KVM: nVMX: Cache host_rsp on a per-VMCS basis

Currently, host_rsp is cached on a per-vCPU basis, i.e. it's stored in
struct vcpu_vmx.  In non-nested usage the caching is for all intents
and purposes 100% effective, e.g. only the first VMLAUNCH needs to
synchronize VMCS.HOST_RSP since the call stack to vmx_vcpu_run() is
identical each and every time.  But when running a nested guest, KVM
must invalidate the cache when switching the current VMCS as it can't
guarantee the new VMCS has the same HOST_RSP as the previous VMCS.  In
other words, the cache loses almost all of its efficacy when running a
nested VM.

Move host_rsp to struct vmcs_host_state, which is per-VMCS, so that it
is cached on a per-VMCS basis and restores its 100% hit rate when
nested VMs are in play.

Note that the host_rsp cache for vmcs02 essentially "breaks" when
nested early checks are enabled as nested_vmx_check_vmentry_hw() will
see a different RSP at the time of its VM-Enter.  While it's possible
to avoid even that VMCS.HOST_RSP synchronization, e.g. by employing a
dedicated VM-Exit stack, there is little motivation for doing so as
the overhead of two VMWRITEs (~55 cycles) is dwarfed by the overhead
of the extra VMX transition (600+ cycles) and is a proverbial drop in
the ocean relative to the total cost of a nested transtion (10s of
thousands of cycles).

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 6 insertions, 18 deletions

diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index 23a2c1b91389..0e67649e39ce 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -1980,17 +1980,6 @@ static void prepare_vmcs02_early(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
 		prepare_vmcs02_early_full(vmx, vmcs12);
 
 	/*
-	 * HOST_RSP is normally set correctly in vmx_vcpu_run() just before
-	 * entry, but only if the current (host) sp changed from the value
-	 * we wrote last (vmx->host_rsp).  This cache is no longer relevant
-	 * if we switch vmcs, and rather than hold a separate cache per vmcs,
-	 * here we just force the write to happen on entry.  host_rsp will
-	 * also be written unconditionally by nested_vmx_check_vmentry_hw()
-	 * if we are doing early consistency checks via hardware.
-	 */
-	vmx->host_rsp = 0;
-
-	/*
 	 * PIN CONTROLS
 	 */
 	exec_control = vmcs12->pin_based_vm_exec_control;
@@ -2754,8 +2743,11 @@ static int nested_vmx_check_vmentry_hw(struct kvm_vcpu *vcpu)
 
 	asm(
 		"sub $%c[wordsize], %%" _ASM_SP "\n\t" /* temporarily adjust RSP for CALL */
+		"cmp %%" _ASM_SP ", %c[host_state_rsp](%[loaded_vmcs]) \n\t"
+		"je 1f \n\t"
 		__ex("vmwrite %%" _ASM_SP ", %[HOST_RSP]") "\n\t"
-		"mov %%" _ASM_SP ", %c[host_rsp](%% " _ASM_CX ")\n\t"
+		"mov %%" _ASM_SP ", %c[host_state_rsp](%[loaded_vmcs]) \n\t"
+		"1: \n\t"
 		"add $%c[wordsize], %%" _ASM_SP "\n\t" /* un-adjust RSP */
 
 		/* Check if vmlaunch or vmresume is needed */
@@ -2771,11 +2763,10 @@ static int nested_vmx_check_vmentry_hw(struct kvm_vcpu *vcpu)
 
 		CC_SET(be)
 	      : ASM_CALL_CONSTRAINT, CC_OUT(be) (vm_fail)
-	      : "c"(vmx),
-		[HOST_RSP]"r"((unsigned long)HOST_RSP),
+	      :	[HOST_RSP]"r"((unsigned long)HOST_RSP),
 		[loaded_vmcs]"r"(vmx->loaded_vmcs),
 		[launched]"i"(offsetof(struct loaded_vmcs, launched)),
-		[host_rsp]"i"(offsetof(struct vcpu_vmx, host_rsp)),
+		[host_state_rsp]"i"(offsetof(struct loaded_vmcs, host_state.rsp)),
 		[wordsize]"i"(sizeof(ulong))
 	      : "cc", "memory"
 	);
@@ -3912,9 +3903,6 @@ void nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 exit_reason,
 		vmx_flush_tlb(vcpu, true);
 	}
 
-	/* This is needed for same reason as it was needed in prepare_vmcs02 */
-	vmx->host_rsp = 0;
-
 	/* Unpin physical memory we referred to in vmcs02 */
 	if (vmx->nested.apic_access_page) {
 		kvm_release_page_dirty(vmx->nested.apic_access_page);