Re: [PATCH] genirq: Discard cpumask of irqaffinity= boot parameter if parsing fails

[Radu Rendec]

On Sat, 2026-08-22 at 18:05 +0200, Lukas Beckmann wrote:
> irq_affinity_setup() ignores the return value of cpulist_parse(). This
> can result in using a "half parsed" cpulist. E.g. on a system with 8
> cores, the parameter "irqaffinity=0,1,6-10,4" would result in using only
> CPU 0 and 1, since the parsing fails at 6-10. Other boot parameters
> which take a cpulist (e.g. isolcpus=, workqueue.unbound_cpus=) already
> handle this by falling back to a default.
> 
> Check the return value of cpulist_parse() and let
> init_irq_default_affinity() set the default in early_irq_init()
> later, if parsing fails.
> 
> Signed-off-by: Lukas Beckmann <lbckmnn@xxxxxxxxxxx>
> ---
>  kernel/irq/irqdesc.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
> index 3a818f07a101..7cbcc3fb9e6d 100644
> --- a/kernel/irq/irqdesc.c
> +++ b/kernel/irq/irqdesc.c
> @@ -28,7 +28,15 @@ static struct lock_class_key irq_desc_lock_class;
>  static int __init irq_affinity_setup(char *str)
>  {
>  	alloc_bootmem_cpumask_var(&irq_default_affinity);
> -	cpulist_parse(str, irq_default_affinity);
> +	if (cpulist_parse(str, irq_default_affinity) < 0) {
> +		/*
> +		 * Clear the mask so that init_irq_default_affinity()
> +		 * defaults it later.
> +		 */
> +		cpumask_clear(irq_default_affinity);
> +		pr_warn("irqaffinity: incorrect CPU range, using default\n");
> +		return 1;
> +	}
>  	/*
>  	 * Set at least the boot cpu. We don't want to end up with
>  	 * bugreports caused by random commandline masks
> 
> base-commit: 1ad6d4a722f509128038f9c85c6a122a382179fe

Reviewed-by: Radu Rendec <radu@xxxxxxxxxx>