kvm: vmx: Support INVPCID in shadow paging mode

Implement support for INVPCID in shadow paging mode as well.

Signed-off-by: Junaid Shahid <junaids@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

3 files changed, 112 insertions, 3 deletions

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index c2b4df8a03cd..1c253f15f9cd 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1317,6 +1317,7 @@ int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code,
 		       void *insn, int insn_len);
 void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva);
+void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid);
 void kvm_mmu_new_cr3(struct kvm_vcpu *vcpu, gpa_t new_cr3);
 
 void kvm_enable_tdp(void);
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index d3a04cf6514b..2af15db12ccd 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -5187,6 +5187,24 @@ void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
 }
 EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);
 
+void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid)
+{
+	struct kvm_mmu *mmu = &vcpu->arch.mmu;
+
+	if (pcid == kvm_get_active_pcid(vcpu)) {
+		mmu->invlpg(vcpu, gva);
+		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
+	}
+
+	++vcpu->stat.invlpg;
+
+	/*
+	 * Mappings not reachable via the current cr3 will be synced when
+	 * switching to that cr3, so nothing needs to be done here for them.
+	 */
+}
+EXPORT_SYMBOL_GPL(kvm_mmu_invpcid_gva);
+
 void kvm_enable_tdp(void)
 {
 	tdp_enabled = true;
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 903c130bb811..80be024cb979 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -3065,7 +3065,7 @@ static bool vmx_rdtscp_supported(void)
 
 static bool vmx_invpcid_supported(void)
 {
-	return cpu_has_vmx_invpcid() && enable_ept;
+	return cpu_has_vmx_invpcid();
 }
 
 /*
@@ -6101,8 +6101,6 @@ static void vmx_compute_secondary_exec_control(struct vcpu_vmx *vmx)
 	if (!enable_ept) {
 		exec_control &= ~SECONDARY_EXEC_ENABLE_EPT;
 		enable_unrestricted_guest = 0;
-		/* Enable INVPCID for non-ept guests may cause performance regression. */
-		exec_control &= ~SECONDARY_EXEC_ENABLE_INVPCID;
 	}
 	if (!enable_unrestricted_guest)
 		exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST;
@@ -8756,6 +8754,97 @@ static int handle_invvpid(struct kvm_vcpu *vcpu)
 	return kvm_skip_emulated_instruction(vcpu);
 }
 
+static int handle_invpcid(struct kvm_vcpu *vcpu)
+{
+	u32 vmx_instruction_info;
+	unsigned long type;
+	bool pcid_enabled;
+	gva_t gva;
+	struct x86_exception e;
+	struct {
+		u64 pcid;
+		u64 gla;
+	} operand;
+
+	if (!guest_cpuid_has(vcpu, X86_FEATURE_INVPCID)) {
+		kvm_queue_exception(vcpu, UD_VECTOR);
+		return 1;
+	}
+
+	vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO);
+	type = kvm_register_readl(vcpu, (vmx_instruction_info >> 28) & 0xf);
+
+	if (type > 3) {
+		kvm_inject_gp(vcpu, 0);
+		return 1;
+	}
+
+	/* According to the Intel instruction reference, the memory operand
+	 * is read even if it isn't needed (e.g., for type==all)
+	 */
+	if (get_vmx_mem_address(vcpu, vmcs_readl(EXIT_QUALIFICATION),
+				vmx_instruction_info, false, &gva))
+		return 1;
+
+	if (kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e)) {
+		kvm_inject_page_fault(vcpu, &e);
+		return 1;
+	}
+
+	if (operand.pcid >> 12 != 0) {
+		kvm_inject_gp(vcpu, 0);
+		return 1;
+	}
+
+	pcid_enabled = kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE);
+
+	switch (type) {
+	case INVPCID_TYPE_INDIV_ADDR:
+		if ((!pcid_enabled && (operand.pcid != 0)) ||
+		    is_noncanonical_address(operand.gla, vcpu)) {
+			kvm_inject_gp(vcpu, 0);
+			return 1;
+		}
+		kvm_mmu_invpcid_gva(vcpu, operand.gla, operand.pcid);
+		return kvm_skip_emulated_instruction(vcpu);
+
+	case INVPCID_TYPE_SINGLE_CTXT:
+		if (!pcid_enabled && (operand.pcid != 0)) {
+			kvm_inject_gp(vcpu, 0);
+			return 1;
+		}
+
+		if (kvm_get_active_pcid(vcpu) == operand.pcid) {
+			kvm_mmu_sync_roots(vcpu);
+			kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
+		}
+
+		/*
+		 * If the current cr3 does not use the given PCID, then nothing
+		 * needs to be done here because a resync will happen anyway
+		 * before switching to any other CR3.
+		 */
+
+		return kvm_skip_emulated_instruction(vcpu);
+
+	case INVPCID_TYPE_ALL_NON_GLOBAL:
+		/*
+		 * Currently, KVM doesn't mark global entries in the shadow
+		 * page tables, so a non-global flush just degenerates to a
+		 * global flush. If needed, we could optimize this later by
+		 * keeping track of global entries in shadow page tables.
+		 */
+
+		/* fall-through */
+	case INVPCID_TYPE_ALL_INCL_GLOBAL:
+		kvm_mmu_unload(vcpu);
+		return kvm_skip_emulated_instruction(vcpu);
+
+	default:
+		BUG(); /* We have already checked above that type <= 3 */
+	}
+}
+
 static int handle_pml_full(struct kvm_vcpu *vcpu)
 {
 	unsigned long exit_qualification;
@@ -8959,6 +9048,7 @@ static int (*const kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = {
 	[EXIT_REASON_XSAVES]                  = handle_xsaves,
 	[EXIT_REASON_XRSTORS]                 = handle_xrstors,
 	[EXIT_REASON_PML_FULL]		      = handle_pml_full,
+	[EXIT_REASON_INVPCID]                 = handle_invpcid,
 	[EXIT_REASON_VMFUNC]                  = handle_vmfunc,
 	[EXIT_REASON_PREEMPTION_TIMER]	      = handle_preemption_timer,
 };