Re: [Xen-devel] [PATCH] x86/vm_event: fix race between vmx_vmexit_handler() and vm_event_resume()

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 27.04.17 at 10:11, <rcojocaru@xxxxxxxxxxxxxxx> wrote:
> On 04/27/17 11:01, Jan Beulich wrote:
>>>>> On 27.04.17 at 09:22, <rcojocaru@xxxxxxxxxxxxxxx> wrote:
>>> --- a/xen/arch/x86/hvm/hvm.c
>>> +++ b/xen/arch/x86/hvm/hvm.c
>>> @@ -473,6 +473,39 @@ static bool hvm_get_pending_event(struct vcpu *v, 
>>> struct x86_event *info)
>>>      return hvm_funcs.get_pending_event(v, info);
>>>  }
>>>  
>>> +static void hvm_vm_event_set_registers(struct vcpu *v)
>>> +{
>>> +    ASSERT(v == current);
>>> +
>>> +    if ( v->arch.vm_event->set_gprs )
>> 
>> I think we will want an unlikely() here. We anyway can only hope for
>> the compiler to always inline this function, such that non-VM-event
>> setups don't get penalized by the extra call here. Strictly speaking
>> the function doesn't belong into this file ...
> 
> Should I move it to the x86 vm_event code?

If you do, then you'll want to have an inline wrapper with the if()
in it, so the actual call at the machine level would be avoided in the
common case.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel