CCS/AM3359: Problem with interrupts

Hello Lalindra, thanks for your reply!

I'm familar with the RTOS implementation od the GPIOs, in fact i have the interrupts already working with RTOS but my application is very time critical sample aquisition via SPI and i try to keep the interrupt latency as short as possible. The RTOS managed interrupts are to slow thats why i want to use the starter-ware implementations in a first step.

I think with my GPIO definitions is everything ok. I can switch my LEDs on an off, read pin values in and even my interrupt triggers the right values in the right registers. The program-counter just doesn't jump to my isr (or at least it looks this way).

I hope someone encountered a similar problem and can help me with this.

There is also the following thread (e2e.ti.com/.../242268) , where a similar problem with the timer interrupt. They solved it with modifing some .cmd files. I think this wasn't without the GNU compiler because my program has only a .lds file

This is my exact code (maybe that helps also).

#include "soc_AM335x.h"
#include "interrupt.h"
#include "beaglebone.h"
#include "gpio_own.h"
#include "pin_mux.h"
#include "hw_control_AM335x.h"

static void gpio1Isr(void)
{
    //clear interrupt
	GPIOPinIntClear(SOC_GPIO_0_REGS,0,14);
	GPIOPinIntClear(SOC_GPIO_0_REGS,1,14);
	
	while(true); //dead loop to catch the pc if the interrupt gets triggerd

    GPIOPinIntEnable(SOC_GPIO_0_REGS,0,14);
    GPIOPinIntEnable(SOC_GPIO_0_REGS,1,14);
    GPIOPinIntWakeUpEnable(SOC_GPIO_0_REGS,0,14);
}

int main(void)
{
    /* Enabling functional clocks for GPIO1 instance. */
    GPIO0ModuleClkConfig();
    GPIO1ModuleClkConfig();

    /* Enabling the GPIO module 0. */
    GPIOModuleEnable(SOC_GPIO_0_REGS);
    GPIOModuleReset(SOC_GPIO_0_REGS);

    /* Enabling the GPIO module 1. */
    GPIOModuleEnable(SOC_GPIO_1_REGS);
    GPIOModuleReset(SOC_GPIO_1_REGS);


    GpioPinMuxSetup(GPIO_1_22,CONTROL_CONF_MUXMODE(7)); // USER LED 1
    GpioPinMuxSetup(GPIO_1_23,CONTROL_CONF_MUXMODE(7)); // USER LED 2
    GpioPinMuxSetup(GPIO_1_24,CONTROL_CONF_MUXMODE(7)); // USER LED 3


    GpioPinMuxSetup(GPIO_0_14,CONTROL_CONF_GPMC_AD_CONF_GPMC_AD_RXACTIVE|0<<CONTROL_CONF_GPMC_AD_CONF_GPMC_AD_PUDEN_SHIFT|1<<CONTROL_CONF_GPMC_AD_CONF_GPMC_AD_PUTYPESEL_SHIFT|CONTROL_CONF_MUXMODE(7)); // Interrupt in
    GPIODirModeSet(SOC_GPIO_0_REGS,14,GPIO_DIR_INPUT);  // interrupt in



    GPIODirModeSet(SOC_GPIO_1_REGS,22,GPIO_DIR_OUTPUT); // USER LED 1
    GPIODirModeSet(SOC_GPIO_1_REGS,23,GPIO_DIR_OUTPUT); // USER LED 2
    GPIODirModeSet(SOC_GPIO_1_REGS,24,GPIO_DIR_OUTPUT); // USER LED 3
    GPIOPinWrite(SOC_GPIO_1_REGS,22,GPIO_PIN_LOW); // USER LED 1
    GPIOPinWrite(SOC_GPIO_1_REGS,23,GPIO_PIN_LOW); // USER LED 2
    GPIOPinWrite(SOC_GPIO_1_REGS,24,GPIO_PIN_LOW); // USER LED 3


    IntAINTCInit();

    IntMasterIRQEnable();

