Re: [PATCH 01/10 v4] Input: ep93xx_keypad: Fix platform_get_irq's error checking

[Dmitry Torokhov]

On Mon, Nov 20, 2017 at 8:29 AM, Russell King - ARM Linux
<linux@xxxxxxxxxxxxxxx> wrote:
> On Mon, Nov 20, 2017 at 09:56:21PM +0530, Arvind Yadav wrote:
>> The platform_get_irq() function returns negative if an error occurs.
>> zero or positive number on success. platform_get_irq() error checking
>> for zero is not correct.
>>
>> Signed-off-by: Arvind Yadav <arvind.yadav.cs@xxxxxxxxx>
>> ---
>> changes in v2 :
>>               Return keypad->irq insted of -ENXIO.
>> changes in v3 :
>>               Add failure case '<= 0' instead of '< 0'. IRQ0 is not valid.
>> changes ib v4 :
>>               Return -ENXIO insted of keypad->irq. Which was not correct in v3.
>>
>>  drivers/input/keyboard/ep93xx_keypad.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/input/keyboard/ep93xx_keypad.c b/drivers/input/keyboard/ep93xx_keypad.c
>> index f77b295..c6eb31a 100644
>> --- a/drivers/input/keyboard/ep93xx_keypad.c
>> +++ b/drivers/input/keyboard/ep93xx_keypad.c
>> @@ -257,7 +257,7 @@ static int ep93xx_keypad_probe(struct platform_device *pdev)
>>       }
>>
>>       keypad->irq = platform_get_irq(pdev, 0);
>> -     if (!keypad->irq) {
>> +     if (keypad->irq <= 0) {
>>               err = -ENXIO;
>
> Still not right.
>
>                 err = keypad->irq < 0 ? keypad->irq : -ENXIO;
>
> would have the right effect.
>
> However, I disagree with the idea of making platform_get_irq() return
> an error if there is a zero irq specified as suggested in a previous
> review - zero means "not present" and that is not necessarily an error -
> it's up to the driver to decide whether it can proceed without an
> interrupt, or whether it should error out on the probe function.

We used to return 0 from platform_get_irq() in early 2.6 days on
errors but not anymore and I do not think keeping treating 0 there as
special makes much sense.

We already return -ENXIO for missing resources, and I do not think
that anyone actually defines IRQ resource with 0 number and attaches
it to devices in home that drivers would recognize it as missing
interrupt and work without it. Also, in case of DT we do not recognize
IRQ 0 as valid and try to fetch it from other sources (static board
data), I think ACPI is similar.

I think we should also return -ENXIO in case of res->start == 0 and
leave it to the driver to check and decide if it wants to deal with
missing interrupt. Most of them do not.


Thanks.

-- 
Dmitry