RE: [Xen-devel] [Question] vcpu-set before or after xen_pause_requested

["Liu, Jinsong" <jinsong.liu@xxxxxxxxx>]

Ian Jackson wrote:
> Liu, Jinsong writes ("RE: [Xen-devel] [Question] vcpu-set before or
> after xen_pause_requested"): 
>> How about this solution:
> ...
>> In this way, old PV guest can still work in old way (no bad than
>> before) with warning message, suggesting user to update their PV
>> driver.
> 
> But, in new guests which are malfunctioning, things will not be
> handled properly: the timeout will trigger and it will seem to have
> worked.  (Consider a guest which is paused or is starved of CPU
> because the host is very busy.)  There is also no good value for the
> timeout.
> 
> I would suggest using a different xenstore key name for the new
> protocol.  That way you don't get the feature with old drivers but at
> least you don't make malfunctions worse with the new ones.
> 

Ian,

I'm not quite clear your idea. Per my understanding, a new xenstore key name 
cannot avoid the issue of old key, like timeout value, malfunction guest, ...
Currently I have no enough resource to solve all problems, so how about you 
solve old issue that PV driver has, and then I will update my HVM vcpu-set 
patch based on the new protocol?

Thanks,
Jinsong
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel