I've implemented in Blackfin BF561 coreB FreeRTOS from the code:
http://www.freertos.org/index.html?http://interactive.freertos.org/forums/79366-analog-devices
I converted to bare metal as elf executable using gcc.
I'm close but I have a runtime error that I cannot figuer out. When the scheduler start and try to trigger the first task, the memory pointer get lost and it doesn't start the first stack, instead in start a function inside the first task and get lost when going out of the function.
this is the corresponding debug log:
COREB: end setup LED
COREB: handler declared
COREB: Initialise New TCB:NewTCB address: 3d01000
COREB: TopofStask: 0, pxTopOfStack = 3d0263c
COREB: pxTaskCode =3c033a0, pvParameters = 0
COREB: returned pxNewTCB->pxTopOfStack = 3d02588
COREB: task created:
COREB: top of stack: 3d02588
COREB: GenericListItem: 0
COREB: Event ListItem: 9
COREB: Priority: 1
COREB: start of stack: 3d02000
COREB: Task Name: BootTas
COREB: TCB number: 0
COREB: Task Tag: 0
COREB: Add the idle task at the lowest priority
COREB: Initialise New TCB:NewTCB address: 3d03000
COREB: TopofStask: 0, pxTopOfStack = 3d0431c
COREB: pxTaskCode =3c0295c, pvParameters = 0
COREB: returned pxNewTCB->pxTopOfStack = 3d04268
COREB: task created:
COREB: top of stack: 3d04268
COREB: GenericListItem: 0
COREB: Event ListItem: a
COREB: Priority: 0
COREB: start of stack: 3d04000
COREB: Task Name: IDLE
COREB: TCB number: 1
COREB: Task Tag: 0
COREB: end Add the idle task at the lowest priority
COREB: if xReturn == 1, and xReturn = 1
COREB: before disable interupt
COREB: after disable interupt
COREB: before xPortStartScheduler
COREB: start xPortStartScheduler fn before set core timer
COREB: after ContextSwitch interupt flag
COREB: before prvSetupTimerInterrupt
COREB: after prvSetupTimerInterupt
COREB: Task Switch context called
COREB: The scheduler is running
COREB: trace switched out TCB:ff700bf8
COREB: name of switch out task:efore xPortStartScheduler
COREB: before Task first check for stack overflow
COREB: Task first check for stack overflow called
COREB: Task second check for stack overflow called
COREB: before call get owner of next entry
COREB: get owner of next entry:
COREB: current TCB 3d01000
COREB: pxReadyTasksLists[ uxTopReadyPriority ] = 1
COREB: TCB content:
COREB: top of stack: 3d02588
COREB: GenericListItem: 0
COREB: Event ListItem: 9
COREB: Priority: 1
COREB: start of stack: 3d02000
COREB: Task Name: BootTas
COREB: TCB number: 0
COREB: Task Tag: 0
COREB: trace switched in:BootTas
COREB: write trace to buffer
COREB: task increment tick: 1
COREB: end of app init
COREB: execption 2b addr ff700be4
COREB: coreb dump stack
COREB: found fp: ff700b64
and this is the discution I started on the Analog Device forum:
http://ez.analog.com/message/38669#38669
Ok I finally resolve the issue. the problem was that the assembler code was not return from the vContext code interruption function with the RTI, instead it was go straight through to the next assembler code function which was the timer interrupt which had the same problem and thus the stack pointer was then going to the next address which happen to be the starting address for the app_init function.....
May the problem was due to the fact that the RTI statement was made in a MACRO called by the vContextSwitch function, I'm not sure.
I fixed it by converting those assembler function to C function and now the stack return from interrupt work properly and the problem is fixed. Note that additional error was found after that which was forgotten of hook task loader function in the scheduler when calling vContextSwhitch as well as the tag of the hook function that was never recorded in the context of the task.
All fixed now.
Best Regards,
William
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With