Hi Quentin, On Tue, 31 Jan 2023 at 03:54, Quentin Schulz <quentin.sch...@theobroma-systems.com> wrote: > > Hi Kever, > > On 1/31/23 03:53, Kever Yang wrote: > > Hi > > > > I think I do this modify for those soc not have TPL, or else it > > will cause the CI build error. > > > > > > TPL/ddr binary is mandatory in all rockchip SoCs now, but the mainline > > U-Boot may not > > > > able to provide the the available TPL, will need ddr init binary from > > rockchip rkbin repository > > > > instead. > > > > So the policy is clear, the binary output on mainline U-Boot rockchip > > platform: > > > > - TPL available: u-boot-rockchip.bin with TPL, SPL, ATF, U-Boot proper; > > > > - TPL not available: u-boot.itb with ATF, U-Boot proper. > > > > idbloader.bin generate by the > > TPL(ddr init binary instead) and SPL via mkimage cmd; > > > > If I understand correctly, the TPL might not be generated by U-Boot but > externally provided by Rockchip in binary form? For those cases, the > defconfig is expected to only request an SPL build I assume. However, > idbloader.bin (from rockchip, U-Boot TPL replacement) shall still be > passed with U-Boot SPL to mkimage? > > In that case, I guess we could have a Kconfig option like > CONFIG_EXTERNAL_TPL dependent on !TPL and have the user pass a TPL > external path environment variable (EXTERNAL_TPL) the same way we do for > BL31 or TEE in cmd_binman. > > Then we should be able to adapt rockchip-u-boot.dtsi to manage this and > still build a valid u-boot-rockchip.bin, e.g. replace > https://source.denx.de/u-boot/u-boot/-/blob/master/arch/arm/dts/rockchip-u-boot.dtsi#L23 > #ifdef CONFIG_TPL > multiple-data-files; > > u-boot-tpl { > }; > #endif > with > #if defined(CONFIG_TPL) || defined(CONFIG_EXTERNAL_TPL) > multiple-data-files; > # ifdef CONFIG_TPL > u-boot-tpl { > }; > # else > blob-ext { > filename = "${EXTERNAL_TPL}" > }; > # endif > #endif > > or something more appropriate for binman, like adapting u-boot-tpl to > handle external blobs for example.
That sounds right to me. But I would like a new entry type for this (rockchip-tpl ?), as we do with BL31, etc. Also, avoid the #idef around multiple-data-files - it is just easier to always enable that. Regards, Simon
