RE: [Xen-ia64-devel] Xen is now runnnig at Bull

["Dong, Eddie" <eddie.dong@xxxxxxxxx>]

Magenheimer, Dan (HP Labs Fort Collins) wrote:
>> When the guest rid is not used anymore, the physical rid should be
>> garbaged. A simple way to determine when the guest rid is not used
>> can 
>> be done by
>> looking at the VHPT table and machine TLB side.
>> (Yes, a reference count needs to be introduced for each guest
>> rid to track the
>> entries in VHPT, and a special indicator in guest tlb to indicating
>> if there are entries in physical TLB)
> 
> But once a rid is put in the machine TLB, there's no way
> to guarantee that it won't be used sometime in the future
> except by flushing the entire TLB, correct?  So it can't
> be garbage collected except after a flush.
When a TLB is inserted into physical machine, yes SW don't know if the
translation is there or not, but the SW know if the guest translation is
purged
or not. If the guest tlb is there and a machine tlb is ever inserted, it
is thought
 as the tlb is in machine side. 
In this way, the max number of physical rids the guest is using  is the
size of
guest TLB. But typically, a guest rid has hundreds of guest TLBs if the
process
is active, so the actual number of physical rid consumed is realitive
small.

> 
>> Is it time for me to share what we thought of for rid virtualization
>> in much details?
> 
> Although I greatly enjoy these technical conversations, I
> guess I have to admit that supporting >64 domains is way
> down on the priority list compared to all the other
> functionality on our plate, so perhaps we should delay
> further discussion of rid virtualization until the
> future?

Agree partially. We depend this on the requirement of real deployment
 coming from like Bull and other companies.

> 
> Dan

Eddie

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel