Claude review: drm/ci: move qualcomm baremetal jobs to lava

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Commit:** `031c0133ee360`

Re-enables the `msm:apq8016` and `msm:apq8096` jobs (removing the dot prefix added in patch 3) and converts them from bare-metal (Google Freedreno farm) to LAVA-based jobs (Collabora farm).

Key changes:

```yaml
-    DEVICE_TYPE: apq8016-sbc-usb-host
+    DEVICE_TYPE: dragonboard-410c
     DRIVER_NAME: msm
-    BM_DTB: https://${PIPELINE_ARTIFACTS_BASE}/arm64/${DEVICE_TYPE}.dtb
+    DTB: apq8016-sbc-usb-host
+    FARM: collabora
     GPU_VERSION: apq8016
-    BM_KERNEL_EXTRA_ARGS: clk_ignore_unused
-    RUNNER_TAG: google-freedreno-db410c
-  script:
-    - ./install/bare-metal/fastboot.sh || exit $?
+    KERNEL_IMAGE_NAME: "Image.gz"
+    KERNEL_IMAGE_TYPE: ""
+    RUNNER_TAG: mesa-ci-x86-64-lava-dragonboard-410c
+    LAVA_FIRMWARE: qcom-lava
```

**Analysis:**

1. The `DEVICE_TYPE` changes from the kernel DT name (`apq8016-sbc-usb-host`) to the LAVA device name (`dragonboard-410c`), while the DTB is now specified separately via `DTB: apq8016-sbc-usb-host`. This is the correct LAVA convention.

2. The old `BM_KERNEL_EXTRA_ARGS: clk_ignore_unused` for apq8016 is dropped. The commit message doesn't mention whether this is still needed under LAVA. The original comment said "disabling unused clocks congests with the MDSS runtime PM trying to disable those clocks and causes boot to fail." If this issue still exists, it would manifest as boot failures in CI, which would be caught immediately. Given this has been tested (pipeline link in cover letter), this is presumably no longer needed or handled differently.

3. The old `BM_KERNEL_EXTRA_ARGS: maxcpus=2` for apq8096 is also dropped. Same reasoning applies.

4. `CONFIG_QCOM_SPMI_VADC=y` is added to `arm64.config`, which is needed for the apq8016 board to function under LAVA.

5. Both jobs get `parallel: 3` and `LAVA_FIRMWARE: qcom-lava`.

6. New expected failures added: `core_setmaster@master-drop-set-user,Fail` and `msm/msm_mapping@memptrs,Fail` for both boards, plus cursor legacy failures for apq8016.

**Verdict:** Clean conversion from bare-metal to LAVA. The dropped kernel args are the only point worth noting, but given the series has been tested in CI, this is acceptable.

---

---
Generated by Claude Code Patch Reviewer