[bug report] hwmon: (core) New hwmon registration API

[bug report] hwmon: (core) New hwmon registration API

[Dan Carpenter]

Hello Guenter Roeck,

This is a semi-automatic email about new static checker warnings.

The patch d560168b5d0f: "hwmon: (core) New hwmon registration API" 
from Aug 26, 2015, leads to the following Smatch complaint:

drivers/hwmon/hwmon.c:624 __hwmon_device_register()
	 error: we previously assumed 'dev' could be null (see line 602)

drivers/hwmon/hwmon.c
   601		hdev->parent = dev;
   602		hdev->of_node = dev ? dev->of_node : NULL;
                                ^^^
I gotta be honest, I have no idea why my scripts say this is a new
warning when it's 2 years old.  This check feels like it could be
deleted and perhaps that's why it didn't complain up to now.  It tries
to ignore extra harmless checks.

   603		hwdev->chip = chip;
   604		dev_set_drvdata(hdev, drvdata);
   605		dev_set_name(hdev, HWMON_ID_FORMAT, id);
   606		err = device_register(hdev);
   607		if (err)
   608			goto free_hwmon;
   609	
   610		if (chip && chip->ops->read &&
   611		    chip->info[0]->type == hwmon_chip &&
   612		    (chip->info[0]->config[0] & HWMON_C_REGISTER_TZ)) {
   613			const struct hwmon_channel_info **info = chip->info;
   614	
   615			for (i = 1; info[i]; i++) {
   616				if (info[i]->type != hwmon_temp)
   617					continue;
   618	
   619				for (j = 0; info[i]->config[j]; j++) {
   620					if (!chip->ops->is_visible(drvdata, hwmon_temp,
   621								   hwmon_temp_input, j))
   622						continue;
   623					if (info[i]->config[j] & HWMON_T_INPUT)
   624						hwmon_thermal_add_sensor(dev, hwdev, j);
                                                                         ^^^
Dereference.

   625				}
   626			}

regards,
dan carpenter

Re: [bug report] hwmon: (core) New hwmon registration API

[Guenter Roeck]

On 01/31/2017 02:49 AM, Dan Carpenter wrote:
> Hello Guenter Roeck,
>
> This is a semi-automatic email about new static checker warnings.
>
> The patch d560168b5d0f: "hwmon: (core) New hwmon registration API"
> from Aug 26, 2015, leads to the following Smatch complaint:
>
> drivers/hwmon/hwmon.c:624 __hwmon_device_register()
> 	 error: we previously assumed 'dev' could be null (see line 602)
>
> drivers/hwmon/hwmon.c
>    601		hdev->parent = dev;
>    602		hdev->of_node = dev ? dev->of_node : NULL;
>                                 ^^^
> I gotta be honest, I have no idea why my scripts say this is a new
> warning when it's 2 years old.  This check feels like it could be
> deleted and perhaps that's why it didn't complain up to now.  It tries
> to ignore extra harmless checks.
>

Nevertheless it is a good warning. hwmon_thermal_add_sensor() calls devm_kzalloc()
with dev as argument.

Guenter

>    603		hwdev->chip = chip;
>    604		dev_set_drvdata(hdev, drvdata);
>    605		dev_set_name(hdev, HWMON_ID_FORMAT, id);
>    606		err = device_register(hdev);
>    607		if (err)
>    608			goto free_hwmon;
>    609	
>    610		if (chip && chip->ops->read &&
>    611		    chip->info[0]->type == hwmon_chip &&
>    612		    (chip->info[0]->config[0] & HWMON_C_REGISTER_TZ)) {
>    613			const struct hwmon_channel_info **info = chip->info;
>    614	
>    615			for (i = 1; info[i]; i++) {
>    616				if (info[i]->type != hwmon_temp)
>    617					continue;
>    618	
>    619				for (j = 0; info[i]->config[j]; j++) {
>    620					if (!chip->ops->is_visible(drvdata, hwmon_temp,
>    621								   hwmon_temp_input, j))
>    622						continue;
>    623					if (info[i]->config[j] & HWMON_T_INPUT)
>    624						hwmon_thermal_add_sensor(dev, hwdev, j);
>                                                                          ^^^
> Dereference.
>
>    625				}
>    626			}
>
> regards,
> dan carpenter
>