This patch corrects the initial unmapping of unused l2 memory
to unmap until the end of virtual address space instead
of unmapping until the end of l2 physical memory.
Signed-off-by: Jaroslaw Stelter <Jaroslaw.Stelter@intel.com>
This patch introduces sys_mm_drv_move_region implementation
which stops usage of the sys_mm_drv_simple_move_region.
The sys_mm_drv_simple_move_region is unsuitable becuase
it iterates through physical pages in linear fashion.
The new implementation queries the tlb for each virtual
page of the requested region to find the physical page
to be remapped.
Signed-off-by: Jaroslaw Stelter <Jaroslaw.Stelter@intel.com>
Mayfly by design uses a memq for storage of its jobs. The memq
requires head and tail to track the content. It is considered
empty is head equals tail.
When memq instance is initialized then there is a new link
instance stored in head and tail, nevertheless the memq is
still empty.
When new job is enqueued to a memq, the inilial link is used
to store the job. New link, provided by enqueue call, is stored
in the tail for future enqueue.
When enqueued job was served and is dequeued, the link it was
assigned to is returned and stored in the job object.
That link will be used in future for call to enqueue.
Now lets consider a situation when we are just after initalization.
Some default initial link is in empty memq. We enqueue and dequeue
a job. After dequeue, the job object stores the initial link object.
The one that was put into the memq during initialization.
Next Bluetooth stack is disabled and enabled again.
The job is enqueued again, but it still stores the initial link
address. After enqueue the memq head points to initial link object,
that stores new job. Tail points to link deliveded by enqueue call,
that is also the initial link object. The memq is considered to be
empty, nevertheless there was a successful enqueue operation.
The issue is casued by lack of re-initialization of a job object
on init. In most cases these objects are static members of some
functions, hence there is no re-initialization after bt_disable
and bt_enable calls.
The problem is fixed by re-initialization of mayfly only once
on bt_enable() call. Then it doesn't matter what links are stored
in dequeued objects and there is no need to re-initialized mayfly
job objects.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
The problem is visible after call to bt_disable when there was
a connection. On a next bt_enable call when hc_tx_thread starts
it is possible that the conn_change object internals are not
re-initialized entirely and the function fails during k_poll call.
The commit changes explicit assignment conn_change object to
a k_poll_signal_init that does proper initialization.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
While execution of bt_disable there is missing de-initialization
of per_adv_sync_pool. If there was a sync in progress, (flags set
to BT_PER_ADV_SYNC_SYNCING) before bt_disable, then when the stack
is re-enabled by bt_enable new sync create fails.
The commit fixes the issue by new function that clears flags for
all per_adv_sync_pool entries. The function is executed by bt_disable.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Migrate tests in connectionless_cte_tx to new ZTEST API.
Remove main.c source file because is unncessary when automatic
unit tests registration is available. Also former use of
predicate to select unit tests for execution is removed
because the tests are split into two distinct tests suites
that has its't own setup and teardown phases.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Migrate tests in connectionless_cte_rx to new ZTEST API.
Remove main.c source file because is unncessary when automatic
unit tests registration is available.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Migrate tests in connectionless_cte_chains to new ZTEST API.
Remove main.c source file because is unncessary when automatic
unit tests registration is available.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Migrate tests in connection_cte_tx_params to new ZTEST API.
Remove main.c source file because is unncessary when automatic
unit tests registration is available.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Migrate tests in connected_cte_req to new ZTEST API.
Remove main.c source file because is unncessary when automatic
unit tests registration is available.
Signed-off-by: Piotr Pryga <piotr.pryga@nordicsemi.no>
Simplify spi_cs_control by removing anonymous struct around the
"gpio" field. Update SPI_CS_CONTROL_PTR_DT to match this change.
Signed-off-by: Kumar Gala <galak@kernel.org>
All in tree users are using the gpio_dt_spec so we can remove the
older gpio struct elements that have been deprecated for some
time.
Signed-off-by: Kumar Gala <galak@kernel.org>
The SSD16xx driver needs to wait for the chip to complete commands
before issuing new commands. The datasheet doesn't specify which
commands cause the chip to raise the busy pin. The current driver has
ssd16xx_busy_wait() calls after commands that are known to take a long
time. This is very fragile and scatters a lot of boiler plate around
the driver.
Include an ssd16xx_busy_wait() call in ssd16xx_write_cmd() instead of
after commands that are suspected to take a long time. This makes the
driver more robust and is more in line with the expectations set out
in the data sheet.
Signed-off-by: Andreas Sandberg <andreas@sandberg.uk>
There are cases when attributes of mapped virtual memory need
to be updated. E.g. in case there is loadable library/module
code loaded to the l2 memory then memory needs to be read-write.
After the code is loaded and is ready to be executed then
attributes of mapped memory should be updated to
read-only/executable without loosing memory contents.
Signed-off-by: Jaroslaw Stelter <Jaroslaw.Stelter@intel.com>
Removes deprecated conf files from the counter/alarm
sample and adds overlays to enable timer nodes in use.
Signed-off-by: Glauber Maroto Ferreira <glauber.ferreira@espressif.com>
Refactors counter driver code, bringing:
- more readability through the removal of abused macros
- better use of the devicetree, by moving peripheral node's
info defined on macros and Kconfigs to dts bindings
- less dependence from dynamic allocations, making counter
instances fully statically-allocated
- better use of DT_INST_* macros
Signed-off-by: Glauber Maroto Ferreira <glauber.ferreira@espressif.com>
- adds properties 'group', 'index' and 'prescaler'.
- updates board's dts to include those properties.
Signed-off-by: Glauber Maroto Ferreira <glauber.ferreira@espressif.com>
The specific hardware instance is already got from
device_is_available function, we don't need to
iterate all hardwares again.
