On 12/05/2012 11:20 PM, Roland Stigge wrote:

Hi Wolfgang,

On 05/12/12 20:01, Wolfgang Grandegger wrote:

quoted

quoted

+	for (i = 0; i < block->ngpio; i++) {
+		status = gpio_request(block->gpio[i], "gpioblock dev");

You could use the name of the GPIO block.

OK.

quoted

quoted

+		if (status)
+			goto err1;
+
+		irq = gpio_to_irq(block->gpio[i]);
+		if (irq >= 0 &&
+		    !test_bit(FLAG_IS_OUT, &gpio_desc[block->gpio[i]].flags) &&
+		    !gpio_block_is_irq_duplicate(block, i)) {
+			status = request_irq(irq, gpio_block_irq_handler,
+					     IRQF_TRIGGER_FALLING,
+					     block->name, block);
+			if (status)
+				goto err2;
+
+			block->irq_controlled = true;
+		}
+	}

There is no need to request IRQs if "O_NONBLOCK" is specified.

Sure? Regarding this, I found: "The poll() function shall not be
affected by the O_NONBLOCK flag." [1]

Ah, didn't know that.

quoted

I observed that the read returns once immediately (without blocking)
after reboot. I did not look into that yet.

Didn't happen to me. Can you tell how this can be reproduced?

I think there is an interrupt pending for some reason. It might not be
the case with your setup.

Feel free to add my:

"Tested by: Wolfgang Grandegger [off-list ref]".

Will do some more tests when time permits, especially with mixed GPIO
input and output.

Wolfgang.