Re: [PATCH v7 51/52] i386/tdx: Validate phys_bits against host value

[Paolo Bonzini <pbonzini@redhat.com>]

On 2/2/25 15:39, Xiaoyao Li wrote:
> On 2/1/2025 2:27 AM, Paolo Bonzini wrote:
>> On Fri, Jan 24, 2025 at 2:40 PM Xiaoyao Li <xiaoyao.li@intel.com> wrote:
>>>
>>> For TDX guest, the phys_bits is not configurable and can only be
>>> host/native value.
>>>
>>> Validate phys_bits inside tdx_check_features().
>>
>> Hi Xiaoyao,
>>
>> to avoid
>>
>> qemu-kvm: TDX requires guest CPU physical bits (48) to match host CPU
>> physical bits (52)
>>
>> I need options like
>>
>> -cpu host,phys-bits=52,guest-phys-bits=52,host-phys-bits-limit=52,- 
>> kvm-asyncpf-int
>>
>> to start a TDX guest, is that intentional?
> 
> "-cpu host" should be sufficient and should not hit the error.
> 
> why did you get "guest CPU physical bits (48)"?

Nevermind - I got it from the RHEL machine types.  The fix (which does not have
to be included upstream, though I would not complain) is

diff --git a/target/i386/kvm/tdx.c b/target/i386/kvm/tdx.c
index e0ab41bcb7b..7aca8219405 100644
--- a/target/i386/kvm/tdx.c
+++ b/target/i386/kvm/tdx.c
@@ -444,6 +444,7 @@ static void tdx_cpu_instance_init(X86ConfidentialGuest *cg, CPUState *cpu)
      X86CPU *x86cpu = X86_CPU(cpu);
  
      object_property_set_bool(OBJECT(cpu), "pmu", false, &error_abort);
+    object_property_set_bool(OBJECT(cpu), "host-phys-bits-limit", 0, &error_abort);
  
      /* invtsc is fixed1 for TD guest */
      object_property_set_bool(OBJECT(cpu), "invtsc", true, &error_abort);

but it also needs the patch at
https://lore.kernel.org/qemu-devel/20250203114132.259155-1-pbonzini@redhat.com/T/.

With these two changes, QEMU accepts "-cpu host" just fine.

Paolo