Re: [RFC PATCH 2/4] irqchip/riscv-intc: Support large non-standard hwirq number

["Clément Léger" <cleger@rivosinc.com>]

On 07/09/2023 04:16, Yu Chien Peter Lin wrote:
> Currently, the implementation of the RISC-V INTC driver uses the
> interrupt cause as hwirq and has a limitation of supporting a
> maximum of 64 hwirqs. However, according to the privileged spec,
> interrupt cause >= 16 are defined for platform use.
> 
> This limitation prevents us from fully utilizing the available
> local interrupt sources. Additionally, the hwirqs used on RISC-V
> are sparse, with only interrupt numbers 1, 5 and 9 (plus Sscofpmf
> or T-Head's PMU irq) being currently used for supervisor mode.
> 
> The patch switches to using irq_domain_create_tree() which
> creates the radix tree map, allowing us to handle a larger
> number of hwirqs.
> 
> Signed-off-by: Yu Chien Peter Lin <peterlin@andestech.com>
> Reviewed-by: Charles Ci-Jyun Wu <dminus@andestech.com>
> Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
> 
> ---
> There are 3 hwirqs of local interrupt source exceed 64 defined in
> AX45MP datasheet [1] Table 56: AX45MP-1C scause Value After Trap:
> - 256+16 Slave port ECC error interrupt (S-mode)
> - 256+17 Bus write transaction error interrupt (S-mode)
> - 256+18 Performance monitor overflow interrupt(S-mode)
> 
> [1] http://www.andestech.com/wp-content/uploads/AX45MP-1C-Rev.-5.0.0-Datasheet.pdf
> ---
>  drivers/irqchip/irq-riscv-intc.c | 10 ++++------
>  1 file changed, 4 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-riscv-intc.c b/drivers/irqchip/irq-riscv-intc.c
> index 4adeee1bc391..76e1229c45de 100644
> --- a/drivers/irqchip/irq-riscv-intc.c
> +++ b/drivers/irqchip/irq-riscv-intc.c
> @@ -24,8 +24,8 @@ static asmlinkage void riscv_intc_irq(struct pt_regs *regs)
>  {
>  	unsigned long cause = regs->cause & ~CAUSE_IRQ_FLAG;
>  
> -	if (unlikely(cause >= BITS_PER_LONG))
> -		panic("unexpected interrupt cause");
> +	if (!irq_find_mapping(intc_domain, cause))
> +		panic("unexpected interrupt cause: %ld", cause);

Hi Yu,

It seems like generic_handle_domain_irq() returns -EINVAL if provided
with NULL (which will happen if the interrupt does not have a mapping
due to __irq_resolve_mapping returning NULL) so maybe it is possible to
remove this check (since __irq_resolve_mapping() is also called by
generic_handle_domain_irq()) and panic if generic_handle_domain_irq()
returns -EINVAL ? That would avoid calling __irq_resolve_mapping() twice
for really rare cases.

Clément

>  
>  	generic_handle_domain_irq(intc_domain, cause);
>  }
> @@ -117,8 +117,8 @@ static int __init riscv_intc_init_common(struct fwnode_handle *fn)
>  {
>  	int rc;
>  
> -	intc_domain = irq_domain_create_linear(fn, BITS_PER_LONG,
> -					       &riscv_intc_domain_ops, NULL);
> +	intc_domain = irq_domain_create_tree(fn, &riscv_intc_domain_ops,
> +					     NULL);
>  	if (!intc_domain) {
>  		pr_err("unable to add IRQ domain\n");
>  		return -ENXIO;
> @@ -132,8 +132,6 @@ static int __init riscv_intc_init_common(struct fwnode_handle *fn)
>  
>  	riscv_set_intc_hwnode_fn(riscv_intc_hwnode);
>  
> -	pr_info("%d local interrupts mapped\n", BITS_PER_LONG);
> -
>  	return 0;
>  }
>