KVM: x86: Don't adjust guest's CPUID.0x12.1 (allowed SGX enclave XFRM)
authorSean Christopherson <seanjc@google.com>
Wed, 3 May 2023 16:08:37 +0000 (09:08 -0700)
committerPaolo Bonzini <pbonzini@redhat.com>
Sun, 21 May 2023 08:05:51 +0000 (04:05 -0400)
Drop KVM's manipulation of guest's CPUID.0x12.1 ECX and EDX, i.e. the
allowed XFRM of SGX enclaves, now that KVM explicitly checks the guest's
allowed XCR0 when emulating ECREATE.

Note, this could theoretically break a setup where userspace advertises
a "bad" XFRM and relies on KVM to provide a sane CPUID model, but QEMU
is the only known user of KVM SGX, and QEMU explicitly sets the SGX CPUID
XFRM subleaf based on the guest's XCR0.

Reviewed-by: Kai Huang <kai.huang@intel.com>
Tested-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230503160838.3412617-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/cpuid.c

index 123bf8b97a4b21098a8271320f711cecf2eae5ec..0c9660a07b233ff404147fa6acda76668241f7ee 100644 (file)
@@ -253,7 +253,6 @@ static void __kvm_update_cpuid_runtime(struct kvm_vcpu *vcpu, struct kvm_cpuid_e
                                       int nent)
 {
        struct kvm_cpuid_entry2 *best;
-       u64 guest_supported_xcr0 = cpuid_get_supported_xcr0(entries, nent);
 
        best = cpuid_entry2_find(entries, nent, 1, KVM_CPUID_INDEX_NOT_SIGNIFICANT);
        if (best) {
@@ -292,21 +291,6 @@ static void __kvm_update_cpuid_runtime(struct kvm_vcpu *vcpu, struct kvm_cpuid_e
                                           vcpu->arch.ia32_misc_enable_msr &
                                           MSR_IA32_MISC_ENABLE_MWAIT);
        }
-
-       /*
-        * Bits 127:0 of the allowed SECS.ATTRIBUTES (CPUID.0x12.0x1) enumerate
-        * the supported XSAVE Feature Request Mask (XFRM), i.e. the enclave's
-        * requested XCR0 value.  The enclave's XFRM must be a subset of XCRO
-        * at the time of EENTER, thus adjust the allowed XFRM by the guest's
-        * supported XCR0.  Similar to XCR0 handling, FP and SSE are forced to
-        * '1' even on CPUs that don't support XSAVE.
-        */
-       best = cpuid_entry2_find(entries, nent, 0x12, 0x1);
-       if (best) {
-               best->ecx &= guest_supported_xcr0 & 0xffffffff;
-               best->edx &= guest_supported_xcr0 >> 32;
-               best->ecx |= XFEATURE_MASK_FPSSE;
-       }
 }
 
 void kvm_update_cpuid_runtime(struct kvm_vcpu *vcpu)