Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 22 Jul 2022 18:33:27 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Vitaly Kuznetsov <vkuznets@redhat.com>
Cc:     kvm@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
        Anirudh Rayabharam <anrayabh@linux.microsoft.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Maxim Levitsky <mlevitsk@redhat.com>,
        linux-hyperv@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 15/25] KVM: VMX: Extend VMX controls macro shenanigans
Message-ID: <YtrtdylmyolAHToz@google.com>
References: <20220714091327.1085353-1-vkuznets@redhat.com>
 <20220714091327.1085353-16-vkuznets@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220714091327.1085353-16-vkuznets@redhat.com>
Precedence: bulk

On Thu, Jul 14, 2022, Vitaly Kuznetsov wrote:
> diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
> index 286c88e285ea..89eaab3495a6 100644
> --- a/arch/x86/kvm/vmx/vmx.h
> +++ b/arch/x86/kvm/vmx/vmx.h
> @@ -467,6 +467,113 @@ static inline u8 vmx_get_rvi(void)
>  	return vmcs_read16(GUEST_INTR_STATUS) & 0xff;
>  }
>  
> +#define __KVM_REQ_VMX_VM_ENTRY_CONTROLS				\
> +	(VM_ENTRY_LOAD_DEBUG_CONTROLS)
> +#ifdef CONFIG_X86_64
> +	#define KVM_REQ_VMX_VM_ENTRY_CONTROLS			\
> +		(__KVM_REQ_VMX_VM_ENTRY_CONTROLS |		\
> +		VM_ENTRY_IA32E_MODE)

This breaks 32-bit builds, but at least we know the assert works!

vmx_set_efer() toggles VM_ENTRY_IA32E_MODE without a CONFIG_X86_64 guard.  That
should be easy enough to fix since KVM should never allow EFER_LMA.  Compile 
tested patch at the bottom.

More problematic is that clang-13 doesn't like the new asserts, and even worse gives
a very cryptic error.  I don't have bandwidth to look into this at the moment, and
probably won't next week either.

ERROR: modpost: "__compiletime_assert_533" [arch/x86/kvm/kvm-intel.ko] undefined!
ERROR: modpost: "__compiletime_assert_531" [arch/x86/kvm/kvm-intel.ko] undefined!
ERROR: modpost: "__compiletime_assert_532" [arch/x86/kvm/kvm-intel.ko] undefined!
ERROR: modpost: "__compiletime_assert_530" [arch/x86/kvm/kvm-intel.ko] undefined!
make[2]: *** [scripts/Makefile.modpost:128: modules-only.symvers] Error 1
make[1]: *** [Makefile:1753: modules] Error 2
make[1]: *** Waiting for unfinished jobs....


> +#else
> +	#define KVM_REQ_VMX_VM_ENTRY_CONTROLS			\
> +		__KVM_REQ_VMX_VM_ENTRY_CONTROLS
> +#endif

EFER.LMA patch, compile tested only.

---
From: Sean Christopherson <seanjc@google.com>
Date: Fri, 22 Jul 2022 18:26:21 +0000
Subject: [PATCH] KVM: VMX: Don't toggle VM_ENTRY_IA32E_MODE for 32-bit
 kernels/KVM

Don't toggle VM_ENTRY_IA32E_MODE in 32-bit kernels/KVM and instead bug
the VM if KVM attempts to run the guest with EFER.LMA=1.  KVM doesn't
support running 64-bit guests with 32-bit hosts.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/vmx/vmx.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index bff97babf381..8623607e596d 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -2894,10 +2894,15 @@ int vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer)
 		return 0;

 	vcpu->arch.efer = efer;
+#ifdef CONFIG_X86_64
 	if (efer & EFER_LMA)
 		vm_entry_controls_setbit(vmx, VM_ENTRY_IA32E_MODE);
 	else
 		vm_entry_controls_clearbit(vmx, VM_ENTRY_IA32E_MODE);
+#else
+	if (KVM_BUG_ON(efer & EFER_LMA, vcpu->kvm))
+		return 1;
+#endif

 	vmx_setup_uret_msrs(vmx);
 	return 0;

base-commit: e22e2665637151a321433b2bb705f5c3b8da40bc
--