12:41 <chewitt> hays petitboot doesn't support extlinux.conf boot

12:42 <chewitt> but n2+ has a hardware switch that allows you to ignore that

12:43 <chewitt> flick the switch to the right and use u-boot on your choice of boot media and should be well

12:43 <chewitt> and 'all' should be well

12:44 <chewitt> last time I asked .. tobetter had zero interest in adding support for extlinux.conf

12:49 <hays> hmm ok thanks.

19:39 <aeroraptor> can anyone here speak to this issue?

19:39 <aeroraptor> https://hub.libre.computer/t/pps-gpio-for-s905x/1503/9

19:39 <aeroraptor> specifically the error: pps-gpio pps@0: failed to map GPIO to IRQ: -6

19:48 <xdarklight> aeroraptor: I've no experience with pps-gpio but I suspect it's due to the fact that it's calling gpiod_to_irq() internally which is not supported by the meson GPIO driver

19:49 <aeroraptor> oh wow, so GPIOs on amlogic chips just... cannot use interrupts?

19:54 <aeroraptor> I don't know if the odroid-c2 uses (used?) a different GPIO driver, but apparently they ended up writing a kernel module

19:54 <aeroraptor> https://github.com/hardkernel/linux/blob/odroidc2-3.14.y/drivers/pps/clients/pps-gpio.c#L58

19:54 <aeroraptor> but I don't know how applicable that is to current stuff

19:55 <xdarklight> GPIOs can use interrupts, but the gpiod_to_irq() API is not supported

19:56 <xdarklight> I don't know if pps-gpio can take an interrupt directly -> that's what we use for other IRQ consumers

19:57 <xdarklight> to give an example, here's how we link the Ethernet PHY's interrupt line (which is routed to GPIOZ_15): https://github.com/torvalds/linux/blob/master/arch/arm64/boot/dts/amlogic/meson-gxm-khadas-vim2.dts#L240-L241

20:01 <aeroraptor> ahh interesting

20:01 <aeroraptor> I think the odroid example using a kernel module is prooooobably the right way to go here

20:04 <xdarklight> looking at the odroid driver: it simply uses gpio_to_irq() instead of gpiod_to_irq() which is basically the same API just with a GPIO number (deprecated) instead of GPIO descriptor. most likely the Amlogic vendor GPIO driver support the GPIO to IRQ translation which is not supported with the mainline driver (yet)

20:06 <xdarklight> there's been a lengthy discussion on the mailing list about supporting the GPIO to IRQ translation upstream as well, both Jerome Brunet and Heiner Kallweit worked on it but there were some issues that they could not resolve

20:12 <narmstrong> There’s not a 1-to-1 or many-to-1 mapping from gpios to an interrupt line, this makes supporting gpiod_to_irq harder since this api has been very badly designed a long time ago, using a DT interrupt property is not the right API to use, the PPS gpio code should be modernized

21:07 <aeroraptor> interesting

21:07 <aeroraptor> I know this sort of thing tends to fall by the wayside, since not a whole lot of folks use PPS inputs on SoCs like these (just like not many folks are using the GPIOCLK outputs)