Re: [PATCH 2/3] kvmclock - the host part.

[Avi Kivity <avi@qumranet.com>]

Glauber de Oliveira Costa wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Avi Kivity escreveu:
>   
> > Glauber de Oliveira Costa wrote:
> >     
> > > This is the host part of kvm clocksource implementation. As it does
> > > not include clockevents, it is a fairly simple implementation. We
> > > only have to register a per-vcpu area, and start writting to it
> > > periodically.
> > >
> > >   
> > >       
> > Missing live migration support  (a way for userspace to read and write
> > the guest clock address).  Should probably be in a separate patch.
> >     
>
> I think it's a matter of issuing a hypercall for reading the clock
> address. It's fair simple, and can be done in a later version of this patch.
> As for writting, the register hypercall itself can be used. It has no
> special side-effects we should care about.
>   

kvm live migration is done (at least thus far) without guest 
involvement.  So you need the host to be able to transfer this state.


>   
> > > @@ -1924,6 +1955,7 @@ out:
> > >          goto preempted;
> > >      }
> > >  
> > > +    kvm_write_guest_time(vcpu);
> > >      post_kvm_run_save(vcpu, kvm_run);
> > >   
> > >       
> > Why here?  Seems like we're leaving the guest for a while at this place.
> >
> > Suggest putting it on top of __vcpu_run(), guarded by a flag, and
> > setting the flag every time we put the vcpu.
> >     
>
> No special preference. It just sounded exity enough to me. I can move to
> where you suggest.
>
>   

It's more than a place, it's a set of rules:

- if the vcpu is migrated, we need a new timebase
- ditto if we're descheduled (well that's the same thing)
- if "some time passes"


--
error compiling committee.c: too many arguments to function

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/