docs(wiki): 2026-11-15 status refresh - DS stage image authored+live (st0=st1=020a0481; DS 0x84000420, two boots); draw-issue mac=0x22 and stores remain; STATE-CURRENT snapshot

1 files changed, 56 insertions, 0 deletions

diff --git a/STATE-CURRENT.md b/STATE-CURRENT.md
index 2fb0241..3cf5c30 100644
--- a/STATE-CURRENT.md
+++ b/STATE-CURRENT.md
@@ -1,5 +1,61 @@
 # Open-IOV Current State
 
+> **Snapshot: 2026-11-15 ~13:55 (SESSION 157 continued; HEAD `692a7eb0` ==
+> `vgpu/master`; the working tree carries only the campaign parameter file
+> `tools/guest-diag/params-r1624.txt`; supersedes the 2026-11-14 ~10:55 entry
+> below).**
+> The graphics front moved two rungs since the last snapshot: the **MME
+> multi-part gate closed** (multi-frame keep-windows; the part-launch poison
+> eliminated; the 2799-word pack completes; `MME-PACK-REARM` fires; the fixed-CB
+> region is aliased; the released calls run macros through to `DRAW_IAB`,
+> `mac=0x22`), and the **DS stage image is authored and live**. The DS gate was
+> pinned precisely: the synthesized prelude's j1 override is overwritten at
+> execution time by the window's own later `0x2054` slot-1 bind pair, so the
+> module now authors a 640-B VTG SPH at `family+0x900` (`SPH-AUTHOR ... first
+> word 020a0481 w6=0000f00f`) and **rewrites the window's own slot-1 pair
+> operand** to that page (`SPH-SLOT1-REWRITE old=0x01e00300 new=0x01e00900`,
+> `692a7eb0`). The DS latch now reads the authored image for both vertex
+> slots: `DS-IMAGE st0 w0=020a0481 code=ff007321 st1 w0=020a0481
+> code=ff007321`, and the DS value fell **`0x86040626 -> 0x84000420`**
+> (`M_PIPE_STAGE TI` + SPH5/SPH10/SPH18) — reproduced on two boots (c749,
+> c754). A hostile second-family window (a 640-B candidate with `w6=0`, whose
+> range overlaps `+0x900`) is guarded: the candidate scan rejects `w6==0` and
+> falls back to the builtin image (c748's `badf1201` poison class).
+>
+> **What remains is the draw-issue call and the stores.** The 728-B draw
+> window's released macros are `0x33/0x39/0x1/0x0`; `DRAW_IAB` (`mac=0x22`) is
+> not among them, so the draws still do not store (`output retirement wait:
+> 0/1030`, `wrote=0`, `px0=0`). c716 ran `0x22` on a 324-B window whose held
+> calls had been released by a cross-lane `MME-PACK-REARM` before the launch;
+> the next levers are the 324-B window class or the window's own call/SA slot
+> census. Fixes along the way: the pump atomic-sleep hoist (the r707e VAS
+> verify moved out of `svc.chan_lock`, with `in_atomic()` guards — c741 showed
+> the `BUG: scheduling while atomic`), the DDX ring-head tracker decode
+> (`sema_ddx_tracker_auth`) plus the `sema_gr_scan_top` raise to 12 GiB, the
+> default-off MME pack carrier, and `timeout -k` in the mission tooling.
+>
+> **X/GL start remains the run-rate limiter.** The per-boot first-generation
+> lottery is frequent: c756 reached authoring/pack-parts only, c757/c758 saw no
+> chain (and c753 none) — the follow-up A/B boots mostly did not survive X plus
+> the draw-synth window. Remaining named gates: the stage type/context bits
+> (SPH5/SPH10/SPH18), the `mac=0x22` call and stores, then the window/screen
+> present.
+>
+> CUDA front unchanged (backgrounded while graphics holds the device; its
+> 10-30 status stands): `41343c63` green (`victory_mail` `VADD bad=0` +
+> `STATUS OK`; `vecadd_loop` 100 launches PASS); the fresh-boot battery's red
+> class is D2H-stage with release-discipline fixes staged; the SESSION 143
+> Vulkan fence regression is not re-measured.
+>
+> Live at check (read-only): host uptime 3 h 53 m (fresh boot 2026-11-15
+> ~10:03); `vgsp` loaded (refcount 0) + mdev
+> `c0ba3c27-7923-4131-a127-759315bd7e72` + QEMU cycle c758 (PID 41318, 12 GiB
+> file-backed guest, kvmfr ivshmem, since 13:27); display-channel polls
+> advancing (X alive); `gl_frames` 2/2 black (`px0=0`); `nvidia-smi` reports no
+> GPU (the claim is held); pstore empty; no Xid/Oops; no reboot during the
+> check. The open-iov.org wiki was refreshed alongside this snapshot (11 pages,
+> revs 24155-24165; 68 pages audited, 0 problems).
+>
 > **Snapshot: 2026-11-14 ~10:55 (SESSION 157 continued; HEAD `339da842` ==
 > `vgpu/master`; working tree clean; supersedes the 2026-11-13 ~11:10 entry
 > below).**