Message ID | 20170915101406.27823-2-cohuck@redhat.com |
---|---|
State | New |
Headers | show |
Series | s390x: pci compat handling | expand |
On 15.09.2017 12:14, 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 > some old machines. > > Create the s390 phb if we need it for compat reasons, even if > we don't provide the zpci facility. > > Fixes: d32bd032d8 ("s390x/ccw: create s390 phb conditionally") > Signed-off-by: Cornelia Huck <cohuck@redhat.com> > --- > hw/s390x/s390-virtio-ccw.c | 6 +++++- > include/hw/s390x/s390-virtio-ccw.h | 1 + > 2 files changed, 6 insertions(+), 1 deletion(-) > > diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c > index 0471407187..7e3148eb4a 100644 > --- a/hw/s390x/s390-virtio-ccw.c > +++ b/hw/s390x/s390-virtio-ccw.c > @@ -269,6 +269,8 @@ static void s390_create_virtio_net(BusState *bus, const char *name) > } > } > > +static S390CcwMachineClass *get_machine_class(void); > + > static void ccw_init(MachineState *machine) > { > int ret; > @@ -288,7 +290,7 @@ static void ccw_init(MachineState *machine) > machine->initrd_filename, "s390-ccw.img", > "s390-netboot.img", true); > > - if (s390_has_feat(S390_FEAT_ZPCI)) { > + if (s390_has_feat(S390_FEAT_ZPCI) || get_machine_class()->pci_compat) { > DeviceState *dev = qdev_create(NULL, TYPE_S390_PCI_HOST_BRIDGE); > object_property_add_child(qdev_get_machine(), > TYPE_S390_PCI_HOST_BRIDGE, > @@ -429,6 +431,7 @@ static void ccw_machine_class_init(ObjectClass *oc, void *data) > s390mc->cpu_model_allowed = true; > s390mc->css_migration_enabled = true; > s390mc->gs_allowed = true; > + s390mc->pci_compat = false; > mc->init = ccw_init; > mc->reset = s390_machine_reset; > mc->hot_add_cpu = s390_hot_add_cpu; > @@ -784,6 +787,7 @@ static void ccw_machine_2_7_class_options(MachineClass *mc) > S390CcwMachineClass *s390mc = S390_MACHINE_CLASS(mc); > > s390mc->cpu_model_allowed = false; > + s390mc->pci_compat = pci_available; > ccw_machine_2_8_class_options(mc); > SET_MACHINE_COMPAT(mc, CCW_COMPAT_2_7); > } > diff --git a/include/hw/s390x/s390-virtio-ccw.h b/include/hw/s390x/s390-virtio-ccw.h > index a9a90c2022..9978c89e90 100644 > --- a/include/hw/s390x/s390-virtio-ccw.h > +++ b/include/hw/s390x/s390-virtio-ccw.h > @@ -41,6 +41,7 @@ typedef struct S390CcwMachineClass { > bool cpu_model_allowed; > bool css_migration_enabled; > bool gs_allowed; > + bool pci_compat; > } S390CcwMachineClass; > > /* runtime-instrumentation allowed by the machine */ > As an alternative, simply diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c index 0f28ebd162..0f22efc3b6 100644 --- a/target/s390x/cpu_models.c +++ b/target/s390x/cpu_models.c @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) return true; } } + if (feat == S390_FEAT_ZPCI) + return true; #endif return 0; } (this is also the way we handle other features without cpu model support -especially if cpu model support is disabled for older machines)
On Fri, 15 Sep 2017 13:57:40 +0200 David Hildenbrand <david@redhat.com> wrote: > As an alternative, simply > > diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c > index 0f28ebd162..0f22efc3b6 100644 > --- a/target/s390x/cpu_models.c > +++ b/target/s390x/cpu_models.c > @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) > return true; > } > } > + if (feat == S390_FEAT_ZPCI) > + return true; Move that out of the CONFIG_KVM #ifdef? (Also, we still have the issue with pci support :/ - depend on pci_available?) > #endif > return 0; > } > > > (this is also the way we handle other features without cpu model support > -especially if cpu model support is disabled for older machines) Now that you mention cpu models (which are disabled on old machines), the failure starts to look plausible :)
