watchthelight 2026-10-21
First run, 2026-10-21 (Legion N-214, XeGfx 0.0.2, macOS 26.7.1, BIOS Hybrid):
- Loads and runs.
kmutil loadthen stages 1 to 4 one by one; the machine stayed up, WindowServer kept running, and the panel never blinked. The self-judging test daemon passed stages 1-3 outright. - Stage 4 is not a real pass. drm and the i915 module initialised (56 init routines), and the display check held (the GGTT entry behind the panel unchanged, pipe A still counting), but i915 never bound the GPU: no probe output and no DRM node. 0.0.3 now fails stage 4 with ENODEV in that case and logs the whole newbus tree, so the next run shows where the hand-off breaks (vgapci, its
drmnchild, or i915's ID match). - Probe values (read-only):
| Field | Value |
|---|---|
| device / revision | 8086:7D67 rev 06 (Arrow Lake-S die) |
| GMD_ID graphics / display | 0x03118004 (12.70) / 0x0380000C (14.0) |
| BAR0 / BAR2 | 16 MiB at 0x6018000000 / 256 MiB at 0x6000000000 |
| stolen memory base | 0x70800000 (the firmware framebuffer) |
MTL_PCODE_STOLEN_ACCESS |
1, so i915's Wa_22018444074 applies: no CPU access to stolen memory through a BAR |
| boot surface | PLANE_SURF 0xFF000000, GGTT entry 0x800003 |
| PCH TCO watchdog | TCO_LOCK set, timer halted: no hardware watchdog on this machine |
Evidence (logs, registry dumps, verdicts) is in the private project repo under evidence/native/N-214. Next run: N-215 with 0.0.3.