On Thu, Mar 31, 2022 at 12:27 AM Hawkins, Nick <nick.hawkins@xxxxxxx> wrote: > On Tue, Mar 29, 2022 at 9:38 PM Hawkins, Nick <nick.hawkins@xxxxxxx>> wrote: > > >> I am in the process of rewriting the timer driver for Linux but have hit a dilemma and I am looking for some direction. The registers that represent the watchdog timer, and timer all lay in the same register region and they are spread out to the point where there are other controls in the same area. > > > >> For instance with our watchdog controls we have: > > > >> @90 the countdown value > >> @96 the configuration > > > >> And for our timer we have: > >> @80 the countdown value > >> @94 the configuration > >> @88 this is actually our timestamp register but is being included in with the timer driver currently to call clocksource_mmio_init. > > > >> What would be your recommendation for this? I was considering creating a gxp-clock that specifically points at the timestamp register but I still have the issue with gxp-timer and gxp-wdt being spread across the same area of registers. > > > I think this is most commonly done using a 'syscon' node, have a look at the files listed by > > I found an example and copied it although I have a couple questions when it comes to actually coding it. Can that be here or should I post these questions in the patch that actually concern the file? > > st: timer@80 { > compatible = "hpe,gxp-timer","syscon","simple-mfd"; > reg = <0x80 0x16>; > interrupts = <0>; > interrupt-parent = <&vic0>; > clocks = <&ppuclk>; > clock-names = "ppuclk"; > clock-frequency = <400000000>; > > watchdog { > compatible = "hpe,gxp-wdt"; > }; > }; I'd have to study the other examples myself to see what is most common. My feeling would be that it's better to either have a "hpe,gxp-timer" parent device with a watchdog child but no syscon, or to have a syscon/simple-mfd parent with both the timer and the watchdog as children. Arnd