Re: [PATCH net-next 3/3] net: ethernet: ti: am65-cpts: implement PTP_SYS_OFFSET_PRECISE
From: James Clark
Date: Thu Nov 19 2026 - 20:27:20 EST
Hi Rob,
Thanks for the review.
On Thu, Nov 19, 2026 at 05:06:14PM -0600, Rob Herring wrote:
> > There is no other user in the tree currently. If in the future something
> > else in the kernel wants to use it, then the solution would be a small
> > driver to mediate this.
>
> Would doing this imply a DT change? If so, your binding is not future
> proof.
I don't think a driver would need a DT change. The property only says
which CPTS input the GTC push event is routed to. It does not say which
bit the GTC exports. The role of the driver would be to mediate the
choice of bit in the GTC_CFG0_PUSHEVT register. The driver would bind
to the ti,am654-gtc node. It would not need a regmap of its own, since
the register is ordinary MMIO, so the node can stay a syscon. The CPTS
driver would use the same phandle to find that driver and ask for the
bit, instead of writing the register itself.
> Not the kernel's job to validate the DT. If something in DT is optional,
> then whether it is missing or wrong doesn't really matter.
[ ... ]
> Is that true for *every* kernel version? Again, better user experience
> IMO to just warn if something is missing and continue on enabling
> whatever you can.
On reflection, I agree. For v2, if the ti,gtc-push lookup fails, I will
make the driver warn and continue without enabling
PTP_SYS_OFFSET_PRECISE.
James