Re: [PATCH v1 3/6] irqchip/gic-v3: support SGI broadcast

[Yu Zhao <yuzhao@google.com>]

On Tue, Oct 29, 2024 at 1:02 PM Marc Zyngier <maz@kernel.org> wrote:
>
> On Fri, 25 Oct 2024 18:31:01 +0100,
> Yu Zhao <yuzhao@google.com> wrote:
> >
> > On Fri, Oct 25, 2024 at 10:15 AM Marc Zyngier <maz@kernel.org> wrote:
> > >
> > > On Fri, 25 Oct 2024 06:07:45 +0100,
> > > Yu Zhao <yuzhao@google.com> wrote:
> > > >
> > > > Hi Marc,
> > > >
> > > > On Tue, Oct 22, 2024 at 9:03 AM Marc Zyngier <maz@kernel.org> wrote:
> > > > >
> > > > > On Mon, 21 Oct 2024 05:22:15 +0100,
> > > > > Yu Zhao <yuzhao@google.com> wrote:
> > > > > >
> > > > > > @@ -1407,6 +1418,13 @@ static void gic_ipi_send_mask(struct irq_data *d, const struct cpumask *mask)
> > > > > >        */
> > > > > >       dsb(ishst);
> > > > > >
> > > > > > +     cpumask_copy(&broadcast, cpu_present_mask);
> > > > >
> > > > > Why cpu_present_mask? I'd expect that cpu_online_mask should be the
> > > > > correct mask to use -- we don't IPI offline CPUs, in general.
> > > >
> > > > This is exactly because "we don't IPI offline CPUs, in general",
> > > > assuming "we" means the kernel, not GIC.
> > > >
> > > > My interpretation of what the GIC spec says ("0b1: Interrupts routed
> > > > to all PEs in the system, excluding self") is that it broadcasts IPIs to
> > > > "cpu_present_mask" (minus the local one). So if the kernel uses
> > > > "cpu_online_mask" here, GIC would send IPIs to offline CPUs
> > > > (cpu_present_mask ^ cpu_online_mask), which I don't know whether it's
> > > > a defined behavior.
> >
> > Thanks for clarifying.
> >
> > > Offline CPUs are not known to the kernel.
> >
> > I assume it wouldn't matter to firmware either, correct? IOW, we
>
> Firmware is on the secure side of the stack.
>
> > wouldn't cause firmware any trouble by letting GIC send IPIs to
> > (cpu_present_mask ^ cpu_online_mask), assuming those two masks can be
> > different on arm64 when hotplug is enabled?
>
> You can't send SGIs from non-secure to secure using ICC_SGI1R_EL1.

Right, this makes sense now.

> You
> would need to use ICC_ASGI1R_EL1, and have secure to allow such
> brokenness via a configuration of GICR_NSACR. Linux doesn't use the
> former, and no sane software touches the latter.
>
>         M.
>
> --
> Without deviation from the norm, progress is not possible.