Skip to content

kboot: hand the iBoot-loaded ISP firmware to Linux on T6041 - #71

Merged
RyanTheTide merged 2 commits into
aurora-silicon:mainfrom
AxxzyWasTaken:t6041-isp-handoff
Nov 6, 2026
Merged

RyanTheTide merged 2 commits into
aurora-silicon:mainfrom
AxxzyWasTaken:t6041-isp-handoff

Conversation

@AxxzyWasTaken

@AxxzyWasTaken AxxzyWasTaken commented Nov 4, 2026 •

Copy link
Copy Markdown
Contributor

Hands the ISP firmware over on T6041 (M4 Max). iBoot loads it and the Linux driver starts it as iBoot left it, but on J616C it moves between boots, so the J616s addresses in the device tree don't hold.

This fills /reserved-memory/isp-firmware-text and isp-firmware-data from this boot's ADT /arm-io/isp0 segment ranges: reg, no-map, and iommu-addresses at the firmware's own addresses. It does nothing if the ISP is disabled or a placeholder is missing, and it refuses an implausible range.

aura#48 does the same job for J614s on T6040, so the two don't overlap.

Testing

  • Booted on a 32-core J616C on macOS 26.6.2 firmware, with linux#931's device tree on a 7.3 kernel. The camera came up as FaceTime HD Camera and a 6-second capture gave a live image, checked by its brightness only.
  • An earlier version has run daily on this machine since 23 October, on a 7.2 tree.
  • Builds with ESP_STAGE2=1 CHAINLOADING=1; clang-format clean. No new host test.

Needs the isp alias in the device tree. On J616C the handoff didn't run until it was added; linux#914 has it.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Merge-train review (aura, M4-series): the change reads well and CI is green, but the commit(s) have no Signed-off-by: line, and a DCO sign-off is a hard requirement before we can merge a contributor's commit. We can't add one for you. Could you amend with git commit -s --amend (or git rebase --signoff origin/main for several commits) and force-push the branch? I'll pick it straight back up.

🤖 Generated with Claude Code

On T6041 iBoot loads the ISP firmware and the Linux driver starts it as
iBoot left it (aurora-silicon/linux#384, Jonathan Goyvaerts' T6040 ISP
support). The TEXT and DATA move between boots, so fill the kernel DT's
/reserved-memory/isp-firmware-text and isp-firmware-data placeholders
from the ADT segment-ranges (reg, no-map, iommu-addresses at the
firmware's own addresses) instead of the m1n1 heap setup, which does not
apply here.

Signed-off-by: AxxzyWasTaken <axxzytte@gmail.com>
dt_set_isp_fwdata() now branches to dt_set_isp_t6041() on T6041, so the
harnesses that compile it need the chip ID and a stub.

Signed-off-by: AxxzyWasTaken <axxzytte@gmail.com>
@AxxzyWasTaken

Copy link
Copy Markdown
Contributor Author

Signed off and force-pushed. No code change.

@RyanTheTide
RyanTheTide merged commit bab8fec into aurora-silicon:main Nov 6, 2026
8 checks passed
@RyanTheTide

Copy link
Copy Markdown
Contributor

Merge train (MTR, 06/11): reviewed. It only runs for chip_id == T6041 with an enabled isp node and the /reserved-memory placeholders, and it validates the segments (16K-aligned, at most 64 MiB) before writing. The integration/aurora J616C tree has no ISP node yet, so until a DT with placeholders lands this is a no-op there, and other SoCs are untouched. CI is green. The hardware proof is the author's J616C camera run.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants