From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755510Ab1HZV0S (ORCPT ); Fri, 26 Aug 2011 17:26:18 -0400 Received: from mga01.intel.com ([192.55.52.88]:27236 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755396Ab1HZV0J (ORCPT ); Fri, 26 Aug 2011 17:26:09 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.68,287,1312182000"; d="scan'208";a="45196508" Subject: Re: MTRR setup failure on Fujitsu Lifebook S761 From: Suresh Siddha Reply-To: Suresh Siddha To: Tony Vroon Cc: "H. Peter Anvin" , LKML , Thomas Gleixner , Ingo Molnar , Tejun Heo , yinghai@kernel.org Date: Fri, 26 Aug 2011 14:29:17 -0700 In-Reply-To: <1314385315.8356.0.camel@localhost> References: <4E579397.6010009@linx.net> <4E57CBBB.3040806@zytor.com> <1314385315.8356.0.camel@localhost> Organization: Intel Corp Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.0.1 (3.0.1-1.fc15) Content-Transfer-Encoding: 7bit Message-ID: <1314394157.4729.23.camel@sbsiddha-desk.sc.intel.com> Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2011-08-26 at 12:01 -0700, Tony Vroon wrote: > On Fri, 2011-08-26 at 09:37 -0700, H. Peter Anvin wrote: > > Hm, I don't see any evidence of the MTRR sanitizer running at all... > > It is almost as if it failed to find an optimal configuration; does that > error path not print anything? MTRR cleanup bails out in mtrr_need_cleanup() if it finds any entry other than WB or UC. And the first entry you had was write-protect. There are couple of issues. a. mtrr_add_page() checks for the type mismatch between the new entry and any existing entry. _ANY_ is wrong and we should have used mtrr_type_lookup() in mtrr_add_page(). mtrr_type_lookup() will look at all the existing MTRR entries and comeup with the right type for a given range. We should use that instead. b. And even after fixing 'a', your bios has already setup all the MTRR's and there are no spare MTRR's available. So we should really sanitize to makeup some space if we can. BTW, before all this: you seem to be using PAT and we really shouldn't depend on the MTRR configuration to get the write-combining behavior if PAT is used. And you seem to say, things were _little_ bit better by using a MTRR script. What do you mean by _little_ ? Do you still see the original issue of hangcheck timer elapsed and it is just that the MTRR warning from drm driver disappeared with your MTRR script? thanks, suresh