From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew Cooper <andrew.cooper3@citrix.com>
Subject: Re: [PATCH RFC 2/2] xen/x86: Introduce XEN_SYSCTL_cpuid
 hypercall
Date: Thu, 27 Feb 2014 12:11:49 +0000
Message-ID: <530F2B85.6060403@citrix.com>
References: <1393499497-9162-1-git-send-email-andrew.cooper3@citrix.com>
	<1393499497-9162-4-git-send-email-andrew.cooper3@citrix.com>
	<530F365B020000780011FD03@nat28.tlf.novell.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
In-Reply-To: <530F365B020000780011FD03@nat28.tlf.novell.com>
List-Unsubscribe: <http://lists.xen.org/cgi-bin/mailman/options/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xen.org>
List-Help: <mailto:xen-devel-request@lists.xen.org?subject=help>
List-Subscribe: <http://lists.xen.org/cgi-bin/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=subscribe>
Sender: xen-devel-bounces@lists.xen.org
Errors-To: xen-devel-bounces@lists.xen.org
To: Jan Beulich <JBeulich@suse.com>
Cc: Keir Fraser <keir@xen.org>, Tim Deegan <tim@xen.org>, Ian Jackson <Ian.Jackson@eu.citrix.com>, Ian Campbell <Ian.Campbell@citrix.com>, Xen-devel <xen-devel@lists.xen.org>
List-Id: xen-devel@lists.xenproject.org

On 27/02/14 11:58, Jan Beulich wrote:
>>>> On 27.02.14 at 12:11, Andrew Cooper <andrew.cooper3@citrix.com> wrote:
>> which permits a toolstack to execute an arbitrary cpuid instruction on a
>> specified physical cpu.
> For one - is it a good idea to expose the unprocessed CPUID to
> any guest code? After all even the Dom0 kernel only gets to see
> processed values, and the fact the without CPUID faulting apps
> in PV guests can inadvertently use the raw values is known to be
> a problem, not a feature.

Any toolstack which uses this specific hypercall to find out information
normally hidden from dom0 using faulting/masking/policy can only shoot
itself.

The usecase is for enumerating the real cache leaves, which are normally
faked up in the policy anyway, so of no use.

>
> And then - if you already have access to control operations, I
> don't think you need the hyypervisor to help you: Limit your
> vCPU's affinity to the particular pCPU you care about, and do
> what you need doing from the kernel (by also setting the
> processes affinity to the particular CPU you could achieve the
> same even from user land).
>
> Jan
>

Having a toolstack rely on being able to pin its vcpus around so some
userspace can enumerate the cache leaves is horrific.

Apart from forcibly messing with a balanced numa setup, what about cpu
pools, or toolstack disaggregation where pinning is restricted?

~Andrew