Re: [PATCH/RFC] soc: bcm2835: Make !RASPBERRYPI_FIRMWARE dummies return failure

[Phil Elwell]

On 08/04/2018 12:28, Stefan Wahren wrote:
> Hi Geert,
> 
> [add Phil]
> 
>> Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> hat am 8. April 2018 um 11:05 geschrieben:
>>
>>
>> If CONFIG_RASPBERRYPI_FIRMWARE=n:
>>
>>     drivers/gpio/gpio-raspberrypi-exp.c: In function ârpi_exp_gpio_get_polarityâ:
>>     drivers/gpio/gpio-raspberrypi-exp.c:71: warning: âget.polarityâ is used uninitialized in this function
>>     drivers/gpio/gpio-raspberrypi-exp.c: In function ârpi_exp_gpio_get_directionâ:
>>     drivers/gpio/gpio-raspberrypi-exp.c:150: warning: âget.directionâ is used uninitialized in this function
>>
>> The dummy firmware interface functions return 0, which means success,
>> causing subsequent code to make use of the never initialized output
>> parameter.
> 
> i think this is more theoritical, because we need to get rpi_firmware first and in this case it's NULL.
> 
>>
>> Fix this by making the dummy functions return an error code (-ENOSYS)
>> instead.
> 
> Anyway i'm okay with the change.
> 
>>
>> Note that this assumes the firmware always fills in the requested data
>> in the CONFIG_RASPBERRYPI_FIRMWARE=y case.
> 
> Unfortunately i don't know.

The firmware overwrites the request code with an error return value. If this appears
and has an error value of zero (RPI_FIRMWARE_STATUS_SUCCESS) then the firmware has
written any expected fields.

> 
>>
>> Fixes: d45f1a563b92dac7 ("staging: vc04_services: fix up rpi firmware functions")
>> Signed-off-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
>> ---
>> Should get.polarity resp. get.direction be preinitialized instead,
>> like is done for get.state?
>>
>> static int rpi_exp_gpio_get(struct gpio_chip *gc, unsigned int off)
>> {
>>         ...
>>         get.state = 0;          /* storage for returned value */
>>
>>         ret = rpi_firmware_property(gpio->fw, RPI_FIRMWARE_GET_GPIO_STATE,
>>                                          &get, sizeof(get));
>>         ...
>> }
>>
>> This would avoid returning uninitialized data if the firmware did not
>> fill in the requested data.
>> ---
>>  include/soc/bcm2835/raspberrypi-firmware.h | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/soc/bcm2835/raspberrypi-firmware.h b/include/soc/bcm2835/raspberrypi-firmware.h
>> index 50df5b28d2c9df6e..8ee8991aa099af3a 100644
>> --- a/include/soc/bcm2835/raspberrypi-firmware.h
>> +++ b/include/soc/bcm2835/raspberrypi-firmware.h
>> @@ -143,13 +143,13 @@ struct rpi_firmware *rpi_firmware_get(struct device_node *firmware_node);
>>  static inline int rpi_firmware_property(struct rpi_firmware *fw, u32 tag,
>>  					void *data, size_t len)
>>  {
>> -	return 0;
>> +	return -ENOSYS;
>>  }
>>  
>>  static inline int rpi_firmware_property_list(struct rpi_firmware *fw,
>>  					     void *data, size_t tag_size)
>>  {
>> -	return 0;
>> +	return -ENOSYS;
>>  }
>>  
>>  static inline struct rpi_firmware *rpi_firmware_get(struct device_node *firmware_node)
>> -- 
>> 2.7.4
>>
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel