Re: [PATCH 1/3] mfd: tps6586x: add version detection

[Stefan Agner <stefan@agner.ch>]

Am 2013-11-27 14:09, schrieb Lee Jones:
> <snip>
> 
>>  static int __remove_subdev(struct device *dev, void *unused)
>>  {
>>  	platform_device_unregister(to_platform_device(dev));
>> @@ -477,6 +486,7 @@ static int tps6586x_i2c_probe(struct i2c_client *client,
>>  {
>>  	struct tps6586x_platform_data *pdata = dev_get_platdata(&client->dev);
>>  	struct tps6586x *tps6586x;
>> +	const char *name = "";
>
> How much memory space does this guarantee?
>
Well, its just an empty string (so, 1). Actually I could assign NULL
here, since we do have a default in the case later on.

> 
>>  	int ret;
>>
>>  	if (!pdata && client->dev.of_node)
>> @@ -487,20 +497,39 @@ static int tps6586x_i2c_probe(struct i2c_client *client,
>>  		return -ENOTSUPP;
>>  	}
>>
>> +	tps6586x = devm_kzalloc(&client->dev, sizeof(*tps6586x), GFP_KERNEL);
>> +	if (tps6586x == NULL) {
> 
> Can you take this oppotunity to change to:
>   if (!tps6586x)
> 
>> +		dev_err(&client->dev, "memory for tps6586x alloc failed\n");
> 
> I'm not keen on -ENOMEM prints, just return -ENOMEM and be done with it.
> 
>> +		return -ENOMEM;
>> +	}
>> +
>>  	ret = i2c_smbus_read_byte_data(client, TPS6586X_VERSIONCRC);
> 
> If you're going to do this, please change 'ret' to 'version'.
> 
>>  	if (ret < 0) {
>>  		dev_err(&client->dev, "Chip ID read failed: %d\n", ret);
>>  		return -EIO;
> 
> Why are you returning an error here when you have a valid enum of:
>   TPS6586X_ANY	= -1,
> 
Hm, when the device is not answering on that request, the probe method
should fail I would say. This means that the device is missing most
likely. However, I should set the device version to TPS6586X_ANY if I
happen to end up in the default case.

>>  	}
>>
>> -	dev_info(&client->dev, "VERSIONCRC is %02x\n", ret);
>> -
>> -	tps6586x = devm_kzalloc(&client->dev, sizeof(*tps6586x), GFP_KERNEL);
>> -	if (tps6586x == NULL) {
>> -		dev_err(&client->dev, "memory for tps6586x alloc failed\n");
>> -		return -ENOMEM;
>> +	tps6586x->version = (enum tps6586x_version)ret;
>> +	switch (ret) {
>> +	case TPS658621A:
>> +		name = "TPS658621A";
>> +		break;
>> +	case TPS658621CD:
>> +		name = "TPS658621C/D";
>> +		break;
>> +	case TPS658623:
>> +		name = "TPS658623";
>> +		break;
>> +	case TPS658643:
>> +		name = "TPS658643";
>> +		break;
>> +	default:
>> +		name = "TPS6586X";
>> +		break;
>>  	}
>>
>> +	dev_info(&client->dev, "Found %s, VERSIONCRC is %02x\n", name, ret);
>> +
> 
> I'd suggest pulling this out of probe() and into a separate subroutine.
> 
> <snip>
Since I will alter the version when I end up in the default case in my
next patch, would you still do a separate subroutine? I think its
somewhat heavily coupled to the probe function.

Sorry missing you on the CC, will do next time :-)

--
Stefan