Re: [PATCH v2 2/5] irqchip/renesas-rzg2l: Add support for RZ/Five SoC

From: Geert Uytterhoeven
Date: Thu Apr 18 2024 - 11:13:28 EST


Hi Prabhakar,

On Thu, Apr 4, 2024 at 3:35 PM Lad, Prabhakar
<prabhakar.csengg@xxxxxxxxx> wrote:
> On Thu, Apr 4, 2024 at 2:31 PM Biju Das <biju.das.jz@xxxxxxxxxxxxxx> wrote:
> > > -----Original Message-----
> > > From: Lad, Prabhakar <prabhakar.csengg@xxxxxxxxx>
> > > On Thu, Apr 4, 2024 at 8:44 AM Biju Das <biju.das.jz@xxxxxxxxxxxxxx> wrote:
> > > > > -----Original Message-----
> > > > > From: Prabhakar <prabhakar.csengg@xxxxxxxxx>
> > > > > 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@xxxxxxxxxxxxxx>
> > > > > ---
> > > > > v1->v2
> > > > > - Added IRQCHIP_MATCH() for RZ/Five
> > > > > - Retaining a copy of OF data in priv
> > > > > - Rebased the changes
> > > > > ---
> > > > > drivers/irqchip/irq-renesas-rzg2l.c | 137
> > > > > +++++++++++++++++++++++++++-
> > > > > 1 file changed, 132 insertions(+), 5 deletions(-)
> > > > >
> > > > > diff --git a/drivers/irqchip/irq-renesas-rzg2l.c
> > > > > b/drivers/irqchip/irq-renesas-rzg2l.c
> > > > > index f6484bf15e0b..6fa8d65605dc 100644
> > > > > --- a/drivers/irqchip/irq-renesas-rzg2l.c
> > > > > +++ b/drivers/irqchip/irq-renesas-rzg2l.c
> > > > > @@ -37,6 +37,8 @@
> > > > > #define TSSEL_SHIFT(n) (8 * (n))
> > > > > #define TSSEL_MASK GENMASK(7, 0)
> > > > > #define IRQ_MASK 0x3
> > > > > +#define IMSK 0x10010
> > > > > +#define TMSK 0x10020
> > > > >
> > > > > #define TSSR_OFFSET(n) ((n) % 4)
> > > > > #define TSSR_INDEX(n) ((n) / 4)
> > > > > @@ -66,15 +68,25 @@ struct rzg2l_irqc_reg_cache {
> > > > > u32 titsr[2];
> > > > > };
> > > > >
> > > > > +/**
> > > > > + * struct rzg2l_irqc_of_data - OF data structure
> > > > > + * @mask_supported: Indicates if mask registers are available */
> > > > > +struct 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_of_data *data;
> > > > > struct irq_fwspec fwspec[IRQC_NUM_IRQ];
> > > > > raw_spinlock_t lock;
> > > > > struct rzg2l_irqc_reg_cache cache;
> > > > > @@ -138,18 +150,102 @@ static void rzg2l_irqc_eoi(struct irq_data *d)
> > > > > irq_chip_eoi_parent(d);
> > > > > }
> > > > >
> > > > > +static void rzg2l_irqc_mask_irq_interrupt(struct rzg2l_irqc_priv *priv,
> > > > > + unsigned int hwirq) {
> > > > > + u32 imsk = readl_relaxed(priv->base + IMSK);
> > > > > + u32 bit = BIT(hwirq - IRQC_IRQ_START);
> > > > > +
> > > > > + writel_relaxed(imsk | bit, priv->base + IMSK); }
> > > > > +
> > > > > +static void rzg2l_irqc_unmask_irq_interrupt(struct rzg2l_irqc_priv *priv,
> > > > > + unsigned int hwirq) {
> > > > > + u32 imsk = readl_relaxed(priv->base + IMSK);
> > > > > + u32 bit = BIT(hwirq - IRQC_IRQ_START);
> > > > > +
> > > > > + writel_relaxed(imsk & ~bit, priv->base + IMSK); }
> > > > > +
> > > > > +static void rzg2l_irqc_mask_tint_interrupt(struct rzg2l_irqc_priv *priv,
> > > > > + unsigned int hwirq) {
> > > > > + u32 tmsk = readl_relaxed(priv->base + TMSK);
> > > > > + u32 bit = BIT(hwirq - IRQC_TINT_START);
> > > > > +
> > > > > + writel_relaxed(tmsk | bit, priv->base + TMSK); }
> > > > > +
> > > > > +static void rzg2l_irqc_unmask_tint_interrupt(struct rzg2l_irqc_priv *priv,
> > > > > + unsigned int hwirq) {
> > > > > + u32 tmsk = readl_relaxed(priv->base + TMSK);
> > > > > + u32 bit = BIT(hwirq - IRQC_TINT_START);
> > > > > +
> > > > > + writel_relaxed(tmsk & ~bit, priv->base + TMSK); }
> > > > > +
> > > > > +/* Must be called while priv->lock is held */ static void
> > > > > +rzg2l_irqc_mask_once(struct rzg2l_irqc_priv *priv, unsigned int
> > > > > +hwirq) {
> > > > > + if (!priv->data->mask_supported)
> > > > > + return;
> > > > > +
> > > > > + if (hwirq >= IRQC_IRQ_START && hwirq <= IRQC_IRQ_COUNT)
> > > > > + rzg2l_irqc_mask_irq_interrupt(priv, hwirq);
> > > > > + else if (hwirq >= IRQC_TINT_START && hwirq < IRQC_NUM_IRQ)
> > > > > + rzg2l_irqc_mask_tint_interrupt(priv, hwirq); }
> > > > > +
> > > > > +static void rzg2l_irqc_mask(struct irq_data *d) {
> > > > > + struct rzg2l_irqc_priv *priv = irq_data_to_priv(d);
> > > > > +
> > > > > + raw_spin_lock(&priv->lock);
> > > > > + rzg2l_irqc_mask_once(priv, irqd_to_hwirq(d));
