Re: [Qemu-devel] [PATCH v3] s390x/ccw: create s390 phb for compat reasons as well

[Cornelia Huck <cohuck@redhat.com>]

On Mon, 18 Sep 2017 14:13:07 +0200
David Hildenbrand <david@redhat.com> wrote:

> On 18.09.2017 14:11, Cornelia Huck wrote:
> > On Mon, 18 Sep 2017 14:03:20 +0200
> > David Hildenbrand <david@redhat.com> wrote:
> >   
> >> On 18.09.2017 10:55, Cornelia Huck wrote:  
> >>> d32bd032d8 ("s390x/ccw: create s390 phb conditionally") made
> >>> registering the s390 pci host bridge conditional on presense
> >>> of the zpci facility bit. Sadly, that breaks migration from
> >>> machines that did not use the cpu model (2.7 and previous).
> >>>
> >>> Create the s390 phb for pre-cpu model machines as well: We can
> >>> tweak s390_has_feat() to always indicate the zpci facility bit
> >>> when no cpu model is available (on 2.7 and previous compat machines).
> >>>
> >>> Fixes: d32bd032d8 ("s390x/ccw: create s390 phb conditionally")
> >>> Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
> >>> Signed-off-by: Cornelia Huck <cohuck@redhat.com>
> >>> ---
> >>>
> >>> v2->v3:
> >>> - no longer RFC (I tested a bit more)
> >>> - removed unrelated hunk
> >>> - more verbose patch description
> >>>
> >>> I'll wait a bit for more acks/reviews and will probably send a pull
> >>> request for s390x tomorrow or so before the amount of queued patches
> >>> gets out of hand...
> >>>
> >>> ---
> >>>  target/s390x/cpu_models.c | 3 +++
> >>>  1 file changed, 3 insertions(+)
> >>>
> >>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c
> >>> index c295e641e6..5169379db5 100644
> >>> --- a/target/s390x/cpu_models.c
> >>> +++ b/target/s390x/cpu_models.c
> >>> @@ -196,6 +196,9 @@ bool s390_has_feat(S390Feat feat)
> >>>              }
> >>>          }
> >>>  #endif
> >>> +        if (feat == S390_FEAT_ZPCI) {
> >>> +            return true;
> >>> +        }
> >>>          return 0;
> >>>      }
> >>>      return test_bit(feat, cpu->model->features);
> >>>     
> >>
> >> 1. cpu->model will always be set for QEMU, so you can move that into the
> >> ifdef, next do the other checks. You can even send a cleanup to remove
> >> the if (kvm_enabled()) check.  
> > 
> > I prefer it the way it is now. There's nothing kvm specific about that
> > bit, and cpu->model always being set is not really obvious.
> >   
> 
> I's already kvm specific as this can never happen with TCG :)
> 
> But whatever you prefer. This is good enough to fix the problem.

Sure, we can revisit that one later. I plan to send a pull request
tomorrow.