u-boot-stm32mp: Fix fit signing#105
u-boot-stm32mp: Fix fit signing#105UVV-gh wants to merge 1 commit intoSTMicroelectronics:scarthgapfrom
Conversation
ST implementation of signing conflicts with the one from oe-core. Signed-off-by: Vyacheslav Yurkov <uvv.mail@gmail.com>
|
Hello, I understand you would like to have a implementation closer to oe-core tools. The proposed patch will probably also break some other mechanism. We will check this. Now, current implementation is supposed to work like this (maybe not like you would wish). Did you follow this wiki for your test ? |
|
For everyone's, please be aware v6.1.1 brings major correction on this use case. see: https://wiki.st.com/stm32mpu/wiki/STM32_MPU_ecosystem_release_note#v6.1.1 |
|
Confirming regression: Signed FIT key selection injects the wrong public key into U-Boot DTB I can reproduce the issue described in this ticket. It matches the “wrong key injected into U-Boot DTB” class of problem and breaks Signed FIT setups (especially with
When building a signed kernel FIT image, the public key injected into the U-Boot DTB does not match the key actually used to sign the FIT configuration ( Typical outcomes:
Result: the build fails (or later boot verification fails) because DTB key injection does not match the signing mode.
The fix should ensure that:
This prevents the DTB “required key” mismatch and aligns with the upstream OE-Core behavior in
|
|
Correction will be delivered on v6.2.0 (Wk08). |
ST implementation of signing conflicts with the one from oe-core from uboot-sign class. This is just a sample fix, which you can use, but I would rather rework it completely by reverting bee6a4b in the first place and use what's already present in oe-core.
Right now the implementation looks like a copy-paste of signing code from uboot-sign class, which also omits a few details and forces users to always set FIT_SIGN_INDIVIDUAL.
Without this fix the build works, but when you try to boot you'll get a verification failed error: