Re: [PATCH v2 0/4] VMX: Properly handle pi descriptor and per-cpu blocking list

[George Dunlap <george.dunlap@citrix.com>]

On 24/06/16 14:42, Wu, Feng wrote:
> 
> 
>> -----Original Message-----
>> From: Dario Faggioli [mailto:dario.faggioli@citrix.com]
>> Sent: Friday, June 24, 2016 6:29 PM
>> To: Wu, Feng <feng.wu@intel.com>; xen-devel@lists.xen.org
>> Cc: keir@xen.org; Tian, Kevin <kevin.tian@intel.com>; jbeulich@suse.com;
>> andrew.cooper3@citrix.com; george.dunlap@eu.citrix.com;
>> konrad.wilk@oracle.com
>> Subject: Re: [PATCH v2 0/4] VMX: Properly handle pi descriptor and per-cpu
>> blocking list
>>
>> On Fri, 2016-06-24 at 06:33 +0000, Wu, Feng wrote:
>>>> -----Original Message-----
>>>> From: Dario Faggioli [mailto:dario.faggioli@citrix.com]
>>>> In your case, AFAICUI, it's:
>>>>  - the vCPU should block, waiting for an event
>>>>  - the event is _not_ arrived, so we indeed should block
>>>>  - we do *not* block, due to some other reason
>>>>
>>>> That does not look right to me... what about the fact that we
>>>> wanted to
>>>> actually wait for the event? :-O
>>> I understand your point. In my understanding, currently, vcpu_block()
>>> is
>>> for guest's HLT operation, which means, guest has nothing to do. In
>>> this case, even we return (not blocking), seems the function is not
>>> broken.
>>>
>> So, basically, you're saying that the vcpu has nothing to do, and in
>> fact it executed an HLT instruction, and you want to let it continue to
>> run? :-O
> 
> Here is my understanding, if the guest has nothing to do, it will
> call HLT, and Xen hypervisor will call vcpu_block(), if we don't
> block the vCPU and return to guest, guest will continue to run
> HLT if it still has nothing to do. 

As it happens, most operating systems will handle this situation just
fine; but I'm not sure this is something we want to rely on.

 -George

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