09:54 <Kwiboo> macromorgan: CounterPillow: current upstream TF-A only have a buffer for 128 KiB fdt, so that may need to be increased if symbols is included in dtb?

09:54 <Kwiboo> see https://github.com/ARM-software/arm-trusted-firmware/blob/master/plat/rockchip/common/params_setup.c#L40-L41

10:00 <qschulz> Kwiboo: I guess we could also strip the symbols before passing it to TF-A?

10:34 <qschulz> well except if TF-A passes the DTB to U-Boot proper as well... which I guess is what's the plan with this bloblist thing?

14:36 <Kwiboo> qschulz: I think my reasoning was wrong, it would not be the linux compiled dtb that is used by u-boot and gets passed to TF-A, it should be the control fdt included in FIT that is used/passed to TF-A and U-Boot proper

14:38 <Kwiboo> U-Boot proper may then as part of a bootflow replace the control FDT with content from a .dtb-file on filesystem, this replaced FDT would then be passed to OS

14:42 <qschulz> Kwiboo: there's been some push (from Google I think?) to use the exact same DTB binary for every part of the boot process

17:24 <macromorgan> on the atags front... should I simply hard-code the address to where the memory ones are (do we think it's that stable) or should I instead iterate through the atags memory to find the sdram header?