Signed-off-by: Chen Peng1 <peng1.chen@intel.com>
Allow to use RTU server sample on any board that has supported
USB device controller. Although it is only a point to point
connection and does not represent a bus, it can, apart from
testing the server implementation, also be used practically
for example to control relays or to read ADC values via
USB connection without implementing custom USB class or driver.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
We cannot implement configure API because there is
no notification of configuration changes provided
for the Abstract Control Model and the UART controller
is only emulated.
However, it allows us to use CDC ACM UART together with
subsystems like Modbus which require configure API for
real controllers.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Clean up unused pinmux function 'npcx_pinctrl_mux_configure()' and
related 'NPCX_DT_IO_ALT*' macro functions.
Please refer https://issuetracker.google.com/234861079 for more detail.
Signed-off-by: Mulin Chao <mlchao@nuvoton.com>
Move the CAN bus network driver from drivers/can to drivers/net as it
implements a network driver, not a CAN controller driver.
Use a separate Kconfig for enabling the CAN bus network driver instead of
piggybacking on the SocketCAN Kconfig. This allows for other
(e.g. out-of-tree) SocketCAN transports.
Signed-off-by: Henrik Brix Andersen <hebad@vestas.com>
Implement support for display blanking by inhibiting display refresh
while blanking is on. Writes to the display are still sent to the
internal RAM in the display controller. Once blanking is turned off,
the contents of the RAM are written to the display. This behavior is
makes it possible to compose a frame buffer in the display controller
without flickering.
Other e-paper drivers such as the GD7965 already implement this
behavior.
Signed-off-by: Andreas Sandberg <andreas@sandberg.uk>
Struct bt_mesh_send_cb.end is called from the mesh advertiser thread.
bt_mesh_reset now requires more stack size, thus can't be called from
the advertiser thread. k_work instance requires less memory than needed
for bt_mesh_reset.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
Upon bt_mesh_reset call, all mesh modules has to clear the privisioning
and configuration data. None of the modules store or erase in settings
subsystem immediately, but put this on the mesh settings work. After the
settings work is scheduled, all stored data will be removed and the
device will eventually be unprovisioned. Until then, the device is not
completely unprovisioned, thus calling bt_mesh_prov_enable,
bt_mesh_provision (and bt_mesh_cdb_create) should not be allowed.
Struct bt_mesh_prov has a reset callback stating that after this
callback is called, the device has been reset and can be re-provisioned
again. Also, bt_mesh_reset API description states that after calling
bt_mesh_reset API, the device needs to reenable the provisioning layer
to be provisioned again. But this is not correct since the settings has
to be cleared before the device can be reprovisionined.
This commit makes bt_mesh_reset flush pending settings so that the
device can be reprovisioned immediately and the API will behave as
written in the description.
The same applies to bt_mesh_cdb_clear.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
Adding command queue mode can reduce the time between each byte to
improve the I2C bus clock stretching during I2C transaction.
I2C command queue mode of it8xxx2 can support I2C APIs including:
i2c_write(), i2c_read(), i2c_burst_read.
Test:
1. tests\drivers\i2c\i2c_api --> pass
2. Reading 16 bytes of data through i2c_burst_read() can reduce
0.72ms(2.54ms->1.82ms) compared to the original pio mode when the
frequency is 100KHz.
3. krabby platform can boot normally.
Signed-off-by: Tim Lin <tim2.lin@ite.corp-partner.google.com>
This ensures that the compiler will have moved the stack pointer below the
stack area where the message will be constructed. Otherwise, the message
can be smashed by an interrupt handler while it is being built.
This bug was found on qemu_cortex_a53 using SDK 0.14.2 with gcc 10.3.0 when
building the samples/subsys/logging/syst/sample.logger.syst.deferred test
under twister using picolibc:
Signed-off-by: Keith Packard <keithp@keithp.com>
Simplify the implementation by using gpio_dt_spec.
As part of this use logical values of 0/1 for gpio_pin_set_dt
and handle any inversions as GPIO flags that might be required
at a board level (in dts config).
Signed-off-by: Kumar Gala <galak@kernel.org>
Implement support for half duplex communication in the bit bang SPI
driver. The SPI driver will use the MOSI pin is for both TX and RX
operations when using half duplex mode.
In half-duplex mode, the driver configures the MOSI pin as an input
pin for input only transactions. Transactions that are bidirectional
are forbidden. After an SPI transaction, the MOSI pin is left as an
input if it was an RX transaction or an output after a TX
transaction. Like before, the MOSI pin is initialized as an
(inactive) output pin when the bus is initialized.
Signed-off-by: Andreas Sandberg <andreas@sandberg.uk>
"dev" is already defined as a static variable,
it should not be redefined locally
Fix regression introduced by commit
2a8e3fe82d
Signed-off-by: Alexandre Bourdiol <alexandre.bourdiol@st.com>
This commits is restricting the execution of the
samples/modules/canopennode to board configuration where
the nvs_sector_size is less than 0x10000.
When the CONFIG_CANOPENNODE_STORAGE is selected, the
settings_subsys_init do not accept
nvs_sector_size > UINT16_MAX.
Then all the stm32h723/h743/h745/h750 cannot run the
sample.modules.canopennode testcase
Signed-off-by: Francois Ramu <francois.ramu@st.com>
Add API function for getting the supported capabilities of a CAN
controller. This allows for writing portable CAN applications.
Signed-off-by: Henrik Brix Andersen <hebad@vestas.com>
Switch from using a driver-specific, compile-time devicetree one-shot
property to supporting the newly added CAN_MODE_ONE_SHOT flag for
enabling/disabling one-shot mode at run-time.
Signed-off-by: Henrik Brix Andersen <hebad@vestas.com>
