diff mbox series

[1/1] s390x/ais: for 2.10 stable: disable ais facility

Message ID 20170921151446.40015-2-borntraeger@de.ibm.com
State New
Headers show
Series 2.10-stable only: disable ais facility | expand

Commit Message

Christian Borntraeger Sept. 21, 2017, 3:14 p.m. UTC
The migration interface for ais was introduced with kernel 4.13
but the capability itself had been active since 4.12. As migration
support is considered necessary lets disable ais in the 2.10
stable version. A proper fix and re-enablement will be done
for qemu 2.11.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
---
 target/s390x/kvm.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

Comments

David Hildenbrand Sept. 21, 2017, 3:18 p.m. UTC | #1
On 21.09.2017 17:14, Christian Borntraeger wrote:
> The migration interface for ais was introduced with kernel 4.13
> but the capability itself had been active since 4.12. As migration
> support is considered necessary lets disable ais in the 2.10
> stable version. A proper fix and re-enablement will be done
> for qemu 2.11.
> 
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> Reviewed-by: Cornelia Huck <cohuck@redhat.com>
> ---
>  target/s390x/kvm.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
> index c4c5791..eb0dbb3 100644
> --- a/target/s390x/kvm.c
> +++ b/target/s390x/kvm.c
> @@ -308,8 +308,13 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
>          }
>      }
>  
> -    /* Try to enable AIS facility */
> -    kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
> +    /*
> +     * The migration interface for ais was introduced with kernel 4.13
> +     * but the capability itself had been active since 4.12. As migration
> +     * support is considered necessary let's disable ais in the 2.10
> +     * machine.
> +     */
> +    /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
>  
>      qemu_mutex_init(&qemu_sigp_mutex);
>  
> 

Reviewed-by: David Hildenbrand <david@redhat.com>
Cornelia Huck Sept. 21, 2017, 3:21 p.m. UTC | #2
On Thu, 21 Sep 2017 17:14:46 +0200
Christian Borntraeger <borntraeger@de.ibm.com> wrote:

> The migration interface for ais was introduced with kernel 4.13
> but the capability itself had been active since 4.12. As migration
> support is considered necessary lets disable ais in the 2.10
> stable version. A proper fix and re-enablement will be done
> for qemu 2.11.

Also now with qemu-devel on cc:

So, should I apply to s390-next? Or can this be applied just to stable?

> 
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> Reviewed-by: Cornelia Huck <cohuck@redhat.com>
> ---
>  target/s390x/kvm.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
> index c4c5791..eb0dbb3 100644
> --- a/target/s390x/kvm.c
> +++ b/target/s390x/kvm.c
> @@ -308,8 +308,13 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
>          }
>      }
>  
> -    /* Try to enable AIS facility */
> -    kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
> +    /*
> +     * The migration interface for ais was introduced with kernel 4.13
> +     * but the capability itself had been active since 4.12. As migration
> +     * support is considered necessary let's disable ais in the 2.10
> +     * machine.
> +     */
> +    /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
>  
>      qemu_mutex_init(&qemu_sigp_mutex);
>
Michael Roth Sept. 21, 2017, 4:50 p.m. UTC | #3
Quoting Cornelia Huck (2017-09-21 10:21:42)
> On Thu, 21 Sep 2017 17:14:46 +0200
> Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> 
> > The migration interface for ais was introduced with kernel 4.13
> > but the capability itself had been active since 4.12. As migration
> > support is considered necessary lets disable ais in the 2.10
> > stable version. A proper fix and re-enablement will be done
> > for qemu 2.11.
> 
> Also now with qemu-devel on cc:
> 
> So, should I apply to s390-next? Or can this be applied just to stable?

Ideally we would get it in master or at least a branch that'll
eventually get pulled so we can reference the upstream commit ID.
It's not set in stone, but usually only patches specifically ported
to stable are applied directly.

> 
> > 
> > Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> > Reviewed-by: Cornelia Huck <cohuck@redhat.com>
> > ---
> >  target/s390x/kvm.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
> > index c4c5791..eb0dbb3 100644
> > --- a/target/s390x/kvm.c
> > +++ b/target/s390x/kvm.c
> > @@ -308,8 +308,13 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
> >          }
> >      }
> >  
> > -    /* Try to enable AIS facility */
> > -    kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
> > +    /*
> > +     * The migration interface for ais was introduced with kernel 4.13
> > +     * but the capability itself had been active since 4.12. As migration
> > +     * support is considered necessary let's disable ais in the 2.10
> > +     * machine.
> > +     */
> > +    /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
> >  
> >      qemu_mutex_init(&qemu_sigp_mutex);
> >  
>
Michael Roth Sept. 21, 2017, 8:26 p.m. UTC | #4
Quoting Michael Roth (2017-09-21 11:50:28)
> Quoting Cornelia Huck (2017-09-21 10:21:42)
> > On Thu, 21 Sep 2017 17:14:46 +0200
> > Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> > 
> > > The migration interface for ais was introduced with kernel 4.13
> > > but the capability itself had been active since 4.12. As migration
> > > support is considered necessary lets disable ais in the 2.10
> > > stable version. A proper fix and re-enablement will be done
> > > for qemu 2.11.
> > 
> > Also now with qemu-devel on cc:
> > 
> > So, should I apply to s390-next? Or can this be applied just to stable?
> 
> Ideally we would get it in master or at least a branch that'll
> eventually get pulled so we can reference the upstream commit ID.
> It's not set in stone, but usually only patches specifically ported
> to stable are applied directly.

