Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

Edvin — Wed, 11 Jun 2025 09:28:33 GMT

Feel free to provide a sample showing these issues, and how to reproduce it. I have not seen it myself.

Best regards,

Edvin

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

atsteve — Tue, 10 Jun 2025 18:58:53 GMT

I can now confirm that the issue appears to lie within the new UART driver.

I had the opportunity to rewrite the code and switch to the asynchronous UART API. This allowed me to choose between the legacy UART driver (zephyr/drivers/serial/uart_nrfx_uarte.c) and the new driver (zephyr/drivers/serial/uart_nrfx_uarte2.c) using the CONFIG_UART_NRFX_UARTE_LEGACY_SHIM option.

When the new driver is in use, I lose incoming data from the UART. When the old driver is in use, everything functions as before. The only difference between these two builds is the CONFIG_UART_NRFX_UARTE_LEGACY_SHIM option.

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

Edvin — Fri, 14 Mar 2025 21:22:08 GMT

Not that I am aware of.

BR,

Edvin

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

erickcinco — Thu, 13 Mar 2025 21:46:03 GMT

Has anyone else reported issues with UART and devices that lack flow control with NCS 2.8.0+?

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

Edvin — Wed, 19 Feb 2025 13:20:19 GMT

Hello,

Sorry for the late reply.

Do you have any samples both for 2.7.0 and 2.8.0 that shows the difference in behavior, so that I can run them on a DK and analyze the pins using a logic analyzer?

Best regards,

Edvin

RE: Reduced performance of interrupt driven serial UART reception in NCS 2.8.0 vs 2.7.0

atsteve — Tue, 18 Feb 2025 17:15:43 GMT

I would just like to add some additional information:

I have also ensured that the High Frequency 32mhz crystal oscillator (HFXO) is selected and running during serial communications using the on-off manager method described in this devzone post: Selecting HFCLK as clock source for hardware timer when using Zephyr

I confirmed that the crystal is indeed selected and running by monitoring the NRF_CLOCK->HFCLKSTAT register and observing the value of 0x00010001 which indicates that the HFXO is selected and running.