From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756065Ab0HXVjz (ORCPT ); Tue, 24 Aug 2010 17:39:55 -0400 Received: from terminus.zytor.com ([198.137.202.10]:48068 "EHLO mail.zytor.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755993Ab0HXVjx (ORCPT ); Tue, 24 Aug 2010 17:39:53 -0400 Message-ID: <4C743C0F.2010101@zytor.com> Date: Tue, 24 Aug 2010 14:39:27 -0700 From: "H. Peter Anvin" User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.8) Gecko/20100806 Fedora/3.1.2-1.fc13 Thunderbird/3.1.2 MIME-Version: 1.0 To: Henrique de Moraes Holschuh CC: Jack Steiner , mingo@elte.hu, tglx@linutronix.de, lenb@kernel.org, linux-acpi@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [RFC] - Mapping ACPI tables as CACHED References: <20100722152220.GA18290@sgi.com> <20100724001449.GA9618@khazad-dum.debian.net> In-Reply-To: <20100724001449.GA9618@khazad-dum.debian.net> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 07/23/2010 05:14 PM, Henrique de Moraes Holschuh wrote: > > Well, as it was raised in this thread, ACPI tables are likely to be near RAM > regions used for IPC with the firmware or SMBIOS, and we have no idea of the > kind of crap that could happen if we enable caching on those areas. > I'm really not sure I buy that argument -- at least not on x86: if that is the case, then when PAT is off (and we fall down to MTRR-only control) then we'd have the same failures. If we mark them cacheable and the MTRRs say uncachable, then we will *still* not cache them (since MTRR UC overrides PAT WB -- in fact "PAT off" really just means ALL the pagetables are marked WB.) In that sense it is probably *safer* to map them WB, since the firmware if it uses page tables at all is extremely likely to have all the cache control bits at zero (meaning WB) -- and if it doesn't use page tables, they are functionally zero by default (MTRR control only.) So I think it'd be safer to map them cacheable -- regardless of if we want to copy them to RAM or not. -hpa