Re: [PATCH] irqchip/aslint-sswi: request IO memory resource

[Thomas Gleixner <tglx@kernel.org>]

On Fri, Jan 16 2026 at 04:42, Chris Mason wrote:
> On Tue,  9 Dec 2025 16:23:33 +0200 Vladimir Kondratiev <vladimir.kondratiev@mobileye.com> wrote:
>
>> Make an aclint_sswi instance visible in the resource list,
>> i.e. /proc/iomem
>> 
>
> Hi everyone,
>
> I ran tip master through my AI review prompts, and this one was flagged.
> This is a lot of words for NULL vs ERR_PTR, but it looks right to me:
>
>> diff --git a/drivers/irqchip/irq-aclint-sswi.c b/drivers/irqchip/irq-aclint-sswi.c
>> index fee30f3bc5ac2..325501f34a9bc 100644
>> --- a/drivers/irqchip/irq-aclint-sswi.c
>> +++ b/drivers/irqchip/irq-aclint-sswi.c
>> @@ -109,7 +109,7 @@ static int __init aclint_sswi_probe(struct fwnode_handle *fwnode)
>>  	if (!is_of_node(fwnode))
>>  		return -EINVAL;
>>
>> -	reg = of_iomap(to_of_node(fwnode), 0);
>> +	reg = of_io_request_and_map(to_of_node(fwnode), 0, NULL);
>>  	if (!reg)
>>  		return -ENOMEM;
>
> Should this error check be updated? of_iomap() returns NULL on failure,
> but of_io_request_and_map() returns IOMEM_ERR_PTR() on failure which is
> non-NULL.
>
> When of_io_request_and_map() fails, the condition (!reg) will be false
> and the function continues with an invalid ERR_PTR value. This gets passed
> to aclint_sswi_parse_irq() which stores (reg + offset) into per_cpu
> sswi_cpu_regs, leading to a crash when IPIs are later sent.
>
> Other callers of of_io_request_and_map() in drivers/irqchip/ use IS_ERR()
> for the error check. For example, sun6i_r_intc_init() in irq-sun6i-r.c:
>
>     base = of_io_request_and_map(node, 0, NULL);
>     if (IS_ERR(base)) {
>         pr_err("%pOF: Failed to map MMIO region\n", node);
>         return PTR_ERR(base);
>     }

That's correct. Does anyone care to send a patch?

Thanks,

        tglx