From: Marc Zyngier <maz at kernel.org>
> On 2020-08-03 07:22, Mark-PK Tsai wrote:
> > Add mt58xx interrupt controller support using hierarchy irq
> > domain.
> >
> > Signed-off-by: Mark-PK Tsai <mark-pk.tsai at mediatek.com>
> > ---
> > drivers/irqchip/Kconfig | 7 ++
> > drivers/irqchip/Makefile | 1 +
> > drivers/irqchip/irq-mt58xx.c | 196 +++++++++++++++++++++++++++++++++++
> > 3 files changed, 204 insertions(+)
> > create mode 100644 drivers/irqchip/irq-mt58xx.c
> >
> > diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
> > index 216b3b8392b5..00453af78be0 100644
> > --- a/drivers/irqchip/Kconfig
> > +++ b/drivers/irqchip/Kconfig
> > @@ -572,4 +572,11 @@ config LOONGSON_PCH_MSI
> > help
> > Support for the Loongson PCH MSI Controller.
> >
> > +config MT58XX_IRQ
> > + bool "MT58XX IRQ"
> > + select IRQ_DOMAIN
> > + select IRQ_DOMAIN_HIERARCHY
> > + help
> > + Support Mediatek MT58XX Interrupt Controller.
> > +
> > endmenu
> > diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
> > index 133f9c45744a..5062e9bfa92d 100644
> > --- a/drivers/irqchip/Makefile
> > +++ b/drivers/irqchip/Makefile
> > @@ -111,3 +111,4 @@ obj-$(CONFIG_LOONGSON_HTPIC) +=
> > irq-loongson-htpic.o
> > obj-$(CONFIG_LOONGSON_HTVEC) += irq-loongson-htvec.o
> > obj-$(CONFIG_LOONGSON_PCH_PIC) += irq-loongson-pch-pic.o
> > obj-$(CONFIG_LOONGSON_PCH_MSI) += irq-loongson-pch-msi.o
> > +obj-$(CONFIG_MT58XX_IRQ) += irq-mt58xx.o
> > diff --git a/drivers/irqchip/irq-mt58xx.c
> > b/drivers/irqchip/irq-mt58xx.c
> > new file mode 100644
> > index 000000000000..e45ad023afa6
> > --- /dev/null
> > +++ b/drivers/irqchip/irq-mt58xx.c
> > @@ -0,0 +1,196 @@
> > +// SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause)
> > +/*
> > + * Copyright (c) 2020 MediaTek Inc.
> > + * Author Mark-PK Tsai <mark-pk.tsai at mediatek.com>
> > + */
> > +
> > +#include <linux/interrupt.h>
> > +#include <linux/io.h>
> > +#include <linux/irq.h>
> > +#include <linux/irqchip.h>
> > +#include <linux/irqdomain.h>
> > +#include <linux/of.h>
> > +#include <linux/of_address.h>
> > +#include <linux/of_irq.h>
> > +#include <linux/slab.h>
> > +
> > +#define INTC_MASK 0x0
> > +#define INTC_EOI 0x20
> > +
> > +struct mtk_intc_chip_data {
> > + char *name;
> > + struct irq_chip chip;
>
> There is no need to embed a full struct irqchip per controller, see
> below.
We want to distinguish which controller the device interrupts are belong to
by "cat /proc/interrupts".
And if all the controller share the same struct, the name field will be the
same.
Do you have suggestion for this?
>
> > + unsigned int irq_start, nr_irqs;
> > + void __iomem *base;
> > +
> };
> > +
> > +static void mtk_poke_irq(struct irq_data *d, u32 offset)
> > +{
> > + struct mtk_intc_chip_data *cd = irq_data_get_irq_chip_data(d);
> > + void __iomem *base = cd->base;
> > + u8 index = (u8)irqd_to_hwirq(d);
>
> Why the restrictive type? Why isn't unsigned int good enough?
You're right, unsigned int is ok.
I'll fix it in patch v2.
>
> > + u16 val, mask;
> > +
> > + mask = 1 << (index % 16);
> > + val = readw_relaxed(base + offset + (index / 16) * 4) | mask;
> > + writew_relaxed(val, base + offset + (index / 16) * 4);
>
> RMW without locking, you will end-up with corruption.
