Skip to content

scripts/dtc: apple-dart-pd-check: let a new DTB inherit its SoC's known gaps - #2560

Merged
RyanTheTide merged 1 commit into
integration/aurorafrom
dtcheck/dart-pd-inherited
Nov 20, 2026
Merged

RyanTheTide merged 1 commit into
integration/aurorafrom
dtcheck/dart-pd-inherited

Conversation

@RyanTheTide

Copy link
Copy Markdown
Contributor

Problem. The dt-checks step "DART power domains, base vs head" compares findings per DTB name. A DTB that the base revision does not have reports every gap it contains as new, even the ones it inherits. On #2525 the new firmware-profile variant t8112-j413-fw27 includes t8112-j413.dts, so J413's existing /soc/iommu@24e808000 gap (one of the known ones) shows up as (new) and fails the job.

Change. For a DTB the baseline lacks, its findings are looked up in the baseline DTBs of the same SoC (the root apple,tNNNN compatible), matching node path, DART compatible and problem. A match is reported as (inherited from <board>) and does not fail; anything else the new DTB adds is still new. DTBs that already exist in the baseline keep the strict per-board comparison. The summary line gains an inherited by new DTBs count. A self-test (scripts/dtc/apple-dart-pd-check-test.py, dtc + pylibfdt fixtures) now runs in the job.

Testing.

Still to run. This PR's own dt-checks job; then a re-run of #2525's dt-checks once this is merged.

Self-test cases
  • unchanged tree passes
  • new variant with the same gap as its board: inherited, passes
  • new variant with a gap of its own: that gap fails as new, the inherited one does not
  • new DTB of a different SoC with a gap at the same path: new, fails
  • new gap on a board that exists in the baseline: fails, even when a sibling board of the same SoC has it
  • same node and compatible but a different problem (disabled power domain): new, fails
  • without --baseline-dtb-dir every gap fails

Assisted-by: LLM (Claude:Opus 5.5)

…wn gaps

The DART power-domain check compares findings per DTB name, so a DTB
the base revision does not have reports every existing gap as new. A
firmware-profile variant such as t8112-j413-fw27 includes
t8112-j413.dts and so repeats J413's known iommu@24e808000 gap under a
new name, which fails the check although nothing changed in the DART
nodes.

For a DTB the baseline lacks, look its findings up in the baseline
DTBs of the same SoC (root "apple,tNNNN" compatible), matching node
path, DART compatible and problem. A match is reported as inherited and
does not fail; anything else the new DTB adds is still new. DTBs that
exist in the baseline keep the strict per-board comparison.

Add a self-test with synthetic fixtures (the J413 fw27 case, a variant
with a gap of its own, a new DTB of another SoC, a new gap on an
existing board, a changed problem on the same node) and run it in the
dt-checks job.

Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
@RyanTheTide
RyanTheTide merged commit c92ed85 into integration/aurora Nov 20, 2026
2 checks passed
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.

1 participant