Other Parts Discussed in Thread: EK-TM4C1294XL
Tool/software: Code Composer Studio
SUMMARY: While debugging, when I return from an assembly function with the code bx lr (with lr holding 0xfffffff9), the session is interrupted and CCS returns an error saying: No source available for "0xfffffff8". It should start the return from ISR process (because of the 0xfffffff9 code) but looks like CCS Debugger can't handle this. So it tries to execute code from 0xfffffff8.
DETAILS:
I'm working on the thread management of a little operating system. To create threads I need to define their initial stack frame. The initial stack frame I'm creating is such that after a context switch a new thread will start executing as if it was returning from an ISR. Something like this:
#define PSR_RESET_VALUE 0x01000000 #define RFI_CODE 0xFFFFFFF9 // return from interrupt /* -------------------------- */ /* CREATE INITIAL STACK FRAME */ *(--stack_pointer) = PSR_RESET_VALUE; // PSR *(--stack_pointer) = (uint32_t) routine; // PC (routine is the function executed by the thread) *(--stack_pointer) = (uint32_t) destroy; // LR (destroy is the function that should be executed if routine returns) // each GPR is initialized with its number just // for debugging purposes *(--stack_pointer) = 0x12121212; // R12 *(--stack_pointer) = 0x03030303; // R3 *(--stack_pointer) = 0x02020202; // R2 *(--stack_pointer) = 0x01010101; // R1 *(--stack_pointer) = 0x00000000; // R0 *(--stack_pointer) = 0x11111111; // R11 *(--stack_pointer) = 0x10101010; // R10 *(--stack_pointer) = 0x09090909; // R9 *(--stack_pointer) = 0x08080808; // R8 *(--stack_pointer) = 0x07070707; // R7 *(--stack_pointer) = 0x06060606; // R6 *(--stack_pointer) = 0x05050505; // R5 *(--stack_pointer) = 0x04040404; // R4 // simulates a return from an ISR, so the registers // are popped from the stack correctly *(--stack_pointer) = RFI_CODE;
After the stack frame preparation is finished the stack pointer of the new thread will point to RFI_CODE (0xfffffff9). I will pop that code into LR before returning from context switch. So I was hoping the hardware would unstack everything correctly and the new thread would start executing. The code for context switch is here:
switch_context:
cpsid i ; disable interrupts
push {r4-r11} ; push callee-save registers
push {lr} ; lr is the address to return next time thread executes
str sp, [r0] ; save hardware sp on current thread tcb.sp (r0 is the address of current thread tcb.sp)
ldr sp, [r1] ; load hardware sp with new thread tcb.sp (r1 is the address of new thread tcb.sp)
pop {lr} ; pop address to return (in the first execution lr will be 0xfffffff9
pop {r4-r11} ; pop callee-save registers
cpsie i ; enable interrupts
bx lr ; continue to execute from where thread was preempted before (in the first execution return from ISR)
The problem is already stated on the summary. When the processor returns from switch_context it should unstack the core registers's contents, as if it was returning from an ISR, and continues to execute. Instead, the debugging session is stopped and CCS says No source available for "0xfffffff8". I don't know why this is not working as I expected and I don't know why it goes to 0xfffffff8 instead of returning from ISR.
In this question a community member says this can be a debugger problem. Once there is no code at address 0xfffffff9, the debugger gets confused. If this is true, how can I continue debugging my code?
MORE INFO:
- Windows 7 64-bit, Service Pack 1
- CCS 7.4.0.00015
- LaunchPad EK-TM4C1294XL
- Stellaris In-Circuit Debug Interface
