From mboxrd@z Thu Jan 1 00:00:00 1970 From: Glauber Costa Subject: Re: [patch 09/18] KVM: x86: introduce facility to support vsyscall pvclock, via MSR Date: Fri, 2 Nov 2012 14:21:05 +0400 Message-ID: <50939E91.2080807@parallels.com> References: <20121024131340.742340256@redhat.com> <20121024131621.707068244@redhat.com> <508E9697.2000003@parallels.com> <508EC089.5030409@goop.org> <20121029184015.GA30422@amt.cnet> <508FA054.6080207@redhat.com> <20121031031219.GA9913@amt.cnet> Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit Cc: Avi Kivity , Jeremy Fitzhardinge , , , , , To: Marcelo Tosatti Return-path: Received: from mx2.parallels.com ([64.131.90.16]:54595 "EHLO mx2.parallels.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933559Ab2KBKVU (ORCPT ); Fri, 2 Nov 2012 06:21:20 -0400 In-Reply-To: <20121031031219.GA9913@amt.cnet> Sender: kvm-owner@vger.kernel.org List-ID: On 10/31/2012 07:12 AM, Marcelo Tosatti wrote: > On Tue, Oct 30, 2012 at 11:39:32AM +0200, Avi Kivity wrote: >> On 10/29/2012 08:40 PM, Marcelo Tosatti wrote: >>> On Mon, Oct 29, 2012 at 10:44:41AM -0700, Jeremy Fitzhardinge wrote: >>>> On 10/29/2012 07:45 AM, Glauber Costa wrote: >>>>> On 10/24/2012 05:13 PM, Marcelo Tosatti wrote: >>>>>> Allow a guest to register a second location for the VCPU time info >>>>>> >>>>>> structure for each vcpu (as described by MSR_KVM_SYSTEM_TIME_NEW). >>>>>> This is intended to allow the guest kernel to map this information >>>>>> into a usermode accessible page, so that usermode can efficiently >>>>>> calculate system time from the TSC without having to make a syscall. >>>>>> >>>>>> Signed-off-by: Marcelo Tosatti >>>>> Can you please be a bit more specific about why we need this? Why does >>>>> the host need to provide us with two pages with the exact same data? Why >>>>> can't just do it with mapping tricks in the guest? >>>> >>>> In Xen the pvclock structure is embedded within a pile of other stuff >>>> that shouldn't be mapped into guest memory, so providing for a second >>>> location allows it to be placed whereever is convenient for the guest. >>>> That's a restriction of the Xen ABI, but I don't know if it affects KVM. >>>> >>>> J >>> >>> It is possible to share the data for KVM in theory, but: >>> >>> - It is a small amount of memory. >>> - It requires aligning to page size (the in-kernel percpu array >>> is currently cacheline aligned). >>> - It is possible to modify flags separately for userspace/kernelspace, >>> if desired. >>> >>> This justifies the duplication IMO (code is simple and clean). >>> >> >> What would be the changes required to remove the duplication? If it's >> just page alignment, then is seems even smaller. In addition we avoid >> expanding the ABI again. > > This would require changing the kernel copy from percpu data, which > there is no guarantee is linear (necessary for fixmap mapping), to > dynamically allocated (which in turn can be tricky due to early boot > clock requirement). > > Hum, no thanks. > You allocate it using bootmemory for vsyscall anyway. If they are strictly in the same physical location, you are not allocating anything extra.