Re: [PATCH v6 4/6] clocksource/drivers/timer-riscv: Use per-CPU timer interrupt

From: Anup Patel
Date: Sun May 31 2020 - 01:57:30 EST


On Sat, May 30, 2020 at 5:11 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
>
> On 2020-05-30 11:07, Anup Patel wrote:
> > Instead of directly calling RISC-V timer interrupt handler from
> > RISC-V local interrupt conntroller driver, this patch implements
> > RISC-V timer interrupt as a per-CPU interrupt using per-CPU APIs
> > of Linux IRQ subsystem.
> >
> > Signed-off-by: Anup Patel <anup.patel@xxxxxxx>
> > Reviewed-by: Atish Patra <atish.patra@xxxxxxx>
> > ---
> > arch/riscv/include/asm/irq.h | 2 --
> > drivers/clocksource/timer-riscv.c | 41 ++++++++++++++++++++++++++++---
> > drivers/irqchip/irq-riscv-intc.c | 8 ------
> > 3 files changed, 38 insertions(+), 13 deletions(-)
> >
> > diff --git a/arch/riscv/include/asm/irq.h
> > b/arch/riscv/include/asm/irq.h
> > index a9e5f07a7e9c..9807ad164015 100644
> > --- a/arch/riscv/include/asm/irq.h
> > +++ b/arch/riscv/include/asm/irq.h
> > @@ -10,8 +10,6 @@
> > #include <linux/interrupt.h>
> > #include <linux/linkage.h>
> >
> > -void riscv_timer_interrupt(void);
> > -
> > #include <asm-generic/irq.h>
> >
> > #endif /* _ASM_RISCV_IRQ_H */
> > diff --git a/drivers/clocksource/timer-riscv.c
> > b/drivers/clocksource/timer-riscv.c
> > index c4f15c4068c0..1fe847983f50 100644
> > --- a/drivers/clocksource/timer-riscv.c
> > +++ b/drivers/clocksource/timer-riscv.c
> > @@ -12,8 +12,11 @@
> > #include <linux/cpu.h>
> > #include <linux/delay.h>
> > #include <linux/irq.h>
> > +#include <linux/irqdomain.h>
> > #include <linux/sched_clock.h>
> > #include <linux/io-64-nonatomic-lo-hi.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/of_irq.h>
> > #include <asm/smp.h>
> > #include <asm/sbi.h>
> >
> > @@ -39,6 +42,7 @@ static int riscv_clock_next_event(unsigned long
> > delta,
> > return 0;
> > }
> >
> > +static unsigned int riscv_clock_event_irq;
> > static DEFINE_PER_CPU(struct clock_event_device, riscv_clock_event) =
> > {
> > .name = "riscv_timer_clockevent",
> > .features = CLOCK_EVT_FEAT_ONESHOT,
> > @@ -74,30 +78,36 @@ static int riscv_timer_starting_cpu(unsigned int
> > cpu)
> > struct clock_event_device *ce = per_cpu_ptr(&riscv_clock_event, cpu);
> >
> > ce->cpumask = cpumask_of(cpu);
> > + ce->irq = riscv_clock_event_irq;
> > clockevents_config_and_register(ce, riscv_timebase, 100, 0x7fffffff);
> >
> > - csr_set(CSR_IE, IE_TIE);
> > + enable_percpu_irq(riscv_clock_event_irq,
> > + irq_get_trigger_type(riscv_clock_event_irq));
> > return 0;
> > }
> >
> > static int riscv_timer_dying_cpu(unsigned int cpu)
> > {
> > - csr_clear(CSR_IE, IE_TIE);
> > + disable_percpu_irq(riscv_clock_event_irq);
> > return 0;
> > }
> >
> > /* called directly from the low-level interrupt handler */
> > -void riscv_timer_interrupt(void)
> > +static irqreturn_t riscv_timer_interrupt(int irq, void *dev_id)
> > {
> > struct clock_event_device *evdev = this_cpu_ptr(&riscv_clock_event);
> >
> > csr_clear(CSR_IE, IE_TIE);
> > evdev->event_handler(evdev);
> > +
> > + return IRQ_HANDLED;
> > }
> >
> > static int __init riscv_timer_init_dt(struct device_node *n)
> > {
> > int cpuid, hartid, error;
> > + struct device_node *child;
> > + struct irq_domain *domain;
> >
> > hartid = riscv_of_processor_hartid(n);
> > if (hartid < 0) {
> > @@ -115,6 +125,23 @@ static int __init riscv_timer_init_dt(struct
> > device_node *n)
> > if (cpuid != smp_processor_id())
> > return 0;
> >
> > + domain = NULL;
> > + for_each_child_of_node(n, child) {
> > + domain = irq_find_host(child);
> > + if (domain)
> > + break;
> > + }
>
> This is a bit clumsy, and probably better written as:
>
> child = of_get_compatible(n, "riscv,cpu-intc");
> if (!child) { error out }
> domain = irq_find_host(child);

I thought of not hard-coding INTC compatible string here but
both RISC-V INTC and RISC-V Timer are RISC-V specific so
I guess it's simpler to use of_get_compatible() directly.

I will update.

>
> It would be even better if each CPU would have its per-CPU interrupt
> controller designated as such (with an interrupt-parent property),
> meaning that you wouldn't have to do anything at all.
>
> Too late for that anyway.

Yes, too late now.

Regards,
Anup