Kicksecure ISO live amnesic (USER Session) boot fails on novacustom-v540tu; works on x230 and nv4x_adl

[tlaurion]

Before filing a bug report

Follow-up to #2008 and #2186. #2186 merged in master.

Hardware details

• Board: novacustom-v540tu (failing). Working: x230 (EOL_x230-maximized), novacustom-nv4x_adl.
• GPU: iGPU-only (Intel Meteor Lake, device ID 7d55).
• Heads/Dasharo: Dasharo Heads-v0.2.1-3144-g7850f7a (DMI BIOS string), vendor-installed by Novacustom.
• ROM: vendor (Dasharo) image, maximized.

Please describe the problem

Describe the bug

#2186 ("initrd: boot dd-written hybrid ISOs in mount-usb and usb-autoboot") fixes #2008 ("kicksecure iso doesn't boot under Heads") on x230 and novacustom-nv4x_adl. On novacustom-v540tu the Kicksecure ISO fails to boot only in live-amnesic mode ("LIVE Mode | USER Session | disposable use"): the machine auto-reboots before/at the LXQt desktop.

The failure is specific to the live-amnesic USER session. Non-live-amnesic modes boot fine on the same machine: "LIVE Mode | SYSMAINT Session" (sysmode) boots to console, and the other grub entries are unaffected.

To Reproduce

1. Boot the Kicksecure 18 live ISO on novacustom-v540tu via Heads (dd-written hybrid ISO path from initrd: boot dd-written hybrid ISOs in mount-usb and usb-autoboot #2186).
2. Select "LIVE Mode | USER Session | disposable use".
3. The machine auto-reboots before the desktop.
4. Select "LIVE Mode | SYSMAINT Session" instead: boots to console fine. Both entries work on x230 and novacustom-nv4x_adl.

Expected behavior

Live amnesic USER session boots to the LXQt desktop with autologin, as on x230 and nv4x_adl.

Additional context

Full sysmode capture (the mode that was successfully logged, since live-amnesic reboots before a capture is possible): https://gist.github.com/tlaurion/f9d3070cd0126a190fe52eec8732c302 (kicksecure_sysmode_v540tu.log).

The two modes pass an identical kernel cmdline except the session target (graphical.target + autologin vs systemd.unit=sysmaint-boot.target + noautologin), so the failure is at graphical/display bring-up, not the overlay. The v540tu is the only machine whose i915 driver does not initialize cleanly. Relevant lines from the gist above:

• L1926: i915 0000:00:02.0: [drm] Found METEORLAKE (device ID 7d55) display version 14.00
• L2360: i915 0000:00:02.0: [drm] [ENCODER:249:DDI B/PHY B] failed to retrieve link info, disabling eDP
• L2393-L2437: panel power-sequencing WARN + Call Trace:
  • L2394: i915 0000:00:02.0: [drm] drm_WARN_ON(intel_dp->pps.vdd_wakeref)
  • L2395: WARNING: CPU: 8 PID: 1012 at drivers/gpu/drm/i915/display/intel_pps.c:755 intel_pps_vdd_on_unlocked+0x32b/0x340 [i915]
  • L2402: RIP: 0010:intel_pps_vdd_on_unlocked+0x32b/0x340 [i915]
  • L2414: Call Trace: ... drm_dp_dpcd_probe / intel_dp_detect / intel_dp_set_edid
• L3565: i915 0000:00:02.0: [drm] *ERROR* GT1: GSC proxy component didn't bind within the expected timeout
• L3566: i915 0000:00:02.0: [drm] *ERROR* GT1: GSC proxy handler failed to init

x230 and nv4x_adl show clean i915 init (zero kernel WARNINGs).

Also noted: the v540tu final ISO cmdline retains intel_iommu=on while x230/nv4x_adl strip it, traced to a per-machine config.user override (config-gui "Debug/Output" menu) rather than a board-config difference. Unlikely the cause (no DMAR faults on v540tu), but worth checking.

This looks like an upstream i915/Meteor Lake display issue surfaced by the graphical autologin session. Next steps: reproduce in plain Debian trixie, and capture a live-amnesic boot over the Heads serial console to confirm the failure at graphical.target.

[tlaurion]

@cc @macpijan @adrelanos Let me need if you need more info

[tlaurion]

Next steps: reproduce in plain Debian trixie, and capture a live-amnesic boot over the Heads serial console to confirm the failure at graphical.target.

@adrelanos a bit bored trying to find proper parameters to remove from kernel cmdline and what to add: let me know and I can provide live+amnesiac logs if i can do sudo there with +single from point of failing on v540tu which otherwise cause automatic reboot.

[tlaurion]

Closed by error