Posted on July 15, 2016 at 14:35Hi All,
I'm using STM32L475, STM32CubeMX 4.15.1, STM32L4 firmware library 1.5.1 and the code generated by STM32CubeMX. When initializing ADC1 and ADC2 the ADC2 initialization, MX_ADC2_Init(), sometimes calls ErrorHandler() because HAL_ADCEx_MultiModeConfigChannel() is returning and error.
HAL_ADCEx_MultiModeConfigChannel(), stm32l4xx_hal_adc_ex.c, is called with hadc->Instance = ADC2; therefore, ADC_MULTI_SLAVE(hadc, &tmphadcSlave) sets tmphadcSlave->Instance = NULL. When ADC_IS_CONVERSION_ONGOING_REGULAR(&tmphadcSlave) is called a NULL pointer, tmphadcSlave->Instance->CR, is dereferenced and leads to unexpected results.
The NULL pointer deference can be fixed by changing macro ADC_IS_CONVERSION_ONGOING_REGULAR, stm32l4xx_hal_adc.h, to
#define ADC_IS_CONVERSION_ONGOING_REGULAR(__HANDLE__) \
(( !(__HANDLE__)->Instance || (((__HANDLE__)->Instance->CR) & ADC_CR_ADSTART) == RESET \
) ? RESET : SET)
With this change ADC_IS_CONVERSION_ONGOING_REGULAR will return RESET when Instance is NULL.
HAL_ADCEx_MultiModeConfigChannel() also calls macro ADC_IS_ENABLE(&tmphadcSlave). ADC_IS_ENABLE should also check for NULL before using tmphadcSlave->Instance.
#define ADC_IS_ENABLE(__HANDLE__) \
(( (__HANDLE__)->Instance && \
((((__HANDLE__)->Instance->CR) & (ADC_CR_ADEN | ADC_CR_ADDIS)) == ADC_CR_ADEN) && \
((((__HANDLE__)->Instance->ISR) & ADC_FLAG_RDY) == ADC_FLAG_RDY) \
) ? SET : RESET)
Is this analysis and resolution correct? I believe returning RESET is appropriate when Instance = NULL. Also, updating the macros seemed cleaner than adding NULL checks to the source; but this is debatable.
Thanks and Regards,
...doug
