USART_DMACmd test

[ledvinap]

Can you test this code? its USART_DMACmd + minor code cleanup ...

[jflyper]

First commit: I still get lots of underruns, and configurator is complaining about various errors once again.

Trying the second one...

Ah, I already had this removed with the 1st commit. Same symptom.

[ledvinap]

And the error is FEIF only ? or are other flags set too ?

[jflyper]

Ah, I have to re-install the debug counter again to check other errors.

But there's something else that is very interesting.

I added the reenable code again, and guess what, I only see occasional FEIF then.

I will push the branch separately.

[ledvinap]

Using pins for trace (+saleae) may be helpful ... it may reveal when the error occurs ..

(https://github.com/ledvinap/cleanflight/blob/local/src/main/drivers/pin_debug.h, but that ais a bit of overkill)

[ledvinap]

'reenable code' ? What do you mean?

[jflyper]

At the beginning of the uartTryStartTxDMA:

        if (s->txDMAStream->NDTR)
            goto reenable;

then jumps to USART_DMACmd(DISABLE).

[jflyper]

This is the working branch.

[ledvinap]

Yes, enabling DMA again should mask the problem (probably even without corrupting data, FIFO error is raised befor UART register is written). But what is causing the buffer underflow?

[ledvinap]

What happens when DMA priority is increased? https://github.com/jflyper/cleanflight/blob/0cc47aa1a6377adf3dd30888584ac0308120c851/src/main/drivers/serial_uart.c#L162

[ledvinap]

(sorry, I have to leave today)

[jflyper]

That's for rx... you meant tx, right?

[jflyper]

Thanks for helping me!!! 👍 I'll get some sleep also. 😴

[ledvinap]

Yes, Tx channel ...

[ledvinap]

Pushed few changes, mostly cosmetic. Removed code that disables USART DMA before enabling DMA