Re: [PATCH 3/9] irqchip/loongson-pch-pic: Add ACPI init support

[Huacai Chen]

Hi, Marc,

On Thu, Aug 12, 2021 at 9:28 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
>
> On Thu, 12 Aug 2021 13:23:27 +0100,
> Huacai Chen <chenhuacai@xxxxxxxxx> wrote:
> >
> [...]
>
> > > > > +struct fwnode_handle *pch_pic_acpi_init(struct fwnode_handle *parent,
> > > > > +                                     struct acpi_madt_bio_pic *acpi_pchpic)
> > > > > +{
> > > > > +     int count;
> > > > > +     struct pch_pic *priv;
> > > > > +     struct irq_domain *parent_domain;
> > > > > +
> > > > > +     priv = kzalloc(sizeof(*priv), GFP_KERNEL);
> > > > > +     if (!priv)
> > > > > +             return NULL;
> > > > > +
> > > > > +     raw_spin_lock_init(&priv->pic_lock);
> > > > > +     priv->base = ioremap(acpi_pchpic->address, acpi_pchpic->size);
> > > > > +     if (!priv->base)
> > > > > +             goto free_priv;
> > > > > +
> > > > > +     priv->domain_handle = irq_domain_alloc_fwnode(priv->base);
> > > > > +     if (!priv->domain_handle) {
> > > > > +             pr_err("Unable to allocate domain handle\n");
> > > > > +             goto iounmap_base;
> > > > > +     }
> > > > > +
> > > > > +     priv->ht_vec_base = acpi_pchpic->gsi_base;
> > > > > +     count = ((readq(priv->base) >> 48) & 0xff) + 1;
> > > > > +     parent_domain = irq_find_matching_fwnode(parent, DOMAIN_BUS_ANY);
> > > > > +     if (!parent_domain) {
> > > > > +             pr_err("Failed to find the parent domain\n");
> > > > > +             goto iounmap_base;
> > > > > +     }
> > > > > +
> > > > > +     priv->pic_domain = irq_domain_create_hierarchy(parent_domain, 0,
> > > > > +                                             count, priv->domain_handle,
> > > > > +                                             &pch_pic_domain_ops, priv);
> > > > > +
> > > > > +     if (!priv->pic_domain) {
> > > > > +             pr_err("Failed to create IRQ domain\n");
> > > > > +             goto iounmap_base;
> > > > > +     }
> > > > > +
> > > > > +     pch_pic_reset(priv);
> > > > > +     pch_pic_priv[nr_pch_pics++] = priv;
> > > > > +
> > > > > +     register_syscore_ops(&pch_pic_syscore_ops);
> > > > > +
> > > > > +     return priv->domain_handle;
> > > > > +
> > > > > +iounmap_base:
> > > > > +     iounmap(priv->base);
> > > > > +free_priv:
> > > > > +     kfree(priv);
> > > > > +
> > > > > +     return NULL;
> > > > > +}
> > > > > +
> > > > > +#endif
> > > >
> > > > A lot of this code is common with its OF counterpart. How about making
> > > > this logic common?
> > > OK, let me think about.
> > Though pch_pic_acpi_init() is similar to pch_pic_of_init(), but it is
> > difficult to make a common function, because we cannot prepare
> > everything before the common function. For example, ioremap() can be
> > the common part, but pch_pic_acpi_init() should get the vector count
> > after ioremap(). If we use an argument to distinguish the caller in
> > the common function, the complexity increases and seems no benefits.
>
> All firmware implementations allocate private data structures, irq
> domains, map MMIO regions, etc. All that can be common. We even have
> APIs that deal with both firmware interfaces.
>
> As for the 'read the vector count from the HW', what does it have to
> do with driving the HW using DT or ACPI? The HW doesn't *know*. If you
> are conflating HW changes and firmware interfaces, then you have
> already lost.
OK, I know, thanks.

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