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

[Gleb Natapov <gleb@redhat.com>]

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.

> >
> >> >
> >> >>               return 1;
> >> >> +     }
> >> >>       /*
> >> >>        * No need to treat VMX capability MSRs specially: If we don't handle
> >> >>        * them, handle_wrmsr will #GP(0), which is correct (they are readonly)
> >> >> @@ -5595,6 +5603,15 @@ static int handle_vmon(struct kvm_vcpu *vcpu)
> >> >>               skip_emulated_instruction(vcpu);
> >> >>               return 1;
> >> >>       }
> >> >> +
> >> >> +#define VMXON_NEEDED_FEATURES \
> >> > Avi asked to use const u64 here.
> >> >
> >> >> +       (FEATURE_CONTROL_LOCKED | FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX)
> >> >> +     if ((vmx->nested.msr_ia32_feature_control & VMXON_NEEDED_FEATURES)
> >> >> +                     != VMXON_NEEDED_FEATURES) {
> >> >> +             kvm_inject_gp(vcpu, 0);
> >> >> +             return 1;
> >> >> +     }
> >> >> +
> >> >>       if (enable_shadow_vmcs) {
> >> >>               shadow_vmcs = alloc_vmcs();
> >> >>               if (!shadow_vmcs)
> >> >> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> >> >> index e8ba99c..2d4eb8e 100644
> >> >> --- a/arch/x86/kvm/x86.c
> >> >> +++ b/arch/x86/kvm/x86.c
> >> >> @@ -850,7 +850,8 @@ static u32 msrs_to_save[] = {
> >> >>  #ifdef CONFIG_X86_64
> >> >>       MSR_CSTAR, MSR_KERNEL_GS_BASE, MSR_SYSCALL_MASK, MSR_LSTAR,
> >> >>  #endif
> >> >> -     MSR_IA32_TSC, MSR_IA32_CR_PAT, MSR_VM_HSAVE_PA
> >> >> +     MSR_IA32_TSC, MSR_IA32_CR_PAT, MSR_VM_HSAVE_PA,
> >> >> +     MSR_IA32_FEATURE_CONTROL
> >> >>  };
> >> >>
> >> >>  static unsigned num_msrs_to_save;
> >> >> --
> >> >> 1.7.9.5
> >> >
> >> > --
> >> >                         Gleb.
> >
> > --
> >                         Gleb.
> 
> 
> 
> --
> Arthur Chunqi Li
> Department of Computer Science
> School of EECS
> Peking University
> Beijing, China

--
			Gleb.