From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mark Lord Subject: Re: Linux 2.6.17: IRQ handler mismatch in serial code? Date: Mon, 19 Jun 2006 17:52:21 -0400 Message-ID: <44971C95.7050700@rtr.ca> References: <20060619180658.58945.qmail@web52908.mail.yahoo.com> <20060619184706.GH3479@flint.arm.linux.org.uk> <4496FEC2.8050903@rtr.ca> <1150751212.2871.44.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from rtr.ca ([64.26.128.89]:60604 "EHLO mail.rtr.ca") by vger.kernel.org with ESMTP id S964922AbWFSVwX (ORCPT ); Mon, 19 Jun 2006 17:52:23 -0400 In-Reply-To: <1150751212.2871.44.camel@localhost.localdomain> Sender: linux-serial-owner@vger.kernel.org List-Id: linux-serial@vger.kernel.org To: Alan Cox Cc: Chris Rankin , linux-kernel@vger.kernel.org, linux-serial@vger.kernel.org Alan Cox wrote: > Ar Llu, 2006-06-19 am 15:45 -0400, ysgrifennodd Mark Lord: >> If the drivers are written "correctly", they shouldn't grab the IRQ >> until someone actually opens the device. Which means they should be >> able the share the IRQ, so long as both devices are not in use (open) >> at the same time. > > This is not the case for ISA bus. Most ISA hardware is physically unable > to share and the drivers for such hardware intentionally grab the IRQ at > load time to avoid it being mis-reused. Eh? The vast majority of ISA bus devices have open-collector IRQ lines, which means that two devices on the same IRQ are "wire-ORed" (but they really use edges). There's no problem whatsoever with this in most setups. I even implemented full support for it in the Linux IDE driver, way back when, so that multiple cards could be installed on the same IRQ, and would work just fine so long as access was serialized. This was the one of the original motivations for the IDE "serialize=" parameter. Cheers