Re: [PATCH v3] irqchip: xilinx: Add support for multiple instances

From: Thomas Gleixner
Date: Thu Feb 13 2020 - 07:01:29 EST


Mubin,

Mubin Usman Sayyed <mubin.usman.sayyed@xxxxxxxxxx> writes:

> From: Mubin Sayyed <mubin.usman.sayyed@xxxxxxxxxx>
>
> This patch adds support for multiple instances of

git grep 'This patch' Documentation/process/submitting-patches.rst

> xilinx interrupt controller. Below configurations are
> supported by driver,
>
> - peripheral->xilinx-intc->xilinx-intc->gic
> - peripheral->xilinx-intc->xilinx-intc

This is really not much of an explanation.

> Signed-off-by: Anirudha Sarangi <anirudha.sarangi@xxxxxxxxxx>
> Signed-off-by: Mubin Sayyed <mubin.usman.sayyed@xxxxxxxxxx>

This Signed-off-by chain is incorrect. See chapter 11 and 12 in the same
document.

> @@ -38,29 +38,32 @@ struct xintc_irq_chip {
> void __iomem *base;
> struct irq_domain *root_domain;
> u32 intr_mask;
> + struct irq_chip *intc_dev;
> + u32 nr_irq;
> };
>
> -static struct xintc_irq_chip *xintc_irqc;
> +static struct xintc_irq_chip *primary_intc;
>
> -static void xintc_write(int reg, u32 data)
> +static void xintc_write(struct xintc_irq_chip *irqc, int reg, u32 data)
> {
> if (static_branch_unlikely(&xintc_is_be))
> - iowrite32be(data, xintc_irqc->base + reg);
> + iowrite32be(data, irqc->base + reg);
> else
> - iowrite32(data, xintc_irqc->base + reg);
> + iowrite32(data, irqc->base + reg);
> }
>
> -static unsigned int xintc_read(int reg)
> +static u32 xintc_read(struct xintc_irq_chip *irqc, int reg)
> {
> if (static_branch_unlikely(&xintc_is_be))
> - return ioread32be(xintc_irqc->base + reg);
> + return ioread32be(irqc->base + reg);
> else
> - return ioread32(xintc_irqc->base + reg);
> + return ioread32(irqc->base + reg);
> }
>
> static void intc_enable_or_unmask(struct irq_data *d)
> {
> - unsigned long mask = 1 << d->hwirq;
> + unsigned long mask = BIT(d->hwirq);
> + struct xintc_irq_chip *irqc = irq_data_get_irq_chip_data(d);

Please order your local variables in reverse fir tree order:

struct xintc_irq_chip *irqc = irq_data_get_irq_chip_data(d);
unsigned long mask = BIT(d->hwirq);

which is the preferred coding style in this subsystem and way simpler to
read.

> static void intc_mask_ack(struct irq_data *d)
> {
> - unsigned long mask = 1 << d->hwirq;
> + unsigned long mask = BIT(d->hwirq);
> + struct xintc_irq_chip *irqc = irq_data_get_irq_chip_data(d);

Ditto.

> pr_debug("irq-xilinx: disable_and_ack: %ld\n", d->hwirq);
> - xintc_write(CIE, mask);
> - xintc_write(IAR, mask);
> + xintc_write(irqc, CIE, mask);
> + xintc_write(irqc, IAR, mask);
> }
> +static unsigned int xintc_get_irq_local(struct xintc_irq_chip *irqc)
> +{
> + u32 hwirq;
> + unsigned int irq = 0;

Same.

> + hwirq = xintc_read(irqc, IVR);
> + if (hwirq != -1U)
> + irq = irq_find_mapping(irqc->root_domain, hwirq);
> +
> + pr_debug("irq-xilinx: hwirq=%d, irq=%d\n", hwirq, irq);

Are these pr_debugs all over the please really required? I can
understand that you use them for development, but are they useful once
the stuff works?

> + return irq;
> +}
> +
> unsigned int xintc_get_irq(void)
> {
> - unsigned int hwirq, irq = -1;
> + u32 hwirq;
> + unsigned int irq = -1;

See above.

> - hwirq = xintc_read(IVR);
> + hwirq = xintc_read(primary_intc, IVR);
> if (hwirq != -1U)
> - irq = irq_find_mapping(xintc_irqc->root_domain, hwirq);
> + irq = irq_find_mapping(primary_intc->root_domain, hwirq);
>
> pr_debug("irq-xilinx: hwirq=%d, irq=%d\n", hwirq, irq);
>
> @@ -138,12 +164,14 @@ static const struct irq_domain_ops xintc_irq_domain_ops = {
> static void xil_intc_irq_handler(struct irq_desc *desc)
> {
> struct irq_chip *chip = irq_desc_get_chip(desc);
> + struct xintc_irq_chip *irqc =
> + irq_data_get_irq_handler_data(&desc->irq_data);

Please avoid these ugly line breaks and put the initialization of the
variable in to the code below the declaration.

> /* Turn on the Master Enable. */
> - xintc_write(MER, MER_HIE | MER_ME);
> - if (!(xintc_read(MER) & (MER_HIE | MER_ME))) {
> + xintc_write(irqc, MER, MER_HIE | MER_ME);
> + if (!(xintc_read(irqc, MER) & (MER_HIE | MER_ME))) {
> static_branch_enable(&xintc_is_be);

I see it's existing logic, but this lacks a comment how it's determined
that xintc is big endian. Looks like some weird "write works?"
probing. Why?

> + xintc_write(irqc, MER, MER_HIE | MER_ME);

So this writes MER_HIE | MER_ME into MER

> + if (!(xintc_read(irqc, MER) & (MER_HIE | MER_ME))) {

but this checks just whether ONE of the bits is set. Shouldn't it check
for MER == (MER_HIE | MER_ME), i.e. read back what was written?

Thanks,

tglx