From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 15 Aug 2001 20:07:52 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 15 Aug 2001 20:07:41 -0400 Received: from samba.sourceforge.net ([198.186.203.85]:27910 "HELO lists.samba.org") by vger.kernel.org with SMTP id ; Wed, 15 Aug 2001 20:07:24 -0400 From: Paul Mackerras MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <15227.3613.529481.433816@cargo.ozlabs.ibm.com> Date: Thu, 16 Aug 2001 10:04:45 +1000 (EST) To: Alan Cox Cc: linux-kernel@vger.kernel.org (Linux Kernel) Subject: Re: PCMCIA IDE_CS in 2.4.7 In-Reply-To: In-Reply-To: X-Mailer: VM 6.75 under Emacs 20.7.2 Reply-To: paulus@samba.org Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Alan Cox writes: > Gunther posted this patch ages ago which seems to solve the problem I found that the first part (adding outb(0x02, ctl_base)) was necessary but the second part (outb(0, ...)) was not. Setting nIEN early on seems safe to me, but it was not clear to me that enabling IRQs (clearing nIEN) just before the ide_request_irq call was safe - fortunately it doesn't seem to be necessary, I presume the IDE code clears nIEN when it wants to start getting interrupts. On my powerbook I also use a patch which makes sure that the chipset gets set to ide_pci so that the IDE code will let me share the irq - the pcmcia controller here has just one interrupt that is used both for card interrupts and for controller interrupts. I'm not sure whether that came from Gunther or someone else. Paul.