The battery charger worker had only ~10% free
stack size so it was a possibility to stack overflow.
After increasing the stack by 512 bytes the worker
has ~40% free stack space.
* Added mechanism enabling CPU to
enter WFI mode when the OS is
in idle, what results in large
power consumption reduction.
* Added mechanism to switch SDRAM to
self-refresh mode before entering
WFI, what resulted in further power
consumption reduction.
* Fix of the issue that source clock for PWM
module was improperly assumed to be derived
from AHB_CLK, while in reality it is
derived from IPG_CLK, what resulted in
module generating signal with 4 times
lower frequency than the configured
one.
* Cleanups.
To allow the automation of gathering logs and swapping out OS images during tests a MSC reboot
endpoint was added. Together with a change to ecoboot, this allows the sending of a "reboot to
MSC" command. The deivce will then reboot and enter MSC mode.
Multiple fixes of clock switching related
stability issues:
* added RC oscillator hysteresis as in NXP example;
* changed DCDC converter config;
* configure PLL2 to be able to run on any CPU
frequency level;
* added switching to 1.275V (overdrive) voltage
when applying any clock change above 12MHz as
well as LDO or bandgap switching, as done in
Mbed OS' lpm.c for RT1050;
* changed BMCR AXI queues weighs for SDRAM
in JLink scripts to disable operations
reordering, as it is known to cause data
integrity issues;
* extracted some code to separate files;
* smaller or bigger code cleanups.
The secure RTC can lock and the clock is not updated.
To prevent this situation we reset LP registers (except for
timestamps and alarms) and clear the LVD flag. Then we
enable again LP SRTC.
Next part of Harmony random resets fixes:
* added RC OSC startup delay as in newest lpm.c;
* changed order of clock source, oscillator and
LDO switching;
* removed connecting internal DCDC load resistor,
as it is only required to speed up converter
startup after it was turned off completely;
* changed DCDC operation mode to continuous
conduction;
* set low VDD_SOC_IN voltage back to 950mV as
suggested in RM.
* added switching DCDC converter mode to
discontinuous conduction for two lowest CPU
clock frequencies, as using CCM mode resulted
in very high current consumption that would
shorten Harmony's life on battery
significantly.
Fixed file access via MTP even when phone is not unlocked.
Now access is granted when the phone is unlocked by the user entering
a passcode. If the phone is not passcode protected (passcode is nor set)
then access to the files is always possible via MTP.
* Fix of the issue that manifested after
running MuditaOS on Harmony via JLink,
which resulted in OS freezing when
trying to set time in normally booted
(via the bootloader) device.
* Minor RTC code cleanup.
* Partial T6 JLink script cleanup.
* Fixed issues with USB charging port detection
* Fixed memory leak in usb_stack/OS integration layer
* Fixed incorrect handling of software timer
* Minor cleaning/refactor.
[BH-1595] Turned off USB/PC charging
Charging enabled only for a dedicated charger. Users won't be able
to charger while connected to the PC.
Remove mechanics enabling to perform reboot
to MSC mode by sending request to one of the
endpoints via serial port, as it is not
implemented in bootloader and may potentially
be unsafe.
Redesign of backup and restore endpoints in the
service-desktop to delegate responsibility for
execution of the actual backup and restore process
onto the Updater.
Getting information about device colour and changing it to number.
Pass information about device colour to usb_stack.
Set a bcdDevice with the corresponding value.
Updated FreeRTOS source code to 10.4.6.
Moved FreeRTOS sources to the third-party directory.
Added necessary changes to the CMake configuration.
Split FreeRTOSConfig.h. From now, each board has its own.
Added missing log headers.
Minor refactor of the module-os cmake.
Fixed stack overflows in bell application main and time service.
Gets data from freertos and prints on frequency change depending
if it's important. Gathering is not costly, printing is though.
For less intrusive checks I would rather disable names gathering
as in worst case scenario it hangs rtos context switching till
thread id is found.
Added atexit functionality. All global destructors and functions
registered with atexit() should be called at program exit.
It was possbile to reuse original newlib implementation with
MALLOC_PROVIDED flag set.
Made sure logger destructructor is called as last one.
Due to mudita OS legacy, it was not possible to simply fix dependencies
in global objectes and hence there are neither global objects
destructors nor functions registered with atexit() called when exit()
is called.
Possibly it will be changed later.
Add support for Reboot and Power off for the
_platform_exit code.
Add _platform_exit mode handle
Signed-off-by: Lucjan Bryndza <lucjan.bryndza@mudita.com>
