AM625: AM62X GPIO Interrupt Configuration Issue: Failed to Set Trigger Mode

Hello,

There is limited support we can give for custom applications since this is not code from TI and is beyond our support scope.

First, you can check the gpiolib.c driver for the gpiod_to_irq function: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/drivers/gpio/gpiolib.c?h=ti-linux-6.6.y#n3396

As you mentioned, the output is Invalid argument which looks to be caused by this line: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/drivers/gpio/gpiolib.c?h=ti-linux-6.6.y#n3414

You can add debug statements to see which condition in the if statement is being met, resulting in the error.

Second, From searching the error message, you can find the Linux driver that prints the error: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/kernel/irq/manage.c?h=ti-linux-6.6.y#n954

Looks like line 979 set the 'ret' variable used by the switch statement, but does not enter any of the case blocks which results into the default case. You could try adding some debug statements within that function to see how the variables are being set.

The 'irq_set_trigger' function itself gets called by other functions within the same driver, so you can also check those functions.

There is a section within the legacy API about IRQs that might help: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/Documentation/driver-api/gpio/legacy.rst?h=ti-linux-6.6.y#n351

Just to keep expectations clear, TI cannot provide E2E support for custom applications.

Best Regards,

Anshu

After tracing the code in the following location of AM62x  SDK:

ti-processor-sdk-linux-am62xx-evm-09.00.00.03\board-support\ti-linux-kernel\drivers\gpio\gpio-davinci.c

I came across the following function:

static int gpio_irq_type_unbanked(struct irq_data *data, unsigned trigger)
{
struct davinci_gpio_controller *d;
struct davinci_gpio_regs __iomem *g;
u32 mask, i;

d = (struct davinci_gpio_controller *)irq_data_get_irq_handler_data(data);
g = (struct davinci_gpio_regs __iomem *)d->regs[0];
for (i = 0; i < MAX_INT_PER_BANK; i++)
if (data->irq == d->irqs[i])
break;

if (i == MAX_INT_PER_BANK)
return -EINVAL;

mask = __gpio_mask(i);

if (trigger & ~(IRQ_TYPE_EDGE_FALLING | IRQ_TYPE_EDGE_RISING))
return -EINVAL;

writel_relaxed(mask, (trigger & IRQ_TYPE_EDGE_FALLING)
? &g->set_falling : &g->clr_falling);
writel_relaxed(mask, (trigger & IRQ_TYPE_EDGE_RISING)
? &g->set_rising : &g->clr_rising);

return 0;
}
From my understanding, this function only supports edge-triggered interrupts:

IRQ_TYPE_EDGE_FALLING
IRQ_TYPE_EDGE_RISING

However, it seems that level-triggered interrupts (e.g., IRQ_TYPE_LEVEL_HIGH and IRQ_TYPE_LEVEL_LOW) are not supported. Given the hardware capabilities of the AM62x platform, I expected it to support both edge- and level-triggered interrupts.

Support for level-triggered interrupts is essential for certain use cases, such as monitoring GPIO inputs that remain active for extended periods, where high-level or low-level states must be detected. It would significantly enhance the flexibility and functionality of the GPIO controller.

Could you please confirm if this is an intentional limitation? If not, could you provide guidance on how to enable IRQ_TYPE_LEVEL_HIGH and IRQ_TYPE_LEVEL_LOW triggers?

Your assistance and insights on this matter would be highly appreciated.

Best regards,
Oliver

Hi Oliver,

I'll check internally on this. Please ping this thread if you have not received a response by Wednesday next week.

Thanks,

Anshu

Hi Anshu,

I appreciate your assistance. Since it's already past Wednesday, I wanted to kindly follow up on this matter to check if there has been any update from your side.

Thank you again for your help!

Best regards,
Oliver

Hi Oliver,

Its a holiday this week for TI India so I've had no response from the developer. I'll ask them again next week when they return.

Thanks,

Anshu

Hi Anshu,

        Thank you for your assistance, and I look forward to hearing back from you.

Best regards,
Oliver

Hi Oliver,

I've asked for a response again. I will update the thread when I have a reply.

Thanks,

Anshu