Camera testcases are skipping on rb3gen2-core-kit

[qcom-anilyada]

[Libcamera_cam] - No sensors reported by 'cam -l' - marking SKIP
[Camera_RDI_FrameCapture] SKIP – Media node not found

Latest Job: https://lava.infra.foundries.io/scheduler/job/376472

while on wrynose branch it is passing

When tracing back, the issue was observed around the below PRs:

qualcomm-linux/meta-qcom#2712

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29155835384 [SKIPPED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29056606175 [PASSED]

qualcomm-linux/meta-qcom#2684

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28843072885 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28853513248 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28867146019 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29246527224 [SKIPPED]

qualcomm-linux/meta-qcom#2711

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29074060595 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29223322233 [SKIPPED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29247504880 [SKIPPED]

Note:

• I think the script may need an update so it can distinguish whether camera sensor detection is coming from a H/W perspective or a S/W perspective.
• In the current scenario, even though this is a S/W issue, the logs still make it look like a H/W or setup issue because they show cam -l not detected.

[qcom-anilyada]

created issue in meta-qcom as well: qualcomm-linux/meta-qcom#3002

[smuppand]

Anil (@qcom-anilyada) I don’t think this is a test script issue. The three named PRs do not modify camera support. Please double-check the setup and confirm whether the sensors are connected properly. If they are, is the camx DTB loading correctly? Have you checked the uefi/ efivars to verify?

[smuppand]

The last passing meta-qcom commit used lava-test-plans 62f75dda. The first skipping commit changed only the lava-test-plans revision to ed4ac266 via meta-qcom PR #2708.

That update changed rb3gen2 from:

{% extends "devices/rb3gen2-core-kit" %}

to:

{% extends "devices/qcs6490-rb3gen2" %}

The old camera path conditionally performed:

cp .../cdt_vision_kit.bin .../cdt.bin

when ENABLE_CAMERA_POSTPROCESS=true. The camera test plan explicitly sets this flag.

The new QDL template does not perform that camera-specific CDT substitution. Consequently, rb3gen2 likely boots with the default CDT rather than the Vision Kit CDT, so camera DT nodes, sensors and /dev/media* never appear.

This also explains why wrynose passes: it remains pinned to the old lava-test-plans revision 62f75dda, while master migrated to QDL.

The currently pinned lava-test-plans revision still has the same problem: the old camera-aware template remains present but is no longer inherited by the meta-qcom rb3gen2 device.

Testkit reporting problem

The test messages are misleading:

• cam -l reporting zero cameras is always converted to SKIP.
• Missing /dev/media* is always converted to SKIP.

For a LAVA job explicitly tagged has-camera, those conditions mean the expected camera stack failed to materialize. They should be FAIL with diagnostic classification such as:

FAIL: Camera hardware expected, but no runtime camera devices were registered

SKIP should be reserved for targets where camera hardware is genuinely not expected.

Recommended fixes

1. Fix lava-test-plans first by restoring the Vision Kit CDT selection in the QDL deployment path.

2. Pass an explicit CAMERA_EXPECTED=1 parameter from camera-tagged LAVA jobs.

3. Add a shared camera applicability helper that distinguishes:

   • camera not expected → SKIP
   • camera expected but DT/media/sensor absent → FAIL
   • camera present but capture fails → FAIL

4. Capture evidence before returning:

   • DT model, compatible and enabled sensor nodes
   • /dev/media*, /dev/video*, and media topology
   • CAMSS driver binding
   • complete cam -l output and exit status
   • captured dmesg errors