Marcelo Tosatti wrote: > On Tue, May 11, 2010 at 06:29:48PM +0800, Xu, Dongxiao wrote: >> From: Dongxiao Xu <[email protected]> >> >> SDM suggests VMXON should be called before VMPTRLD, and VMXOFF >> should be called after doing VMCLEAR. >> >> Therefore in vmm coexistence case, we should firstly call VMXON >> before any VMCS operation, and then call VMXOFF after the >> operation is done. >> >> Signed-off-by: Dongxiao Xu <[email protected]> >> --- >> arch/x86/kvm/vmx.c | 38 +++++++++++++++++++++++++++++++------- >> 1 files changed, 31 insertions(+), 7 deletions(-) >> >> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c >> index c536b9d..dbd47a7 100644 >> --- a/arch/x86/kvm/vmx.c >> +++ b/arch/x86/kvm/vmx.c >> @@ -168,6 +168,8 @@ static inline struct vcpu_vmx *to_vmx(struct >> kvm_vcpu *vcpu) >> >> static int init_rmode(struct kvm *kvm); >> static u64 construct_eptp(unsigned long root_hpa); >> +static void kvm_cpu_vmxon(u64 addr); >> +static void kvm_cpu_vmxoff(void); >> >> static DEFINE_PER_CPU(struct vmcs *, vmxarea); >> static DEFINE_PER_CPU(struct vmcs *, current_vmcs); >> @@ -786,8 +788,11 @@ static void vmx_vcpu_load(struct kvm_vcpu >> *vcpu, int cpu) { struct vcpu_vmx *vmx = to_vmx(vcpu); >> u64 tsc_this, delta, new_offset; >> + u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); >> >> - if (vmm_exclusive && vcpu->cpu != cpu) >> + if (!vmm_exclusive) >> + kvm_cpu_vmxon(phys_addr); >> + else if (vcpu->cpu != cpu) >> vcpu_clear(vmx); >> >> if (per_cpu(current_vmcs, cpu) != vmx->vmcs) { >> @@ -833,8 +838,10 @@ static void vmx_vcpu_load(struct kvm_vcpu >> *vcpu, int cpu) static void vmx_vcpu_put(struct kvm_vcpu *vcpu) >> { >> __vmx_load_host_state(to_vmx(vcpu)); >> - if (!vmm_exclusive) >> + if (!vmm_exclusive) { >> __vcpu_clear(to_vmx(vcpu)); >> + kvm_cpu_vmxoff(); >> + } >> } >> >> static void vmx_fpu_activate(struct kvm_vcpu *vcpu) >> @@ -1257,9 +1264,11 @@ static int hardware_enable(void *garbage) >> FEATURE_CONTROL_LOCKED | >> FEATURE_CONTROL_VMXON_ENABLED); >> write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug >> safe */ - kvm_cpu_vmxon(phys_addr); >> >> - ept_sync_global(); >> + if (vmm_exclusive) { >> + kvm_cpu_vmxon(phys_addr); >> + ept_sync_global(); >> + } >> >> return 0; > > The documentation recommends usage of INVEPT all-context after > execution of VMXON and prior to execution of VMXOFF. Is it not > necessary?
After adding the patch, when vCPU is scheduled in a CPU, it will call tlb_flush() to invalidate the EPT and VPID cache/tlb for the vCPU. Therefore the correctness for KVM is guaranteed. Thanks, Dongxiao-- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
