[RFC PATCH 3/3] KVM: x86: Use KVM_BUG/KVM_BUG_ON to handle bugs that are fatal to the VM

[Sean Christopherson]

On Thu, Sep 24, 2020 at 02:34:14PM +0200, Vitaly Kuznetsov wrote:
> Sean Christopherson <sean.j.christopherson at intel.com> writes:
> > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> > index 6f9a0c6d5dc5..810d46ab0a47 100644
> > --- a/arch/x86/kvm/vmx/vmx.c
> > +++ b/arch/x86/kvm/vmx/vmx.c
> > @@ -2250,7 +2250,7 @@ static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg)
> >  		vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & guest_owned_bits;
> >  		break;
> >  	default:
> > -		WARN_ON_ONCE(1);
> > +		KVM_BUG_ON(1, vcpu->kvm);
> >  		break;
> >  	}
> >  }
> > @@ -4960,6 +4960,7 @@ static int handle_cr(struct kvm_vcpu *vcpu)
> >  			return kvm_complete_insn_gp(vcpu, err);
> >  		case 3:
> >  			WARN_ON_ONCE(enable_unrestricted_guest);
> > +
> >  			err = kvm_set_cr3(vcpu, val);
> >  			return kvm_complete_insn_gp(vcpu, err);
> >  		case 4:
> > @@ -4985,14 +4986,13 @@ static int handle_cr(struct kvm_vcpu *vcpu)
> >  		}
> >  		break;
> >  	case 2: /* clts */
> > -		WARN_ONCE(1, "Guest should always own CR0.TS");
> > -		vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS));
> > -		trace_kvm_cr_write(0, kvm_read_cr0(vcpu));
> > -		return kvm_skip_emulated_instruction(vcpu);
> > +		KVM_BUG(1, vcpu->kvm, "Guest always owns CR0.TS");
> > +		return -EIO;
> >  	case 1: /*mov from cr*/
> >  		switch (cr) {
> >  		case 3:
> >  			WARN_ON_ONCE(enable_unrestricted_guest);
> > +
> 
> Here, were you intended to replace WARN_ON_ONCE() with KVM_BUG_ON() or
> this is just a stray newline added?

I think it's just a stray newline.  At one point I had converted this to a
KVM_BUG_ON(), but then reversed direction because it's not fatal to the guest,
i.e. KVM should continue to function even though it's spuriously intercepting
CR3 loads.

Which, rereading this patch, completely contradicts the KVM_BUG() for CLTS.

That's probably something we should sort out in this RFC: is KVM_BUG() only
to be used if the bug is fatal/dangerous, or should it be used any time the
error is definitely a KVM (or hardware) bug.

> >  			val = kvm_read_cr3(vcpu);
> >  			kvm_register_write(vcpu, reg, val);
> >  			trace_kvm_cr_read(cr, val);
> > @@ -5330,7 +5330,9 @@ static int handle_ept_misconfig(struct kvm_vcpu *vcpu)
> >  
> >  static int handle_nmi_window(struct kvm_vcpu *vcpu)
> >  {
> > -	WARN_ON_ONCE(!enable_vnmi);
> > +	if (KVM_BUG_ON(!enable_vnmi, vcpu->kvm))
> > +		return -EIO;
> > +
> >  	exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING);
> >  	++vcpu->stat.nmi_window_exits;
> >  	kvm_make_request(KVM_REQ_EVENT, vcpu);
> > @@ -5908,7 +5910,8 @@ static int vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
> >  	 * below) should never happen as that means we incorrectly allowed a
> >  	 * nested VM-Enter with an invalid vmcs12.
> >  	 */
> > -	WARN_ON_ONCE(vmx->nested.nested_run_pending);
> > +	if (KVM_BUG_ON(vmx->nested.nested_run_pending, vcpu->kvm))
> > +		return -EIO;
> >  
> >  	/* If guest state is invalid, start emulating */
> >  	if (vmx->emulation_required)
> > @@ -6258,7 +6261,9 @@ static int vmx_sync_pir_to_irr(struct kvm_vcpu *vcpu)
> >  	int max_irr;
> >  	bool max_irr_updated;
> >  
> > -	WARN_ON(!vcpu->arch.apicv_active);
> > +	if (KVM_BUG_ON(!vcpu->arch.apicv_active, vcpu->kvm))
> > +		return -EIO;
> > +
> >  	if (pi_test_on(&vmx->pi_desc)) {
> >  		pi_clear_on(&vmx->pi_desc);
> >  		/*
> > @@ -6345,7 +6350,7 @@ static void handle_external_interrupt_irqoff(struct kvm_vcpu *vcpu)
> >  	gate_desc *desc;
> >  	u32 intr_info = vmx_get_intr_info(vcpu);
> >  
> > -	if (WARN_ONCE(!is_external_intr(intr_info),
> > +	if (KVM_BUG(!is_external_intr(intr_info), vcpu->kvm,
> >  	    "KVM: unexpected VM-Exit interrupt info: 0x%x", intr_info))
> >  		return;
> >  
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index 17f4995e80a7..672eb5142b34 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -8363,6 +8363,10 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> >  	bool req_immediate_exit = false;
> >  
> >  	if (kvm_request_pending(vcpu)) {
> > +		if (kvm_check_request(KVM_REQ_VM_BUGGED, vcpu)) {
> 
> Do we want to allow userspace to continue executing the guest or should
> we make KVM_REQ_VM_BUGGED permanent by replacing kvm_check_request()
> with kvm_test_request()?

In theory, it should be impossible to reach this again as "r = -EIO" will
bounce this out to userspace, the common checks to deny all ioctls() will
prevent reinvoking KVM_RUN.

> > +			r = -EIO;
> > +			goto out;
> > +		}
> >  		if (kvm_check_request(KVM_REQ_GET_VMCS12_PAGES, vcpu)) {
> >  			if (unlikely(!kvm_x86_ops.nested_ops->get_vmcs12_pages(vcpu))) {
> >  				r = 0;
> 
> -- 
> Vitaly
>