From mboxrd@z Thu Jan 1 00:00:00 1970 From: Radim =?utf-8?B?S3LEjW3DocWZ?= Subject: Re: [RFC PATCH 2/2] KVM: x86: add support of kvm-clock stablity in L2 Date: Mon, 10 Jul 2017 15:25:27 +0200 Message-ID: <20170710132526.GA28283@potion> References: <1498647301-130851-1-git-send-email-dplotnikov@virtuozzo.com> <1498647301-130851-3-git-send-email-dplotnikov@virtuozzo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: kvm@vger.kernel.org, pbonzini@redhat.com, den@virtuozzo.com, rkagan@virtuozzo.com To: Denis Plotnikov Return-path: Received: from mx1.redhat.com ([209.132.183.28]:41766 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932096AbdGJNZb (ORCPT ); Mon, 10 Jul 2017 09:25:31 -0400 Content-Disposition: inline In-Reply-To: <1498647301-130851-3-git-send-email-dplotnikov@virtuozzo.com> Sender: kvm-owner@vger.kernel.org List-ID: 2017-06-28 13:55+0300, Denis Plotnikov: > Get rid of complex shadow monotonic timekeeper support in KVM. > Extend and use timekeeper infrastructure instead. > > Make kvm-clock stable in L2 using the changed timekeeper I see this patch as two logical changes: 1) refactoring that drops host time caching (gtod) 2) stable kvmclock for L2 when L1 is using kvmclock Doing it in two patches would definitely be nicer and would allow us to focus on the (less controversial) L2 enablement. About (1), the removed pvclock_gtod_notify() also verifid whether the host is using TSC and disabled master clock otherwise. Other changes don't explain why it's not needed anymore -- is it because any changes that make TSC unusable would be caught elsewhere? About (2), it seems that there is no event in case the kvmclock is not reliable anymore. We probably don't need it now, but it would be good to have it thought out -- would that use a new notifier? Thanks.