Skip to content

Conversation

@petejohanson
Copy link

Needed for wired split to work well on nRF52 without impacting BLE timing critical work handling.

pres2300 and others added 30 commits May 23, 2025 21:26
When rx_timeout is set to a sufficiently small value,
rx_timeout_slab could potentially get set to a greater
than necessary value that causes spurious UART_RX_RDY
events.

Fixes zephyrproject-rtos#62828

Signed-off-by: Jacob Preston <jacob.preston@synapse.com>
ISR function prototype requires const void *.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Rework Kconfig to improve handling of multiple UART instances by
using Kconfig template file.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
The ISR prototype used when building without the
interrupt driven UART was not matching the
signature for interrupt handlers, which results in
build warnings.
Let's fix it.

Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Instead of getting the hardcoded address from the DT structure
use its symbolic name which will be resolved by the nRF HAL
definitions to the same value.

This allows the GPIO peripherals' addresses to be redefined
for the simulated targets.

Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
For simulation, we cannot get the UART regiter address
for the pinctrl config structure from DT, as that
cannot match the one allocated at build time.
So let's override it at runtime with the correct address
which is stored in the UART config structure.

Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
While waiting for the UART to be ready in ISR
mode, for simulation only, add a tiny delay per
iteration of the busy wait loops to allow
time to pass.
This Z_SPIN_DELAY is an empty macro for any
other target than simulation.

Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Increase the busy wait time granularity while runing in simulation
for poll mode. In this mode, the driver spends a *lot* of time
waiting for the UART to be done. The smallest frame time is
~10µs @ 1Mbps, so busywaiting in very small increments
is very wastefull as we keep shuting down and turning on the simulated
MCU.
Let's increase the busy wait time increments of the driver to 3 micros,
which should not change much the behaviour.

Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
UART_RX_RDY event can be generated from UARTE interrupt or k_timer
handler. When ENDRX event occurs then k_timer is stopped (it can
be restarted if there is another buffer provided). However, if UARTE
interrupt priority is higher than k_timer priority (RTC is used
underneath) then k_timer handler may still be executed later.
K_timer notifies new bytes based on RXDRDY HW event which is
counter by the TIMER (using PPI). It may happen that RXDRDY
event arrives due to byte received into RX FIFO but since there is
not buffer provided it stays in that FIFO. Given all this, it
was possible that RX_RDY event was reported from ENDRX UARTE event,
timer was stopped but because UARTE interrupt had higher priority
timer handler is executed after UARTE interrupt is handled. In
timer handler TIMER counter reports more bytes and calls
UART_RX_RDY event with null buffer and non-zero amount of bytes.

Fixed by generating UART_RX_RDY event only if RX buffer is not
null.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
- avoid to use undefined macros in #if expressions

Signed-off-by: Hess Nathan <nhess@baumer.com>
Legacy shim takes less flash so it should be a first choice on cores
with less code memory (like RISCV cores on nrf54h20). Adding new
instances support to the legacy shim.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
When low power mode is enabled then whenever UARTE is not active,
driver attempts to put the peripheral into the lowest power state
by stopping and disabling UARTE. However, it did not put pins into
sleep state which could lead to increased current consumption.
Adding pins state handling to the low power mode.

Pins are put into sleep state only if CONFIG_PM_DEVICE=y.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
This is needed to avoid warnings about uninitialized
structure member, which was added in nrfx 3.6.

Signed-off-by: Nikodem Kastelik <nikodem.kastelik@nordicsemi.no>
Some UARTE instances may be clocked at higher speeds than 16MHz, so the
baudrate setting needs to be scaled accordingly. This patch parses the
`clocks` property and obtains the clock-frequency property of the
associated clock, assuming it is a fixed-clock. We should ideally have a
proper clock control subsystem where frequency can be queried using an
API, but we're far from there.

Signed-off-by: Gerard Marull-Paretas <gerard@teslabs.com>
When runtime configuration is not enable then we can determine at
compile time what are the hardware settings. Function which
translates zephyr values to nRF register values is not needed as
macros can do that at compile time.

