I am about to finish porting FreeRTOS v8.0.1 to the Texas Instruments
CC2538 SoC (Cortex M3 + 802.15.4 radio) for an IoT platform named
OpenMote (http://www.openmote.com/). The basics are already working just
fine (I have a demo with tasks and semaphores using SysTicks), but I am
stuck trying to get tickless support to work. I developed my own
vPortSetupTimerInterrupt, vPortSuppressTicksAndSleep and
SleepTimerHandler functions based on those of the Atmel SAM examples but
replacing the ASF functions from the Atmel library with those of the
Texas Instruments library (which by the way is not as great since it
does no comply to CMSIS). I have also installed the SleepTimerHandler to
the interrupt table.
If I try my RTC code outside of FreeRTOS it works just fine, the uC goes
to low-power mode until the RTC expires and the process is repatead on
and on without problems. However, when I integrate it with FreeRTOS it
stops working. It looks like the interrupt is fired the first time but
does not continue to fire any more. The FreeRTOSConfig.h looks like the
define configCPUCLOCKHZ 16000000
define configSYSTICKCLOCKHZ 32768
define configUSETICKLESSIDLE 1
define configTICKRATEHZ ( ( TickType_t ) 128 )
#define configPRIO_BITS __NVIC_PRIO_BITS
#define configPRIO_BITS 4 /* 15 priority
define configLIBRARYLOWESTINTERRUPT_PRIORITY 0x0F
define configLIBRARYMAXSYSCALLINTERRUPTPRIORITY 10
define configKERNELINTERRUPTPRIORITY (
configLIBRARYLOWESTINTERRUPTPRIORITY << (8 - configPRIOBITS) )
define configMAXSYSCALLINTERRUPT_PRIORITY (
configLIBRARYMAXSYSCALLINTERRUPTPRIORITY << (8 - configPRIO_BITS) )
If I disable the configUSETICKLESSIDLE and configure the appropriate
values for the clocks then the demo works just fine, so I asume the
vector table and the interrupt priorities are fine.
Any other ideas where I could look at?
Are you using a critical section inside your SuppressTicksAndSleep function? If so, are you using the taskENTER/EXIT_CRITICAL() macros, or are you globally disabling and enabling interrupts by setting the appropriate bit in the status register? You must use the latter for it to work...
The taskENTER/EXIT_CRITICAL() macros don't disable interrupts on the Cortex-M core, they only mask off up to a certain interrupt priority. If the RTC interrupt is masked off by this priority then it can't bring the Cortex-M out of low power mode.
On the other hand, by way of some magic in the Cortex-M core, if you leave the interrupt mask such that it does not mask off any interrupts, but instead globally disable interrupts, the RTC interrupt will be able to bring the Cortex-M core out of low power mode, although the interrupt won't actually be executed until interrupts are globally enabled again.