> Please store the address calculation in a temporaty variable to make it
> more readable
>
Thanks for the comment, I will fix it in pacth v2.
> > +
> }
> > +
> > +static void mtk_clear_irq(struct irq_data *d, u32 offset)
> > +{
> > + struct mtk_intc_chip_data *cd = irq_data_get_irq_chip_data(d);
> > + void __iomem *base = cd->base;
> > + u8 index = (u8)irqd_to_hwirq(d);
> > + u16 val, mask;
> > +
> > + mask = 1 << (index % 16);
> > + val = readw_relaxed(base + offset + (index / 16) * 4) & ~mask;
> > + writew_relaxed(val, base + offset + (index / 16) * 4);
>
> Same comments.
>
You're right, unsigned int is ok.
I'll fix it in patch v2.
> > +
> }
> > +
> > +static void mtk_intc_mask_irq(struct irq_data *d)
> > +{
> > + mtk_poke_irq(d, INTC_MASK);
> > + irq_chip_mask_parent(d);
> > +
> }
> > +
> > +static void mtk_intc_unmask_irq(struct irq_data *d)
> > +{
> > + mtk_clear_irq(d, INTC_MASK);
> > + irq_chip_unmask_parent(d);
> > +
> }
> > +
> > +static void mtk_intc_eoi_irq(struct irq_data *d)
> > +{
> > + mtk_poke_irq(d, INTC_EOI);
> > + irq_chip_eoi_parent(d);
> > +
> }
> > +
> > +static struct irq_chip mtk_intc_chip = {
> > + .irq_mask = mtk_intc_mask_irq,
> > + .irq_unmask = mtk_intc_unmask_irq,
> > + .irq_eoi = mtk_intc_eoi_irq,
> > + .irq_get_irqchip_state = irq_chip_get_parent_state,
> > + .irq_set_irqchip_state = irq_chip_set_parent_state,
> > + .irq_set_affinity = irq_chip_set_affinity_parent,
> > + .irq_set_vcpu_affinity = irq_chip_set_vcpu_affinity_parent,
>
> How about retrigger?
>
What is retrigger means?
To be honest, I just try to direct all the irqchip ops implemented in
/drivers/irqchip/irq-gic.c to gic driver.
But "irq_set_vcpu_affinity" is not used in our projects now.
Should I remove ".irq_set_vcpu_affinity" here?
> > + .irq_set_type = irq_chip_set_type_parent,
> > + .flags = IRQCHIP_SET_TYPE_MASKED |
> > + IRQCHIP_SKIP_SET_WAKE |
> > + IRQCHIP_MASK_ON_SUSPEND,
> > +
> };
> > +
> > +static int mt58xx_intc_domain_translate(struct irq_domain *d,
> > + struct irq_fwspec *fwspec,
> > + unsigned long *hwirq,
> > + unsigned int *type)
> > +{
> > + if (is_of_node(fwspec->fwnode)) {
> > + if (fwspec->param_count != 3)
> > + return -EINVAL;
> > +
> > + /* No PPI should point to this domain */
> > + if (fwspec->param[0] != 0)
> > + return -EINVAL;
> > +
> > + *hwirq = fwspec->param[1];
> > + *type = fwspec->param[2] & IRQ_TYPE_SENSE_MASK;
> > + return 0;
> > +
> }
> > +
> > + return -EINVAL;
> > +
> }
> > +
> > +static int mt58xx_intc_domain_alloc(struct irq_domain *domain,
> > unsigned int virq,
> > + unsigned int nr_irqs, void *data)
> > +{
> > + int i;
> > + irq_hw_number_t hwirq;
> > + struct irq_fwspec parent_fwspec, *fwspec = data;
> > + struct mtk_intc_chip_data *cd = (struct mtk_intc_chip_data
> > *)domain->host_data;
> > +
> > + /* Not GIC compliant */
> > + if (fwspec->param_count != 3)
> > + return -EINVAL;
> > +
> > + /* No PPI should point to this domain */
> > + if (fwspec->param[0])
> > + return -EINVAL;
> > +
> > + if (fwspec->param[1] >= cd->nr_irqs)
> > + return -EINVAL;
> > +
