From mboxrd@z Thu Jan 1 00:00:00 1970 From: Keir Fraser Subject: Re: [PATCH] x86, cpuidle: remove assertion on X86_FEATURE_TSC_RELIABLE Date: Tue, 17 May 2011 08:58:16 +0100 Message-ID: References: <625BA99ED14B2D499DC4E29D8138F1505C9BBF8192@shsmsx502.ccr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <625BA99ED14B2D499DC4E29D8138F1505C9BBF8192@shsmsx502.ccr.corp.intel.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: "Tian, Kevin" , Dan Magenheimer , Jan Beulich Cc: xen devel List-Id: xen-devel@lists.xenproject.org On 17/05/2011 01:50, "Tian, Kevin" wrote: >> IIRC, this is not true on a multi-socket motherboard. Even though each >> socket >> has NONSTOP_TSC, they are using different crystals, correct? >> > > it's true that sockets may use different crystals, and NONSTOP_TSC has nothing > to say > synchronization among sockets/cores. So it really depends on how you define a > 'reliable': > is it reliable enough to be a Xen time source, or reliable enough to > passthrough to the > guest? I'll need to check current assumption and your previous discussions > first before > saying anything inappropriate. :-) Yes, Dan is right, RELIABLE_TSC means something more than just NONSTOP_TSC and CONSTANT_TSC. It means that: 1. TSCs do not stop in deep sleep (NONSTOP_TSC) 2. TSCs do not change rate with core frequency (CONSTANT_TSC) 3. Further, that all TSCs system wide run at the same rate at all times, in perfect sync (not represented by any other cpu feature flag). -- Keir