> > > > > + raw_spin_unlock(&priv->lock);
> > > > > + irq_chip_mask_parent(d);
> > > > > +}
> > > > > +
> > > > > +/* Must be called while priv->lock is held */ static void
> > > > > +rzg2l_irqc_unmask_once(struct rzg2l_irqc_priv *priv, unsigned int
> > > > > +hwirq) {
> > > > > + if (!priv->data->mask_supported)
> > > > > + return;
> > > > > +
> > > > > + if (hwirq >= IRQC_IRQ_START && hwirq <= IRQC_IRQ_COUNT)
> > > > > + rzg2l_irqc_unmask_irq_interrupt(priv, hwirq);
> > > > > + else if (hwirq >= IRQC_TINT_START && hwirq < IRQC_NUM_IRQ)
> > > > > + rzg2l_irqc_unmask_tint_interrupt(priv, hwirq); }
> > > > > +
> > > > > +static void rzg2l_irqc_unmask(struct irq_data *d) {
> > > > > + struct rzg2l_irqc_priv *priv = irq_data_to_priv(d);
> > > > > +
> > > > > + raw_spin_lock(&priv->lock);
> > > > > + rzg2l_irqc_unmask_once(priv, irqd_to_hwirq(d));
> > > > > + raw_spin_unlock(&priv->lock);
> > > > > + irq_chip_unmask_parent(d);
> > > > > +}
> > > > > +
> > > > > static void rzg2l_tint_irq_endisable(struct irq_data *d, bool
> > > > > enable) {
> > > > > + struct rzg2l_irqc_priv *priv = irq_data_to_priv(d);
> > > > > unsigned int hw_irq = irqd_to_hwirq(d);
> > > > >
> > > > > if (hw_irq >= IRQC_TINT_START && hw_irq < IRQC_NUM_IRQ) {
> > > > > - struct rzg2l_irqc_priv *priv = irq_data_to_priv(d);
> > > > > u32 offset = hw_irq - IRQC_TINT_START;
> > > > > u32 tssr_offset = TSSR_OFFSET(offset);
> > > > > u8 tssr_index = TSSR_INDEX(offset);
> > > > > u32 reg;
> > > > >
> > > > > raw_spin_lock(&priv->lock);
> > > > > + if (enable)
> > > > > + rzg2l_irqc_unmask_once(priv, hw_irq);
> > > > > + else
> > > > > + rzg2l_irqc_mask_once(priv, hw_irq);
> > > > > reg = readl_relaxed(priv->base + TSSR(tssr_index));
> > > > > if (enable)
> > > > > reg |= TIEN << TSSEL_SHIFT(tssr_offset); @@
> > > > > -157,6 +253,13 @@ static void rzg2l_tint_irq_endisable(struct irq_data *d, bool enable)
> > > > > reg &= ~(TIEN << TSSEL_SHIFT(tssr_offset));
> > > > > writel_relaxed(reg, priv->base + TSSR(tssr_index));
> > > > > raw_spin_unlock(&priv->lock);
> > > > > + } else {
> > > > > + raw_spin_lock(&priv->lock);
> > > > > + if (enable)
> > > > > + rzg2l_irqc_unmask_once(priv, hw_irq);
> > > > > + else
> > > > > + rzg2l_irqc_mask_once(priv, hw_irq);
> > > > > + raw_spin_unlock(&priv->lock);
> > > > > }
> > > > > }
> > > > >
> > > > > @@ -324,8 +427,8 @@ static struct syscore_ops rzg2l_irqc_syscore_ops
> > > > > = { static const struct irq_chip irqc_chip = {
> > > > > .name = "rzg2l-irqc",
> > > > > .irq_eoi = rzg2l_irqc_eoi,
> > > > > - .irq_mask = irq_chip_mask_parent,
> > > > > - .irq_unmask = irq_chip_unmask_parent,
> > > > > + .irq_mask = rzg2l_irqc_mask,
> > > > > + .irq_unmask = rzg2l_irqc_unmask,
> > > >
> > > > I feel this will be clean, if we have
> > > >
> > > > static const struct irq_chip rzg2l_irqc_chip = {
> > > > .name = "rzg2l-irqc",
> > > > ...
> > > > .irq_mask = irq_chip_mask_parent,
> > > > .irq_unmask = irq_chip_unmask_parent,
> > > > ....
> > > > };
> > > >
> > > > static const struct irq_chip rzfive_irqc_chip = {
> > > > .name = "rzfive-irqc",
> > > > ...
> > > > .irq_mask = rzfive_irqc_mask,
> > > > .irq_unmask = rzfive_irqc_unmask,
> > > > ....
> > > > };
> > > >
> > > > And passing this in rzg2l_irqc_init() and rzfive_irqc_init(), see
> > > > below
> > > >
> > > > return rzg2l_irqc_init_helper(node, parent, & rzg2l_irqc_chip); return
> > > > rzg2l_irqc_init_helper(node, parent, & rzfive_irqc_chip);
> > > >
> > > If we do the above we are stuck with "struct irq_chip" as data, for further upcoming SoCs (for
> > > example RZ/V2H) which have more features we need to pass custom data to handle these features.
> >
> > That time device data can be extended like below
> >
> > struct rz_g2l_irq_chip {
> > struct irq_chip;
> > void *data; /* custom data */
> > }
> >
> Ok, but i'll wait for Geert to come back on this as Geert suggested to
> me to do it this way.

I agree with Biju.

Having separate irq_chips lets us avoid taking the spinlock on RZ/G2L.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68korg

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