From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266627AbUI0Kmg (ORCPT ); Mon, 27 Sep 2004 06:42:36 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266631AbUI0Kmg (ORCPT ); Mon, 27 Sep 2004 06:42:36 -0400 Received: from cantor.suse.de ([195.135.220.2]:10925 "EHLO Cantor.suse.de") by vger.kernel.org with ESMTP id S266627AbUI0Kme (ORCPT ); Mon, 27 Sep 2004 06:42:34 -0400 Date: Mon, 27 Sep 2004 12:41:47 +0200 From: Andi Kleen To: "Mallick, Asit K" , "Siddha, Suresh B" , tom.l.nguyen@intel.com, linux-kernel@vger.kernel.org Subject: [PATCH] Fix x86-64 properly with MSI & Suresh's change Message-ID: <20040927104147.GE3532@wotan.suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Together with Suresh's recent LH workaround: this patch makes x86-64 compile again with MSI on. i386 uses an CPU number, x86-64 an CPU mask for MSI_TARGET_CPUS and that didn't work very well. I must admit I don't fully understand how MSI irq affinity is supposed to work (why do you always redirect to the current CPU?), but this matches i386 which is presumably the best tested MSI platform. Signed-off-by: Andi Kleen diff -u linux/include/asm-x86_64/msi.h-o linux/include/asm-x86_64/msi.h --- linux/include/asm-x86_64/msi.h-o 2004-09-24 13:04:06.000000000 +0200 +++ linux/include/asm-x86_64/msi.h 2004-09-27 12:19:56.000000000 +0200 @@ -7,10 +7,11 @@ #define ASM_MSI_H #include +#include #define LAST_DEVICE_VECTOR 232 #define MSI_DEST_MODE MSI_LOGICAL_MODE #define MSI_TARGET_CPU_SHIFT 12 -#define MSI_TARGET_CPU TARGET_CPUS +#define MSI_TARGET_CPU hard_smp_processor_id() #endif /* ASM_MSI_H */