From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263910AbTJFAPQ (ORCPT ); Sun, 5 Oct 2003 20:15:16 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263911AbTJFAPQ (ORCPT ); Sun, 5 Oct 2003 20:15:16 -0400 Received: from e2.ny.us.ibm.com ([32.97.182.102]:4313 "EHLO e2.ny.us.ibm.com") by vger.kernel.org with ESMTP id S263910AbTJFAPN (ORCPT ); Sun, 5 Oct 2003 20:15:13 -0400 From: James Cleverdon Reply-To: jamesclv@us.ibm.com Organization: IBM LTC To: Zwane Mwaikambo Subject: Re: [PATCH] Dynamic irq_vector allocation for 2.6.0-test6-mm Date: Sun, 5 Oct 2003 17:14:31 -0700 User-Agent: KMail/1.5 Cc: Andrew Morton , linux-kernel@vger.kernel.org References: <200310031807.20650.jamesclv@us.ibm.com> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200310051714.31044.jamesclv@us.ibm.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Friday 03 October 2003 9:18 pm, Zwane Mwaikambo wrote: > On Fri, 3 Oct 2003, James Cleverdon wrote: > > irq_vector is indexed by a value that can be as large as the sum of all > > the RTEs in all the I/O APICs. On a 32-way x445 or a 16-way x440 with > > PCI expansion boxes, the static array will overflow. > > What is number of interrupt sources on that 32x? That depends on the number of PCI cards plugged into the slots and how many of the four interrupt lines (/INTA - /INTD) are used. However it doesn't matter. Even if no cards are present, the array will still overflow because the I/O APIC init routines assign a vector to each RTE and run off the end of irq_vector. -- James Cleverdon IBM xSeries Linux Solutions {jamesclv(Unix, preferred), cleverdj(Notes)} at us dot ibm dot comm