Re: [PATCH RFC V5 3/5] xen: Force-enable relevant MSR events; optimize the number of sent MSR events

[Razvan Cojocaru <rcojocaru@bitdefender.com>]

On 08/08/2014 05:34 PM, Jan Beulich wrote:
>>>> On 06.08.14 at 17:58, <rcojocaru@bitdefender.com> wrote:
>> --- a/xen/arch/x86/mm/mem_event.c
>> +++ b/xen/arch/x86/mm/mem_event.c
>> @@ -600,6 +600,9 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
>>              rc = mem_event_enable(d, mec, med, _VPF_mem_access, 
>>                                      HVM_PARAM_ACCESS_RING_PFN,
>>                                      mem_access_notification);
>> +
>> +            if ( rc == 0 && hvm_funcs.enable_intro_msr_interception )
>> +                hvm_funcs.enable_intro_msr_interception(d);
> 
> Isn't the sequence of operations wrong here (leaving a window in
> time where mem events are already enabled but the necessary MSRs
> aren't being intercepted yet? Or was it that guests are being paused
> while all this takes place?

I've looked into how this is implemented on the libxc side, and it would
appear that xc_mem_event_enable() (in xc_mem_event.c) does explicitly
pause the domain before calling xc_mem_event_control() (which ends up in
the code above), so it should not be a problem.


Thanks,
Razvan Cojocaru