Exiting the Kernel

wdawson61 wrote on Friday, August 03, 2012:

Hi.  I have an arm cortex m3 (lpc1758) freeRTOS application that is a bootloader,  It lives in low memory, runs on a FreeRTOS task , gets pristine hw out of reset then either programs some other memory in the system from a USB flash drive and drives a nifty LCD display OR executes a completely separate image with its own kernel and views on how things should be done.  This OTHER body of code attempts to kill-off  FreeRTOS’s ISRs (PendSV, Systick, SVCALL), then installs it’s own set of handlers by moving the vector table, etc.  This has been working until recently.  Now I’m getting an InvPC Usage fault from the Cortex M3 when the other kernel tries to its context switching magic.  If I trace the UsageFault via the stack frame, it points to FreeRTOS’s vPortYieldFromISR.  Even though the OTHER body of code attempts to clear the pendsv and svcall pending bits the NVIC.  My question is:

How can I ensure that FreeRTOS is indeed done and will not have stacked exceptions to return to / from  or any other other pending exceptions before I yank the exception vector table out from under it?

Any help would be greatly appreciated as this is a difficult issue (for me) to debug!!


davedoors wrote on Friday, August 03, 2012:

If you have actually moved the interrupt vector table, then the only way a FreeRTOS interrupt handler could execute after would be if you had copied the FreeRTOS handlers into your new vector table, or you moved the vector table from an interrupt that had nested with an already existing FreeRTOS handler.