KVM/VMX: Add hv tlb range flush support
authorLan Tianyu <Tianyu.Lan@microsoft.com>
Thu, 6 Dec 2018 13:21:07 +0000 (21:21 +0800)
committerPaolo Bonzini <pbonzini@redhat.com>
Fri, 21 Dec 2018 10:28:39 +0000 (11:28 +0100)
This patch is to register tlb_remote_flush_with_range callback with
hv tlb range flush interface.

Signed-off-by: Lan Tianyu <Tianyu.Lan@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/vmx/vmx.c

index 3ef444de0c5ead31115d56a0cd12b5a8efc67753..d1c818a877dfe209d4a8158d88d322d9b2f3ada0 100644 (file)
@@ -421,7 +421,34 @@ static void check_ept_pointer_match(struct kvm *kvm)
        to_kvm_vmx(kvm)->ept_pointers_match = EPT_POINTERS_MATCH;
 }
 
-static int vmx_hv_remote_flush_tlb(struct kvm *kvm)
+int kvm_fill_hv_flush_list_func(struct hv_guest_mapping_flush_list *flush,
+               void *data)
+{
+       struct kvm_tlb_range *range = data;
+
+       return hyperv_fill_flush_guest_mapping_list(flush, range->start_gfn,
+                       range->pages);
+}
+
+static inline int __hv_remote_flush_tlb_with_range(struct kvm *kvm,
+               struct kvm_vcpu *vcpu, struct kvm_tlb_range *range)
+{
+       u64 ept_pointer = to_vmx(vcpu)->ept_pointer;
+
+       /*
+        * FLUSH_GUEST_PHYSICAL_ADDRESS_SPACE hypercall needs address
+        * of the base of EPT PML4 table, strip off EPT configuration
+        * information.
+        */
+       if (range)
+               return hyperv_flush_guest_mapping_range(ept_pointer & PAGE_MASK,
+                               kvm_fill_hv_flush_list_func, (void *)range);
+       else
+               return hyperv_flush_guest_mapping(ept_pointer & PAGE_MASK);
+}
+
+static int hv_remote_flush_tlb_with_range(struct kvm *kvm,
+               struct kvm_tlb_range *range)
 {
        struct kvm_vcpu *vcpu;
        int ret = -ENOTSUPP, i;
@@ -431,29 +458,26 @@ static int vmx_hv_remote_flush_tlb(struct kvm *kvm)
        if (to_kvm_vmx(kvm)->ept_pointers_match == EPT_POINTERS_CHECK)
                check_ept_pointer_match(kvm);
 
-       /*
-        * FLUSH_GUEST_PHYSICAL_ADDRESS_SPACE hypercall needs the address of the
-        * base of EPT PML4 table, strip off EPT configuration information.
-        * If ept_pointer is invalid pointer, bypass the flush request.
-        */
        if (to_kvm_vmx(kvm)->ept_pointers_match != EPT_POINTERS_MATCH) {
                kvm_for_each_vcpu(i, vcpu, kvm) {
-                       u64 ept_pointer = to_vmx(vcpu)->ept_pointer;
-
-                       if (!VALID_PAGE(ept_pointer))
-                               continue;
-
-                       ret |= hyperv_flush_guest_mapping(
-                               ept_pointer & PAGE_MASK);
+                       /* If ept_pointer is invalid pointer, bypass flush request. */
+                       if (VALID_PAGE(to_vmx(vcpu)->ept_pointer))
+                               ret |= __hv_remote_flush_tlb_with_range(
+                                       kvm, vcpu, range);
                }
        } else {
-               ret = hyperv_flush_guest_mapping(
-                               to_vmx(kvm_get_vcpu(kvm, 0))->ept_pointer & PAGE_MASK);
+               ret = __hv_remote_flush_tlb_with_range(kvm,
+                               kvm_get_vcpu(kvm, 0), range);
        }
 
        spin_unlock(&to_kvm_vmx(kvm)->ept_pointer_lock);
        return ret;
 }
+static int hv_remote_flush_tlb(struct kvm *kvm)
+{
+       return hv_remote_flush_tlb_with_range(kvm, NULL);
+}
+
 #endif /* IS_ENABLED(CONFIG_HYPERV) */
 
 /*
@@ -7554,8 +7578,11 @@ static __init int hardware_setup(void)
 
 #if IS_ENABLED(CONFIG_HYPERV)
        if (ms_hyperv.nested_features & HV_X64_NESTED_GUEST_MAPPING_FLUSH
-           && enable_ept)
-               kvm_x86_ops->tlb_remote_flush = vmx_hv_remote_flush_tlb;
+           && enable_ept) {
+               kvm_x86_ops->tlb_remote_flush = hv_remote_flush_tlb;
+               kvm_x86_ops->tlb_remote_flush_with_range =
+                               hv_remote_flush_tlb_with_range;
+       }
 #endif
 
        if (!cpu_has_vmx_ple()) {