Graphics test failure on lemans Ridesx devices

[bhargav0610]

We are seeing graphics test failure even though display test case is passing and display connected

kmscube is unable to successfully initialize legacy DRM modesetting on the MSM driver after Weston had previously acquired DRM master. Although the logs clearly show a connected and enabled display on card1 (DP 3 @ 1920×1080), atomic modesetting is not supported, forcing kmscube to rely on legacy KMS, which is sensitive to both DRM master ownership and correct DRM card selection. After stopping Weston, DRM master is not cleanly or reliably handed over, and kmscube likely opens the wrong DRM device leading it to report “no connected connector” despite the display being present and valid

Lava job: LAVA | Scheduler | Jobs | 185555

[WARN] 2026-03-13 15:35:33 - display_is_cpu_renderer gbm not supported, falling back to auto (may touch Wayland/Weston).[INFO] 2026-03-13 15:35:33 - ---------------- EGLINFO pipeline detection (select one) ----------------[INFO] 2026-03-13 15:35:33 - EGLINFO: Pipeline type: GPU (hardware)[INFO] 2026-03-13 15:35:33 - EGLINFO: Pipeline=WAYLAND platform:[INFO] 2026-03-13 15:35:33 - EGLINFO: EGL driver name: msm[INFO] 2026-03-13 15:35:33 - EGLINFO: GL_VENDOR: freedreno[INFO] 2026-03-13 15:35:33 - EGLINFO: GL_RENDERER: FD663[INFO] 2026-03-13 15:35:33 - ---------------- End EGLINFO pipeline detection --------------------------[INFO] 2026-03-13 15:35:33 - Weston is running, stopping it so kmscube can modeset (DRM master)[INFO] 2026-03-13 15:35:33 - Running kmscube with --count=999 ...[FAIL] 2026-03-13 15:35:33 - KMSCube : Execution failed (rc=255) — see ./KMSCube_run.logno connected connector!failed to initialize legacy DRM[INFO] 2026-03-13 15:35:33 - Restoring Weston after failure+ true+ /home/lava-185555/0/tests/1_KMSCube/Runner/utils/send-to-lava.sh KMSCube.res<<<LAVA_SIGNAL_TESTCASE TEST_CASE_ID=KMSCube RESULT=FAIL>>>+ set +xReceived signal: TEST_CASE_ID=KMSCube RESULT=FAIL

Thanks,
Bhargav

[smuppand]

@bhargav0610 I triggered a job using the same build as the one referenced in the previously shared LAVA job, and I did not encounter the issue.

https://lava.infra.foundries.io/results/188287

[smuppand]

Observation from the recent failure job.
The monitor is detected correctly. The failure is most likely because kmscube is opening the wrong DRM card.

From the logs..
card1-DP-3 status=connected enabled=enabled type=DP modes=52 first=1920x1080 cur=1920x1080@60
Connected display (sysfs): card1-DP-3(DP,1920x1080)

It does not pass /dev/dri/card1 to kmscube. So kmscube auto-selects a DRM device, and on multi-card platforms it may pick /dev/dri/card0 or another card with no connected real connector. Then kmscube reports

no connected connector!
failed to initialize legacy DRM