Hi Kever, On Tue, 7 Jul 2020 at 06:01, Kever Yang <[email protected]> wrote: > > Hi Simon, > > On 2020/6/26 上午9:12, Simon Glass wrote: > > Hi Heiko, > > > > On Fri, 19 Jun 2020 at 04:45, Heiko Stuebner <[email protected]> wrote: > >> From: Heiko Stuebner <[email protected]> > >> > >> If the newly added fit-generator key-options are found, append needed > >> signature nodes to all generated image blocks, so that they can get > >> signed when mkimage later compiles the .itb from the generated .its. > >> > >> Signed-off-by: Heiko Stuebner <[email protected]> > >> --- > >> arch/arm/mach-rockchip/make_fit_atf.py | 57 +++++++++++++++++++++++++- > >> 1 file changed, 56 insertions(+), 1 deletion(-) > > We have to stop the growth of these ad-hoc scripts or U-Boot with > > become a tangled mess of vendor-specific scripts. > > > > I sent a series to convert more of rockchip to use binman. Can you > > please take a look at that? > > > > It deletes the fit_spl_optee.sh script. We need to delete make_fit_atf as > > well. > > The fit_spl_optee.sh is much more simple because it use optee.bin > directly for u-boot.itb, while make_fit_atf > > needs to parse bl31.elf and get the start address for different blob. > Any suggestion on how this > > could be implement in binman?
See for example elf.GetSymbolAddress() which reads an ELF symbol value. This is used in the u-boot-spl-bss-pad entry. It should be pretty simple. You can create new entry types (in the etypes/) dir to do all sorts of things fairly easily. Regards, Simon