> > + hwirq = fwspec->param[1];
> > + for (i = 0; i < nr_irqs; i++)
> > + irq_domain_set_hwirq_and_chip(domain, virq + i, hwirq + i,
> > + &cd->chip,
> > + domain->host_data);
> > +
> > + parent_fwspec = *fwspec;
> > + parent_fwspec.fwnode = domain->parent->fwnode;
> > + parent_fwspec.param[1] = cd->irq_start + hwirq;
> > + return irq_domain_alloc_irqs_parent(domain, virq, nr_irqs,
> > &parent_fwspec);
> > +
> }
> > +
> > +static const struct irq_domain_ops mt58xx_intc_domain_ops = {
> > + .translate = mt58xx_intc_domain_translate,
> > + .alloc = mt58xx_intc_domain_alloc,
> > + .free = irq_domain_free_irqs_common,
> > +
> };
> > +
> > +int __init
> > +mt58xx_intc_of_init(struct device_node *dn, struct device_node
> > *parent)
> > +{
> > + static int nr_intc;
> > + struct irq_domain *domain, *domain_parent;
> > + struct mtk_intc_chip_data *cd;
> > + unsigned int irq_start, irq_end;
> > +
> > + domain_parent = irq_find_host(parent);
> > + if (!domain_parent) {
> > + pr_err("mt58xx-intc: interrupt-parent not found\n");
> > + return -EINVAL;
> > +
> }
> > +
> > + cd = kzalloc(sizeof(*cd), GFP_KERNEL);
> > + if (!cd)
> > + return -ENOMEM;
> > +
> > + cd->chip = mtk_intc_chip;
> > + if (of_property_read_u32_index(dn, "mediatek,irqs-map-range", 0,
> > &irq_start) ||
> > + of_property_read_u32_index(dn, "mediatek,irqs-map-range", 1,
> > &irq_end)) {
> > + kfree(cd);
> > + return -EINVAL;
> > +
> }
> > +
> > + if (of_property_read_bool(dn, "mediatek,intc-no-eoi"))
> > + cd->chip.irq_eoi = irq_chip_eoi_parent;
>
> No. Just add a flag to your chip data structure, and check for this
> flag in your irq_eoi callback. Or provide two distinct irq_chip
> structures that will only differ by the irq_eoi method.
Thanks for the comment, I will modify it in patch v2.
>
> > +
> > + cd->irq_start = irq_start;
> > + cd->nr_irqs = irq_end - irq_start + 1;
> > + cd->chip.name = kasprintf(GFP_KERNEL, "mt58xx-intc-%d", nr_intc++);
>
> Neither. That's not useful, and is a waste of memory. Stick to constant
> names in the irq_chip structure.
>
Actually we have multiple irq controller in our SoCs.
And if we use the constant names in irq_chip structure, the information in
"/proc/interrupts" will be hard to understand because all the irqchip name
is the same.
Do you have any suggestion for this?
> > + if (!cd->chip.name) {
> > + kfree(cd);
> > + return -ENOMEM;
> > +
> }
> > +
> > + cd->base = of_iomap(dn, 0);
> > + if (!cd->base) {
> > + kfree(cd->chip.name);
> > + kfree(cd);
> > + return -ENOMEM;
> > +
> }
> > +
> > + domain = irq_domain_add_hierarchy(domain_parent, 0, cd->nr_irqs,
> > + dn, &mt58xx_intc_domain_ops, cd);
> > + if (!domain) {
> > + kfree(cd->chip.name);
> > + iounmap(cd->base);
> > + kfree(cd);
> > + return -ENOMEM;
> > +
> }
> > +
> > + return 0;
> > +
> }
> > +
> > +IRQCHIP_DECLARE(mt58xx_intc, "mediatek,mt58xx-intc",
> > mt58xx_intc_of_init);
>
> On a side note, the merge window has just opened. Please refrain from
> reposting this until -rc1.
Got it, and thanks for your comments.
I'll update the patch and post it after -rc1.