From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261244AbUCKNHt (ORCPT ); Thu, 11 Mar 2004 08:07:49 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261234AbUCKNHt (ORCPT ); Thu, 11 Mar 2004 08:07:49 -0500 Received: from zero.aec.at ([193.170.194.10]:30982 "EHLO zero.aec.at") by vger.kernel.org with ESMTP id S261244AbUCKNHs (ORCPT ); Thu, 11 Mar 2004 08:07:48 -0500 To: Boszormenyi Zoltan cc: linux-kernel@vger.kernel.org Subject: Re: x86_64 IOMMU question References: <1yxkz-6ct-31@gated-at.bofh.it> From: Andi Kleen Date: Thu, 18 Mar 2004 00:14:09 +0100 In-Reply-To: <1yxkz-6ct-31@gated-at.bofh.it> (Boszormenyi Zoltan's message of "Thu, 11 Mar 2004 13:20:15 +0100") Message-ID: User-Agent: Gnus/5.110002 (No Gnus v0.2) Emacs/21.2 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Boszormenyi Zoltan writes: > is it possible to use the IOMMU to help 32 bit devices > that limit their capabilities with pci_set_dma_mask()? > E.g. the emu10k1 limits itself under 256MB. Can the IOMMU > pass the data to/from the card from/to above 256MB? It can only remap to the AGP aperture, which is usually just below the 4GB boundary. In theory you could move the aperture to a very low address and remap to that (see arch/x86_64/kernel/aperture.c), but that would waste memory. -Andi