On Monday, March 21, 2016 8:39 AM, Ian Abbott wrote:
> On 18/03/16 19:38, H Hartley Sweeten wrote:
>> This driver has a lot of checkpatch.pl issues:
>> total: 0 errors, 71 warnings, 27 checks, 1222 lines checked
>>
>> There is also a lot of cruft that bloats the driver and makes it harder
>> to follow.
>>
>> This series fixes all the checkpatch.pl issues:
>> total: 0 errors, 0 warnings, 0 checks, 944 lines checked
>>
>> And, even better, removes the cruft:
>>
>> text data bss dec hex filename
>> 10249 344 0 10593 2961 ni_660x.o.original
>> 4719 344 0 5063 13c7 ni_660x.o
>>
<snip>
> It all looks fine, apart from some breakage in patches 30 and 31 related
> to the calls to ni_gpct_device_destroy().
Gah... Your correct.
I think ni_gpct_device_destroy() should be changed to:
void ni_gpct_device_destroy(struct ni_gpct_device *counter_dev)
{
- if (!counter_dev->counters)
+ if (!counter_dev)
return;
kfree(counter_dev->counters);
kfree(counter_dev);
}
Then the NULL checks can be removed from the drivers.
> I'm not entirely sure what the memory barrier moved by patch 27 is
> for - it's probably redundant.
I'm not sure either. But having is after the spin_lock_irqsave() didn't
make any sense.
ni_mio_common.c: ni_E_interrupt() has similar smp_mb() with a
comment so I made it the same here.
If you don't think they are necessary I don't have any issues with
removing them.
Hartley
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
