On Mon, Dec 06, 2021 at 05:11:38PM -0700, Simon Glass wrote:
[snip]
Changes in v7:
- Fix 'linst' typo
- Switch to using an empty file for qemu virt
- Switch to using an empty file for qemu virt
- Switch to using an empty file for qemu virt
- Reword commit message
- Bring in files from Linux instead
- Bring in files from Linux instead
- Rewrite the commit message
- Use 'empty' instead of 'fake'
- Use 'empty' instead of 'fake'
- Switch to using an empty file for qemu virt
- Bring in files from Linux instead
- Fix 'Is' typo in commit message
- Drop FDTSRC_PASSAGE for now
- Drop FDTSRC_PASSAGE for now
- Add new patch to add a Kconfig for boards with a prior stage
Alright, so I know François said to move to empty rather than dumping/"fake" dts file for this go-round. But I still don't like it. What Andre said vs the highbank patch is right. You develop against the binding.
Further, you see the problem of "it's board-specific how to update a platform dtb" having an answer of "hammer it in to U-Boot". The follow-up problem of "it's only going to get worse" has another solution to it I believe, which is that you're supposed to use capsule updates for this, if you're SystemReady compliant and able to ship the DTB on-device (I assume François or someone else can chime in here if I'm wrong).
I have more to say, but I think I need to reply that to the doc patch that adds your rationale for how DT should work in U-Boot.