KVM: vmx: rename __vmx_load_host_state() and vmx_save_host_state()
authorSean Christopherson <sean.j.christopherson@intel.com>
Mon, 23 Jul 2018 19:32:44 +0000 (12:32 -0700)
committerPaolo Bonzini <pbonzini@redhat.com>
Mon, 6 Aug 2018 15:59:12 +0000 (17:59 +0200)
Now that the vmx_load_host_state() wrapper is gone, i.e. the only
time we call the core functions is when we're actually about to
switch between guest/host, rename the functions that handle lazy
state switching to vmx_prepare_switch_to_{guest,host}_state() to
better document the full extent of their functionality.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/vmx.c

index 72c3925..ab27f97 100644 (file)
@@ -2661,7 +2661,7 @@ static unsigned long segment_base(u16 selector)
 }
 #endif
 
-static void vmx_save_host_state(struct kvm_vcpu *vcpu)
+static void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
 {
        struct vcpu_vmx *vmx = to_vmx(vcpu);
 #ifdef CONFIG_X86_64
@@ -2735,7 +2735,7 @@ static void vmx_save_host_state(struct kvm_vcpu *vcpu)
                                   vmx->guest_msrs[i].mask);
 }
 
-static void __vmx_load_host_state(struct vcpu_vmx *vmx)
+static void vmx_prepare_switch_to_host(struct vcpu_vmx *vmx)
 {
        if (!vmx->loaded_cpu_state)
                return;
@@ -2938,7 +2938,7 @@ static void vmx_vcpu_put(struct kvm_vcpu *vcpu)
 {
        vmx_vcpu_pi_put(vcpu);
 
-       __vmx_load_host_state(to_vmx(vcpu));
+       vmx_prepare_switch_to_host(to_vmx(vcpu));
 }
 
 static bool emulation_required(struct kvm_vcpu *vcpu)
@@ -6099,8 +6099,8 @@ static void vmx_set_constant_host_state(struct vcpu_vmx *vmx)
 #ifdef CONFIG_X86_64
        /*
         * Load null selectors, so we can avoid reloading them in
-        * __vmx_load_host_state(), in case userspace uses the null selectors
-        * too (the expected case).
+        * vmx_prepare_switch_to_host(), in case userspace uses
+        * the null selectors too (the expected case).
         */
        vmcs_write16(HOST_DS_SELECTOR, 0);
        vmcs_write16(HOST_ES_SELECTOR, 0);
@@ -10565,9 +10565,9 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
         * The sysexit path does not restore ds/es, so we must set them to
         * a reasonable value ourselves.
         *
-        * We can't defer this to vmx_load_host_state() since that function
-        * may be executed in interrupt context, which saves and restore segments
-        * around it, nullifying its effect.
+        * We can't defer this to vmx_prepare_switch_to_host() since that
+        * function may be executed in interrupt context, which saves and
+        * restore segments around it, nullifying its effect.
         */
        loadsegment(ds, __USER_DS);
        loadsegment(es, __USER_DS);
@@ -11668,7 +11668,8 @@ static void prepare_vmcs02_full(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
         * Set host-state according to L0's settings (vmcs12 is irrelevant here)
         * Some constant fields are set here by vmx_set_constant_host_state().
         * Other fields are different per CPU, and will be set later when
-        * vmx_vcpu_load() is called, and when vmx_save_host_state() is called.
+        * vmx_vcpu_load() is called, and when vmx_prepare_switch_to_guest()
+        * is called.
         */
        vmx_set_constant_host_state(vmx);
 
@@ -13707,7 +13708,7 @@ static struct kvm_x86_ops vmx_x86_ops __ro_after_init = {
        .vcpu_free = vmx_free_vcpu,
        .vcpu_reset = vmx_vcpu_reset,
 
-       .prepare_guest_switch = vmx_save_host_state,
+       .prepare_guest_switch = vmx_prepare_switch_to_guest,
        .vcpu_load = vmx_vcpu_load,
        .vcpu_put = vmx_vcpu_put,