Repository navigation
Zephyr signed fit uboot cortex a - #3790
Merged
jasonrandrews merged 14 commits intoOct 5, 2026
Merged
jasonrandrews merged 14 commits into
jasonrandrews merged 14 commits into
Conversation
… Cortex-A New Learning Path under embedded-and-microcontrollers, with the TI AM62L EVM as the worked example: sign a Zephyr image into a FIT, build the public key into U-Boot through CONFIG_DEVICE_TREE_INCLUDES, add a boot command that starts Zephyr only after bootm verifies it, and prove it on the board with a wrong-key and a tampered image. Adds Roy Jamil to the contributors list.
…sal tests out, redo diagrams - Page 3 now uses Workbench for Zephyr in VS Code: import the AArch64 toolchain, add a Zephyr 4.4.2 workspace (Texas Instruments template), create and build the application, with annotated screenshots of each wizard and of the build. - New page 7 holds the optional tests: second image signed with key-b, tampered copy, fit_check_sign on both, and the two refusals on the board. Pages 3-6 keep only the trusted image; the production review is now page 8. - Page 1 shows a generic Cortex-A boot chain first, then the AM62L one; the diagrams use Arial and the AM62L labels match the binman packaging (BL1+TIFS, BL31+OP-TEE+SPL). - Prerequisites point at the Workbench for Zephyr Learning Path; further reading links the three Ac6 courses, and page 8 ends with a short pointer to them.
- Cut about a quarter of the prose on every page: tool-option explanations, repeated justifications and asides are gone; the reasons that prevent mistakes stay, one sentence each. - The four Workbench for Zephyr screenshots no longer show the Windows user name in paths, and the activity bar is cropped out.
- The fit_check_sign, mkimage and board-run outputs are complete again: no line is replaced by an ellipsis. - Every output block is followed by a "Lines to look for" call-out that names the lines carrying the verdict, then the explanation of those lines as it stood before the prose was lightened.
- env.sh opens with a board block (BOARD, ZEPHYR_ADDR, FIT_ADDR, BOOT_DEV, UBOOT_DEFCONFIG) and a vendor SDK block; the .its, the CONFIG_PREBOOT fragment and the make lines take their values from it. The expanded fragment is byte-identical to the former literal one. - Each page keeps a generic section first and marks the TI-specific parts (SDK install, card image, boot switches, console, binman make variables) as AM62L EVM sections. HS-FS and HS-SE are TI's names for the development and production security states, given once on the first page. - The porting section on the last page is a checklist that maps onto the env.sh blocks. - Review fixes: the ROM/partition claims are no longer stated for every Cortex-A SoC; any key passes in the development state; the wrong-key mkimage output is shown in full; timestamps agree across pages.
- The boot log, the trusted boot and the two refusals on pages 6 and 7 now reproduce what the EVM printed (Tera Term recording of 11 September): TF-A and SPL lines, the U-Boot banner with SoC AM62LX SR1.1 HS-FS, the real read times and FIT timestamp, and Zephyr's second-core line. - The demo-only echo lines of the recording are replaced by the echo the Learning Path's own boot command prints. - Page 8 quotes the same banner line.
…boxes - The range brackets in both diagrams now open downwards, towards the boxes they group, and start and end exactly on those boxes' edges; the blue bracket no longer starts in the gap before the Zephyr box. - Bracket labels are centred on their bracket and no wider than it. - The generic diagram names who checks each box under that box, as the AM62L one does, instead of under the gaps between boxes. - The grey labels say what TI's documentation says: before the key is fused, any key passes.
- The prerequisites link the extension's Visual Studio Marketplace page. - The setup page links both, in the same sentence form as the published Workbench for Zephyr Learning Path. - The Zephyr build page links the Workbench for Zephyr site where the walkthrough starts.
- About 30% less prose; every command and output a reader runs or checks is still there. - Each idea is explained once: the security states on the production page, the required-signature rule on the boot chain page, the fail-closed && chain on the U-Boot page. - The boot switch section keeps the SW3 setting; the full pincount setting is a link in the troubleshooting list. - env.sh carries a comment per board value instead of two paragraphs. - Removed the branch-byte check, which the header magic check covers, the mkimage -l tip and a duplicate signature-node block.
The Learning Path needed a board. It now runs on the TI AM62L EVM or in QEMU, as one generic story with tabs where the two targets differ. - page 1 gains a comparison table and tabs the target-specific chain - page 2 writes env-am62l.sh or env-qemu.sh and takes the U-Boot source from the TI SDK or from the U-Boot project - pages 4 and 5 use UBOOT_CC, and page 5 tabs how the key reaches the control device tree: CONFIG_DEVICE_TREE_INCLUDES on the EVM, a dumped QEMU device tree merged with signature.dtsi and passed with EXT_DTB - 6-boot-the-board.md becomes 6-boot-the-target.md, "Boot the target": the vendor card image or a 64 MiB disk image, the board or QEMU - page 7 uses BOOT_IMG, so its commands are identical on both targets - the Add Application screenshot gains a QEMU call-out Page 8 drops the "what you've proved" framing: a Summary section says what U-Boot checks with key-a and what the stages below it accept, the boot-path audit table is gone, and the closing section says what to change next instead of listing courses. The FIT signature diagram loses its failure-modes box, which repeated messages the test page already shows. The QEMU track was run end to end with unpatched U-Boot 2025.07: trusted boot, wrong-key refusal and tampered-image refusal.
Every page now presents QEMU first and the AM62L EVM as the same work on hardware: the comparison table, all nine tab blocks, the environment lines, the prerequisites and the board lists. Where an output block is the board's, the text now says so, and page 3 warns that the Workbench screenshots were captured with the EVM selected. Also corrects the name of TI's tool: the AM62L uses Keywriter Lite.
anupras-mohapatra-arm
added
ready_to_publish
jasonrandrews
added
publish
and removed
ready_to_publish
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Before submitting a pull request for a new Learning Path, please review Create a Learning Path
Please do not include any confidential information in your contribution. This includes confidential microarchitecture details and unannounced product information.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the Creative Commons Attribution 4.0 International License.