Re: [PATCH] ioctl conversion

["Stoyan Gaydarov" <stoyboyker@gmail.com>]

On Fri, Jul 11, 2008 at 3:08 AM, Arnd Bergmann <arnd@arndb.de> wrote:
> On Friday 11 July 2008, Stoyan Gaydarov wrote:
>> -       if (_IOC_TYPE(cmd) != SNAPSHOT_IOC_MAGIC)
>> +       lock_kernel();
>> +
>> +       if (_IOC_TYPE(cmd) != SNAPSHOT_IOC_MAGIC) {
>> +               unlock_kernel();
>>                 return -ENOTTY;
>> -       if (_IOC_NR(cmd) > SNAPSHOT_IOC_MAXNR)
>> +       }
>> +       if (_IOC_NR(cmd) > SNAPSHOT_IOC_MAXNR) {
>> +               unlock_kernel();
>>                 return -ENOTTY;
>> -       if (!capable(CAP_SYS_ADMIN))
>> +       }
>> +       if (!capable(CAP_SYS_ADMIN)) {
>> +               unlock_kernel();
>>                 return -EPERM;
>> +       }
>>
>>         data = filp->private_data;
>
> The more common way to express this is to end the function with
>
> out:
>        unlock_kernel();
>        return ret;
> }

This would work normally but there are three early returns and the
rest are just switch statements that set the return code.

>
> and then jump to that label in the error case. This makes it
> much easier to verify that you haven't missed a cased.
>
>        Arnd <><
>