I've written a (working) kernel module which uses returns a gpio_chip instance to specify how reads and writes (using the module-provided GPIOs) should be handled.
static struct gpio_chip template_chip = { .label = "my_module", .owner = THIS_MODULE, .get_direction = my_module_gpio_get_direction, .direction_input = my_module_gpio_direction_in, .get = my_module_gpio_get, .direction_output = my_module_gpio_direction_out, .set = my_module_gpio_set, .to_irq = my_module_gpio_to_irq, .can_sleep = true,};static int my_module_gpio_probe(struct platform_device *pdev){ struct my_module *my_module = dev_get_drvdata(pdev->dev.parent); struct my_module_platform_data *pdata = dev_get_platdata(my_module->dev); struct my_module_gpio_data *my_module_gpio; int ret; printk(KERN_INFO "my_module_gpio_probe\n"); my_module_gpio = devm_kzalloc(&pdev->dev, sizeof(*my_module_gpio), GFP_KERNEL); if (my_module_gpio == NULL) return -ENOMEM; my_module_gpio->my_module = my_module; my_module_gpio->gpio_chip = template_chip; my_module_gpio->gpio_chip.ngpio = NUM_GPIOS; my_module_gpio->gpio_chip.parent = &pdev->dev; ...I now want my (software generated) GPIOs to be able to produce an interrupt when my module changes the GPIO values.
I want to support both the old /sys/class/gpio interface and the new chardev interface.
While googling, I've found lots of examples of how to consume interrupts in a kernel module.
What do I need to do to produce interrupts for my module's GPIOs?