LPC1768: Attaching callback with only one CAN interface defined causes hardfault

[ral613]

The Environment

• Mbed LPC1768 on an Embedded Artists base board.
• LPCXpresso v5.2.4 [Build 2122].
• Using CAN2, LEDs and serial I/O interfaces.

The Problem

Hardfault interrupt on LPC1768 when attaching a callback function to a CAN object and only one CAN object is defined.

How to Reproduce

Run the following app on an mbed LPC1768 . Note the blue LEDs of death as HardFault_Handler() is called.

#include "mbed.h"

BusOut myleds(LED1, LED2, LED3, LED4);
CAN can2(p30, p29);
bool got_message = false;

extern "C" void HardFault_Handler ()
{
    error("Hit HardFault handler!\r\n");
}

void can_callback()
{
    got_message = true;
}

int main()
{
    CANMessage msg;

    printf("main()\n");
    can2.attach(can_callback, CAN::RxIrq);
    int i = 0;
    while(1) {
        printf("loop()\n");
        if (got_message)
        {
            got_message = false;
            can2.read(msg);
            printf("Message received: %d\n", msg.data[0]);
        }
        wait(0.2);
        myleds = i++;
        i %= 16;
    }
}

The Workaround

Create a second unused CAN object

CAN can1(p9, p10);

and run the app again. The app now works OK.

The Cause

Checking IER on both CAN peripheral modules when only one of them is enabled causes a hard fault.

The error is in function can_irq_set(), at line 167 of file …/mbed/targets/hal/TARGET_NXP/TARGET_LPC176x/can_api.c.

[mbednotifications]

I assume that this can be fixed by first checking if the CAN peripheral is
active,

Something similar is happening here:
// Have to check that the CAN block is active before reading the Interrupt
// Control Register, or the mbed hangs
void can_irq_n() {
uint32_t icr;

if(LPC_SC->PCONP & (1 << 13)) {
    icr = LPC_CAN1->ICR & 0x1FF;
    can_irq(icr, 0);
}

if(LPC_SC->PCONP & (1 << 14)) {
    icr = LPC_CAN2->ICR & 0x1FF;
    can_irq(icr, 1);
}

}

I don't have the hardware available now but I could look at it later this
week if nobody cares to beat me to it.

On Friday, September 6, 2013 11:55:20 AM UTC-7, Richard Low wrote:

The Environment

• Mbed LPC1768 on an Embedded Artists base board.
• LPCXpresso v5.2.4 [Build 2122].
• Using CAN2, LEDs and serial I/O interfaces.

The Problem

Hardfault interrupt on LPC1768 when attaching a callback function to a CAN
object and only one CAN object is defined.
How to Reproduce

Run the following app on an mbed LPC1768 . Note the blue LEDs of death as
HardFault_Handler() is called.

#include "mbed.h"

BusOut myleds(LED1, LED2, LED3, LED4);
CAN can2(p30, p29);
bool got_message = false;

extern "C" void HardFault_Handler ()
{
error("Hit HardFault handler!\r\n");
}

void can_callback()
{
got_message = true;
}

int main()
{
CANMessage msg;

printf("main()\n");
can2.attach(can_callback, CAN::RxIrq);
int i = 0;
while(1) {
    printf("loop()\n");
    if (got_message)
    {
        got_message = false;
        can2.read(msg);
        printf("Message received: %d\n", msg.data[0]);
    }
    wait(0.2);
    myleds = i++;
    i %= 16;
}

}

The Workaround

Create a second unused CAN object

CAN can1(p9, p10);

and run the app again. The app now works OK.
The Cause

Checking IER on both CAN peripheral modules when only one of them is
enabled causes a hard fault.

The error is in function can_irq_set(), at line 167 of file
…/mbed/targets/hal/TARGET_NXP/TARGET_LPC176x/can_api.chttps://github.com/mbedmicro/mbed/blob/master/libraries/mbed/targets/hal/TARGET_NXP/TARGET_LPC176X/can_api.c
.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/56
.

[mbednotifications]

Fixed and verified that it works by modifying the 'can_interrupt' testcase.
#63

On Sunday, September 8, 2013 7:15:23 PM UTC-7, Joris wrote:

I assume that this can be fixed by first checking if the CAN peripheral is
active,

Something similar is happening here:
// Have to check that the CAN block is active before reading the Interrupt
// Control Register, or the mbed hangs
void can_irq_n() {
uint32_t icr;

if(LPC_SC->PCONP & (1 << 13)) {
    icr = LPC_CAN1->ICR & 0x1FF;
    can_irq(icr, 0);
}

if(LPC_SC->PCONP & (1 << 14)) {
    icr = LPC_CAN2->ICR & 0x1FF;
    can_irq(icr, 1);
}

}

I don't have the hardware available now but I could look at it later this
week if nobody cares to beat me to it.

On Friday, September 6, 2013 11:55:20 AM UTC-7, Richard Low wrote:

The Environment

• Mbed LPC1768 on an Embedded Artists base board.
• LPCXpresso v5.2.4 [Build 2122].
• Using CAN2, LEDs and serial I/O interfaces.

The Problem

Hardfault interrupt on LPC1768 when attaching a callback function to a
CAN object and only one CAN object is defined.
How to Reproduce

Run the following app on an mbed LPC1768 . Note the blue LEDs of death as
HardFault_Handler() is called.

#include "mbed.h"

BusOut myleds(LED1, LED2, LED3, LED4);
CAN can2(p30, p29);
bool got_message = false;

extern "C" void HardFault_Handler ()
{
error("Hit HardFault handler!\r\n");
}

void can_callback()
{
got_message = true;
}

int main()
{
CANMessage msg;

printf("main()\n");
can2.attach(can_callback, CAN::RxIrq);
int i = 0;
while(1) {
    printf("loop()\n");
    if (got_message)
    {
        got_message = false;
        can2.read(msg);
        printf("Message received: %d\n", msg.data[0]);
    }
    wait(0.2);
    myleds = i++;
    i %= 16;
}

}

The Workaround

Create a second unused CAN object

CAN can1(p9, p10);

and run the app again. The app now works OK.
The Cause

Checking IER on both CAN peripheral modules when only one of them is
enabled causes a hard fault.

The error is in function can_irq_set(), at line 167 of file
…/mbed/targets/hal/TARGET_NXP/TARGET_LPC176x/can_api.chttps://github.com/mbedmicro/mbed/blob/master/libraries/mbed/targets/hal/TARGET_NXP/TARGET_LPC176X/can_api.c
.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/56
.