AST2500 Watchdog Timer Device Tree Entry

[Joel Stanley]

Hey Nolan,

On Fri, Feb 17, 2017 at 5:31 AM, Nolan Hergert <nolan.hergert at gmail.com> wrote:
> To Joel in particular,

Feel free to put me in To if you want my attention. Your email bubbles
up the to of my to-read list when you do so.

> Thanks for making a WDT driver for the Aspeed chips already! It makes diving
> into these things a lot easier.

No worries :) I'm glad our hard work is paying off.
>
> I didn't understand some of the things in the aspeed-g5.dtsi file though:
>
> wdt1: wdt at 1e785000 {
> compatible = "aspeed,ast2500-wdt";
> reg = <0x1e785000 0x1c>;
> };
>
> wdt2: wdt at 1e785020 {
> compatible = "aspeed,ast2500-wdt";
> reg = <0x1e785020 0x1c>;
> status = "disabled";
> };
>
> wdt3: wdt at 1e785040 {
> compatible = "aspeed,wdt";
> reg = <0x1e785074 0x1c>;
> status = "disabled";
> };
>
>
>   * Register lengths are 0x1C instead of 0x20. The skipped register in the
> datasheet for AST2500 is determining which hw modules to reset, which sounds
> like something we'd want to configure.'

You're correct. We are ignoring this feature of the ast2500 watchdog
IP with the current driver. I believe this an oversight; we should add
the functionality when matching against the "aspeed,ast2500-wdt"
compatible string.

>   * wdt3 compatible string is different.
>   * wdt3 register offset appears wrong.

That's for picking these up. I suspect no one has tried to enable WDT3
so we have not hit the issue. These are both bugs in the device tree.

If you're in a position to contribute fixes to any of these issues I
encourage you to submit patches. The fixes should go upstream first,
and then we can backport the patches to the openbmc tree.

I've got some documentation about how our kernel development process
works. Please have a read. Any and all feedback is welcome!

 https://github.com/openbmc/linux/wiki/DevelopmentProcess
 https://github.com/openbmc/linux/wiki/SubmittingPatches

Cheers,

Joel