KVM: x86/speculation: Disable Fill buffer clear within guests

The enumeration of MD_CLEAR in CPUID(EAX=7,ECX=0).EDX{bit 10} is not an
accurate indicator on all CPUs of whether the VERW instruction will
overwrite fill buffers. FB_CLEAR enumeration in
IA32_ARCH_CAPABILITIES{bit 17} covers the case of CPUs that are not
vulnerable to MDS/TAA, indicating that microcode does overwrite fill
buffers.

Guests running in VMM environments may not be aware of all the
capabilities/vulnerabilities of the host CPU. Specifically, a guest may
apply MDS/TAA mitigations when a virtual CPU is enumerated as vulnerable
to MDS/TAA even when the physical CPU is not. On CPUs that enumerate
FB_CLEAR_CTRL the VMM may set FB_CLEAR_DIS to skip overwriting of fill
buffers by the VERW instruction. This is done by setting FB_CLEAR_DIS
during VMENTER and resetting on VMEXIT. For guests that enumerate
FB_CLEAR (explicitly asking for fill buffer clear capability) the VMM
will not use FB_CLEAR_DIS.

Irrespective of guest state, host overwrites CPU buffers before VMENTER
to protect itself from an MMIO capable guest, as part of mitigation for
MMIO Stale Data vulnerabilities.

Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>

1 files changed, 69 insertions, 0 deletions

diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 4fa216acadce..6e8fb36bc49a 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -229,6 +229,9 @@ static const struct {
 #define L1D_CACHE_ORDER 4
 static void *vmx_l1d_flush_pages;
 
+/* Control for disabling CPU Fill buffer clear */
+static bool __read_mostly vmx_fb_clear_ctrl_available;
+
 static int vmx_setup_l1d_flush(enum vmx_l1d_flush_state l1tf)
 {
 	struct page *page;
@@ -360,6 +363,60 @@ static int vmentry_l1d_flush_get(char *s, const struct kernel_param *kp)
 	return sprintf(s, "%s\n", vmentry_l1d_param[l1tf_vmx_mitigation].option);
 }
 
+static void vmx_setup_fb_clear_ctrl(void)
+{
+	u64 msr;
+
+	if (boot_cpu_has(X86_FEATURE_ARCH_CAPABILITIES) &&
+	    !boot_cpu_has_bug(X86_BUG_MDS) &&
+	    !boot_cpu_has_bug(X86_BUG_TAA)) {
+		rdmsrl(MSR_IA32_ARCH_CAPABILITIES, msr);
+		if (msr & ARCH_CAP_FB_CLEAR_CTRL)
+			vmx_fb_clear_ctrl_available = true;
+	}
+}
+
+static __always_inline void vmx_disable_fb_clear(struct vcpu_vmx *vmx)
+{
+	u64 msr;
+
+	if (!vmx->disable_fb_clear)
+		return;
+
+	rdmsrl(MSR_IA32_MCU_OPT_CTRL, msr);
+	msr |= FB_CLEAR_DIS;
+	wrmsrl(MSR_IA32_MCU_OPT_CTRL, msr);
+	/* Cache the MSR value to avoid reading it later */
+	vmx->msr_ia32_mcu_opt_ctrl = msr;
+}
+
+static __always_inline void vmx_enable_fb_clear(struct vcpu_vmx *vmx)
+{
+	if (!vmx->disable_fb_clear)
+		return;
+
+	vmx->msr_ia32_mcu_opt_ctrl &= ~FB_CLEAR_DIS;
+	wrmsrl(MSR_IA32_MCU_OPT_CTRL, vmx->msr_ia32_mcu_opt_ctrl);
+}
+
+static void vmx_update_fb_clear_dis(struct kvm_vcpu *vcpu, struct vcpu_vmx *vmx)
+{
+	vmx->disable_fb_clear = vmx_fb_clear_ctrl_available;
+
+	/*
+	 * If guest will not execute VERW, there is no need to set FB_CLEAR_DIS
+	 * at VMEntry. Skip the MSR read/write when a guest has no use case to
+	 * execute VERW.
+	 */
+	if ((vcpu->arch.arch_capabilities & ARCH_CAP_FB_CLEAR) ||
+	   ((vcpu->arch.arch_capabilities & ARCH_CAP_MDS_NO) &&
+	    (vcpu->arch.arch_capabilities & ARCH_CAP_TAA_NO) &&
+	    (vcpu->arch.arch_capabilities & ARCH_CAP_PSDP_NO) &&
+	    (vcpu->arch.arch_capabilities & ARCH_CAP_FBSDP_NO) &&
+	    (vcpu->arch.arch_capabilities & ARCH_CAP_SBDR_SSDP_NO)))
+		vmx->disable_fb_clear = false;
+}
+
 static const struct kernel_param_ops vmentry_l1d_flush_ops = {
 	.set = vmentry_l1d_flush_set,
 	.get = vmentry_l1d_flush_get,
@@ -2252,6 +2309,10 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			ret = kvm_set_msr_common(vcpu, msr_info);
 	}
 
+	/* FB_CLEAR may have changed, also update the FB_CLEAR_DIS behavior */
+	if (msr_index == MSR_IA32_ARCH_CAPABILITIES)
+		vmx_update_fb_clear_dis(vcpu, vmx);
+
 	return ret;
 }
 
@@ -4553,6 +4614,8 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
 	kvm_make_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu);
 
 	vpid_sync_context(vmx->vpid);
+
+	vmx_update_fb_clear_dis(vcpu, vmx);
 }
 
 static void vmx_enable_irq_window(struct kvm_vcpu *vcpu)
@@ -6777,6 +6840,8 @@ static noinstr void vmx_vcpu_enter_exit(struct kvm_vcpu *vcpu,
 		 kvm_arch_has_assigned_device(vcpu->kvm))
 		mds_clear_cpu_buffers();
 
+	vmx_disable_fb_clear(vmx);
+
 	if (vcpu->arch.cr2 != native_read_cr2())
 		native_write_cr2(vcpu->arch.cr2);
 
@@ -6785,6 +6850,8 @@ static noinstr void vmx_vcpu_enter_exit(struct kvm_vcpu *vcpu,
 
 	vcpu->arch.cr2 = native_read_cr2();
 
+	vmx_enable_fb_clear(vmx);
+
 	guest_state_exit_irqoff();
 }
 
@@ -8185,6 +8252,8 @@ static int __init vmx_init(void)
 		return r;
 	}
 
+	vmx_setup_fb_clear_ctrl();
+
 	for_each_possible_cpu(cpu) {
 		INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu));