On Thu, May 25, 2017 at 03:53:10PM +0200, Paolo Bonzini wrote:
>
>
> On 25/05/2017 12:28, Andrew Jones wrote:
> > +int gic_init(void)
> > +{
> > + if (gicv2_init())
> > + gic_common_ops = &gicv2_common_ops;
> > + else if (gicv3_init())
> > + gic_common_ops = &gicv3_common_ops;
> > + return gic_version();
> > +}
> > +
>
> Of the other uses of gic_version(), this one is the ugliest:
>
> if (gic_version() == 2)
> base = gicv2_dist_base();
> else
> base = gicv3_redist_base();
>
> writel(val, base + GICD_ICACTIVER);
>
> Could you change this writel to a new op gicv[23]_eoi or something like
> that?
That use is just for an irq handler specific to a single unit test,
and at this time I don't think we need an op for that writel, as
that's the only place we currently write that register.
I do have some plans for irq handler refactoring in arm/gic.c, in
order to prepare for more tests. I'll clean this up a bit when I
do that.
Thanks,
drew
