On Mon, 4 Jan 2021, Sudeep Holla wrote:
On Mon, Dec 21, 2020 at 10:38:52AM -0700, Rob Herring wrote:
On Fri, Dec 18, 2020 at 9:43 PM François Ozog francois.ozog@linaro.org wrote:
Hi
I assume this needs to be analyzed from System Device Tree perspective: https://trustedfirmware-a.readthedocs.io/en/latest/components/psa-ffa-manife...
That's not what we're reviewing upstream[1].
Thanks Rob for including me. I wasn't aware of such formal proposal of bindings from the firmware/system perspective. I agree and happy to re-use the bindings if required.
However, the idea with bindings for OSPM(mainly kernel with no hypervisor) was to keep it minimal or even zero(which is not possible as the specification lacks any mechanism to get the UUID list for the partitions.
Also, where are these system/firmware DT bindings getting reviewed ? There are so many bindings used in some of these firmware/systems without much(even zero) review of them on a list, so I am bit worried to use them as is. Should they not get reviewed on some list if not existing device-tree mailing list. Sorry if they are on some list which I am not yet part of, happy to get added to them.
They have been discussed during the System Device Tree calls and on the system-dt mailing list: https://lists.openampproject.org/mailman/listinfo/system-dt. We have recordings and minutes of the meetings. Have a look at this video for an overview on System Device Tree: https://connect.linaro.org/resources/san19/san19-115/.
Most of the times System Device Tree is not used at runtime and Linux is typically not the primary target. System-DT is most often used at build time / configuration time before the target board is turned on. A tool named "Lopper" (https://github.com/devicetree-org/lopper) generates multiple traditional Device Trees from a single System Device Tree.
Things don't always look exactly the same in System Device Tree compared to regular Device Trees. So, we don't have to necessarely reuse the openamp,domain bindings as they are today. However, it would be good to align them where it makes sense to reduce complexity. We have gone through a similar exercise for the RemoteProc bindings and I think it improved both System-DT and regular DT as a result.
I would like to hear a bit more about your use-case to plot together the best course of action.
Cheers,
Stefano