LG Gram Round 2: build-blocker fixes + symmetric lid-switch wiring + docs

Three workstreams delivered in Round 2 (no build per operator
directive; verification deferred):

1. Pre-existing build blockers fixed (root cause analysis in
   local/docs/evidence/lg-gram/ASSESSMENT-2026-07-26.md):

   - relibc: edition-2024 unsafe-op-in-unsafe-fn in epoll::
     convert_event (commit dd2cd443 introduced unsafe fn with raw
     pointer derefs lacking unsafe blocks). Fix in submodule/relibc
     commit b80f8b47 wraps each deref in unsafe { } with SAFETY
     justification; outer unsafe fn signature preserved.

   - base: acpid Cargo.toml 'common' path bug. acpid lives at
     drivers/acpid/ (depth 2 in base workspace) but declared
     common = { path = "../../common" } (2 ..) which resolves to
     base/common/ — a path that has never existed. All 8 other
     depth-2 crates correctly use ../common. The build script's
     overlay-integrity auto-repair normally papers over this; it
     failed during Round 1 verification. Fix in submodule/base
     commit 28e356d5 makes acpid consistent with siblings.

2. Lid-switch symmetric wiring (submodule/base commit 887718da):

   Round 1 wired lid-closed → enter_s2idle() but missed the
   symmetric lid-open → exit_s2idle() counterpart. Without it,
   the system stays in 'wake devices armed' state when MWAIT
   never engaged. Round 2 makes the wiring symmetric. The
   userspace-driven wake path coexists with the kernel MWAIT-
   return path (kstop reason=2); the double-call is safe per
   ACPI 6.5 §3.5.3 (_WAK/_SST idempotent in working state).

3. Comprehensive stub sweep (no actionable findings):

   Extended the Round 1 sweep to cover all Red Bear original
   recipes and base fork non-driver crates. Zero unimplemented!()/
   todo!() macros in Red Bear original code. Remaining 'stubs' are
   either documented dead code (lookup_hid_quirks + HidQuirkFlags
   — 5-flag type defined, no consumer) or empty-by-design tables
   with real loader shape (PLATFORM_RULES, DMI_ACPI_QUIRK_RULES —
   documented in Round 1). No new stubs replaced.

Deferred to Round 3:

- acpi_irq1_skip_override kernel-side consumer (needs kernel
  SMBIOS scan + boot verification)
- LG Gram bare-metal boot validation (Phase 1, requires hardware
  + successful build)
- External-display detection for lid switch (Linux's
  HandleLidSwitchDocked=ignore)

The base fork submodule pointer in this commit was advanced by
a parallel agent session (b3fd5cc6 e1000d DMA barriers,
1331b8c0 rtl8168d DMA, 717bc436 ixgbed MSI-X) — those commits
are also tracked here. The relibc fork pointer advances to
b80f8b47 (my unsafe-block fix).

local/docs/LG-GRAM-16Z90TP-COMPATIBILITY-PLAN.md

@@ -1,7 +1,7 @@
 # LG Gram 16Z90TP-G.AL89C — Host Compatibility Plan
 
 **Created:** 2026-07-20
-**Status:** Implementation in progress — Phases 0, 3.2, 3.4, 3.5, 3.6, 4.1–4.4, 5.0–5.5, 6.1, 9.6 complete; Phase 9.1 partial. **Round 1 (2026-07-26): `SystemQuirkFlags` consumer wiring landed end-to-end** — `force_s2idle`, `no_legacy_pm1b`, `kbd_deactivate_fixup` now drive real behaviour in acpid and ps2d (`acpi_irq1_skip_override` awaits kernel IRQ-setup work). `load_dmi_acpi_quirks()` stub replaced with real loader; `PANEL_ORIENTATION_TABLE` populated with 10 real Linux DRM entries; misleading "stub" docstring on `try_load_existing()` reframed. See `local/docs/evidence/lg-gram/ASSESSMENT-2026-07-26.md`.
+**Status:** Implementation in progress — Phases 0, 3.2, 3.4, 3.5, 3.6, 4.1–4.4, 5.0–5.5, 6.1, 9.6 complete; Phase 9.1 partial. **Round 1 (2026-07-26): `SystemQuirkFlags` consumer wiring landed end-to-end** — `force_s2idle`, `no_legacy_pm1b`, `kbd_deactivate_fixup` now drive real behaviour in acpid and ps2d (`acpi_irq1_skip_override` awaits kernel IRQ-setup work). `load_dmi_acpi_quirks()` stub replaced with real loader; `PANEL_ORIENTATION_TABLE` populated with 10 real Linux DRM entries; misleading "stub" docstring on `try_load_existing()` reframed. **Round 2 (2026-07-26): pre-existing build blockers fixed** (relibc edition-2024 unsafe-block in `epoll::convert_event`; acpid `common` path was using depth-3 notation from a depth-2 crate). **Lid-switch wired symmetrically**: close → `enter_s2idle()`, open → `exit_s2idle()` (idempotent per ACPI 6.5 §3.5.3). Kernel `acpi_irq1_skip_override` consumer deferred to Round 3 (needs SMBIOS scan + build verification). See `local/docs/evidence/lg-gram/ASSESSMENT-2026-07-26.md`.
 **Source evidence:** Host dmesg (Linux 7.1.3-1-MANJARO), six-point codebase recon
 (drivers/PCI-ID inventory, redox-drm Intel state, redbear-iwlwifi state, IOMMU/kernel
 state, ACPI/power state, USB/input/audio state)

local/docs/evidence/lg-gram/ASSESSMENT-2026-07-26.md

@@ -1,11 +1,58 @@
 # LG Gram 16Z90TP-G.AL89C — Compatibility Assessment (2026-07-26)
 