On 15.09.2017 15:07, Cornelia Huck wrote: > On Fri, 15 Sep 2017 13:57:40 +0200 > David Hildenbrand <david@redhat.com> wrote: > >> As an alternative, simply >> >> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c >> index 0f28ebd162..0f22efc3b6 100644 >> --- a/target/s390x/cpu_models.c >> +++ b/target/s390x/cpu_models.c >> @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) >> return true; >> } >> } >> + if (feat == S390_FEAT_ZPCI) >> + return true; > > Move that out of the CONFIG_KVM #ifdef? > Should not make a difference I think, cpu models are not shielded off for TCG I think (maybe we should fix that, but doesn't look that easy - they have no host model which would allow cpu->model = NULL). > (Also, we still have the issue with pci support :/ - depend on > pci_available?) hm .... > >> #endif >> return 0; >> } >> >> >> (this is also the way we handle other features without cpu model support >> -especially if cpu model support is disabled for older machines) > > Now that you mention cpu models (which are disabled on old machines), > the failure starts to look plausible :) >
On Fri, 15 Sep 2017 15:09:02 +0200 David Hildenbrand <david@redhat.com> wrote: > On 15.09.2017 15:07, Cornelia Huck wrote: > > On Fri, 15 Sep 2017 13:57:40 +0200 > > David Hildenbrand <david@redhat.com> wrote: > > > >> As an alternative, simply > >> > >> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c > >> index 0f28ebd162..0f22efc3b6 100644 > >> --- a/target/s390x/cpu_models.c > >> +++ b/target/s390x/cpu_models.c > >> @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) > >> return true; > >> } > >> } > >> + if (feat == S390_FEAT_ZPCI) > >> + return true; > > > > Move that out of the CONFIG_KVM #ifdef? > > > > Should not make a difference I think, cpu models are not shielded off > for TCG I think (maybe we should fix that, but doesn't look that easy - > they have no host model which would allow cpu->model = NULL). It just looks odd to have it inside the define... > > > (Also, we still have the issue with pci support :/ - depend on > > pci_available?) > > hm .... Should we maybe disallow compat machines if CONFIG_PCI is off, and think about something for the 2.11 machine? Otherwise, I cannot see that end well :/ But I'll just draw up a v2 for now.
On 15.09.2017 15:22, Cornelia Huck wrote: > On Fri, 15 Sep 2017 15:09:02 +0200 > David Hildenbrand <david@redhat.com> wrote: > >> On 15.09.2017 15:07, Cornelia Huck wrote: >>> On Fri, 15 Sep 2017 13:57:40 +0200 >>> David Hildenbrand <david@redhat.com> wrote: >>> >>>> As an alternative, simply >>>> >>>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c >>>> index 0f28ebd162..0f22efc3b6 100644 >>>> --- a/target/s390x/cpu_models.c >>>> +++ b/target/s390x/cpu_models.c >>>> @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) >>>> return true; >>>> } >>>> } >>>> + if (feat == S390_FEAT_ZPCI) >>>> + return true; >>> >>> Move that out of the CONFIG_KVM #ifdef? >>> >> >> Should not make a difference I think, cpu models are not shielded off >> for TCG I think (maybe we should fix that, but doesn't look that easy - >> they have no host model which would allow cpu->model = NULL). > > It just looks odd to have it inside the define... sure, you can then also just move it into kvm_enable() next to the other ones. > >> >>> (Also, we still have the issue with pci support :/ - depend on >>> pci_available?) >> >> hm .... > > Should we maybe disallow compat machines if CONFIG_PCI is off, and > think about something for the 2.11 machine? Otherwise, I cannot see > that end well :/ If they required PCI, than that is the only way to do it. > > But I'll just draw up a v2 for now. >
