Add a driver for the Arm Ethos-U NPU, including a Devicetree entry for
the mps3_an547 platform.
Signed-off-by: Kristofer Jonsson <kristofer.jonsson@arm.com>
Signed-off-by: Fredrik Knutsson <fredrik.knutsson@arm.com>
It was previously written as regular members of a struct, which
allows the C compiler to do things the way it wants. On ARM64, gcc
would typically write field by pairs (`STP`), which
would generate aborts.
By using `sys_write32`, we force it the right way.
Signed-off-by: Henri Xavier <datacomos@huawei.com>
After a firmware update, DNS needs to be reconfigured.
Fix debug log for AT command strings.
Signed-off-by: Ryan Erickson <ryan.erickson@lairdconnect.com>
GPDMA should work with disabled interrupts in ctrl_lo
This also helps with stability in SOF scenarios
Signed-off-by: Adrian Bonislawski <adrian.bonislawski@intel.com>
The sensor uses the ALERT terminology (pin can be configured to trigger
on certain events such as conversion ready or overvoltage alerts). The
"IRQ" name is not clear.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Specify the units of the current LSB in microamperes, so that we can
measure low maximum currents. Right now it was specified in
milliamperes, but ignored and always hardcoded to 1mA in the driver.
This makes the driver pretty much useless when the maximum current to be
measured is in a range of e.g. 20-50mA.
This patch also removes some unnecessary ifdeffery: since we write the
calibration register, we can always provice measurements with the right
units. It is also wrong to provide sensor readings that do not match
with the units specified by the channel. After this change voltage is
always reported in V, current in A and power in W.
Note that power measurement had the current LSB hardcoded in the
calculation (assuming 1mA/LSB), this has been fixed as well.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Following dtschema practices, add phyisical units to the shunt resistor
value: milliohms.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
With the recent introduction of DT_HAS* Kconfig helpers, such build
asserts are highly unlikely, so just remove them.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Driver had Doxygen stubs for sensor interface calls. First, most were
incomplete, outdated or just wrong. Second, implementations just need to
make sure to adhere to the error codes documented in the public
interface (sensor API in this case).
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
- Add many missing includes so that headers are self-contained
- Sort includes properly (locals first, stdc, zephyr, etc.)
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Right now we had a mix of int (signed) and uint32_t (fixed-width
unsigned integer). Use int32_t for all cases, both values may be either
positive/negative, and we have a defined range when using fixed-width
integers.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The "pmic" driver was in reality a driver for NXP PCA9420 PMIC. There's
no "universal PMIC". While the driver may work for other NXP PMICs, it
is clearly not generic for other vendors PMIC.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The regulator API supports all voltage/current ops, so there's no need
to guard them under "PMIC". If a particular driver doesn't support the
operations it will just return -ENOSYS and the command will fail.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The previous "consumer" API was designed without API ops, meaning only a
single implementation (singleton) was possible. This was inconsistent
with all other APIs, and, a problem in systems where more than a single
PMIC IC is present.
Note that all "consumer" ops are optional, so this won't break any
existing drivers.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
In Zephyr all regulator APIs are consumer APIs. There's no point in
adding a separate header file. Our current implementation is also
inconsistent with Linux, where enable/disable are also part of the
consumer.h API.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Fail with error if any of flashdisk partition assumptions are not met:
* uniform page size through the whole partition
* flashdisk starts at page boundary
* flashdisk ends at page boundary
Read-only flashdisks are not subject to above conditions because the
cache buffer is not used for read-only flashdisks.
The checks can be disabled via Kconfig option to save code space.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
Force cache-size to 0 and treat flashdisk as read-only when backing
partition has read-only flag set. This allows users to save RAM when the
application does not write to the flashdisk, e.g. when a predefined FAT
filesystem is used.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
In usb_dc_ep_disable() mark an endpoint as unoccupied in addition to
being disabled. This allows the endpoint to get enabled properly
with a subsequent call to usb_dc_ep_enable()
See: zephyrproject-rtos/zephyr#51685
Signed-off-by: Milind Paranjpe <mparanjpe@yahoo.com>
Right now it is possible that some devices define 0 pinctrl states in
devicetree, because pinctrl-N entries may still be optional for backward
compatibility. If the programmer makes a mistake and forgets them,
application could experience runtime crashes because
pinctrl_lookup_states assumes you have at least one state, so it does
not perform any bounds checking. Change the while condition so that it
is skipped if states count is zero.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The can_kvaser_pci_data_##inst type was supposed to be
can_kvaser_pci_data and not can_kvaser_pci_config. This got unnoticed
because there were no struct members to pre-initialize, and a reference
to the struct was stored using a void pointer.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
The NPCX driver did use deprecated voltage macros. Define vendor
specific flags instead.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The pin get function was using deprecated macros. Note that the set
implementation already uses vendor specific flags.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The buf_status register is 32 bit wide but was saved in a
uint8_t. This caused some buffers never to be handled which
results in the pico getting stuck in the interrupt handler.
