Re: [PATCH 3/3] hwmon: lenovo-ec-sensors: Use devm_request_region for automatic cleanup

[Mark Pearson]

Hi Guenter,

On Wed, May 13, 2026, at 9:36 PM, Guenter Roeck wrote:
> On 5/13/26 18:14, Kean wrote:
>> Replace manual request_region()/release_region() with
>> devm_request_region(). This lets the device-managed framework
>> handle I/O region lifetime automatically and fixes:
>> 
>> - A double release_region() when probe fails after acquiring the
>>    I/O region: the probe error path releases it, and then
>>    lenovo_ec_init() releases it again on the same error path.
>> 
>> - A release-after-free in lenovo_ec_exit() where release_region()
>>    was called after platform_device_unregister(), which has already
>>    released the I/O region via the platform device removal path.
>> 
>> - Missing release_region() in lenovo_ec_probe() on the DMI match
>>    failure path, which leaked the I/O region.
>> 
>> Remove all manual release_region() calls that are now handled
>> automatically by the devm framework.
>> 
>> Signed-off-by: Kean <rh_king@xxxxxxx>
>> 
> Why this empty line ?
>
>> Reviewed-by: Mark Pearson <mpearson-lenovo@xxxxxxxxx>
>
> How do I know that this review really happened ?

The reviewed by's are correct - I can confirm I did review these patches before Kean pushed them upstream. We did some internal review first to discuss the issues he identified.

We sometimes take this approach with the platform/x86 patches. We can do it separately next time if preferred here.

Mark

>
>> ---
>>   drivers/hwmon/lenovo-ec-sensors.c | 13 +++++--------
>>   1 file changed, 5 insertions(+), 8 deletions(-)
>> 
>> diff --git a/drivers/hwmon/lenovo-ec-sensors.c b/drivers/hwmon/lenovo-ec-sensors.c
>> index b0f2a04ce679..ea74bddbad5a 100644
>> --- a/drivers/hwmon/lenovo-ec-sensors.c
>> +++ b/drivers/hwmon/lenovo-ec-sensors.c
>> @@ -519,8 +519,8 @@ static int lenovo_ec_probe(struct platform_device *pdev)
>>   	if (!ec_data)
>>   		return -ENOMEM;
>>   
>> -	if (!request_region(IO_REGION_START, IO_REGION_LENGTH, "LNV-WKS")) {
>> -		pr_err(":request fail\n");
>> +	if (!devm_request_region(dev, IO_REGION_START, IO_REGION_LENGTH, "LNV-WKS")) {
>> +		dev_err(dev, "Failed to request I/O region.\n");
>>   		return -EIO;
>>   	}
>>   
>> @@ -541,7 +541,6 @@ static int lenovo_ec_probe(struct platform_device *pdev)
>>   	    (inb_p(MCHP_EMI0_EC_DATA_BYTE1) != 'C') ||
>>   	    (inb_p(MCHP_EMI0_EC_DATA_BYTE2) != 'H') ||
>>   	    (inb_p(MCHP_EMI0_EC_DATA_BYTE3) != 'P')) {
>> -		release_region(IO_REGION_START, IO_REGION_LENGTH);
>>   		return -ENODEV;
>>   	}
>>   
>> @@ -579,7 +578,8 @@ static int lenovo_ec_probe(struct platform_device *pdev)
>>   		lenovo_ec_chip_info.info = lenovo_ec_hwmon_info_p8;
>>   		break;
>>   	default:
>> -		release_region(IO_REGION_START, IO_REGION_LENGTH);
>> +		dev_err(dev, "Unsupported platform type %ld\n",
>> +			(long)dmi_id->driver_data);
>
> This is not documented in the commit message and, on top of that, pointless.
> It isn't even noise, it is just pointless (the default case can not be reached).
>
>>   		return -ENODEV;
>>   	}
>>   
>> @@ -608,10 +608,8 @@ static int __init lenovo_ec_init(void)
>>   		platform_create_bundle(&lenovo_ec_sensors_platform_driver,
>>   				       lenovo_ec_probe, NULL, 0, NULL, 0);
>>   
>> -	if (IS_ERR(lenovo_ec_sensors_platform_device)) {
>> -		release_region(IO_REGION_START, IO_REGION_LENGTH);
>> +	if (IS_ERR(lenovo_ec_sensors_platform_device))
>>   		return PTR_ERR(lenovo_ec_sensors_platform_device);
>> -	}
>>   
>>   	return 0;
>>   }
>> @@ -619,7 +617,6 @@ module_init(lenovo_ec_init);
>>   
>>   static void __exit lenovo_ec_exit(void)
>>   {
>> -	release_region(IO_REGION_START, IO_REGION_LENGTH);
>>   	platform_device_unregister(lenovo_ec_sensors_platform_device);
>>   	platform_driver_unregister(&lenovo_ec_sensors_platform_driver);
>>   }