From owner-freebsd-hackers  Wed Feb 19 16:54:28 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id QAA01691
          for hackers-outgoing; Wed, 19 Feb 1997 16:54:28 -0800 (PST)
Received: from etinc.com (et-gw-fr1.etinc.com [204.141.244.98])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id QAA01659;
          Wed, 19 Feb 1997 16:53:09 -0800 (PST)
Received: from ntws (ntws.etinc.com [204.141.95.142]) by etinc.com (8.8.3/8.6.9) with SMTP id TAA06147; Wed, 19 Feb 1997 19:57:04 -0500 (EST)
Message-Id: <3.0.32.19970219195057.00af2590@etinc.com>
X-Sender: dennis@etinc.com
X-Mailer: Windows Eudora Pro Version 3.0 (32)
Date: Wed, 19 Feb 1997 19:51:01 -0500
To: Luigi Rizzo <luigi@labinfo.iet.unipi.it>, matt@lkg.dec.com (Matt Thomas)
From: dennis <dennis@etinc.com>
Subject: Re: if_de bug found!
Cc: hackers@freebsd.org, dg@freebsd.org
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

At 11:30 AM 2/19/97 +0100, Luigi Rizzo wrote:
>After long and painful debugging I found the bug with the 21140A driver
>(if_de.c) from netbsd. The driver has a workaround for a bug of the
>21140A when buffers overflow, and stops the receiver and resets the
>board when this occurs. This is done in the driver (around line 3230)
>by reading from CSR8 into "misses".
>
>Matt said this is an 'extremely rare' event, on my board, though, it
>turned out to happen at the end of every interrupt... A look at
>the 21140A reference manuals (on line on the DEC site) shows that the 3
>high bits of CSR8 are set, so that they should be ignored. After doing
>this, the driver works like a charm.
>
>Probably the correct fix is to trigger the workaround only when b16 of
>CSR8 is set. Matt ?
>
>Thanks to Matt for writing the driver and suggesting that the event is
>rare, and to DEC for making the hardware manuals available on line.
>
>I'll submit a PR shortly. This might be a 2.2 candidate, if the if_de
>driver in 2.2 has the same workaround. David, can you check ?

Does this fix the -AC problem?

Will something that lets use use -AC cards in 2.1.7 be made available? 

Dennis