This optimization saves almost 400 bytes of code.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Code size can significantly reduced (220 bytes per instance) if data
structure (stored in RAM) is not initilized (can be moved to .bss).
Data for asynchronous API had two fields which can easily be moved
to the configuration structure (which is in ROM) because they do not
change duing runtime.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
The initialization of the UARTE is attempting to handle potential
improper handover from the bootloader. This handling is additional
complexity which should be fixed in the bootloader or whatever
component is failing to deinit the UART properly.

This commit removes the handling of improper handover from the
bootloader, while additionally clearing the ERROR flag when suspending
the UARTE component.

Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
When interrupt driven instance is using only TX then low power mode
is automatically enabled and then TXSTOPPED interrupt is disabled in
idle so uarte_nrfx_irq_tx_ready_complete() was returning wrong value.
Adding a dedicate flag which holds information is user enabled TX
interrupt.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
When rx_flush() was called with NULL buffer it was an indication to
drop data from the FIFO. However, it is still important to get the
correct amount of dropped data because when PPI+TIMER are used to
count bytes those flushed bytes are also counted (because each
byte generates RXDRDY event). If those bytes are not counted then
total amount of reported bytes is not correctly aligned with bytes
counted by TIMER.

Reworking rx_flush() to correctly count number of dropped bytes and
then add this flushed bytes to the total amount of RX bytes
reported to the user in RXTO event handler.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
we should be checking for config, not data.

Fixes zephyrproject-rtos#78114

Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Use short which is available on some devices.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Split async control block structure into tx and rx structures.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
This is a leftover from pre-pinctrl era and no longer makes sense.
Driver always manages gpio through pinctrl.

Support removed from uart and uarte shims.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Use nrfx_gppi as abstraction over (D)PPI.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Update nrfx drivers with the minimum number of bytes that can be sent
in a single call to `uart_fifo_fill`.

Signed-off-by: Jordan Yates <jordan@embeint.com>
Refactor RX asynchronous API function to use a pointer to the RX
async data structure instead of top level data structure pointer.
It improves readability with more concise code.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Rework driver to support new way of asynchronous RX handling.
Previously RX was handled in two modes: using RXDRDY interrupt for byte
counting or TIMER + PPI. Both modes had flaws. RXDRDY interrupt mode
could miscalculated amount of received bytes when interrupt was not
handled on time. Data was not lost but was not reported on time that
could lead to issues. PPI+TIMER mode requires additional resources
thus it was not the default mode. Often user was not aware of that
option and was expiriencing driver RX faults.

New RX mode is switching buffers when there is new data (RXDRDY event
not set for given amount of time). It does not require additional
resources to get precise byte counting. Additionally, this is in line
with new UARTE feature (RX frame timeout) which is present in nRF54X
devices. The behavior of the driver is the same for legacy devices
and new one. For legacy devices k_timer periodic interrupts are used
to check if there are any new bytes and it is not needed when RX frame
timeout is present.

Improved RX mode is enabled by default
(CONFIG_UART_NRFX_UARTE_ENHANCED_RX=y) but legacy modes are still
available though not recommended to be used.

Note that new RX mode (CONFIG_UART_NRFX_UARTE_ENHANCED_RX=y) behaves
a bit different because timeout always triggers switch of buffers
which means that there will be no UART_RX_RDY events with non-zero
offset. It also means that every UART_RX_RDY will be followed by
UART_RX_BUF_RELEASED.

After rework, driver is recommended to be used for all platforms as it
performs much better and takes much less code than the second UART shim
available for Nordic devices.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Frame timeout is a hardware feature present in newer versions
of UARTE (e.g. in NRF54X platforms) for detecting idle state
on RX line and ending RX after configurable timeout.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Various API tweaks to account for cherry-pickes and Nordic HAL updates.

Signed-off-by: Peter Johanson <peter@peterjohanson.com>
Signed-off-by: Peter Johanson <peter@peterjohanson.com>
@petejohanson petejohanson merged commit 628a0d8 into zmkfirmware:v3.5.0+zmk-fixes Jun 16, 2025
7 of 10 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

10 participants