On Tue, Dec 01, 2015 at 01:25:50PM +0100, Christophe Ricard wrote: > For example during an i2c_device_probe where an i2c slave device > describe in devicetree has an interrupts property. > i2c_device_probe (drivers/i2c/i2c-core.c), retrieves irq property from > of_irq_get which will looks for an "interrupts" property in > of_irq_parse_of (drivers/of/irq.c). > of_irq_get will then call irq_create_mapping (kernel/irq/irqdomain.c) > which will set the irq_type retrieved during the interrupts node > parsing. Found it now thanks. > This will allow from an i2c slave drivers to configure an interrupt > handler matching the exact devicetree data for the interrupts property > of the i2c slave node. Makes sense. > Now for the same kind of i2c driver using acpi description, the GpioInt > polarity/type is at the moment never kept in the irq property. > It is possible to check that following about the same path... > i2c_device_probe (drivers/i2c/i2c-core.c), retrieves irq property from > acpi_dev_gpio_irq_get but does not save the irq_type. > This would allow not to have to use an additional gpio field and all > the configuration step to configure the gpio interrupt correctly in a > device driver and taking a real benefit of the GpioInt acpi keyword > compare to GpioIo keyword. > Most the of the drivers based on acpi description retrieve gpio number > to assign an interrupt and a fix polarity. I believe my patchset > proposal would improve this and allow to > be much closer with devicetree. > Do you see any issue with this ? No, but I wonder if it would be better to do this in acpi_dev_gpio_irq_get() instead of acpi_find_gpio() which gets called everytime a GPIO is looked up? -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html