Re: [PATCH 2/5] irqchip/renesas-rzg2l: Add support for RZ/Five SoC

[Geert Uytterhoeven <geert@linux-m68k.org>]

Hi Prabhakar,

On Wed, Jan 31, 2024 at 7:36 PM Lad, Prabhakar
<prabhakar.csengg@gmail.com> wrote:
> On Tue, Jan 30, 2024 at 11:38 AM Geert Uytterhoeven
> <geert@linux-m68k.org> wrote:
> > On Mon, Jan 29, 2024 at 4:16 PM Prabhakar <prabhakar.csengg@gmail.com> wrote:
> > > From: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
> > >
> > > The IX45 block has additional mask registers (NMSK/IMSK/TMSK) as compared
> > > to the RZ/G2L (family) SoC.
> > >
> > > Introduce masking/unmasking support for IRQ and TINT interrupts in IRQC
> > > controller driver. Two new registers, IMSK and TMSK, are defined to
> > > handle masking on RZ/Five SoC. The implementation utilizes a new data
> > > structure, `struct rzg2l_irqc_data`, to determine mask support for a
> > > specific controller instance.
> > >
> > > Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
> >
> > > --- a/drivers/irqchip/irq-renesas-rzg2l.c
> > > +++ b/drivers/irqchip/irq-renesas-rzg2l.c
> > > @@ -66,15 +68,25 @@ struct rzg2l_irqc_reg_cache {
> > >         u32     titsr[2];
> > >  };
> > >
> > > +/**
> > > + * struct rzg2l_irqc_data - OF data structure
> > > + * @mask_supported: Indicates if mask registers are available
> > > + */
> > > +struct rzg2l_irqc_data {
> >
> > This structure has the same name as the single static struct
> > rzg2l_irqc_priv instance, which is confusing.
> >
> Agreed, I will rename it to rzg2l_irqc_of_data
>
> > > +       bool    mask_supported;
> > > +};
> > > +
> > >  /**
> > >   * struct rzg2l_irqc_priv - IRQ controller private data structure
> > >   * @base:      Controller's base address
> > > + * @data:      OF data pointer
> > >   * @fwspec:    IRQ firmware specific data
> > >   * @lock:      Lock to serialize access to hardware registers
> > >   * @cache:     Registers cache for suspend/resume
> > >   */
> > >  static struct rzg2l_irqc_priv {
> > >         void __iomem                    *base;
> > > +       const struct rzg2l_irqc_data    *data;
> >
> > Replacing this by a bool would avoid a pointer dereference in each user,
> > and allows you to make rzg2l_irqc_data etc. __initconst.
> >
> Do you mean just add "bool mask_supported" here and get rid of struct
> rzg2l_irqc_data ? Can you please elaborate here..

Either add "bool mask_supported" here, or add a copy of the full
struct rzg2l_irqc_data (see below).

>
> > >         struct irq_fwspec               fwspec[IRQC_NUM_IRQ];
> > >         raw_spinlock_t                  lock;
> > >         struct rzg2l_irqc_reg_cache     cache;
> >
> > > @@ -371,9 +475,23 @@ static int rzg2l_irqc_parse_interrupts(struct rzg2l_irqc_priv *priv,
> > >         return 0;
> > >  }
> > >
> > > +static const struct rzg2l_irqc_data rzfive_irqc_data = {
> > > +       .mask_supported = true,
> > > +};
> > > +
> > > +static const struct rzg2l_irqc_data rzg2l_irqc_default_data = {
> > > +       .mask_supported = false,
> > > +};
> > > +
> > > +static const struct of_device_id rzg2l_irqc_matches[] = {
> > > +       { .compatible = "renesas,r9a07g043f-irqc", .data = &rzfive_irqc_data },
> > > +       { }
> > > +};
> > > +
> > >  static int rzg2l_irqc_init(struct device_node *node, struct device_node *parent)
> > >  {
> > >         struct irq_domain *irq_domain, *parent_domain;
> > > +       const struct of_device_id *match;
> > >         struct platform_device *pdev;
> > >         struct reset_control *resetn;
> > >         int ret;
> > > @@ -392,6 +510,12 @@ static int rzg2l_irqc_init(struct device_node *node, struct device_node *parent)
> > >         if (!rzg2l_irqc_data)
> > >                 return -ENOMEM;
> > >
> > > +       match = of_match_node(rzg2l_irqc_matches, node);
> > > +       if (match)
> > > +               rzg2l_irqc_data->data = match->data;
> > > +       else
> > > +               rzg2l_irqc_data->data = &rzg2l_irqc_default_data;
> >
> > Instead of matching a second time, I'd rather add a second
> > IRQCHIP_MATCH() entry with a different init function, passing the
> > actual rzg2l_irqc_data pointer.
> >
> OK, or rather just pass true/false instead of rzg2l_irqc_of_data pointer.?

Yes, that would be fine for me, too.
It all depends on whether you plan to add, or see a need for adding,
more flags or other fields in the future (and even for flags, you could
combine them in an unsigned long).

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds