FreeRTOS for nRF51822

# FreeRTOS Flash demo port to nrf51822+softdevice ## Description This is a port of the simple FreeRTOS blinky demo. Critical region handling has been updated to use the softdevice api. Due to a lack of SysTick hardware in the nrf51 the tick timer is implemented via RTC1. RTC1 auses the LFCLK; so we don't need HFCLK running most of the time. The vApplicationIdleHook is implemented to call sd_app_evt_wait(). So it ought to sleep as well as can be expected. ## Questionable Decisions + Todos I've cleared the portDISABLE_INTERRUPT macros. Most of FreeRTOS seems to use the critical region code which ought to be right... but there may be some code that still believes that it's disabling interrupts directly (when it is not). CPSID/CPSIE Fun: If CPSID is called in for an extended period of time, then the radio will miss it's deadlines. Eventually, the correct way to handle these may be to store/restore a mask of enabled application interrupts (excluding those blocked or restricted by the sd) and sprinkle it around as needed. You would think that the sd_nvic_critical_region_enter would do something like this, however in SDK 6.0.0 there is only a stub, leaving us to wonder how this is implemented in the softdevice blob. port.c(ulSetInterruptMaskFromISR): Removed cpsid before the branch. There was no complimentary cspie in the vClearInterruptMaskFromISR. These functions seem to set and clear PRIMASK directly, which is a no-no for nrf51 with sd running. I've cleared the macros by which these functions are referenced with no ill-effects. They are left in the code in case some day it proves a better idea to clear / set ICSR as mentioned above. port.c(xPortPendSVHandler): I commented out the cspid/cspie bookending the branch to vTaskSwitchContext. This would seem like a "bad idea", but the logic in that function has nothing to do with the actual context switch. It's more of a hook for stack overflow testing and the trace framework. Could be bad news for task stats and the like. No practical effect on my projects yet. I've reduced the default call stack size to the 1596B required for softdevice operation (0x600 down from 0xC00). Task stacks are allocated from the heap, but if you add a lot of functionality outside tasks (anything running in Handler mode or using MSP...BLE event handlers, perhaps ) then you'll need to increase the stack in startup_nrf51.s (and decrease the configTOTAL_HEAP_SIZE in FreeRTOSConfig.h). There's only ~3kB of RAM left after the s120 softdevice has it's way (it uses 10K+1.5K stack); so memory will be tight. Careful application design is needed. Patches welcome, this code is PoC quality (at best). ## Prerequisites - GCC ARM Toolchain (arm-eabi-none-gcc, et. al) in your path - NRF SDK installed locally and the path specified as SDK_PATH in the Makefile - NRF Softdevice loaded onto the development board. - Tested with s120v1, NRF51822 SDK 6.1.0, and arm gcc 4.8 2014q2 ## Usage 1. Edit Makefile as needed. You'll also want to correct the LED_0 definition in main.c to correspond to your development boards LED. 2. Flash softdevice 3. `make flash` ## Credits - Build environment based heavily on [hlnd/nrf51-pure-gcc-setup](https://github.com/hlnd/nrf51-pure-gcc-setup) - Heavily influenced by the existing FreeRTOS Flash Demo and the existing FreeRTOS ARM_CM0 code.