Signed-off-by: Lukas Gehreke <lk.gehreke@gmail.com>
Some modems supports multiple access technologies.
The access technology of the serving cell can be extracted from the
+CEREG command, which is already sent to the modem.
The parse function is extended to also parse the access technology value.
Fixes https://github.com/zephyrproject-rtos/zephyr/issues/51153.
Signed-off-by: Jeppe Odgaard <jeppe.odgaard@prevas.dk>
This moves the CMD_SET_DLF command macro to public header
so application can use it without weird include path to
include the "private" header file under drivers/serial.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Ensure that PECI block is enabled in the EC Subsystem by clearing
the PECI_DIS (peci disable) register
Signed-off-by: Jay Vasanth <jay.vasanth@microchip.com>
Instead of calling __HAL_RCC_OSPIM_CLK_ENABLE() to enable the OSPI
manager clock, we now use a new clock binding in the dts.
In order to avoid confusion between the different clocks, the driver
is modified to select the clock based on their names instead of indexes.
Signed-off-by: Guillaume Gautier <guillaume.gautier-ext@st.com>
Remove a set of Kconfig symbols for configuring the RTC on STM32
that are no longer useful
Signed-off-by: Guillaume Gautier <guillaume.gautier-ext@st.com>
After a soft-reset of the host controller the flash device is not reset.
This can cause the flash device to still be busy writing or erasing when
the host controller boots and initializes the flash. The often results
in errors and the flash device not being initialized.
This fix polls the status register until the WIP flag is off before
initializing the flash device.
Fixes#51713
Signed-off-by: Gregers Gram Rygg <gregers.gram.rygg@nordicsemi.no>
In order to configure domain clock, clock_control_configure should be
used instead of clock_control_on which is only useful for bus clock gating.
Signed-off-by: Erwan Gouriou <erwan.gouriou@linaro.org>
Move functions defined in the Cryptographic toolbox of the Bluetooth
specification inside their own files in the following folder:
`zephyr/subsys/bluetooth/bt_crypto`. The functions were previously
implemented in `zephyr/subsys/bluetooth/host/smp.c`.
The cryptographic toolbox functions can now be accessed from outside of the
host.
In addition to that, tests for each cryptographic toolbox functions have
been added.
Signed-off-by: Théo Battrel <theo.battrel@nordicsemi.no>
Fixed how data is read in mcp23sxx_read_port_regs
The data is now stored in an array to get the
actual data from spi
Signed-off-by: Daniel Skatt <daniel.skatt@gmail.com>
Channel SENSOR_CHAN_GAUGE_FULL_CHARGE_CAPACITY was returned in place of
SENSOR_CHAN_GAUGE_FULL_AVAIL_CAPACITY, due to their matching names.
Ensuring that the channel names are unique fixes the issue.
Signed-off-by: Marco Argiolas <marco.argiolas@ftpsolutions.com.au>
CC setting algorithm is handling a case when CC is too soon (next
tick from now). It was setting CC to one tick further in the future
if that was detected. Step was repeated if counter incremented during
setting CC and CC was behind the counter because of risk of setting
CC too late. In certain scenarios we might spend a lot of time in
that loop, especially if optimization is turned off. Test shown that
loop was executed dozens of time (up. to 700us). To prevent
prolonged execution whenever CC setting fails we set CC to one more
tick further in future.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
The reference count, dp->sref, is never incremented for the SSP DAI as
we use the runtime PM API's to probe/remove the DAI during
resume/suspend. Therefore, use the DAI state to check if the DAI is
active in any direction before proceesing to set the config. This fixes
the IO errors seen during simultaenous playback and capture on the same
SSP.
Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
