From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750929AbVLOTAt (ORCPT ); Thu, 15 Dec 2005 14:00:49 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750934AbVLOTAt (ORCPT ); Thu, 15 Dec 2005 14:00:49 -0500 Received: from smarthost2.sentex.ca ([205.211.164.50]:24278 "EHLO smarthost2.sentex.ca") by vger.kernel.org with ESMTP id S1750929AbVLOTAs (ORCPT ); Thu, 15 Dec 2005 14:00:48 -0500 From: "Stuart MacDonald" To: "'Alan Cox'" , "'Meelis Roos'" Cc: Subject: RE: Serial: bug in 8250.c when handling PCI or other level triggers Date: Thu, 15 Dec 2005 14:00:31 -0500 Organization: Connect Tech Inc. Message-ID: <000901c601a9$d243fe50$294b82ce@stuartm> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 In-Reply-To: <1134599362.25663.72.camel@localhost.localdomain> Importance: Normal Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org From: On Behalf Of Alan Cox > I don't think so. The bug as such is something I can only see being > triggerable either by a virtual machine or by something like serious > noise on the signal lines (eg put a 10Khz carrier on the > carrier detect > line) We found and patched this bug in one of our products. The patch was to raise the loop counter to something more appropriate for our hardware. The condition: the not-to-speedy embedded CPU and all ports in use. The interrupt handler would hit the loop limit because the combination of all ports running meant usually there was one port that needed servicing, upping the loop count by one. ..Stu