From mboxrd@z Thu Jan 1 00:00:00 1970 From: "H. Peter Anvin" Subject: Re: [Bisected] 3.7-rc1 can't resume (still present in 3.9) Date: Fri, 12 Jul 2013 16:45:52 -0700 Message-ID: <51E09530.8030300@zytor.com> References: <20130218155439.GA902@onny> <20130223131856.GA682@onny> <51293A8B.5060107@zytor.com> <20130320143257.GA473@onny> <20130501170118.GB488@onny> <518151DD.106@zytor.com> <51DDC972.3010005@student.kit.edu> <51DDF4DF.4000902@zytor.com> <51E092FC.1030200@student.kit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from terminus.zytor.com ([198.137.202.10]:42502 "EHLO mail.zytor.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753289Ab3GLXqY (ORCPT ); Fri, 12 Jul 2013 19:46:24 -0400 In-Reply-To: <51E092FC.1030200@student.kit.edu> Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: =?UTF-8?B?Q2hyaXN0aWFuIFPDvG5rZW5iZXJn?= Cc: Jonas Heinrich , "len.brown@intel.com" , "pavel@ucw.cz" , "tglx@linutronix.de" , "mingo@redhat.com" , "x86@kernel.org" , "rjw@sisk.pl" , "linux-kernel@vger.kernel.org" , "linux-pm@vger.kernel.org" On 07/12/2013 04:36 PM, Christian S=C3=BCnkenberg wrote: >=20 > Jonas tried your patch and it fixes suspend/resume on his T43, althou= gh > IMHO the safest approach would be to just add an exception for > Vendor=3D=3DIntel && Family=3D=3D6 && Model=3D=3D13, or more generall= y Vendor=3D=3DIntel > && !supports_long_mode, as the same erratum also warns about wrmsr > possibly not triggering a GP either. > Anyways, at least on this specific MSR with the Pentium M Jonas teste= d, > it behaved correctly on every try, so I'd say your patch does the tri= ck, > thank you very much! >=20 Using vendor matches is not really a great way to deal with things that can better be handled analytically. If WRMSR doesn't fault, it is not a problem... > As a side note, I found a similar erratum #33 in "Pentium=C2=AE Proce= ssor > Specification Update" for Intel P54C (Family 5, Model 2), which would= , > supposed there are P54C systems with ACPI sleep/resume support, resul= t > in MSR 0 (P5_MC_ADDR) to be saved and restored instead of nonexistent= EFER. Doesn't really matter, as we'd only read that one after an #MC. -hpa