08:01 <austriancoder> NishanthMenon: the trick with the register to detect single or dual core A53 works. Would I make sense to add generic support in U-Boot to patch the dtb before jumping to linux - somewhere in am642_init.c? This would remove the need for different device-trees or overlays to support the same hw layout with a single or dual core A53.

10:24 <NishanthMenon> austriancoder: I think a generic function might be helpful for "features". What I was hoping was to apply x or y or z overlay than have bootloader directly monkey with properties.

10:25 <NishanthMenon> I know Tom is trying to crack down to keep UBoot code as simple as possible and we are questioning how eeprom code itself has gotten so complex..

10:26 <NishanthMenon> My rationale for.overlays has been increasing dependence on UBoot for fixups and variety of bootloader's beyond UBoot in the ecosystem's ability to do dtb surgery like what UBoot does..

12:03 <austriancoder> NishanthMenon: how would the overlay flow would look like? Use default dtb and depending on some board detection load some extra overlay? I think this forces us to create a new version of the bootloader when we model a new layout. Atm we select a fit configuration based on board detection.

12:04 <austriancoder> Or would the kernel be a place to read the register and bring only available cores up?

12:04 <austriancoder> It costs about 5 seconds boot time with a dtb with two cores on a single core system.

12:07 <NishanthMenon> austriancoder: u-boot does'nt care about number of A53s. our designs and chip spin offs are careful about that. so the superset dtb will work to start a53 up, to start off the kernel, two flows possible: bootloader has two choices: a) dtb surgery. b) apply overlay (could include additional changes depending on type of board, all in one shot)..

12:07 <NishanthMenon> today u-boot can do dtb surgery AND apply overlays on top.

12:21 <austriancoder> NishanthMenon: we can include two dtbs in the fit - one for single and one for dual core. Would work too. So do you think that this is the better way to do it instead of dtb surgery? We are highly interested in upstreaming any possible U-Boot patch.

13:23 <NishanthMenon> austriancoder: I believe that was exactly the approach that iot2050 took for AM65

13:25 <austriancoder> NishanthMenon: okay.. will look into the multiple dtb approach

13:25 <NishanthMenon> austriancoder: cool..

13:26 <NishanthMenon> austriancoder: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/ti if you look at the iot2050 variants, I think your platform could use a simpler approach

13:31 <NishanthMenon> context, though: at the time of iot2050 integration, we did'nt have overlays as something accepted in k.org back then..