Interrupt Latency 2.6.33.7 rt30

Interrupt Latency 2.6.33.7 rt30

[Christian Kraus]

Hello
I have wrote a GPIO-driver, that toggle the value of a GPIO when a 
interrupt is comming from an another GPIO.
When I measure the latency between the interrupt and the toggled Pin 
it's about ten times higher than without the rt-patch.
(without Patch = 8-11 us
  with Patch = 30-110 us)

Can anybody help me please?

I have a Beagleboard with an Omap-Arm running.

Here are the changes in gpio_keys.c:

static irqreturn_t gpio_keys_isr(int irq, void *dev_id)
{
     struct gpio_button_data *bdata = dev_id;
     struct gpio_keys_button *button = bdata->button;

     BUG_ON(irq != gpio_to_irq(button->gpio));

     if (button->debounce_interval)
     {
         mod_timer(&bdata->timer,
             jiffies + msecs_to_jiffies(button->debounce_interval));



     }
     else
     {
     //    schedule_work(&bdata->work);

         if(gpio_get_value(158)==0)
         {
             gpio_set_value(158, 1);
         }
         else
         {
             gpio_set_value(158, 0);
         }


     }

     return IRQ_HANDLED;
}


Thx Christian

Re: Interrupt Latency 2.6.33.7 rt30

[Uwe Kleine-König]

On Thu, Aug 25, 2011 at 11:03:33PM +0200, Christian Kraus wrote:
> Hello
> I have wrote a GPIO-driver, that toggle the value of a GPIO when a
> interrupt is comming from an another GPIO.
> When I measure the latency between the interrupt and the toggled Pin
> it's about ten times higher than without the rt-patch.
> (without Patch = 8-11 us
>  with Patch = 30-110 us)
It's expected and known that an rt kernel has a worse average latency.
Having said that I cannot say if the increase you see is "too much".

The best option you have is asking ftrace what happens in these 30-110
us after making sure you have a decent sched_clock implementation.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Interrupt Latency 2.6.33.7 rt30

[Thomas Gleixner]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 919 bytes --]

On Fri, 26 Aug 2011, Uwe Kleine-König wrote:

> On Thu, Aug 25, 2011 at 11:03:33PM +0200, Christian Kraus wrote:
> > Hello
> > I have wrote a GPIO-driver, that toggle the value of a GPIO when a
> > interrupt is comming from an another GPIO.
> > When I measure the latency between the interrupt and the toggled Pin
> > it's about ten times higher than without the rt-patch.
> > (without Patch = 8-11 us
> >  with Patch = 30-110 us)
> It's expected and known that an rt kernel has a worse average latency.

Nonsense.

> Having said that I cannot say if the increase you see is "too much".
> 
> The best option you have is asking ftrace what happens in these 30-110
> us after making sure you have a decent sched_clock implementation.

The problem is simply that the interrupt is forced threaded, so it
depends on the scheduler when it goes on the CPU, while on !RT the
code runs in hard interrupt context.

Thanks,

	tglx