Error Handling

Nano-RK uses the convention that system calls return -1 upon error and 1 upon success.

NRK_OK 1

NRK_ERROR -1

_nrk_errno_set

void _nrk_errno_set( NRK_ERRNO error_code )
Parameters: NRK_ERRNO error code to set
Return Values: none

This function should be used by user defined system calls to set an error number.

nrk_errno_get

uint8_t nrk_errno_get( )
Parameters: none
Return Values: uint8_t error code last set for current task

Once a function returns failure (-1), this function can be used to get a more descriptive error code.

Kernel-Errors

Kernel errors are triggered when various OS services fail. There are various flags in nrk_cfg.h that govern how the system responds to errors:

  • NRK_REPORT_ERRORS
    • Print error messages. Usually this should be disabled in final deployments.
  • NRK_HALT_ON_ERROR
    • Halt execution and print a single error message
  • NRK_HALT_AND_LOOP_ON_ERROR
    • Halt execution and prints error in a loop
  • NRK_LOG_ERROR
    • Not yet implemented. Will store error log to MMC card.
  • NRK_WATCHDOG
    • Enable the watchdog timer in case a kernel service hangs during a deployment
  • NRK_REBOOT_ON_ERROR
    • This must be used with the watchdog timer to reboot the system on errors 
// NRK_REPORT_ERRORS will cause the kernel to print out information about
// missed deadlines or reserve violations
#define NRK_REPORT_ERRORS

// NRK_HALT_ON_ERRORS will cause the kernel to freeze on errors so that
// it is easier to see debugging messages.
#define NRK_HALT_ON_ERROR

// NRK_HALT_AND_LOOP_ON_ERRORS will cause the kernel to freeze on errors but continue
// to print the panic message in a loop so that the node can be plugged into a terminal
// in order to indentify the problem.
#define NRK_HALT_AND_LOOP_ON_ERROR

*NRK ERROR( task_id ): ERROR_MESSAGE

task_id is the ID of the task that caused the error. 0 means the kernel had an internal error.

  • "Task Stack Overflow"
    • The canary value in the identified task was over written. Try making the task stack size larger.
  • "Reserve Error in Scheduler"
  • "Task Reserve Violated"
    • Increase the reservation for the task, or disable a task reserve by setting it to 0
  • "Scheduler Missed Wakeup"
    • See Timer Overflow
  • "Duplicated Task ID"
    • There is a problem with the way the tasks are being configured
  • "Unexpected Restart"
  • "Idle or Kernel Stack Overflow"
  • "Extra Task started, is nrk_cfg.h ok?"
    • Most likely NRK_MAX_TASKS in nrk_cfg.h needs to be larger
  • "Low Voltage"
  • "Unhandled Interrupt Vector"
    • An interrupt mask was set that should not have been.
    • The code executed passed the end of a task (no while() loop inside task etc)
  • "Timer Overflow"
    • This is likely a problem in the kernel. It can happen if the fuses get cleared such that the ASYNC clock is now operating differently. It can also happen if NRK_SLEEP_WAKEUP_TIME in nrk_platform_time.h is set too low. NRK_SLEEP_WAKEUP_TIME is the max number of ms required for the processor to wake from deep sleep.
  • "Device Driver Error"
  • "Failed to create Signal"
  • "Failed to create Semaphore"
    • An internal OS related operation failed to create a semaphore. This is probably because NRK_MAX_RESOURCE_CNT needs to be increased for some functionality the application is trying to use.
  • "Kernel function not implemented"
  • "UNKOWN"

nrk_kernel_error_add

void nrk_kernel_error( Kernel_Panic_ID, uint8_t task_ID )
Parameters: Kernel_Panic_ID is a #define kernel panic id
Parameters: uint8_t task_ID is the PID of the offending task
Return Values: none

This function is called from within kernel code to post kernel panics. This can be used by applications in cases of a hoplessly fatal error (this should rarely be the case though). 

nrk_kernel_error_add( NRK_SIGNAL_CREATE_ERROR ,nrk_cur_task_TCB->task_ID);

Kernel-Watchdog-Timer

For final deployments, you may wish to enable extra protection against the system halting by using a watchdog timer. After including NRK_WATCHDOG in the nrk_cfg.h file, the system watchdog timer will be enabled at bootup and set each time the Nano-RK scheduler executes. Under normal operation, application tasks will be bounded by their reservations. If a part of the OS fails to exit within 8 seconds, the system will reboot. Upon restart, a watchdog kernel panic is triggered. When NRK_HALT_ON_ERROR or NRK_HALT_AND_LOOP_ON_ERROR is enabled, the watchdog timer is disabled upon an error so as to not interfere with the normal error printing operation. 

// Enable the watchdog as a protective measure
// This will only activate if the scheduler fails.
#define NRK_WATCHDOG

| Contents | Time Management |