Re: [PATCH v3] i2c: QUP based bus driver for Qualcomm MSM chipsets

[Ben Dooks <ben-linux@fluff.org>]

On 09/09/10 20:01, Sundar wrote:
> Hi Kenneth,
> 
> just some error codes and leaks if I am right :)
> 
> On Wed, Sep 8, 2010 at 6:29 AM, Kenneth Heitke <kheitke@codeaurora.org> wrote:
>> +
>> +static int __devinit
>> +qup_i2c_probe(struct platform_device *pdev)
>> +{
>> +
>> +       qup_mem = platform_get_resource_byname(pdev, IORESOURCE_MEM,
>> +                                               "qup_phys_addr");
>> +       if (!qup_mem) {
>> +               dev_err(&pdev->dev, "no qup mem resource?\n");
>> +               return -ENODEV;
> 
> I think this should be -ENXIO instead of -ENODEV?

I think both are inappropriate here, -ENXIO is an IO failure
and -ENODEV is 'device is not here'. However, some people don't
like the use of things like -ENOENT (filesystem errors) for this
sort of situation. ENOTSUP might be something you could use.

As a note, both ENXIO and ENODEV will be taken by the device
framework as a device is not there at-all and not display any
error to the user (which is annoying if you then forget to
do anything about it)

As a note, you may want to have a look at devres to offload the
tracking of the requested and mapped resources.