On 08/16/2012 09:29 PM, Rafał Miłecki wrote:
> 2012/8/16 Arend van Spriel <arend@broadcom.com>:
>> On 08/16/2012 07:39 PM, Rafał Miłecki wrote:
>>>
>>> 2012/8/16 Florian Fainelli<florian@openwrt.org>:
>>>>>
>>>>>>> +void __init bcm47xx_gpio_init(void)
>>>>>>> +{
>>>>>>> + int err;
>>>>>>> +
>>>>>>> + switch (bcm47xx_bus_type) {
>>>>>>> +#ifdef CONFIG_BCM47XX_SSB
>>>>>>> + case BCM47XX_BUS_TYPE_SSB:
>>>>>>> + bcm47xx_gpio_count = ssb_gpio_count(&bcm47xx_bus.ssb);
>>>>>>> +#endif
>>>>>>> +#ifdef CONFIG_BCM47XX_BCMA
>>>>>>> + case BCM47XX_BUS_TYPE_BCMA:
>>>>>>> + bcm47xx_gpio_count =
>>>>>>> bcma_gpio_count(&bcm47xx_bus.bcma.bus);
>>>>>>> +#endif
>>>>>>> + }
>>>>
>>>>>
>>>>> Is this exclusive? Cannot we have both SSB and BCMA on the same device?
>>>
>>> This applies to SoC only, so I believe it's fine. We don't have SoCs
>>> based on BCMA and SSB at the same time.
>>
>>
>> It is indeed more than unlikely for a chip to have two silicon
>> interconnects, which is what SSB and BCMA are. However, it does look
>> suspicious from a code reading perspective. So I general I stick to the rule
>> that each case must have a break and fall-thru are clearly commented.
>
> Ahh, I though question is related to the enum used for bustype. I
> definitely vote for using "break"
I will add the missing break.
Hauke
|