Hi Simon,
On 2/6/25 1:46 PM, Simon Glass wrote:
Hi Quentin,
On Wed, 5 Feb 2025 at 06:00, Quentin Schulz <quentin.sch...@cherry.de> wrote:
Hi Naoki,
On 1/29/25 2:25 PM, FUKAUMI Naoki wrote:
After adding the below configs, binman produces an error.
CONFIG_SPL_DFU
CONFIG_SPL_ENV_SUPPORT
CONFIG_SPL_USB_DWC3_GENERIC
CONFIG_SPL_USB_GADGET
binman: Section '/binman/simple-bin-spi': Symbol '_binman_u_boot_any_prop_size'
in entry '/binman/simple-bin-spi/mkimage/u-boot-spl/u-boot-spl-nodtb':
Entry 'u-boot-any' not found in list
(u-boot-tpl-nodtb,u-boot-tpl-dtb,u-boot-tpl,u-boot-spl-nodtb,u-boot-spl-dtb,u-boot-spl,mkimage,fit,simple-bin-spi)
Add "no-write-symbols;" to u-boot-tpl and u-boot-spl.
I'll be frank with you, I have no clue what's happening :)
I means that simple-bin-spi has a binary inside it, u-boot-spl, which
uses binman_sym_declare(ulong, u_boot_any, size). Most likely it is
the one at the top of common/spl.c
This symbol is filled in (by binman, at build-time) with the size of
the U-Boot binary. If there is no such U-Boot image then it cannot do
this.
Why is it not failing right now? Why does it fail when you enable
CONFIG_SPL_DFU, CONFIG_SPL_ENV_SUPPORT, CONFIG_SPL_USB_DWC3_GENERIC,
CONFIG_SPL_USB_GADGET for Rock-Pi-4? This doesn't generate the
u-boot.itb file we need for simple-bin-spi as we anyway take the fit
blob generated by simple-bin instead.
It would make sense to always fail if this relies on u-boot.itb to be
built for simple-bin-spi as well?
I suggested to use only blobs from simple-bin in simple-bin-spi but
according to
https://docs.u-boot.org/en/latest/develop/package/binman.html#image-dependencies
this is undefined behavior. Maybe we should use a template
(https://docs.u-boot.org/en/latest/develop/package/binman.html#templates)
and rebuild everything for the SPI image as well. I saw such a patch in
your VBE part H series I believe, maybe it makes sense to have this
merged separately?
This feels somewhat wrong though, if I'm not mistaken every binary used
in simple-bin and simple-bin-spi is the same, to the exception of how
the idbloader.img is generated (-T rkspi/rksd), so it doesn't make a lot
of sense to me to go through a complete rebuild of all binaries, instead
of just running mkimage a second time with the same set of binaries. But
if that's the only way to have something with defined behavior AND
working, then so be it.
Cheers,
Quentin
