On Fri, Jan 27, 2023 at 06:55:46PM +0100, Mario Kicherer wrote:
On 2023-01-27 18:10, Tom Rini wrote:
Yes, all of the platforms that define the value (since it roughly means "ROM set this value in something we can check") instead of enum'ing it still compile that file and now fail to build.
Okay, I think I understand your point now. I am not sure what's the best way to proceed here. Should I try to build all targets that contain BOOT_DEVICE_NOR and add a #define for BOOT_DEVICE_NOR2 if my patch really breaks the build? Or should I just add a #define BOOT_DEVICE_NOR2 to every #define BOOT_DEVICE_NOR?
It looks like a change would be necessary in these files then:
arch/arm/include/asm/arch-am33xx/spl.h arch/arm/include/asm/arch-orion5x/spl.h arch/arm/mach-k3/include/mach/j721e_spl.h arch/arm/mach-k3/include/mach/j721s2_spl.h arch/microblaze/include/asm/spl.h arch/powerpc/include/asm/spl.h
Ah, OK. Keep in mind that MMC1/MMC2 are for different physical MMC devices on a given platform. I think this is more like the case where you should be able to override spl_nor_get_uboot_base at the board level to say if you're loading A or B?
Ah yes, it's been a while since I wrote this patch originally. spl_nor_get_uboot_base() chooses between A and B and BOOT_DEVICE_NOR2 signals that I should choose the other one now.
I think you should be able to solve this problem without making further changes upstream, yes.