AVR32 SPI clock frequency bug?

[Martin Guy]

I think I've spotted a bug in the calculation of the SPI clock
frequency on AVR32:

diff --git a/src/platform/avr32/platform.c b/src/platform/avr32/platform.c
index 75a2d65..899a19a 100644
--- a/src/platform/avr32/platform.c
+++ b/src/platform/avr32/platform.c
@@ -213,10 +213,10 @@ int platform_init()
spiopt.modfdis = TRUE;
spiopt.pcs_decode = FALSE;
spiopt.delay = 0;
- spi_initMaster(&AVR32_SPI0, &spiopt, REQ_CPU_FREQ);
+ spi_initMaster(&AVR32_SPI0, &spiopt, REQ_PBA_FREQ);

#if NUM_SPI > 4
- spi_initMaster(&AVR32_SPI1, &spiopt, REQ_CPU_FREQ);
+ spi_initMaster(&AVR32_SPI1, &spiopt, REQ_PBA_FREQ);
#endif

#endif

where
int spi_initMaster(volatile avr32_spi_t *spi, const
spi_master_options_t *opt, U32 pba_hz)
{
...

which I expect would really set a frequency of one quarter of that
which was asked for (and reported) since CPU=60000000 and PBA=15000000

but I don't have any SPI hardware to test this, so before I make this
fairly obvious change, which affects all AVR32 boards, has anyone
tried SPI on any of these boards? It seems a fairly obvious error but
I wouldn't want to break something that was working... if it was
working...

M

[Bogdan Marinescu]

Hi,

I didn't try SPI on AVR32, but your fix is obviously right because of the AVR32 system architecture (as described in the datasheet) and because of this line from spi.c:

int spi_initMaster(volatile avr32_spi_t *spi, const spi_master_options_t *opt, U32 pba_hz)

I'd go ahead with it.

Best,

Bogdan

>
>    M
>

[Marcus]

Yes, the SPI should be based on the PBA clock. It is better to correct this even if it breaks something else that work with wrong frequency.

Regards,
Marcus