On Sat, 19 Oct 2019, Aymeric Vincent wrote:
We track mainline Linux DTS so this local change should be reverted.I'm not against removing this local change if indeed your workaround also works but your sentence is not valid: sometimes the DTS has missing pieces or conveys broken information. As you can see we have quite a handful of such local changes to other DTSs in the same directory.
Our local changes in that directory are for cases where mainline dts is missing features. I don't think we need to add a global timer if the private watchdog timer gives the same functionality.
I think we still need to address the issue of having multiple timers competing and the question is still open. We probably need to use a generic way to choose the main timer, now that there are several possibilities on a given ARM platform.
Agreed that we need to handle dts with multiple timer candidates for hardclock. Do we really need to make it configurable though? What do we lose by just selecting the first one that attaches?