Re: [Xen-devel] [PATCH v1 0/2] xen/arm: maintenance_interrupt SMP fix

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Wed, 29 Jan 2014, Julien Grall wrote:
> On 29/01/14 10:56, Oleksandr Tyshchenko wrote:
> > Hello all,
> > 
> > I just recollected about one hack which we created
> > as we needed to route HW IRQ in domU.
> > 
> > diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
> > index 9d793ba..d0227b9 100644
> > --- a/tools/libxl/libxl_create.c
> > +++ b/tools/libxl/libxl_create.c
> > @@ -989,8 +989,6 @@ static void domcreate_launch_dm(libxl__egc *egc,
> > libxl__multidev *multidev,
> > 
> >           LOG(DEBUG, "dom%d irq %d", domid, irq);
> > 
> > -        ret = irq >= 0 ? xc_physdev_map_pirq(CTX->xch, domid, irq, &irq)
> > -                       : -EOVERFLOW;
> >           if (!ret)
> >               ret = xc_domain_irq_permission(CTX->xch, domid, irq, 1);
> >           if (ret < 0) {
> > diff --git a/xen/arch/arm/vgic.c b/xen/arch/arm/vgic.c
> > index 2e4b11f..b54c08e 100644
> > --- a/xen/arch/arm/vgic.c
> > +++ b/xen/arch/arm/vgic.c
> > @@ -85,7 +85,7 @@ int domain_vgic_init(struct domain *d)
> >       if ( d->domain_id == 0 )
> >           d->arch.vgic.nr_lines = gic_number_lines() - 32;
> >       else
> > -        d->arch.vgic.nr_lines = 0; /* We don't need SPIs for the guest */
> > +        d->arch.vgic.nr_lines = gic_number_lines() - 32; /* We do
> > need SPIs for the guest */
> > 
> >       d->arch.vgic.shared_irqs =
> >           xzalloc_array(struct vgic_irq_rank, DOMAIN_NR_RANKS(d));
> > diff --git a/xen/common/domctl.c b/xen/common/domctl.c
> > index 75e2df3..ba88901 100644
> > --- a/xen/common/domctl.c
> > +++ b/xen/common/domctl.c
> > @@ -29,6 +29,7 @@
> >   #include <asm/page.h>
> >   #include <public/domctl.h>
> >   #include <xsm/xsm.h>
> > +#include <asm/gic.h>
> > 
> >   static DEFINE_SPINLOCK(domctl_lock);
> >   DEFINE_SPINLOCK(vcpu_alloc_lock);
> > @@ -782,8 +783,11 @@ long
> > do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
> >               ret = -EINVAL;
> >           else if ( xsm_irq_permission(XSM_HOOK, d, pirq, allow) )
> >               ret = -EPERM;
> > -        else if ( allow )
> > -            ret = pirq_permit_access(d, pirq);
> > +        else if ( allow ) {
> > +            struct dt_irq irq = {pirq + NR_LOCAL_IRQS,0};
> > +            ret = pirq_permit_access(d, irq.irq);
> > +            gic_route_irq_to_guest(d, &irq, "");
> > +        }
> >           else
> >               ret = pirq_deny_access(d, pirq);
> >       }
> > (END)
> > 
> > It seems, the following patch can violate the logic about routing
> > physical IRQs only to CPU0.
> 
> I forgot the smp_processor_id() in gic_route_irq_to_guest(). As this function
> is only called (for upstream) when dom0 is building, only CPU0 is used.

Right, that's why changing it to cpumask_of(0) shouldn't make any
difference for xen-unstable (it should make things clearer, if nothing
else) but it should fix things for Oleksandr.

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