Hi,
I have observed a spike in DT presentations at events in 2018 and a relative smaller set in 2019.
Many discussions raise a lot of good points, there are lists of things to solve or proposals in various sites but there is no clear path to a widely acceptable solution. It feels like the traditional model of "contribute code first" reaches its limits here: Device Tree is not just a device tree. It is a pivotal piece of information manipulated by too many components from Trusted Firmware A to Mac OSX and U-boot. And there are many legitimate "owners" of pieces of the DT...
We need to find a way to build consensus across projects. Mail discussions are good but I feel this does not fully help getting closer to having up streamed code and specs into projects such as (and not limited to): - Trusted Firmware - OP-TEE - Hafnium - U-Boot - Linux - Android - FreeBSD - DeviceTree - EBBR - Firmware Framework A - Parsec
To that end, what about collectively building a Technical Report with a number of editors in charge of different sections?
An example outline could be: - the problems we try to solve (and what is outside of scope) . board diversity handling (build/boot/runtime) . cape/hat/FPGA handling (build/boot/runtime) . product value chain handling (soc vendor/product vendor/sw vendor...) . trust of DT . device assignment to application or VM - DT evolution . actors (UML meaning) . responsibility boundaries (signatures...) . DTB format evolution . DTBo . DT at build time for many upstream projects - cross project incremental roadmap highlights . introduce DT signature
Of course the Technical Report shall be supported by different PoCs on many topics such as DTB format so that the architecture is future proof and we can start small.
What are your views?
Cheers
FF
boot-architecture@lists.linaro.org