    //    GPIO interrupts
    IntSystemEnable(SYS_INT_GPIOINT0A);
    IntPrioritySet(SYS_INT_GPIOINT0A, 3, AINTC_HOSTINT_ROUTE_IRQ);
    IntRegister(SYS_INT_GPIOINT0A, gpio1Isr);

    // Setting the GPIO_CLK_PIN_NUMBER to raise IRQ at falling edge of input *
    GPIOIntTypeSet(SOC_GPIO_0_REGS,14, GPIO_INT_TYPE_FALL_EDGE);

    GPIOPinIntEnable(SOC_GPIO_0_REGS,0,14);
    GPIOPinIntEnable(SOC_GPIO_0_REGS,1,14);
    GPIOPinIntWakeUpEnable(SOC_GPIO_0_REGS,0,14);


    while(true) // main loop
    {
	    // Blink led (works fine before triggering the external interrupt)
    	owndelay(20000000);
        GPIOPinWrite(SOC_GPIO_1_REGS,24,GPIO_PIN_LOW); // USER LED 3
    	owndelay(20000000);
        GPIOPinWrite(SOC_GPIO_1_REGS,24,GPIO_PIN_HIGH); // USER LED 3
    }
}

I managed it to debug with interrupts and got some new information (but couldn't solve the problem)

in the generated startup_ARMCA8.S are the following assembly instructions

    .section .isr_vector
    .align 4
    .globl  __isr_vector
__isr_vector:
        LDR   pc, [pc,#24]       @ 0x00 Reset
        LDR   pc, [pc,#-8]       @ 0x04 Undefined Instruction
        LDR   pc, [pc,#24]       @ 0x08 Supervisor Call
        LDR   pc, [pc,#-8]       @ 0x0C Prefetch Abort
        LDR   pc, [pc,#-8]       @ 0x10 Data Abort
        LDR   pc, [pc,#-8]       @ 0x14 Not used
        LDR   pc, [pc,#-8]       @ 0x18 IRQ interrupt
        LDR   pc, [pc,#-8]       @ 0x1C FIQ interrupt
        .long  Entry
        .long  0
        .long  SVC_Handler
        .long  0
        .long  0
        .long  0
        .long  0
        .long  0

when an interrupt is triggered, the PC jumps to the instruction with offset 0x18 (which is an infinite loop i think (LDR pc, [pc,#-8])) so it gets caught there. if i modify that to LDR pc, [pc,#16] i can load the address of an ISR_HANDLER similar to the SVC_Handler. In the exceptionhandler.S file is a ISR_HANDLER with save and restore context instructions and an call of INTCCommonIntrHandler.

Now something is missing what do is have to do that my own ISR written in C is executed?

Hi,

To debug it, it helped to work with the external interrupts. I simply strated the debug mode in CCS with always activated interrupts. Then i paused the execution and pushed my Interrupt-Button. After that i did the assembly stepping and my program counter jumped to the interrupt address...

I found some time to include the startup.c and the init.S (which i thik are essentially it - but i already had the other files for MMU funtionality) Now I think I'm facing the same problems as you. The compiler seems to miss some addresses like "=_bss_start" or "=_stack" which are used in the init.S file. Has the liker script to define them? It would help me a lot if you'd still have the liker script which worked out for you.
Thank you for your reply, I think I am on the right way...

Ok I think I got it!
i changed the following things in the linker File

changed PROVIDE(__stack = __StackTop); to PROVIDE(_stack = __StackTop); (Note the underscore)

added the following lines under the .bss: { ...}
PROVIDE(_bss_start = __bss_start__);
PROVIDE(_bss_end = __bss_end__);

I wish TI would provide the correct files right away instead of empty interrupt-handlers which land in endless loops. Setting up CCS so you can do bare-metal programming isnt very comfortable. (Maybe I'm just spoiled by PSoC-Creator ;-) )

Thanks to all!

Hey,

i am glad you found a solution!

Actually i think my workaround was to add an underscore in front of bss_start/end__ everywhere in the linker's file..I hope it is right..At least that helped me overcome the build errors..