[PATCH RFC] target/i386: filter out VMX_PIN_BASED_POSTED_INTR when enabling SynIC

[Vitaly Kuznetsov]

When a multi-vCPU guest is created with hv_synic, secondary vCPUs fail
to initialize with

qemu-system-x86_64: error: failed to set MSR 0x48d to 0xff00000016

This is caused by SynIC enablement on the boot CPU: when we do this
KVM disables apicv for the whole guest so we can't set
VMX_PIN_BASED_POSTED_INTR bit in MSR_IA32_VMX_TRUE_PINBASED_CTLS anymore.
(see nested_vmx_setup_ctls_msrs() in KVM).

This used to work before fine-grained VMX feature enablement because
we were not setting VMX MSRs.

Fix the issue by filtering out VMX_PIN_BASED_POSTED_INTR when enabling
SynIC. We also need to re-order kvm_init_msrs() with hyperv_init_vcpu()
so filtering on secondary CPUs happens before.

Signed-off-by: Vitaly Kuznetsov <address@hidden>
---
RFC: This is somewhat similar to eVMCS breakage and it is likely possible
to fix this in KVM. I decided to try QEMU first as this is a single
control and unlike eVMCS we don't need to keep a list of things to disable.
---
 target/i386/kvm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/target/i386/kvm.c b/target/i386/kvm.c
index 69eb43d796e6..6829b597fdbf 100644
--- a/target/i386/kvm.c
+++ b/target/i386/kvm.c
@@ -1366,6 +1366,7 @@ static Error *hv_no_nonarch_cs_mig_blocker;
 static int hyperv_init_vcpu(X86CPU *cpu)
 {
     CPUState *cs = CPU(cpu);
+    CPUX86State *env = &cpu->env;
     Error *local_err = NULL;
     int ret;
 
@@ -1431,6 +1432,9 @@ static int hyperv_init_vcpu(X86CPU *cpu)
             return ret;
         }
 
+        /* When SynIC is enabled, APICv controls become unavailable */
+        env->features[FEAT_VMX_PINBASED_CTLS] &= ~VMX_PIN_BASED_POSTED_INTR;
+
         if (!cpu->hyperv_synic_kvm_only) {
             ret = hyperv_x86_synic_add(cpu);
             if (ret < 0) {
@@ -1845,13 +1849,13 @@ int kvm_arch_init_vcpu(CPUState *cs)
         has_msr_tsc_aux = false;
     }
 
-    kvm_init_msrs(cpu);
-
     r = hyperv_init_vcpu(cpu);
     if (r) {
         goto fail;
     }
 
+    kvm_init_msrs(cpu);
+
     return 0;
 
  fail:
-- 
2.24.1