Le lun. 22 mars 2021 à 19:24, Joshua Watt <jpewhacker@...> a écrit :
On 3/22/21 9:59 AM, Yann Dirson wrote:
Hi Trevor,
Le lun. 22 mars 2021 à 15:47, Trevor Woerner <twoerner@...> a écrit :
On Mon 2021-03-22 @ 02:42:12 PM, yann.dirson@... wrote:
This supports both the 2GB and 4GB versions of the board. This is not done with 2 different machine definitions since only u-boot has to change between those two configurations, but with a NANOPIM4_HW variable to set in local.conf.
Traditionally in meta-rockchip this is done using two separate machine files with all the common things factored into a common include file. See tinker-board and tinker-board-s as well as all the rock-pi-4* definitions for examples.
I would prefer if the same thing was done here.
Damned that was how I did my first patch, I just felt it was much better this way :(
Digging up that original commit re rerolling.
Wouldn't it be useful to have a standard way to specify such hardware variants, that would be recognized as such by the layer index ?
FWIW, I'd rather u-boot could automatcially detect what variant of the board it's running on and do the right thing (usually, just selecting the proper DTB from the FIT image). The only reason I did not do this on the rock-pi-4 is because there doesn't appear to be a way for u-boot to detect which variant it's on :/
Wait, it's only the dtb used by u-boot which is different in this case, the kernel uses a single dtb for both variants. Those probably get embedded inside the bootloader, in any case they are not those embedded in the fitImage.
I have no clue how easily the two boards could be told one from the other (let alone if that can be done safely), and I feel that would rather be uboot-side work, rather than something to be solved on the yocto side.
Ah, I was referring to the DTB used by the Kernel (which u-boot can switch), not the DTB used by u-boot itself (which AFAIK is compiled into u-boot and cannot be changed).
