Dear Tien Fong,
In message 1520485397.10181.12.camel@intel.com you wrote:
This looks fine as a concept but I am not keen on the implementation.
This patchset has been going through many rounds and a lot of time spending in review, and it is already working and being tested. I still
That's life....
have a lot subsequent patches pending on this patchset. I would suggest to accept this patchset, then we can enhance it to driver model in later.
No. The usual approach is to _first_ clean up the code, _before_ it gets added. there have been just too many cases that interest in cleaning up disappeared quickly once the unclean version was pulled into mainline.
- It should use driver model (only) in U-Boot proper. If there is
some SPL problem then add a specific function or feature for SPL.
We can doing this in later since it is require sometime to figure out and testing.
We can also do it now, and that is better.
- It should not be necessary ti manually init subsystems - driver
model does this for you
This is for initializing storage driver in very early SPL, where loading from storage to configure some critical HW need to done first.
The key problem is that your approach does not scale. We had the same discussion elsewhere.
- You can use the uclass name to find things
Yeah, once it is converted to driver model, we can use the uclass for searching HW info in DTS.
Then please convert it.
Thanks.
Best regards,
Wolfgang Denk