Somehow I missed the "for 2.10 stable" in the subject line. I can apply
this directly if it comes to that, but it seems safer to apply this
upstream as well in the meantime, IMO. If the proper fix didn't make
2.11 for whatever reason we'd still be okay with 2.10.1<->2.11, for
instance.

> 
> > 
> > > 
> > > Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> > > Reviewed-by: Cornelia Huck <cohuck@redhat.com>
> > > ---
> > >  target/s390x/kvm.c | 9 +++++++--
> > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
> > > index c4c5791..eb0dbb3 100644
> > > --- a/target/s390x/kvm.c
> > > +++ b/target/s390x/kvm.c
> > > @@ -308,8 +308,13 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
> > >          }
> > >      }
> > >  
> > > -    /* Try to enable AIS facility */
> > > -    kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
> > > +    /*
> > > +     * The migration interface for ais was introduced with kernel 4.13
> > > +     * but the capability itself had been active since 4.12. As migration
> > > +     * support is considered necessary let's disable ais in the 2.10
> > > +     * machine.
> > > +     */
> > > +    /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
> > >  
> > >      qemu_mutex_init(&qemu_sigp_mutex);
> > >  
> > 
> 
>
Christian Borntraeger Sept. 22, 2017, 6:47 a.m. UTC | #5
On 09/21/2017 10:26 PM, Michael Roth wrote:
> Quoting Michael Roth (2017-09-21 11:50:28)
>> Quoting Cornelia Huck (2017-09-21 10:21:42)
>>> On Thu, 21 Sep 2017 17:14:46 +0200
>>> Christian Borntraeger <borntraeger@de.ibm.com> wrote:
>>>
>>>> The migration interface for ais was introduced with kernel 4.13
>>>> but the capability itself had been active since 4.12. As migration
>>>> support is considered necessary lets disable ais in the 2.10
>>>> stable version. A proper fix and re-enablement will be done
>>>> for qemu 2.11.
>>>
>>> Also now with qemu-devel on cc:
>>>
>>> So, should I apply to s390-next? Or can this be applied just to stable?
>>
>> Ideally we would get it in master or at least a branch that'll
>> eventually get pulled so we can reference the upstream commit ID.
>> It's not set in stone, but usually only patches specifically ported
>> to stable are applied directly.
> 
> Somehow I missed the "for 2.10 stable" in the subject line. I can apply
> this directly if it comes to that, but it seems safer to apply this
> upstream as well in the meantime, IMO. If the proper fix didn't make
> 2.11 for whatever reason we'd still be okay with 2.10.1<->2.11, for
> instance.
> 

So what about the following. We will schedule this disable patch also for
current master. Whenever we have a proper fix we can then revert/fixup the
disable patch. This patch will then be backported to 2.10.1
For 2.11 we then fixup the ais code and also add the "disable ais for compat
machines" patch.

Christian
Cornelia Huck Sept. 22, 2017, 7:24 a.m. UTC | #6
On Fri, 22 Sep 2017 08:47:41 +0200
Christian Borntraeger <borntraeger@de.ibm.com> wrote:

> On 09/21/2017 10:26 PM, Michael Roth wrote:
> > Quoting Michael Roth (2017-09-21 11:50:28)  
> >> Quoting Cornelia Huck (2017-09-21 10:21:42)  
> >>> On Thu, 21 Sep 2017 17:14:46 +0200
> >>> Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> >>>  
> >>>> The migration interface for ais was introduced with kernel 4.13
> >>>> but the capability itself had been active since 4.12. As migration
> >>>> support is considered necessary lets disable ais in the 2.10
> >>>> stable version. A proper fix and re-enablement will be done
> >>>> for qemu 2.11.  
> >>>
> >>> Also now with qemu-devel on cc:
> >>>
> >>> So, should I apply to s390-next? Or can this be applied just to stable?  
> >>
> >> Ideally we would get it in master or at least a branch that'll
> >> eventually get pulled so we can reference the upstream commit ID.
> >> It's not set in stone, but usually only patches specifically ported
> >> to stable are applied directly.  
> > 
> > Somehow I missed the "for 2.10 stable" in the subject line. I can apply
> > this directly if it comes to that, but it seems safer to apply this
> > upstream as well in the meantime, IMO. If the proper fix didn't make
> > 2.11 for whatever reason we'd still be okay with 2.10.1<->2.11, for
> > instance.
> >   
> 
> So what about the following. We will schedule this disable patch also for
> current master. Whenever we have a proper fix we can then revert/fixup the
> disable patch. This patch will then be backported to 2.10.1
> For 2.11 we then fixup the ais code and also add the "disable ais for compat
> machines" patch.

Sounds good. I'll queue this one and send a pull request, just to be on
the safe side.
diff mbox series

Patch

diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
index c4c5791..eb0dbb3 100644
--- a/target/s390x/kvm.c
+++ b/target/s390x/kvm.c
@@ -308,8 +308,13 @@  int kvm_arch_init(MachineState *ms, KVMState *s)
         }
     }
 
-    /* Try to enable AIS facility */
-    kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
+    /*
+     * The migration interface for ais was introduced with kernel 4.13
+     * but the capability itself had been active since 4.12. As migration
+     * support is considered necessary let's disable ais in the 2.10
+     * machine.
+     */
+    /* kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0); */
 
     qemu_mutex_init(&qemu_sigp_mutex);