Re: [Qemu-devel] [PATCH] target-i386: Allow changing of Hypervisor CPUIDs.

[Eduardo Habkost]

On Thu, Sep 06, 2012 at 03:00:47PM -0400, Don Slutz wrote:
> On 09/06/12 14:40, Eduardo Habkost wrote:
> >On Thu, Sep 06, 2012 at 12:28:05PM +0200, Andreas Färber wrote:
> >>Am 30.08.2012 21:20, schrieb Don Slutz:
> >>>This is primarily done so that the guest will think it is running
> >>>under vmware when hypervisor=vmware is specified as a property of a
> >>>cpu.
> >>>
> >>>Also allow this to work in accel=tcg mode.
> >>>
> >>>The new cpu properties hyper_level, hyper_extra, hyper_extra_a, and
> >>>hyper_extra_b can be used to further adjust what the guest sees.
> >>>
> >>>Signed-off-by: Don Slutz <address@hidden>
> >>>---
> >>>  target-i386/cpu.c |  178 
> >>> +++++++++++++++++++++++++++++++++++++++++++++++++++++
> >>>  target-i386/cpu.h |    9 +++
> >>>  target-i386/kvm.c |   33 ++++++++--
> >>>  3 files changed, 214 insertions(+), 6 deletions(-)
> >>Please don't add new CPU command line options without matching QOM
> >>properties. There are patch series in the works that convert the CPU
> >>definitions to QOM subclasses, the parameters then need to be set on the
> >>X86CPU object instance.
> I found:
> http://lists.gnu.org/archive/html/qemu-devel/2012-08/msg00587.html
> 
> Is this the right set?

Yes. But note that there are lots of work in progress in the list (and
work that we haven't submitted to the list yet). You don't necessarily
need to wait for the above to be applied, probably you can just rebase
on top of the "CPU properties" work (that affects your patch more
directly). See below.


> >It seems to be a good idea to wait for the CPU properties series from
> >Igor to be applied, before implementing this. It should make the new
> >code much simpler.
> So Igor has a patch set that does a similar change like above.  I so
> far have not found it on QEMU-DEVEL.  Please provide a pointer to
> this patch set.

I was talking specifically about the "CPU properties" series, at:
http://article.gmane.org/gmane.comp.emulators.qemu/165728

The "CPU model classes" series (URL you sent above) is an additional
series to be applied on top of the "CPU properties" series. But the "CPU
model classes" work probably don't impact your patch, so you can simply
be ready to rebase/resend your patch after Igor sends a new version of
the CPU properties series.

Note that the URLs above are old versions of the work in progress. You
may find newer versions at my github tree[1] or at Igor's tree[2].

Basically once we introduce CPU properties, you'll just need to define
the new properties on X86CPU, without the need to change the CPU model
string parsing code. It should make your patch simpler.


[1] An experimental rebase of Igor's CPU properties series (Igor's
    latest version may look different):
    
https://github.com/ehabkost/qemu-hacks/tree/work/cpu-properties-igor-rebase-v4.3-2012-08-31
    CPU model classes:
    (may look different when I submit to qemu-devel again):
    
https://github.com/ehabkost/qemu-hacks/tree/work/cpu-model-classes-v2.8-2012-08-31
    Lots of work in progress, including multiple branches/series:
    (may look very different when I submit to qemu-devel):
    
https://github.com/ehabkost/qemu-hacks/tree/work/cpuid-refactor-v0.22-2012-08-31
[2] Igor's work in progress branch:
    https://github.com/imammedo/qemu/tree/x86-cpu-properties.WIP


> >
> >>Note that the convention for QOM properties is
> >>descriptive-name-with-dashes rather than shortened_abbrev_with_underscore.
> >>
> I will be re-working this change, and wait (for v2) until master has
> changed to the new way.
>    -Don

-- 
Eduardo