22:19:59 -0700
Received: (from mail@localhost) by gatekeeper.alpharel.com (8.6.8/8.6.6a) id WAA17136; Sun, 30 Jul 1995 22:15:44 -0700
Received: from optisun40.optigfx.com(147.203.1.40) by gatekeeper.alpharel.com via smap (V1.3)
	id sma017134; Sun Jul 30 22:15:40 1995
Received: by optisun40.optigfx.COM (4.1/SMI-4.1.SI1)
	id AA16043; Sun, 30 Jul 95 22:15:38 PDT
Date: Sun, 30 Jul 95 22:15:38 PDT
From: schwarz@alpharel.com (Steve Schwarz)
Message-Id: <9507310515.AA16043@optisun40.optigfx.COM>
To: freebsd-bugs@freebsd.org, freebsd-hackers@freebsd.org,
        freebsd-hardware@freebsd.org, hosokawa@mt.cs.keio.ac.jp,
        jleppek@suw2k.ess.harris.com
Subject: 3c589 card CAN be made to work BNC
Sender: bugs-owner@freebsd.org
Precedence: bulk

Ok.  It's been a long weekend (note the time) but trial and error is
always like that.  I >finally< have wrestled my 3c589 into working BNC.

I went over to the Linux 3c589_cs.c driver and read it (line by line)
against FreeBSD's /usr/src/sys/i386/isa/if_zp.c.  In doing so, I
discovered some things Linux was doing that FreeBSD was not.  This
leads to the following patch to FreeBSD's
/usr/src/sys/i386/isa/if_zp.c.  My original of that source file came
from the 2.0.5 cdrom.  Here is the single diff-section patch:

########################################################################
*** if_zp.c     Mon Jul 31 00:47:22 1995
--- if_zp.c.orig        Sun Jul 30 17:42:19 1995
***************
*** 1069,1082 ****
      if (!(ifp->if_flags & IFF_ALTPHYS) && (sc->ep_connectors & BNC))
{
  #endif
  #ifdef ZP_DEBUG
!       printf("START TRANSCEIVER");
  #endif        /* ZP_DEBUG */
-       GO_WINDOW(0);
-       /* Set the xcvr */
-       outw(BASE + EP_W0_ADDRESS_CFG, 3 << 14);
-         GO_WINDOW(2);
        outw(BASE + EP_COMMAND, START_TRANSCEIVER);
-         GO_WINDOW(1);
        DELAY(1000);
      }
  #if defined(__NetBSD__) || defined(__FreeBSD__)
--- 1069,1077 ----
      if (!(ifp->if_flags & IFF_ALTPHYS) && (sc->ep_connectors & BNC))
{
  #endif
  #ifdef ZP_DEBUG
!       printf("START TRANCEIVER");
  #endif        /* ZP_DEBUG */
        outw(BASE + EP_COMMAND, START_TRANSCEIVER);
        DELAY(1000);
      }
  #if defined(__NetBSD__) || defined(__FreeBSD__)
########################################################################

This patch is not elegant but it is enough to make BNC work without
affecting UTP.

I confess to being a total novice as far as the procedure for
submitting bug fixes.  If I have not given enough information for
someone to handle getting this fix in, please just let me know.

I can't resist quoting the comment from the Linux source file that
made me first think I had found the answer:

/* This is vital: the transceiver used must be set in the resource
   configuration register.  It took me many hours to discover this. */

I know the feeling.

sts