Re: [sean-jc:x86/evmcs_vmptrldst] [KVM] 7fecb27e8d: bunch of kselftests.kvm failures

[Sean Christopherson <seanjc@google.com>]

On Thu, Sep 03, 2026, Sean Christopherson wrote:
> On Thu, Sep 03, 2026, kernel test robot wrote:
> > kernel test robot noticed "kselftests.kvm.hyperv_evmcs.fail" on:
> > 
> > commit: 7fecb27e8d353adda91989eb01b3ae200b846a42 ("KVM: selftests: Always assert that vmwrite() succeeds") https://github.com/sean-jc/linux x86/evmcs_vmptrldst
> > 
> > in testcase: kselftests
> > version: kselftests-x86_64-f5098b6bae76-1_20260727
> > with following parameters:
> > 
> > 	group: kvm
> > 
> > config: x86_64-rhel-9.4-kselftests
> > compiler: gcc-14
> > test machine: 16 threads Intel(R) Core(TM) i7-13620H (Raptor Lake) with 32G memory
> > 
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
> > | Reported-by: kernel test robot <yi1.lai@intel.com>
> > | Closes: https://lore.kernel.org/oe-lkp/202609030248.b516759d-lkp@intel.com
> > 
> > The kernel config and materials to reproduce are available at:
> > https://download.01.org/0day-ci/archive/20260903/202609030248.b516759d-lkp@intel.com
> > 
> > Detailed log:
> > # selftests: kvm: evmcs_smm_controls_test
> > # Random seed: 0x7ad5b4f9
> > # 
> > # ==== Test Assertion Failure ====
> > #   include/x86/vmx.h:434: __vmwrite(encoding, value)
> > #   pid=5240 tid=5240 errno=4 - Interrupted system call
> > #      1	0x0000000000412f61: assert_on_unhandled_exception at processor.c:781
> > #      2	0x0000000000406f86: _vcpu_run at kvm_util.c:1687
> > #      3	 (inlined by) vcpu_run at kvm_util.c:1698
> > #      4	0x0000000000402cc6: main at evmcs_smm_controls_test.c:93
> > #      5	0x00007f0a85d39ca7: ?? ??:0
> > #      6	0x00007f0a85d39d64: ?? ??:0
> > #      7	0x00000000004030c0: _start at ??:?
> > #   vmwrite[0x0] = 0x0 hit VM-Fail
> 
> /facepalm
> 
> I managed to *repeatedly* not update the repository I was using to build selftests.
> Thank you friendly kernel test robot for saving me from major embarassment!

Ok, this is too funny not to tell someone, even if that someone ends up being a
bot.

I thought for sure I messed up the encoding, i.e. that the "vmwrite[0x0]" in the
report was either wrong, or or the encoding was getting clobbered along the way.
Nope, it just so happens that VIRTUAL_PROCESSOR_ID is encoding '0', and the very
first thing done by prepare_vmcs() => init_vmcs_control_fields() is:

	vmwrite(VIRTUAL_PROCESSOR_ID, 0);

The actual bug is stupid simple (and also hilarious/typical):

diff --git a/tools/testing/selftests/kvm/include/x86/vmx.h b/tools/testing/selftests/kvm/include/x86/vmx.h
index 81028c24dc9e..a4807e41fe69 100644
--- a/tools/testing/selftests/kvm/include/x86/vmx.h
+++ b/tools/testing/selftests/kvm/include/x86/vmx.h
@@ -430,7 +430,7 @@ static __always_inline int __vmwrite(u64 encoding, u64 value)
 
 static inline void vmwrite(u64 encoding, u64 value)
 {
-	__GUEST_ASSERT(__vmwrite(encoding, value),
+	__GUEST_ASSERT(!__vmwrite(encoding, value),
 		       "vmwrite[0x%lx] = 0x%lx hit VM-Fail", encoding, value);
 }
 
Just to double check that I didn't botch the formatting, I verified that this:

diff --git a/tools/testing/selftests/kvm/lib/x86/vmx.c b/tools/testing/selftests/kvm/lib/x86/vmx.c
index f32f8319301d..6066d200bcc2 100644
--- a/tools/testing/selftests/kvm/lib/x86/vmx.c
+++ b/tools/testing/selftests/kvm/lib/x86/vmx.c
@@ -251,11 +251,11 @@ static inline void init_vmcs_control_fields(struct vmx_pages *vmx)
 {
 	u32 sec_exec_ctl = 0;
 
+	vmwrite(PIN_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PINBASED_CTLS));
+
 	vmwrite(VIRTUAL_PROCESSOR_ID, 0);
 	vmwrite(POSTED_INTR_NV, 0);
 
-	vmwrite(PIN_BASED_VM_EXEC_CONTROL, rdmsr(MSR_IA32_VMX_TRUE_PINBASED_CTLS));
-
 	if (vmx->eptp_gpa) {
 		u64 eptp = vmx->eptp_gpa | EPTP_WB | EPTP_PWL_4;
 
Yields:

  vmwrite[0x4000] = 0x3f00000016 hit VM-Fail