On Thu, May 12, 2022, Vitaly Kuznetsov wrote: > Maxim Levitsky <mlevitsk@xxxxxxxxxx> writes: > > > Fixes: 1c2361f667f36 ("KVM: x86: Use __try_cmpxchg_user() to emulate atomic accesses") > > Signed-off-by: Maxim Levitsky <mlevitsk@xxxxxxxxxx> > > --- > > arch/x86/kvm/x86.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > Yes, this is the root cause of the TDP mmu leak I was doing debug of in the last week. > > Non working cmpxchg on which TDP mmu relies makes it install two differnt shadow pages > > under same spte. Ewww, as in running a buggy L0 resulted in a CMPXCHG going sideways in L1? That's awful. My apologies :-( > In case the fix is not squashed with 1c2361f667f36, the above should > really go before '---'. > > > > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > > index 8ee8c91fa7625..79cabd3d97d22 100644 > > --- a/arch/x86/kvm/x86.c > > +++ b/arch/x86/kvm/x86.c > > @@ -7329,7 +7329,7 @@ static int emulator_cmpxchg_emulated(struct x86_emulate_ctxt *ctxt, > > goto emul_write; > > > > hva = kvm_vcpu_gfn_to_hva(vcpu, gpa_to_gfn(gpa)); > > - if (kvm_is_error_hva(addr)) > > + if (kvm_is_error_hva(hva)) > > Looks like a typo indeed, so > > Reviewed-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx> Yep, if this doesn't get squashed Reviewed-by: Sean Christopherson <seanjc@xxxxxxxxxx>