Re: [PATCH 6/6] mfd: qcom-pm8008: Tolerate missing interrupt

[Jishnu Prakash <jishnu.prakash@oss.qualcomm.com>]

On 9/18/2026 10:27 PM, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Medium] mfd: qcom-pm8008: bogus IRQ mappings for child cells when client->irq is 0
> --
> 
> commit 2c273914596163e5696becde808ceb33d98f0fe8
> Author: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> 
> mfd: qcom-pm8008: Tolerate missing interrupt
> 

...

> [ ... ]
> 
>> -	/* Needed by GPIO driver. */
>> -	dev_set_drvdata(dev, regmap_irq_get_domain(irq_data));
>> +		/* Needed by GPIO driver. */
>> +		dev_set_drvdata(dev, regmap_irq_get_domain(irq_data));
>> +	}
>>  
>>  	return devm_mfd_add_devices(dev, PLATFORM_DEVID_AUTO, data->mfd_cells,
>> -				data->num_mfd_cells, NULL, 0,
>> -				regmap_irq_get_domain(irq_data));
>> +				    data->num_mfd_cells, NULL, 0,
>> +				    regmap_irq_get_domain(irq_data));
> 
> [Severity: Medium]
> Will this lead to bogus IRQ mappings for child cells when client->irq is 0?
> 
> When the device has no IRQ, irq_data remains NULL, and we pass a NULL domain
> to devm_mfd_add_devices() with an irq_base of 0.
> 
> The MFD cells (such as pm8008_temp_res for qpnp-temp-alarm) specify IRQ
> resources using DEFINE_RES_IRQ() with hardcoded offsets, for example:
> 
> static const struct resource pm8008_temp_res[] = {
> 	DEFINE_RES_MEM(PM8008_TEMP_ALARM_ADDR, 0x100),
> 	DEFINE_RES_IRQ(PM8008_IRQ_TEMP_ALARM),
> };
> 
> When mfd_add_device() processes these IRQ resources and the domain is NULL,
> it falls back to assigning the IRQ as irq_base + start. Since irq_base is 0,
> a child device like the temp alarm will be assigned a global physical Linux
> IRQ matching its offset (e.g., 0 + 5 = 5).
> 
> If a child driver requests this IRQ, could it steal an interrupt from another
> hardware device on the system, leading to unexpected behavior or resource
> conflicts? Should the driver dynamically strip or adjust the IRQ resources
> from the MFD cells when no parent IRQ is present?

This sounds right, I'll make this fix in the next version of this series.

Thanks,
Jishnu

>