Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 25 Sep 2002 05:05:49 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 25 Sep 2002 05:05:48 -0400 Received: from sullivan.realtime.net ([205.238.132.76]:12557 "EHLO sullivan.realtime.net") by vger.kernel.org with ESMTP id ; Wed, 25 Sep 2002 05:05:48 -0400 Date: Wed, 25 Sep 2002 04:11:01 -0500 (CDT) Message-Id: <200209250911.g8P9B1734121@sullivan.realtime.net> From: Milton Miller To: ajm@sgi.com (Alan Mayer) Subject: Re: irqs on large machines (patch) Cc: linux-kernel@vger.kernel.org In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1238 Lines: 31 PowerPC 64 (p690 in particular) had a similar but different problem with NR_IRQS. The hardware gives direct vectors, but the number space is 9 bits to identify the pci host bridge in the drawers, and yet 4 more to identify the source on the pci bus. All of the 9 bits are used at various levels of the hardware for routing, so the global number space is a bit sparse. Each interrupt can be sent to the global queue or a specific processor's queue. Rather than size NR_IRQS for the native index, the hardware numbers are mapped with a simple mapping table that directly maps dense linux irqs to hardware numbers. Thus the linux NR_IRQs is set based on the total number of hardware sources. (And yes, we found we needed /proc/interrupts to be seq_file based, but that is long merged). Perhaps this will give you ideas for other alternatives? This approach could also allow you to assign IO interrupts to a node if your hardware allows. milton -- [I'll look for any replys on the list] - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/