From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261387AbULEUwG (ORCPT ); Sun, 5 Dec 2004 15:52:06 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261384AbULEUwG (ORCPT ); Sun, 5 Dec 2004 15:52:06 -0500 Received: from clock-tower.bc.nu ([81.2.110.250]:23495 "EHLO localhost.localdomain") by vger.kernel.org with ESMTP id S261387AbULEUt7 (ORCPT ); Sun, 5 Dec 2004 15:49:59 -0500 Subject: Re: 2.6.10-rc2-mm4 From: Alan Cox To: Terence Ripperda Cc: Linux Kernel Mailing List In-Reply-To: <20041203215927.GE1709@hygelac> References: <20041203215927.GE1709@hygelac> Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1102275985.9335.22.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Sun, 05 Dec 2004 19:46:27 +0000 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Gwe, 2004-12-03 at 21:59, Terence Ripperda wrote: > I assume you mean traditional pci in this case, but I remain confused. > the pci spec calls for 32-bits of addressing, although there is an > optional extension for 64-bit bus extension pins. I can't speak for other > pci devices, but all of our pci devices are 32-bit. The current DRI drivers don't really deal much with PCI devices. A pure PCI video card on 64bit boxes might be problematic although I'd question the sanity of anyone doing this 8) > > additionally, the pci-express spec defines legacy and non-legacy > devices. legacy devices are only required to address 32-bits, whereas > non-legacy devices are required to handle 64-bit addresses. I'd assumed video card vendors were non-legacy but ok > I certainly understand the concerns with this, although I was led to > believe that recent 2.6 work made the zone balancing much less > expensive. is that not the case? Andrew certainly believes this is. Certainly in 2.4 it was not. > > I can find users for a 512Mb or 1Gb DMA region > > there was some brief discussion of this when we originally discussed > 32-bit addressing issues, but I don't know if a satisfactory solution was > reached. If a 1Gb region was prefered for this reason, that should satisfy > nvidia's needs for 32-bit addressing, but I couldn't speak for any other device > drivers. If the VM can take it and get it right I am all for a 512Mb or 1Gb DMA region to fix the various devices that have 29-31bit DMA issues. If it fixes Nvidia's needs to then fine. Alan