On 09/15/2017 03:37 PM, David Hildenbrand wrote: > On 15.09.2017 15:22, Cornelia Huck wrote: >> On Fri, 15 Sep 2017 15:09:02 +0200 >> David Hildenbrand <david@redhat.com> wrote: >> >>> On 15.09.2017 15:07, Cornelia Huck wrote: >>>> On Fri, 15 Sep 2017 13:57:40 +0200 >>>> David Hildenbrand <david@redhat.com> wrote: >>>> >>>>> As an alternative, simply >>>>> >>>>> diff --git a/target/s390x/cpu_models.c b/target/s390x/cpu_models.c >>>>> index 0f28ebd162..0f22efc3b6 100644 >>>>> --- a/target/s390x/cpu_models.c >>>>> +++ b/target/s390x/cpu_models.c >>>>> @@ -195,6 +195,8 @@ bool s390_has_feat(S390Feat feat) >>>>> return true; >>>>> } >>>>> } >>>>> + if (feat == S390_FEAT_ZPCI) >>>>> + return true; >>>> >>>> Move that out of the CONFIG_KVM #ifdef? >>>> >>> >>> Should not make a difference I think, cpu models are not shielded off >>> for TCG I think (maybe we should fix that, but doesn't look that easy - >>> they have no host model which would allow cpu->model = NULL). >> >> It just looks odd to have it inside the define... > > sure, you can then also just move it into kvm_enable() next to the other > ones. > >> >>> >>>> (Also, we still have the issue with pci support :/ - depend on >>>> pci_available?) >>> >>> hm .... >> >> Should we maybe disallow compat machines if CONFIG_PCI is off, and >> think about something for the 2.11 machine? Otherwise, I cannot see >> that end well :/ > > If they required PCI, than that is the only way to do it. > >> >> But I'll just draw up a v2 for now. >> Sorry for not contributing anything here. I don't have the whole thing in my head, so I can't quite follow without digging into this. I don't think I will be able to join today -- maybe I could have a look next week . Halil > >
diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c index 0471407187..7e3148eb4a 100644 --- a/hw/s390x/s390-virtio-ccw.c +++ b/hw/s390x/s390-virtio-ccw.c @@ -269,6 +269,8 @@ static void s390_create_virtio_net(BusState *bus, const char *name) } } +static S390CcwMachineClass *get_machine_class(void); + static void ccw_init(MachineState *machine) { int ret; @@ -288,7 +290,7 @@ static void ccw_init(MachineState *machine) machine->initrd_filename, "s390-ccw.img", "s390-netboot.img", true); - if (s390_has_feat(S390_FEAT_ZPCI)) { + if (s390_has_feat(S390_FEAT_ZPCI) || get_machine_class()->pci_compat) { DeviceState *dev = qdev_create(NULL, TYPE_S390_PCI_HOST_BRIDGE); object_property_add_child(qdev_get_machine(), TYPE_S390_PCI_HOST_BRIDGE, @@ -429,6 +431,7 @@ static void ccw_machine_class_init(ObjectClass *oc, void *data) s390mc->cpu_model_allowed = true; s390mc->css_migration_enabled = true; s390mc->gs_allowed = true; + s390mc->pci_compat = false; mc->init = ccw_init; mc->reset = s390_machine_reset; mc->hot_add_cpu = s390_hot_add_cpu; @@ -784,6 +787,7 @@ static void ccw_machine_2_7_class_options(MachineClass *mc) S390CcwMachineClass *s390mc = S390_MACHINE_CLASS(mc); s390mc->cpu_model_allowed = false; + s390mc->pci_compat = pci_available; ccw_machine_2_8_class_options(mc); SET_MACHINE_COMPAT(mc, CCW_COMPAT_2_7); } diff --git a/include/hw/s390x/s390-virtio-ccw.h b/include/hw/s390x/s390-virtio-ccw.h index a9a90c2022..9978c89e90 100644 --- a/include/hw/s390x/s390-virtio-ccw.h +++ b/include/hw/s390x/s390-virtio-ccw.h @@ -41,6 +41,7 @@ typedef struct S390CcwMachineClass { bool cpu_model_allowed; bool css_migration_enabled; bool gs_allowed; + bool pci_compat; } S390CcwMachineClass; /* runtime-instrumentation allowed by the machine */
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 some old machines. Create the s390 phb if we need it for compat reasons, even if we don't provide the zpci facility. Fixes: d32bd032d8 ("s390x/ccw: create s390 phb conditionally") Signed-off-by: Cornelia Huck <cohuck@redhat.com> --- hw/s390x/s390-virtio-ccw.c | 6 +++++- include/hw/s390x/s390-virtio-ccw.h | 1 + 2 files changed, 6 insertions(+), 1 deletion(-)