From mboxrd@z Thu Jan  1 00:00:00 1970
From: Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH v6 5/5] x86/kvm: Avoid dynamic allocation of pvclock data
 when SEV is active
Date: Tue, 11 Sep 2018 12:19:13 +0200
Message-ID: <554c6fb5-0de4-5415-9970-0d09325d718b@redhat.com>
References: <1536343050-18532-1-git-send-email-brijesh.singh@amd.com>
 <1536343050-18532-6-git-send-email-brijesh.singh@amd.com>
 <20180910122727.GE21815@zn.tnic>
 <026d5ca5-7b77-de6c-477e-ff39f0291ac0@amd.com>
 <1536586152.11460.40.camel@intel.com>
 <097eb5f5-2cd9-8b08-32c5-d90c8e0cbb6d@amd.com>
 <1536593297.11460.72.camel@intel.com>
 <11618b8b-4d1f-9307-35f0-3c0f0fc856ca@amd.com>
 <20180910164851.GC20286@zn.tnic>
 <3b8b4c9c-b0f6-1e08-3d26-0e146cd7189e@redhat.com>
 <20180911100142.GA11418@zn.tnic>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Cc: Brijesh Singh <brijesh.singh@amd.com>,
        Sean Christopherson <sean.j.christopherson@intel.com>,
        x86@kernel.org, linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Tom Lendacky <thomas.lendacky@amd.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        "H. Peter Anvin" <hpa@zytor.com>,
        =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>
To: Borislav Petkov <bp@suse.de>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20180911100142.GA11418@zn.tnic>
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
List-Id: kvm.vger.kernel.org

On 11/09/2018 12:01, Borislav Petkov wrote:
> On Tue, Sep 11, 2018 at 11:26:21AM +0200, Paolo Bonzini wrote:
>> Usually the kvmclock structs are all the same, but there is support for
>> old machines with inconsistent TSCs (across different sockets typically).
> 
> Would that be a problem, though? Sounds like an "improvement" to me. :-)
> 
> I mean, if we keep using the same TSC across all vCPUs, the guest will
> actually see a single TSC and thus have stable and synchronized TSCs.
> Unlike the host.

That's exactly what kvmclock is for, it provides a stable and
synchronized clock on top of unsynchronized TSCs.  But that's also why
you need one struct per vCPU, at least in the synchronized case.

Paolo

> I.e., the guest will be better than the host! :-)
>