2017-05-10 23:35 GMT+08:00 Paolo Bonzini <pbonzini@xxxxxxxxxx>:
>
>
> On 10/05/2017 12:19, Wanpeng Li wrote:
>> * with old userspace.
>> */
>> - if (xstate_bv & ~kvm_supported_xcr0())
>> + if (xstate_bv & ~kvm_supported_xcr0() ||
>> + mxcsr & ~vcpu->arch.guest_fpu.state.xsave.i387.mxcsr_mask)
>> return -EINVAL;
>> load_xsave(vcpu, (u8 *)guest_xsave->region);
>> } else {
>> - if (xstate_bv & ~XFEATURE_MASK_FPSSE)
>> + if (xstate_bv & ~XFEATURE_MASK_FPSSE ||
>> + mxcsr & ~vcpu->arch.guest_fpu.state.fxsave.mxcsr_mask)
>> return -EINVAL;
>> memcpy(&vcpu->arch.guest_fpu.state.fxsave,
>> guest_xsave->region, sizeof(struct fxregs_state));
>
> Hmm, thinking more about it, maybe use mxcsr_feature_mask instead of
> digging into vcpu->arch.guest_fpu? If you send v2, please remember to
ERROR: "mxcsr_feature_mask" [arch/x86/kvm/kvm.ko] undefined. So we
should dig into vcpu->arch.guest_fpu.
Regards,
Wanpeng Li
