FreeRTOS Task Priority VS Interrupt Priority nRF52840

[Dave_couling]

Hi guys,

I have a 4 thread freeRTOS setup.

1. BLE - priority 4
2. TWI - priority 3
3. LED - priority 2
4. Logging - priority 1

After running for some time my code always gets caught in an infinite loop inside the vListInsert function of freeRTOS, or hardfaults @ PC:0x20003188 PSR: 0x4000000e (PendSV). It occurs like a random race-condition. Typically this indicates an issue with interrupt priorities. Looking thru my Stack trace, On the BLE thread (4), I get a SAADC IRQ (IRQ priority 6 in sdk_config.h) which triggers a callback function to write the value to the Log Thread (1). Here I try and grab a Semaphore which calls the following sequence, xSemaphoreTake() -> xQueueGenericReceive() -> vTaskPlaceOnEventList() -> vListInsert().

As an FYI I'm trying to Port FreeRTOS to the nRF52840 which i realize doesn't have an official support from Nordic yet.  Just want to double check there is nothing unique about the any vector implementation that might propagate to different behaviour in upper levels of an RTOS.  Thanks for your help.

[Julie H]

Keep in mind that if a LOWER priority thread has acquired a resource, it can block a HIGHER priority thread because it cannot release the resource until the higher priority thread gives up. Since you are running the SAADC task when the hang occurs, you need to check your lower priority tasks to see if there is some place where you acquire the same mutex. There's only one lower priority task, and that's the LED one. Check there.

[Dave_couling]

I don't use Mutex, instead I use Binary Semaphores. If grabbing the semaphore fails I simply move on. I have set break points in case the semaphore fails, but that's not hitting. I believe the issue lies in the TWI thread as everyything is fine when I bypass ever starting that thread.

[Julie H]

Do you know where in vListInsert() it is looping? I looked at the source code .. that I believe is correct .. and there is a loop where it is attempting to find an insertion point and there is no check to ensure the list isn't corrupt and the loop will ever terminate.

[Dave_couling]

I'll try and catch it next time. I think the root of the issue is on the TWI thread. Everything is stable if I remove this thread. Though for some reason I am hardfault faulting instead of the vListInsert() loop. The PSR: 0x4000000e indicates a Negative Flag in the PendSV.

[Julie H]

Could be, if the TWI thread is doing something to corrupt the list. If you can point me at the source code for the version of FreeRTOS you're looking at I could stare at it when I've got nothing better to look at ...