We currently allocate the vgic maintenance interrupt by calling into
the low-level irqdomain code. Not only this is unnecessary, but this
is also pretty wrong: we end-up skipping a bunch of irqdesc state
setup

A simple "cat /proc/interrupt" shows how wrong we are, as the
interrupt appears as "Edge" instead of "Level".

Instead, just call the standard irq_create_fwspec_mapping(), which
is the right tool for the job. Duh.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---

Notes:
    This applies to Oliver's kvm-arm64/apple-vgic-mi branch.

 drivers/irqchip/irq-apple-aic.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/irqchip/irq-apple-aic.c b/drivers/irqchip/irq-apple-aic.c
index 09fd52d91e45..76ee7c5e7b7e 100644
--- a/drivers/irqchip/irq-apple-aic.c
+++ b/drivers/irqchip/irq-apple-aic.c
@@ -1201,9 +1201,7 @@ static int __init aic_of_ic_init(struct device_node *node, struct device_node *p
 			},
 		};
 
-		vgic_info.maint_irq = irq_domain_alloc_irqs(irqc->hw_domain,
-							    1, NUMA_NO_NODE,
-							    &mi);
+		vgic_info.maint_irq = irq_create_fwspec_mapping(&mi);
 		WARN_ON(!vgic_info.maint_irq);
 	}
 
-- 
2.34.1

On Wed, 1 Feb 2023 16:40:56 +0000, Marc Zyngier wrote:
> We currently allocate the vgic maintenance interrupt by calling into
> the low-level irqdomain code. Not only this is unnecessary, but this
> is also pretty wrong: we end-up skipping a bunch of irqdesc state
> setup
> 
> A simple "cat /proc/interrupt" shows how wrong we are, as the
> interrupt appears as "Edge" instead of "Level".
> 
> [...]

Applied to kvmarm/next, thanks!

[1/1] irqchip/apple-aic: Correctly map the vgic maintenance interrupt
      https://git.kernel.org/kvmarm/kvmarm/c/ad818e6010ef

--
Best,
Oliver

On 02/02/2023 01.40, Marc Zyngier wrote:
> We currently allocate the vgic maintenance interrupt by calling into
> the low-level irqdomain code. Not only this is unnecessary, but this
> is also pretty wrong: we end-up skipping a bunch of irqdesc state
> setup
> 
> A simple "cat /proc/interrupt" shows how wrong we are, as the
> interrupt appears as "Edge" instead of "Level".
> 
> Instead, just call the standard irq_create_fwspec_mapping(), which
> is the right tool for the job. Duh.
> 
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> ---
> 
> Notes:
>     This applies to Oliver's kvm-arm64/apple-vgic-mi branch.
> 
>  drivers/irqchip/irq-apple-aic.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-apple-aic.c b/drivers/irqchip/irq-apple-aic.c
> index 09fd52d91e45..76ee7c5e7b7e 100644
> --- a/drivers/irqchip/irq-apple-aic.c
> +++ b/drivers/irqchip/irq-apple-aic.c
> @@ -1201,9 +1201,7 @@ static int __init aic_of_ic_init(struct device_node *node, struct device_node *p
>  			},
>  		};
>  
> -		vgic_info.maint_irq = irq_domain_alloc_irqs(irqc->hw_domain,
> -							    1, NUMA_NO_NODE,
> -							    &mi);
> +		vgic_info.maint_irq = irq_create_fwspec_mapping(&mi);
>  		WARN_ON(!vgic_info.maint_irq);
>  	}
>  

Reviewed-by: Hector Martin <marcan@marcan.st>

- Hector