[uracoli-devel] SPI initializaion issue

[Daniel Thiele]

Dear uracoli developers,
during bringing up the muse231 board i just fell over the following issue:
-	When initializing the SPI (ATmega88PA) the SPIF flag is set after  
SPCR = (_BV(SPE) | _BV(MSTR)
-	When executing the first function using SPI, e.g. trx_reg_read() the  
SPIF is still set and accessing SPDR leads to SPIF is cleared and WCOL  
(write collision is set)
-	The following SPI_WAITFOR(); infinitely waits for the SPIF (assuming  
end of SPI transfer) which never occurs

Possible ways to avoid this are:
(1)     Clearing SPIF manually after enabling SPI
(2)     Clearing SPIF (and WCOL) before each SPI access
(3)     Giving timeout to SPI_WAITFOR()

Perhaps this issue only affects ATmega88PA, I cant remember having  
this problems with previous types of MCUs? I would like to discuss the  
way to avoid this. In my opinion, clearing all SPIF and WCOL flags  
after initialization manually is a good approach. But in long terms  
the SPI access functions should be reworked to avoid infinite waiting  
(by software timeouts).

Following the code snippet for SPI initialization (taken from  
boards/muse231.h).

BTW: I would have handled this issue by dropping a bug entry but the  
new savannah password policies make it impossible for me to create a  
valid password (not enough characters, no digits, ...)

Best regards
Daniel


/**
 * @brief inline function for SPI initialization
 */
static inline void SPI_INIT(uint8_t spirate)
{
    /* first configure SPI Port, then SPCR */
    DDR_SPI  |= SPI_MOSI | SPI_SCK | SPI_SS;
    DDR_SPI  &= ~SPI_MISO;
    PORT_SPI |= SPI_SCK | SPI_SS;

    SPCR = (_BV(SPE) | _BV(MSTR));

    SPCR &= ~(_BV(SPR1) | _BV(SPR0) );
    SPSR &= ~_BV(SPI2X);

    SPCR |= (spirate & 0x03);
    SPSR |= ((spirate >> 2) & 0x01);
}