Re: [PATCH v2] kvm: nVMX: Set nested_run_pending in vmx_set_nested_state after checks complete

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, May 8, 2019 at 2:13 PM Aaron Lewis <aaronlewis@xxxxxxxxxx> wrote:
>
> From: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> Date: Wed, May 8, 2019 at 11:13 AM
> To: Aaron Lewis
> Cc: Paolo Bonzini, <linux-kernel@xxxxxxxxxxxxxxx>,
> <kvm@xxxxxxxxxxxxxxx>, Peter Shier
>
> > On Wed, May 08, 2019 at 10:53:12AM -0700, Aaron Lewis wrote:
> > > nested_run_pending is also checked in
> > > nested_vmx_check_vmentry_postreqs
> > > (https://elixir.bootlin.com/linux/v5.1/source/arch/x86/kvm/vmx/nested.c#L2709)
> > > so I think the setting needs to be moved to just prior to that call
> > > with Paolo's rollback along with another for if the prereqs and
> > > postreqs fail.  I put a patch together below:
> >
> > Gah, I missed that usage (also, it's now nested_vmx_check_guest_state()).
> >
> > Side topic, I think the VM_ENTRY_LOAD_BNDCFGS check should be gated by
> > nested_run_pending, a la the EFER check.'
> >
> > > ------------------------------------
> > >
> > > nested_run_pending=1 implies we have successfully entered guest mode.
> > > Move setting from external state in vmx_set_nested_state() until after
> > > all other checks are complete.
> > >
> > > Signed-off-by: Aaron Lewis <aaronlewis@xxxxxxxxxx>
> > > Reviewed-by: Peter Shier <pshier@xxxxxxxxxx>
> > > ---
> > >  arch/x86/kvm/vmx/nested.c | 14 +++++++++-----
> > >  1 file changed, 9 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> > > index 6401eb7ef19c..cf1f810223d2 100644
> > > --- a/arch/x86/kvm/vmx/nested.c
> > > +++ b/arch/x86/kvm/vmx/nested.c
> > > @@ -5460,9 +5460,6 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
> > >   if (!(kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE))
> > >   return 0;
> > >
> > > - vmx->nested.nested_run_pending =
> > > - !!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
> >
> > Alternatively, it might be better to leave nested_run_pending where it
> > is and instead add a label to handle clearing the flag on error.  IIUC,
> > the real issue is that nested_run_pending is left set after a failed
> > vmx_set_nested_state(), not that its shouldn't be set in the shadow
> > VMCS handling.
> >
> > Patch attached, though it's completely untested.  The KVM selftests are
> > broken for me right now, grrr.
> >
> > > -
> > >   if (nested_cpu_has_shadow_vmcs(vmcs12) &&
> > >       vmcs12->vmcs_link_pointer != -1ull) {
> > >   struct vmcs12 *shadow_vmcs12 = get_shadow_vmcs12(vcpu);
> > > @@ -5480,14 +5477,21 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
> > >   return -EINVAL;
> > >   }
> > >
> > > + vmx->nested.nested_run_pending =
> > > + !!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
> > > +
> > >   if (nested_vmx_check_vmentry_prereqs(vcpu, vmcs12) ||
> > > -     nested_vmx_check_vmentry_postreqs(vcpu, vmcs12, &exit_qual))
> > > +     nested_vmx_check_vmentry_postreqs(vcpu, vmcs12, &exit_qual)) {
> > > +     vmx->nested.nested_run_pending = 0;
> > >   return -EINVAL;
> > > + }
> > >
> > >   vmx->nested.dirty_vmcs12 = true;
> > >   ret = nested_vmx_enter_non_root_mode(vcpu, false);
> > > - if (ret)
> > > + if (ret) {
> > > + vmx->nested.nested_run_pending = 0;
> > >   return -EINVAL;
> > > + }
> > >
> > >   return 0;
> > >  }
>
> Here is an update based on your patch.  I ran these changes against
> the test vmx_set_nested_state_test, and it run successfully.
>
> That's correct that we are only concerned about restoring the state of
> nested_run_pending, so it's fine to set it where we do as long as we
> back the state change out before returning if we get an error.
>
> ---------------------------------------------
>
> nested_run_pending=1 implies we have successfully entered guest mode.
> Move setting from external state in vmx_set_nested_state() until after
> all other checks are complete.
>
> Signed-off-by: Aaron Lewis <aaronlewis@xxxxxxxxxx>
> Tested-by: Aaron Lewis <aaronlewis@xxxxxxxxxx>
> Reviewed-by: Peter Shier <pshier@xxxxxxxxxx>
> ---
>  arch/x86/kvm/vmx/nested.c | 20 ++++++++++++++------
>  1 file changed, 14 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 6401eb7ef19c..fe5814df5149 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -5468,28 +5468,36 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
>   struct vmcs12 *shadow_vmcs12 = get_shadow_vmcs12(vcpu);
>
>   if (kvm_state->size < sizeof(kvm_state) + 2 * sizeof(*vmcs12))
> - return -EINVAL;
> + goto error_guest_mode_einval;
>
>   if (copy_from_user(shadow_vmcs12,
>      user_kvm_nested_state->data + VMCS12_SIZE,
> -    sizeof(*vmcs12)))
> - return -EFAULT;
> +    sizeof(*vmcs12))) {
> + ret = -EFAULT;
> + goto error_guest_mode;
> + }
>
>   if (shadow_vmcs12->hdr.revision_id != VMCS12_REVISION ||
>       !shadow_vmcs12->hdr.shadow_vmcs)
> - return -EINVAL;
> + goto error_guest_mode_einval;
>   }
>
>   if (nested_vmx_check_vmentry_prereqs(vcpu, vmcs12) ||
>       nested_vmx_check_vmentry_postreqs(vcpu, vmcs12, &exit_qual))
> - return -EINVAL;
> + goto error_guest_mode_einval;
>
>   vmx->nested.dirty_vmcs12 = true;
>   ret = nested_vmx_enter_non_root_mode(vcpu, false);
>   if (ret)
> - return -EINVAL;
> + goto error_guest_mode_einval;
>
>   return 0;
> +
> +error_guest_mode_einval:
> + ret = -EINVAL;
> +error_guest_mode:
> + vmx->nested.nested_run_pending = 0;
> + return ret;
>  }
>
>  void nested_vmx_vcpu_setup(void)

Sean, I updated this patch based on the one you sent out.  Does it
look good now?



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux