Re: [PATCH v2] KVM: nVMX: Fix read/write to MSR_IA32_FEATURE_CONTROL

[Gleb Natapov <gleb@redhat.com>]

On Sun, Jul 07, 2013 at 05:17:00PM +0800, Arthur Chunqi Li wrote:
> On Sun, Jul 7, 2013 at 4:50 PM, Gleb Natapov <gleb@redhat.com> wrote:
> > On Sun, Jul 07, 2013 at 04:47:05PM +0800, Arthur Chunqi Li wrote:
> >> On Sun, Jul 7, 2013 at 3:20 PM, Gleb Natapov <gleb@redhat.com> wrote:
> >> > On Sun, Jul 07, 2013 at 03:16:08PM +0800, Arthur Chunqi Li wrote:
> >> >> On Sun, Jul 7, 2013 at 2:59 PM, Gleb Natapov <gleb@redhat.com> wrote:
> >> >> > On Fri, Jul 05, 2013 at 11:13:17PM +0800, Arthur Chunqi Li wrote:
> >> >> >> Fix read/write to IA32_FEATURE_CONTROL MSR in nested environment.
> >> >> >>
> >> >> >> This patch simulate this MSR in nested_vmx and the default value is
> >> >> >> 0x0. BIOS should set it to 0x5 before VMXON. After setting the lock
> >> >> >> bit, write to it will cause #GP(0).
> >> >> >>
> >> >> >> Signed-off-by: Arthur Chunqi Li <yzt356@gmail.com>
> >> >> > The patch is heavily based on Nadav's one. You need to preserve
> >> >> > authorship. Add From: and Nadav's Sign-off-by.
> >> >> >
> >> >> >> ---
> >> >> >>  arch/x86/kvm/vmx.c |   25 +++++++++++++++++++++----
> >> >> >>  arch/x86/kvm/x86.c |    3 ++-
> >> >> >>  2 files changed, 23 insertions(+), 5 deletions(-)
> >> >> >>
> >> >> >> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> >> >> >> index 260a919..5e3d44e 100644
> >> >> >> --- a/arch/x86/kvm/vmx.c
> >> >> >> +++ b/arch/x86/kvm/vmx.c
> >> >> >> @@ -373,6 +373,7 @@ struct nested_vmx {
> >> >> >>        * we must keep them pinned while L2 runs.
> >> >> >>        */
> >> >> >>       struct page *apic_access_page;
> >> >> >> +     u64 msr_ia32_feature_control;
> >> >> >>  };
> >> >> >>
> >> >> >>  #define POSTED_INTR_ON  0
> >> >> >> @@ -2277,8 +2278,11 @@ static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
> >> >> >>
> >> >> >>       switch (msr_index) {
> >> >> >>       case MSR_IA32_FEATURE_CONTROL:
> >> >> >> -             *pdata = 0;
> >> >> >> -             break;
> >> >> >> +             if (nested_vmx_allowed(vcpu)){
> >> >> >> +                     *pdata = to_vmx(vcpu)->nested.msr_ia32_feature_control;
> >> >> >> +                     break;
> >> >> >> +             }
> >> >> >> +             return 0;
> >> >> >>       case MSR_IA32_VMX_BASIC:
> >> >> >>               /*
> >> >> >>                * This MSR reports some information about VMX support. We
> >> >> >> @@ -2356,9 +2360,13 @@ static int vmx_set_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
> >> >> >>       if (!nested_vmx_allowed(vcpu))
> >> >> >>               return 0;
> >> >> >>
> >> >> >> -     if (msr_index == MSR_IA32_FEATURE_CONTROL)
> >> >> >> -             /* TODO: the right thing. */
> >> >> >> +     if (msr_index == MSR_IA32_FEATURE_CONTROL){
> >> >> >> +             if (to_vmx(vcpu)->nested.msr_ia32_feature_control
> >> >> >> +                             & FEATURE_CONTROL_LOCKED)
> >> >> >> +                     return 0;
> >> >> >> +             to_vmx(vcpu)->nested.msr_ia32_feature_control = data;
> >> >> > You need to allow setting it from userspace, otherwise reset will not
> >> >> > clear it.
> >> >> As to my memory, if the lock bit of IA32_FEATURE_CONTROL is set, this
> >> >> MSR cannot be set or reset. So what do you mean setting it from
> >> >> userspace?
> >> > It can be reset by power cycling CPU obviously, so when userspace emulates
> >> > power cycle reset it has to be able to zero the MSR. BTW you need to write
> >> > corespondent QEMU patch to handle this MSR during reset and migration.
> >> What is the current call trace when QEMU do system_reset? Will it
> >> trigger vmx_vcpu_reset() in vmx.c?
> > It will not. QEMU inits vcpu state in userspace and write it into the
> > kernel with set_regs/set_srges ioctls.
> So how will this achieve isolation and protection? I mean if QEMU can
> reset this MSR when reboot, can user in OS also reset it? Actually,
> after setting to lock bit, it cannot be set until next reboot.
There is a way to tell if msr write is done by QEMU or a guest OS.

--
			Gleb.