NetBSD Problem Report #56846

From www@netbsd.org  Wed May 18 22:53:15 2022
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits))
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 284DF1A921F
	for <gnats-bugs@gnats.NetBSD.org>; Wed, 18 May 2022 22:53:15 +0000 (UTC)
Message-Id: <20220518225312.EDE641A923A@mollari.NetBSD.org>
Date: Wed, 18 May 2022 22:53:12 +0000 (UTC)
From: jarkur@pro.onet.pl
Reply-To: jarkur@pro.onet.pl
To: gnats-bugs@NetBSD.org
Subject: Hardware lockup on laptops with ToPIC95B Cardbus controller
X-Send-Pr-Version: www-1.0

>Number:         56846
>Category:       port-i386
>Synopsis:       Hardware lockup on laptops with ToPIC95B Cardbus controller
>Confidential:   no
>Severity:       serious
>Priority:       low
>Responsible:    port-i386-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed May 18 22:55:00 +0000 2022
>Originator:     Jaroslaw Kurkiewicz
>Release:        9.2
>Organization:
>Environment:
NetBSD 9.2 NetBSD 9.2 (GENERIC) i386
>Description:
Write buffering (or more accurately PCI Posted Writes I believe) should not be enabled by default on ToPIC95B Cardbus controllers. As a result, system completely locks up under any kind of heavy Cardbus activity. Observed on Toshiba Satellite Pro 440CDT with 3COM Cardbus 3c575CT Ethernet card.

It might also affect ToPIC95.


>How-To-Repeat:
Put heavy load on Cardbus on any old Toshiba laptop with 6.B0 Toshiba BIOS and ToPIC95B controller. Any attempt to, for example, saturate a 100Mbit/s Ethernet link results in irrecoverable hardware lockup.
>Fix:
Fix: Quick fix is to disable PCI Posted Writes. Since datasheet seems to be unobtainable further tests could prove whether it can be reenabled at some stage. I believe the following change in line 765 of pccbb_chipinit() might be enough but I don't know the codebase at all: 

if (sc->sc_chipset != CB_TOPIC95B)
    bcr |= CB_BCR_WRITE_POST_ENABLE;
else
    bcr &= ~CB_BCR_WRITE_POST_ENABLE;

Related Linux kernel bug 55961 should provide sufficient information and inspiration.

Workaround 1: Switch controller to PCIC-compatible mode and avoid using Cardbus cards.

Workaround 2: Do not put heavy load on Cardbus. (Following network card example the problem mostly disappears if the card is switched to 10MBit/s, half-duplex mode).

Home	PR Database Search