KVM: x86: Disallow KVM_SET_SREGS{2} if incoming CR0 is invalid

Reject KVM_SET_SREGS{2} with -EINVAL if the incoming CR0 is invalid,
e.g. due to setting bits 63:32, illegal combinations, or to a value that
isn't allowed in VMX (non-)root mode.  The VMX checks in particular are
"fun" as failure to disallow Real Mode for an L2 that is configured with
unrestricted guest disabled, when KVM itself has unrestricted guest
enabled, will result in KVM forcing VM86 mode to virtual Real Mode for
L2, but then fail to unwind the related metadata when synthesizing a
nested VM-Exit back to L1 (which has unrestricted guest enabled).

Opportunistically fix a benign typo in the prototype for is_valid_cr4().

Cc: stable@vger.kernel.org
Reported-by: syzbot+5feef0b9ee9c8e9e5689@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/000000000000f316b705fdf6e2b4@google.comSigned-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20230613203037.1968489-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

arch/x86/include/asm/kvm-x86-ops.h

@@ -37,6 +37,7 @@ KVM_X86_OP(get_segment)
 KVM_X86_OP(get_cpl)
 KVM_X86_OP(set_segment)
 KVM_X86_OP(get_cs_db_l_bits)
+KVM_X86_OP(is_valid_cr0)
 KVM_X86_OP(set_cr0)
 KVM_X86_OP_OPTIONAL(post_set_cr3)
 KVM_X86_OP(is_valid_cr4)

arch/x86/include/asm/kvm_host.h

@@ -1566,9 +1566,10 @@ struct kvm_x86_ops {
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
+	bool (*is_valid_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
-	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
+	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
 	void (*get_idt)(struct kvm_vcpu *vcpu, struct desc_ptr *dt);

arch/x86/kvm/svm/svm.c

@@ -1786,6 +1786,11 @@ static void sev_post_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 	}
 }
 
+static bool svm_is_valid_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+{
+	return true;
+}
+
 void svm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
@@ -4809,6 +4814,7 @@ static struct kvm_x86_ops svm_x86_ops __initdata = {
 	.set_segment = svm_set_segment,
 	.get_cpl = svm_get_cpl,
 	.get_cs_db_l_bits = svm_get_cs_db_l_bits,
+	.is_valid_cr0 = svm_is_valid_cr0,
 	.set_cr0 = svm_set_cr0,
 	.post_set_cr3 = sev_post_set_cr3,
 	.is_valid_cr4 = svm_is_valid_cr4,

arch/x86/kvm/vmx/vmx.c

@@ -3047,6 +3047,15 @@ static void enter_rmode(struct kvm_vcpu *vcpu)
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	struct kvm_vmx *kvm_vmx = to_kvm_vmx(vcpu->kvm);
 
+	/*
+	 * KVM should never use VM86 to virtualize Real Mode when L2 is active,
+	 * as using VM86 is unnecessary if unrestricted guest is enabled, and
+	 * if unrestricted guest is disabled, VM-Enter (from L1) with CR0.PG=0
+	 * should VM-Fail and KVM should reject userspace attempts to stuff
+	 * CR0.PG=0 when L2 is active.
+	 */
+	WARN_ON_ONCE(is_guest_mode(vcpu));
+
 	vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR);
 	vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES);
 	vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS);
@@ -3236,6 +3245,17 @@ void ept_save_pdptrs(struct kvm_vcpu *vcpu)
 #define CR3_EXITING_BITS (CPU_BASED_CR3_LOAD_EXITING | \
 			  CPU_BASED_CR3_STORE_EXITING)
 
+static bool vmx_is_valid_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+{
+	if (is_guest_mode(vcpu))
+		return nested_guest_cr0_valid(vcpu, cr0);
+
+	if (to_vmx(vcpu)->nested.vmxon)
+		return nested_host_cr0_valid(vcpu, cr0);
+
+	return true;
+}
+
 void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
@@ -5375,18 +5395,11 @@ static int handle_set_cr0(struct kvm_vcpu *vcpu, unsigned long val)
 		val = (val & ~vmcs12->cr0_guest_host_mask) |
 			(vmcs12->guest_cr0 & vmcs12->cr0_guest_host_mask);
 
-		if (!nested_guest_cr0_valid(vcpu, val))
-			return 1;
-
 		if (kvm_set_cr0(vcpu, val))
 			return 1;
 		vmcs_writel(CR0_READ_SHADOW, orig_val);
 		return 0;
 	} else {
-		if (to_vmx(vcpu)->nested.vmxon &&
-		    !nested_host_cr0_valid(vcpu, val))
-			return 1;
-
 		return kvm_set_cr0(vcpu, val);
 	}
 }
@@ -8214,6 +8227,7 @@ static struct kvm_x86_ops vmx_x86_ops __initdata = {
 	.set_segment = vmx_set_segment,
 	.get_cpl = vmx_get_cpl,
 	.get_cs_db_l_bits = vmx_get_cs_db_l_bits,
+	.is_valid_cr0 = vmx_is_valid_cr0,
 	.set_cr0 = vmx_set_cr0,
 	.is_valid_cr4 = vmx_is_valid_cr4,
 	.set_cr4 = vmx_set_cr4,

arch/x86/kvm/x86.c

@@ -906,6 +906,22 @@ int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3)
 }
 EXPORT_SYMBOL_GPL(load_pdptrs);
 
+static bool kvm_is_valid_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+{
+#ifdef CONFIG_X86_64
+	if (cr0 & 0xffffffff00000000UL)
+		return false;
+#endif
+
+	if ((cr0 & X86_CR0_NW) && !(cr0 & X86_CR0_CD))
+		return false;
+
+	if ((cr0 & X86_CR0_PG) && !(cr0 & X86_CR0_PE))
+		return false;
+
+	return static_call(kvm_x86_is_valid_cr0)(vcpu, cr0);
+}
+
 void kvm_post_set_cr0(struct kvm_vcpu *vcpu, unsigned long old_cr0, unsigned long cr0)
 {
 	/*
@@ -952,20 +968,13 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	unsigned long old_cr0 = kvm_read_cr0(vcpu);
 
-	cr0 |= X86_CR0_ET;
-
-#ifdef CONFIG_X86_64
-	if (cr0 & 0xffffffff00000000UL)
+	if (!kvm_is_valid_cr0(vcpu, cr0))
 		return 1;
-#endif
-
-	cr0 &= ~CR0_RESERVED_BITS;
 
-	if ((cr0 & X86_CR0_NW) && !(cr0 & X86_CR0_CD))
-		return 1;
+	cr0 |= X86_CR0_ET;
 
-	if ((cr0 & X86_CR0_PG) && !(cr0 & X86_CR0_PE))
-		return 1;
+	/* Write to CR0 reserved bits are ignored, even on Intel. */
+	cr0 &= ~CR0_RESERVED_BITS;
 
 #ifdef CONFIG_X86_64
 	if ((vcpu->arch.efer & EFER_LME) && !is_paging(vcpu) &&
@@ -11468,7 +11477,8 @@ static bool kvm_is_valid_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
 			return false;
 	}
 
-	return kvm_is_valid_cr4(vcpu, sregs->cr4);
+	return kvm_is_valid_cr4(vcpu, sregs->cr4) &&
+	       kvm_is_valid_cr0(vcpu, sregs->cr0);
 }
 
 static int __set_sregs_common(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs,