Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263693AbTKFQEp (ORCPT ); Thu, 6 Nov 2003 11:04:45 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263688AbTKFQEp (ORCPT ); Thu, 6 Nov 2003 11:04:45 -0500 Received: from pop3.galileo.co.il ([199.203.130.130]:62892 "EHLO mail.galileo.co.il") by vger.kernel.org with ESMTP id S263681AbTKFQEl (ORCPT ); Thu, 6 Nov 2003 11:04:41 -0500 Message-ID: <3FAA712F.2030409@il.marvell.com> Date: Thu, 06 Nov 2003 18:05:03 +0200 From: Mark Mokryn Organization: Marvell Semiconductor User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624 Netscape/7.1 (ax) X-Accept-Language: en-us, en MIME-Version: 1.0 To: linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org, linux-mm@vger.kernel.org Subject: Highmem SCSI driver Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 772 Lines: 20 We are trying to test 64-bit PCI DMA for a SCSI driver on a Xeon box, RH9 2.4.20-8 bigmem kernel, 6GB RAM. The machine shows 6GB in top, we set highmem_io in the driver, PCI DMA mask covers 64-bit range, etc. Of course we're trying to make sure that the system does not create bounce buffers unnecessarily. On a 64-bit box (AMD64) everything works as expected. On the Xeon, no matter what we try, we never see I/Os mapped above 4GB. Any ideas on how we can drive I/Os mapped above 4GB down to our driver? Thanks, -Mark - 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/