Re: [DROP][PATCH] KVM: x86: Fix the #GP(0) and #UD conditions for XSETBV emulation

[Like Xu]

On 17/1/2022 7:18 pm, Paolo Bonzini wrote:
> On 1/17/22 10:44, Like Xu wrote:
> > > Also, the "Fixes" tag is not really correct as the behavior was the same 
> > > before.  Rather, it fixes commit 02d4160fbd76 ("x86: KVM: add xsetbv to the 
> >
> > It seems the original code comes from 81dd35d42c9a ("KVM: SVM: Add xsetbv 
> > intercept").
> > 2acf923e38 ("KVM: VMX: Enable XSAVE/XRSTOR for guest") and 92f9895c146d.
> >
> > > emulator", 2019-08-22).  Checking OSXSAVE is a bug in the emulator path, even 
> > > though it's not needed in the XSETBV vmexit case.
> >
> > The kvm_emulate_xsetbv() has been removed from the emulator path.
> > I'm not really sure why it's not needed in the XSETBV vmexit case. More details ?
>
> Nevermind, I confused AMD (where #UD is generated before checking for 
> exceptions) with Intel where it's unconditional.
>
> So the bug should be there since 2acf923e38 by executing XSETBV with 
> CR4.XSAVE=0.  If so, please include a testcase.

In the testcase "executing XSETBV with CR4.XSAVE=0",

- on the VMX, #UD delivery does not require vm-exit;
- on the SVM, #UD is trapped but goes to the ud_interception() path;

There is no room to run this pointless patch.
Thank you and sorry for the noise.

> Paolo
>
> > > Thanks,
> > >
> > > Paolo
> > >
> > > > +    if ((is_protmode(vcpu) && static_call(kvm_x86_get_cpl)(vcpu) != 0) ||
> > > >           __kvm_set_xcr(vcpu, kvm_rcx_read(vcpu), kvm_read_edx_eax(vcpu))) {
> > > >           kvm_inject_gp(vcpu, 0);
> > > >           return 1;