Re: [PATCH v2 1/3] genirq/irq_sim: Allow both one and two cell bindings

[Marc Zyngier <maz@kernel.org>]

On Wed, 28 Sep 2022 08:32:25 +0100,
Wei Yongjun <weiyongjun1@huawei.com> wrote:
> 
> 
> 
> On 2022/9/26 20:55, Marc Zyngier wrote:
> > On Mon, 26 Sep 2022 07:24:48 -0400,
> > Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> >>
> >> On Mon, Sep 26, 2022 at 10:27 AM Wei Yongjun <weiyongjun@huaweicloud.com> wrote:
> >>>
> >>> From: Wei Yongjun <weiyongjun1@huawei.com>
> >>>
> >>> The IRQ simulator only support one cell binding now, this patch make it
> >>> works with either one or two cell bindings, where the cell values map
> >>> directly to the irq number and irq flags.
> >>>
> >>> Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
> >>> ---
> >>>  kernel/irq/irq_sim.c | 1 +
> >>>  1 file changed, 1 insertion(+)
> >>>
> >>> diff --git a/kernel/irq/irq_sim.c b/kernel/irq/irq_sim.c
> >>> index dd76323ea3fd..73a90b7b6022 100644
> >>> --- a/kernel/irq/irq_sim.c
> >>> +++ b/kernel/irq/irq_sim.c
> >>> @@ -149,6 +149,7 @@ static void irq_sim_domain_unmap(struct irq_domain *domain, unsigned int virq)
> >>>  static const struct irq_domain_ops irq_sim_domain_ops = {
> >>>         .map            = irq_sim_domain_map,
> >>>         .unmap          = irq_sim_domain_unmap,
> >>> +       .xlate          = irq_domain_xlate_onetwocell,
> >>>  };
> >>>
> >>>  /**
> >>> --
> >>> 2.34.1
> >>>
> >>
> >> You'll need Marc's (Cc'ed) Ack here.
> 
> Hi Marc,
> 
> > 
> > The question is what will the simulator code do with this information.
> > Throw it away? What of 3/4/5 cell bindings? I'd rather see the
> 
> The 3/4/5 cell bindings is selience ignored currently.
> 
> > simulator being extended to deal with arbitrary bindings instead of
> > trading a harcoded limit for another one. And also give some
> > semantics to the extra cells.
> 
> Would you means we should allow the users to overwrite the xlate callback
> or overwrite the domain_ops?

Neither. I think the caller should provide an irq_domain_ops structure
at domain creation, with the .xlate member populated, and the irq_sim
code would add its own ops to it.

Providing NULL would ensure we fallback to the existing behaviour.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.