Compared to the RFC, this take addresses Rob's and David's comments. I added an extra patch to check for an explicit #address-cells property in all interrupt provider nodes. Also I tried to filter out unresolved phandle references, so that it just gives up in case this is an overlay DT. Changelog below. =============== The interrupt-map properties can be quite tricky, as they are already hard to read in their source form, and depend on at least two nodes. The first patch adds a separate test for interrupt providers, to avoid redundant warnings for missing properties later. Patch 2/3 is the actual check routine, to verify some features of the map. In particular it should be able to spot missing fields or fields using the wrong number of cells. Most of the fields cannot easily be verified, but at the least the phandle to the interrupt controller and the total number of cells in the property should be valid. It actually find problems in the Linux kernel tree's .dts files. The third patch is more a proof of concept, not sure it should be merged as is, since it deviates from the current output scheme: It adds more output in case a map is considered broken, so the issue can be found much easier. I would be happy with it living on the list for now. Please let me know what you think. Cheers, Andre Changelog RFC .. v1: - add extra test for #interrupt-cells and #address-cells - report expected and actual interrupt-map property length - drop (now) redundant warnings - explicitly check for overlay DT - consider nested interrupt-map properties Andre Przywara (3): checks: Add interrupt provider test checks: Validate interrupt-map properties checks: interrupt-map: Dump entries on error checks.c | 192 ++++++++++++++++++++++++++++++++++++- tests/bad-interrupt-controller.dts | 7 ++ tests/bad-interrupt-map.dts | 21 ++++ tests/run_tests.sh | 2 + 4 files changed, 221 insertions(+), 1 deletion(-) create mode 100644 tests/bad-interrupt-controller.dts create mode 100644 tests/bad-interrupt-map.dts -- 2.14.1
