Patchwork qemu: whitelist kvm pv eoi feature

login
register
mail settings
Submitter Michael S. Tsirkin
Date May 11, 2012, 7:43 a.m.
Message ID <20120511074334.GA1504@redhat.com>
Download mbox | patch
Permalink /patch/158451/
State New
Headers show

Comments

Michael S. Tsirkin - May 11, 2012, 7:43 a.m.
Whitelist kvm pv eoi feature. The feature is enabled
with -cpu kvm64. To disable: -cpu kvm64,-kvm_eoi.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---

Sending a copy to kernel list as this is needed
to test the pv eoi feature recently submitted.

 target-i386/cpuid.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
Eric Blake - May 11, 2012, 12:47 p.m.
[adding libvirt, dropping kernel]

On 05/11/2012 01:43 AM, Michael S. Tsirkin wrote:
> Whitelist kvm pv eoi feature. The feature is enabled
> with -cpu kvm64. To disable: -cpu kvm64,-kvm_eoi.
> 
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> ---
> 
> Sending a copy to kernel list as this is needed
> to test the pv eoi feature recently submitted.
> 
>  target-i386/cpuid.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/target-i386/cpuid.c b/target-i386/cpuid.c
> index 465ea15..c421b19 100644
> --- a/target-i386/cpuid.c
> +++ b/target-i386/cpuid.c
> @@ -75,7 +75,7 @@ static const char *ext3_feature_name[] = {
>  };
>  
>  static const char *kvm_feature_name[] = {
> -    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, NULL, NULL,
> +    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, "kvm_eoi", NULL,

Should libvirt also be recognizing this processor capability in its XML
in order to drive the -kvm_eoi option?
Michael S. Tsirkin - May 13, 2012, 12:02 p.m.
On Fri, May 11, 2012 at 06:47:43AM -0600, Eric Blake wrote:
> [adding libvirt, dropping kernel]
> 
> On 05/11/2012 01:43 AM, Michael S. Tsirkin wrote:
> > Whitelist kvm pv eoi feature. The feature is enabled
> > with -cpu kvm64. To disable: -cpu kvm64,-kvm_eoi.
> > 
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > ---
> > 
> > Sending a copy to kernel list as this is needed
> > to test the pv eoi feature recently submitted.
> > 
> >  target-i386/cpuid.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/target-i386/cpuid.c b/target-i386/cpuid.c
> > index 465ea15..c421b19 100644
> > --- a/target-i386/cpuid.c
> > +++ b/target-i386/cpuid.c
> > @@ -75,7 +75,7 @@ static const char *ext3_feature_name[] = {
> >  };
> >  
> >  static const char *kvm_feature_name[] = {
> > -    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, NULL, NULL,
> > +    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, "kvm_eoi", NULL,
> 
> Should libvirt also be recognizing this processor capability in its XML
> in order to drive the -kvm_eoi option?

I think so.

> -- 
> Eric Blake   eblake@redhat.com    +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
Michael S. Tsirkin - May 13, 2012, 12:05 p.m.
On Sun, May 13, 2012 at 03:02:58PM +0300, Michael S. Tsirkin wrote:
> On Fri, May 11, 2012 at 06:47:43AM -0600, Eric Blake wrote:
> > [adding libvirt, dropping kernel]
> > 
> > On 05/11/2012 01:43 AM, Michael S. Tsirkin wrote:
> > > Whitelist kvm pv eoi feature. The feature is enabled
> > > with -cpu kvm64. To disable: -cpu kvm64,-kvm_eoi.
> > > 
> > > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > > ---
> > > 
> > > Sending a copy to kernel list as this is needed
> > > to test the pv eoi feature recently submitted.
> > > 
> > >  target-i386/cpuid.c |    2 +-
> > >  1 files changed, 1 insertions(+), 1 deletions(-)
> > > 
> > > diff --git a/target-i386/cpuid.c b/target-i386/cpuid.c
> > > index 465ea15..c421b19 100644
> > > --- a/target-i386/cpuid.c
> > > +++ b/target-i386/cpuid.c
> > > @@ -75,7 +75,7 @@ static const char *ext3_feature_name[] = {
> > >  };
> > >  
> > >  static const char *kvm_feature_name[] = {
> > > -    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, NULL, NULL,
> > > +    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, "kvm_eoi", NULL,
> > 
> > Should libvirt also be recognizing this processor capability in its XML
> > in order to drive the -kvm_eoi option?
> 
> I think so.

A follow-up question: does libvirt use -cpu kvm?
How are such features normally handled for
cross-version migration?

> > -- 
> > Eric Blake   eblake@redhat.com    +1-919-301-3266
> > Libvirt virtualization library http://libvirt.org
> > 
> 
>

Patch

diff --git a/target-i386/cpuid.c b/target-i386/cpuid.c
index 465ea15..c421b19 100644
--- a/target-i386/cpuid.c
+++ b/target-i386/cpuid.c
@@ -75,7 +75,7 @@  static const char *ext3_feature_name[] = {
 };
 
 static const char *kvm_feature_name[] = {
-    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, NULL, NULL,
+    "kvmclock", "kvm_nopiodelay", "kvm_mmu", "kvmclock", "kvm_asyncpf", NULL, "kvm_eoi", NULL,
     NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
     NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
     NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,