When lazy save/restore of MSR_KERNEL_GS_BASE was introduced[1], the
MSR was intercepted in all modes and was only restored for the host
when the guest is in 64-bit mode. So at the time, going through the
full host restore prior to accessing MSR_KERNEL_GS_BASE was necessary
to load host state and was not a significant waste of cycles.
Later, MSR_KERNEL_GS_BASE interception was disabled for a 64-bit
guest[2], and then unconditionally saved/restored for the host[3].
As a result, loading full host state is overkill for accesses to
MSR_KERNEL_GS_BASE, and completely unnecessary when the guest is
not in 64-bit mode.
Add a dedicated utility to read/write the guest's MSR_KERNEL_GS_BASE
(outside of the save/restore flow) to minimize the overhead incurred
when accessing the MSR. When setting EFER, only decache the MSR if
the new EFER will disable long mode.
Removing out-of-band usage of vmx_load_host_state() also eliminates,
or at least reduces, potential corner cases in its usage, which in
turn will (hopefuly) make it easier to reason about future changes
to the save/restore flow, e.g. optimization of saving host state.
[1] commit 44ea2b1758d8 ("KVM: VMX: Move MSR_KERNEL_GS_BASE out of the vmx
autoload msr area")
[2] commit 5897297bc228 ("KVM: VMX: Don't intercept MSR_KERNEL_GS_BASE")
[3] commit c8770e7ba63b ("KVM: VMX: Fix host userspace gsbase corruption")
Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
---
arch/x86/kvm/vmx.c | 46 +++++++++++++++++++++++++++++++++++-----------
1 file changed, 35 insertions(+), 11 deletions(-)
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index a56bd335e9a2..80c2537aa1cc 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2678,13 +2678,31 @@ static void __vmx_load_host_state(struct vcpu_vmx *vmx)
load_fixmap_gdt(raw_smp_processor_id());
}
-static void vmx_load_host_state(struct vcpu_vmx *vmx)
+#ifdef CONFIG_X86_64
+static u64 vmx_read_guest_kernel_gs_base(struct vcpu_vmx *vmx)
{
- preempt_disable();
- __vmx_load_host_state(vmx);
- preempt_enable();
+ if (is_long_mode(&vmx->vcpu)) {
+ preempt_disable();
+ if (vmx->loaded_cpu_state)
+ rdmsrl(MSR_KERNEL_GS_BASE,
+ vmx->msr_guest_kernel_gs_base);
+ preempt_enable();
+ }
+ return vmx->msr_guest_kernel_gs_base;
}
+static void vmx_write_guest_kernel_gs_base(struct vcpu_vmx *vmx, u64 data)
+{
+ if (is_long_mode(&vmx->vcpu)) {
+ preempt_disable();
+ if (vmx->loaded_cpu_state)
+ wrmsrl(MSR_KERNEL_GS_BASE, data);
+ preempt_enable();
+ }
+ vmx->msr_guest_kernel_gs_base = data;
+}
+#endif
+
static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)
{
struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu);
@@ -3757,8 +3775,7 @@ static int vmx_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
msr_info->data = vmcs_readl(GUEST_GS_BASE);
break;
case MSR_KERNEL_GS_BASE:
- vmx_load_host_state(vmx);
- msr_info->data = vmx->msr_guest_kernel_gs_base;
+ msr_info->data = vmx_read_guest_kernel_gs_base(vmx);
break;
#endif
case MSR_EFER:
@@ -3858,8 +3875,7 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
vmcs_writel(GUEST_GS_BASE, data);
break;
case MSR_KERNEL_GS_BASE:
- vmx_load_host_state(vmx);
- vmx->msr_guest_kernel_gs_base = data;
+ vmx_write_guest_kernel_gs_base(vmx, data);
break;
#endif
case MSR_IA32_SYSENTER_CS:
@@ -4732,10 +4748,18 @@ static void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer)
return;
/*
- * Force kernel_gs_base reloading before EFER changes, as control
- * of this msr depends on is_long_mode().
+ * MSR_KERNEL_GS_BASE is not intercepted when the guest is in
+ * 64-bit mode as a 64-bit kernel may frequently access the
+ * MSR. This means we need to manually save/restore the MSR
+ * when switching between guest and host state, but only if
+ * the guest is in 64-bit mode. Sync our cached value if the
+ * guest is transitioning to 32-bit mode and the CPU contains
+ * guest state, i.e. the cache is stale.
*/
- vmx_load_host_state(to_vmx(vcpu));
+#ifdef CONFIG_X86_64
+ if (!(efer & EFER_LMA))
+ (void)vmx_read_guest_kernel_gs_base(vmx);
+#endif
vcpu->arch.efer = efer;
if (efer & EFER_LMA) {
vm_entry_controls_setbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE);
--
2.18.0
