[libvirt] [PATCH 1/3] cpu_map.xml: add cmt feature to x86

[Jiri Denemark]

On Thu, Oct 29, 2015 at 14:02:28 +0800, Qiaowei Ren wrote:
> Some Intel processor families (e.g. the Intel Xeon processor E5 v3
> family) introduced CMT (Cache Monitoring Technology) to measure the
> usage of cache by applications running on the platform. This patch
> add it into x86 part of cpu_map.xml.

When sending a series of patches, please use --cover-letter to create a
0/n patch where you describe what the series is trying to achieve. As a
nice side effect , individual patches will be sent as replies to the
cover letter, which is much better than sending 2..n patches as replies
to the first one.

> Signed-off-by: Qiaowei Ren <qiaowei.ren at intel.com>
> ---
>  .gnulib             | 2 +-
>  src/cpu/cpu_map.xml | 3 +++
>  2 files changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/.gnulib b/.gnulib
> index f39477d..106a386 160000
> --- a/.gnulib
> +++ b/.gnulib
> @@ -1 +1 @@
> -Subproject commit f39477dba778e99392948dd3dd19ec0d46aee932
> +Subproject commit 106a3866d01f9dd57ab4f10dbeb0d5a8db73a9f7

As Peter said, this hunk should be removed.

> diff --git a/src/cpu/cpu_map.xml b/src/cpu/cpu_map.xml
> index b9e95cf..14ccbd8 100644
> --- a/src/cpu/cpu_map.xml
> +++ b/src/cpu/cpu_map.xml
> @@ -317,6 +317,9 @@
>      <feature name='rtm'>
>        <cpuid function='0x00000007' ebx='0x00000800'/>
>      </feature>
> +    <feature name='cmt'>
> +      <cpuid function='0x00000007' ebx='0x00001000'/>
> +    </feature>
>      <feature name='rdseed'>
>        <cpuid function='0x00000007' ebx='0x00040000'/>
>      </feature>

This looks like it makes sense, but it really doesn't. This patches
causes libvirt to report cmt feature on host CPUs which support it, but
what's the point since we are apparently not interested in exposing the
feature to guests? Not to mention that even if the host CPU supports
cmt, the host kernel does not have to supported so advertising the CPU
feature doesn't really say whether it's usable or not.

Jirka