[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v4 2/7] xen/arm: SMP support



On Fri, 26 Apr 2013, Ian Campbell wrote:
> On Fri, 2013-04-26 at 11:27 +0100, Stefano Stabellini wrote:
> > On Fri, 26 Apr 2013, Ian Campbell wrote:
> > > On Thu, 2013-04-25 at 19:45 +0100, Stefano Stabellini wrote:
> > > > On Thu, 25 Apr 2013, Ian Campbell wrote:
> > > > > > > > @@ -216,6 +245,8 @@ static int __init xen_guest_init(void)
> > > > > > > >          * is required to use VCPUOP_register_vcpu_info to 
> > > > > > > > place vcpu info
> > > > > > > >          * for secondary CPUs as they are brought up. */
> > > > > > > >         per_cpu(xen_vcpu, 0) = 
> > > > > > > > &HYPERVISOR_shared_info->vcpu_info[0];
> > > > > > > > +       for_each_online_cpu(i)
> > > > > > > > +               xen_secondary_init(i);
> > > > > > > >  
> > > > > > > >         gnttab_init();
> > > > > > > >         if (!xen_initial_domain())
> > > > > > > [...]
> > > > > > > > @@ -244,7 +280,7 @@ static int __init xen_init_events(void)
> > > > > > > >                 return -EINVAL;
> > > > > > > >         }
> > > > > > > >  
> > > > > > > > -       enable_percpu_irq(xen_events_irq, 0);
> > > > > > > > +       on_each_cpu(xen_percpu_enable_events, NULL, 0);
> > > > > > > 
> > > > > > > It feels like there ought to be some sort of per-cpu bringup 
> > > > > > > callback
> > > > > > > which takes care of these dynamically. Maybe that doesn't matter 
> > > > > > > until
> > > > > > > we get vcpu hotplug going?
> > > > > > 
> > > > > > I suspect there isn't one, considering that on_each_cpu is also 
> > > > > > used by 
> > > > > > kvm_vgic_hyp_init, kvm_timer_hyp_init and others.
> > > > > 
> > > > > Could we use cpu_notifiers for this?
> > > > 
> > > > cpu_notifiers are for cpu hotplug, not for secondary cpu bringup
> > > 
> > > Are you sure they don't also trigger during bringup, because the
> > > distinction is a little bit academic...
> > 
> > My mistake, they do run on secondary cpus, but not in our case because
> > xen_guest_init is called *after* cpu_notifiers are called.
> > So, they are too early for Xen.
> 
> Another reason to consider calling xen_guest_init much earlier then
> IMHO. although we can live with the solution you have now I suppose.

Yeah, moving the call to xen_guest_init earlier might be a good idea,
but I wouldn't want to do it in this patch series, at this point of the
Linux release cycle.

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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.