Re: [PATCH v5 03/16] irqchip/riscv-intc: Introduce Andes hart-level interrupt controller

[Anup Patel]

On Wed, Dec 13, 2023 at 9:15 PM Yu-Chien Peter Lin
<peterlin@xxxxxxxxxxxxx> wrote:
>
> On Wed, Dec 13, 2023 at 08:15:28PM +0530, Anup Patel wrote:
> > On Wed, Dec 13, 2023 at 12:35 PM Yu Chien Peter Lin
> > <peterlin@xxxxxxxxxxxxx> wrote:
> > >
> > > Add support for the Andes hart-level interrupt controller. This
> > > controller provides interrupt mask/unmask functions to access the
> > > custom register (SLIE) where the non-standard S-mode local interrupt
> > > enable bits are located.
> > >
> > > To share the riscv_intc_domain_map() with the generic RISC-V INTC and
> > > ACPI, add a chip parameter to riscv_intc_init_common(), so it can be
> > > passed to the irq_domain_set_info() as private data.
> > >
> > > Andes hart-level interrupt controller requires the "andestech,cpu-intc"
> > > compatible string to be present in interrupt-controller of cpu node.
> > > e.g.,
> > >
> > >   cpu0: cpu@0 {
> > >       compatible = "andestech,ax45mp", "riscv";
> > >       ...
> > >       cpu0-intc: interrupt-controller {
> > >           #interrupt-cells = <0x01>;
> > >           compatible = "andestech,cpu-intc", "riscv,cpu-intc";
> > >           interrupt-controller;
> > >       };
> > >   };
> > >
> > > Signed-off-by: Yu Chien Peter Lin <peterlin@xxxxxxxxxxxxx>
> > > Reviewed-by: Charles Ci-Jyun Wu <dminus@xxxxxxxxxxxxx>
> > > Reviewed-by: Leo Yu-Chi Liang <ycliang@xxxxxxxxxxxxx>
> > > ---
> > > Changes v1 -> v2:
> > >   - New patch
> > > Changes v2 -> v3:
> > >   - Return -ENXIO if no valid compatible INTC found
> > >   - Allow falling back to generic RISC-V INTC
> > > Changes v3 -> v4: (Suggested by Thomas [1])
> > >   - Add comment to andes irq chip function
> > >   - Refine code flow to share with generic RISC-V INTC and ACPI
> > >   - Move Andes specific definitions to include/linux/soc/andes/irq.h
> > > Changes v4 -> v5: (Suggested by Thomas)
> > >   - Fix commit message
> > >   - Subtract ANDES_SLI_CAUSE_BASE from d->hwirq to calculate the value of mask
> > >   - Do not set chip_data to the chip itself with irq_domain_set_info()
> > >   - Follow reverse fir tree order variable declarations
> > >
> > > [1] https://patchwork.kernel.org/project/linux-riscv/patch/20231019135723.3657156-1-peterlin@xxxxxxxxxxxxx/
> > > ---
> > >  drivers/irqchip/irq-riscv-intc.c | 53 ++++++++++++++++++++++++++++----
> > >  include/linux/soc/andes/irq.h    | 17 ++++++++++
> > >  2 files changed, 64 insertions(+), 6 deletions(-)
> > >  create mode 100644 include/linux/soc/andes/irq.h
> > >
> > > diff --git a/drivers/irqchip/irq-riscv-intc.c b/drivers/irqchip/irq-riscv-intc.c
> > > index 2fdd40f2a791..0b6bf3fb1dba 100644
> > > --- a/drivers/irqchip/irq-riscv-intc.c
> > > +++ b/drivers/irqchip/irq-riscv-intc.c
> > > @@ -17,6 +17,7 @@
> > >  #include <linux/module.h>
> > >  #include <linux/of.h>
> > >  #include <linux/smp.h>
> > > +#include <linux/soc/andes/irq.h>
> > >
> > >  static struct irq_domain *intc_domain;
> > >
> > > @@ -46,6 +47,31 @@ static void riscv_intc_irq_unmask(struct irq_data *d)
> > >         csr_set(CSR_IE, BIT(d->hwirq));
> > >  }
> > >
> > > +static void andes_intc_irq_mask(struct irq_data *d)
> > > +{
> > > +       /*
> > > +        * Andes specific S-mode local interrupt causes (hwirq)
> > > +        * are defined as (256 + n) and controlled by n-th bit
> > > +        * of SLIE.
> > > +        */
> > > +       unsigned int mask = BIT(d->hwirq - ANDES_SLI_CAUSE_BASE);
> > > +
> > > +       if (d->hwirq < ANDES_SLI_CAUSE_BASE)
> > > +               csr_clear(CSR_IE, mask);
> > > +       else
> > > +               csr_clear(ANDES_CSR_SLIE, mask);
> > > +}
> > > +
> > > +static void andes_intc_irq_unmask(struct irq_data *d)
> > > +{
> > > +       unsigned int mask = BIT(d->hwirq - ANDES_SLI_CAUSE_BASE);
> > > +
> > > +       if (d->hwirq < ANDES_SLI_CAUSE_BASE)
> > > +               csr_set(CSR_IE, mask);
> > > +       else
> > > +               csr_set(ANDES_CSR_SLIE, mask);
> >
> > Clearly, Andes does not have any CSR for:
> > XLEN <= local interrupt <ANDES_SLI_CAUSE_BASE
> > and
> > ANDES_SLI_CAUSE_BASE + XLEN <= local interrupt
>
> Ah, what am I doing here.
> sorry for that silly patch.

This patch is okay only if we can guarantee that
hwirq is within accepted range.

For example, riscv_intc_domain_alloc() can deny
invalid local interrupts.

Regards,
Anup