| Message ID | 1459371583-4824-1-git-send-email-pbonzini@redhat.com |
|---|---|
| State | New |
| Headers | show |
On 03/30/16 22:59, Paolo Bonzini wrote: > KVM does not let you read or write this MSR if the corresponding CPUID > bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX > to be ignored by KVM_SET_MRSS. typo: KVM_SET_MRSS / KVM_SET_MSRS Thanks for fixing this. Laszlo > > One visible symptom is that s3.flat from kvm-unit-tests fails with > CPUs that do not have RDTSCP, because the SMBASE is not reset to > 0x30000 after reset. > > Fixes: c9b8f6b6210847b4381c5b2ee172b1c7eb9985d6 > Cc: qemu-stable@nongnu.org > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> > --- > target-i386/kvm.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/target-i386/kvm.c b/target-i386/kvm.c > index 87ab969..19e2d94 100644 > --- a/target-i386/kvm.c > +++ b/target-i386/kvm.c > @@ -917,6 +917,9 @@ int kvm_arch_init_vcpu(CPUState *cs) > if (env->features[FEAT_1_EDX] & CPUID_MTRR) { > has_msr_mtrr = true; > } > + if (!(env->features[FEAT_8000_0001_EDX] & CPUID_EXT2_RDTSCP)) { > + has_msr_tsc_aux = false; > + } > > return 0; > } >
Am 31.03.2016 um 14:56 schrieb Laszlo Ersek: > On 03/30/16 22:59, Paolo Bonzini wrote: >> KVM does not let you read or write this MSR if the corresponding CPUID >> bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX >> to be ignored by KVM_SET_MRSS. Is it possible that this causes a freeze when migrating vom qemu 2.2.0 to 2.5.1? Thanks, Peter
On 31/03/2016 15:12, Peter Lieven wrote: >> >>> KVM does not let you read or write this MSR if the corresponding CPUID >>> bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX >>> to be ignored by KVM_SET_MRSS. > > Is it possible that this causes a freeze when migrating vom qemu 2.2.0 > to 2.5.1? I wouldn't exclude it if the CPU model is Westmere or earlier. Paolo
> Am 31.03.2016 um 15:23 schrieb Paolo Bonzini <pbonzini@redhat.com>: > > > >> On 31/03/2016 15:12, Peter Lieven wrote: >>>> >>>> KVM does not let you read or write this MSR if the corresponding CPUID >>>> bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX >>>> to be ignored by KVM_SET_MRSS. >> >> Is it possible that this causes a freeze when migrating vom qemu 2.2.0 >> to 2.5.1? > > I wouldn't exclude it if the CPU model is Westmere or earlier. the Host CPU or the Emulated CPU? We emulate Westmere and have also some old blades with Westmere CPUs. anyway, i give this a try. Michael, this should have gone into 2.5.1 i think... Thanks, Peter
On Wed, Mar 30, 2016 at 10:59:42PM +0200, Paolo Bonzini wrote: > KVM does not let you read or write this MSR if the corresponding CPUID > bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX > to be ignored by KVM_SET_MRSS. > > One visible symptom is that s3.flat from kvm-unit-tests fails with > CPUs that do not have RDTSCP, because the SMBASE is not reset to > 0x30000 after reset. > > Fixes: c9b8f6b6210847b4381c5b2ee172b1c7eb9985d6 > Cc: qemu-stable@nongnu.org > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
On 31/03/2016 17:38, Peter Lieven wrote: > > >> Am 31.03.2016 um 15:23 schrieb Paolo Bonzini <pbonzini@redhat.com>: >> >> >> >>> On 31/03/2016 15:12, Peter Lieven wrote: >>>>> >>>>> KVM does not let you read or write this MSR if the corresponding CPUID >>>>> bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX >>>>> to be ignored by KVM_SET_MRSS. >>> >>> Is it possible that this causes a freeze when migrating vom qemu 2.2.0 >>> to 2.5.1? >> >> I wouldn't exclude it if the CPU model is Westmere or earlier. > > the Host CPU or the Emulated CPU? The guest. Paolo > We emulate Westmere and have also some old blades with Westmere CPUs. > > anyway, i give this a try. > > Michael, this should have gone into 2.5.1 i think... > > Thanks, > Peter > >
Am 31.03.2016 um 18:52 schrieb Paolo Bonzini: > > On 31/03/2016 17:38, Peter Lieven wrote: >> >>> Am 31.03.2016 um 15:23 schrieb Paolo Bonzini <pbonzini@redhat.com>: >>> >>> >>> >>>> On 31/03/2016 15:12, Peter Lieven wrote: >>>>>> KVM does not let you read or write this MSR if the corresponding CPUID >>>>>> bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX >>>>>> to be ignored by KVM_SET_MRSS. >>>> Is it possible that this causes a freeze when migrating vom qemu 2.2.0 >>>> to 2.5.1? >>> I wouldn't exclude it if the CPU model is Westmere or earlier. >> the Host CPU or the Emulated CPU? > The guest. I can confirm this seems to fix the migration for emulated Westmere CPUs. Regardless if its a 2.2.0 -> 2.5.1 or 2.5.1 -> 2.5.1 migration. Peter
diff --git a/target-i386/kvm.c b/target-i386/kvm.c index 87ab969..19e2d94 100644 --- a/target-i386/kvm.c +++ b/target-i386/kvm.c @@ -917,6 +917,9 @@ int kvm_arch_init_vcpu(CPUState *cs) if (env->features[FEAT_1_EDX] & CPUID_MTRR) { has_msr_mtrr = true; } + if (!(env->features[FEAT_8000_0001_EDX] & CPUID_EXT2_RDTSCP)) { + has_msr_tsc_aux = false; + } return 0; }
KVM does not let you read or write this MSR if the corresponding CPUID bit is not set. This in turn causes MSRs that come after MSR_TSC_AUX to be ignored by KVM_SET_MRSS. One visible symptom is that s3.flat from kvm-unit-tests fails with CPUs that do not have RDTSCP, because the SMBASE is not reset to 0x30000 after reset. Fixes: c9b8f6b6210847b4381c5b2ee172b1c7eb9985d6 Cc: qemu-stable@nongnu.org Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> --- target-i386/kvm.c | 3 +++ 1 file changed, 3 insertions(+)