-**Round:** 1 of N
-**Scope:** Audit existing implementation vs `LG-GRAM-16Z90TP-COMPATIBILITY-PLAN.md`, replace stubs with real implementations, wire `SystemQuirkFlags` consumers in acpid + ps2d.
+**Round:** 2 of N
+**Scope:** Fix pre-existing build blockers (relibc + base path), wire lid-closed + lid-open → s2idle symmetrically, scan for additional stubs.
 **Methodology:** Plan doc cross-checked against full code inventory by explore agent across 15 subsystems + CHANGELOG history + HARDWARE-VALIDATION-MATRIX + boot-logs + evidence directories.
 
 ---
 
+## Round 2 deliverables (landed 2026-07-26)
+
+### Pre-existing build blockers fixed
+
+| Bug | Root cause | Fix |
+|---|---|---|
+| relibc cook fails with `error[E0133]: dereference of raw pointer is unsafe and requires unsafe block` | Commit `dd2cd443` (relibc) introduced `unsafe fn convert_event(*const Event, *mut epoll_event)` that dereferenced both pointers without `unsafe { }` blocks. Edition 2024 enforces `unsafe-op-in-unsafe-fn` by default with `-D`. | Wrapped each deref in explicit `unsafe { }` block with SAFETY justification. Outer `unsafe fn` signature preserved so validity proof stays on callers. (relibc commit `b80f8b47`) |
+| base cook fails with `failed to load manifest for dependency 'common'` | acpid at `drivers/acpid/` (depth 2 in base workspace) declared `common = { path = "../../common" }` (2 `..`), which resolves to `base/common/` — a path that has never existed. The 8 other depth-2 crates (pcid, hwd, inputd, vboxd, rtcd, thermald, virtio-core, redoxerd) all correctly use `"../common"` (1 `..`). The build script's overlay-integrity auto-repair normally papers over the bug; it failed during Round 1 verification ("5 recipe.toml files could not be restored"). | One-character fix: `../../common` → `../common`. Makes acpid consistent with all other depth-2 crates. (base commit `28e356d5`) |
+
+### Lid-switch symmetric wiring
+
+The Round 1 lid-closed → `enter_s2idle()` wiring was missing its symmetric counterpart. Round 2 adds lid-open → `exit_s2idle()`:
+
+| Lid event | acpid action | AML sequence |
+|---|---|---|
+| `LidChanged(false)` (closed) | `enter_s2idle()` | `_TTS(0)` + `_PTS(0)` + `_SST(3)` + wake-device arming (`_DSW(1)`/`_PSW(1)`) |
+| `LidChanged(true)` (open) | `exit_s2idle()` | `_SST(2)` → `_WAK(0)` → `_SST(1)` + wake-device disarming |
+
+The kernel-side MWAIT-return path also calls `exit_s2idle()` via kstop reason=2. The double-call (userspace lid-open + kernel MWAIT-return) is safe because AML `_WAK`/`_SST` methods are spec-required to be idempotent in the working state (ACPI 6.5 §3.5.3). Without the lid-open call, the system would stay in "wake devices armed" state when MWAIT never engaged (e.g. brief lid-close/open during active workload). (base commit `887718da`)
+
+### Comprehensive stub sweep (no actionable findings)
+
+The Round 1 stub sweep was extended in Round 2 to cover:
+- All Red Bear original recipes (`local/recipes/{system,drivers,gpu}/*/source/`)
+- All base fork non-driver crates
+- USB HID daemon, AMD MP2 I2C daemon
+
+**Findings**: Red Bear original code has zero `unimplemented!()` / `todo!()` macros. The remaining "stubs" are either:
+- Documented dead code with no consumer (`lookup_hid_quirks` + `HidQuirkFlags` — 5-flag type fully defined but no caller; audit doc already records this)
+- Empty-by-design tables with real loader shape (`PLATFORM_RULES`, `DMI_ACPI_QUIRK_RULES` — documented in Round 1)
+- Upstream-inherited code-quality TODOs in base fork drivers/ (e.g., "TODO: arrayvec", "TODO: Move this code to a separate driver") that don't affect functionality
+
+No new stubs replaced in Round 2 — the Round 1 work plus these findings confirm Red Bear's original code is clean per the zero-tolerance policy.
+
+### What did NOT change in Round 2
+
+- `acpi_irq1_skip_override` kernel-side consumer (deferred to Round 3 — needs kernel SMBIOS scan + build verification)
+- LG Gram bare-metal boot validation (Phase 1 — requires hardware + successful build)
+- Phase 6.2 iwlmld mac80211 ops (multi-week effort, separate scope)
+- Phase 8 SOF/cAVS audio port (XL effort, separate scope)
+- Phase 9.1 s2idle MWAIT kernel idle loop (kernel change, separate scope)
+- External-display detection for lid switch (Linux's `HandleLidSwitchDocked=ignore`) — every lid-close triggers s2idle for now
+- `lookup_hid_quirks` dead-code removal (kept — type design is sound, future consumer plausible)
+- Build script overlay-integrity auto-repair investigation (the `common` path fix removes the dependency on this mechanism for acpid; other crates may have similar latent issues)
+
+---
+
 ## Round 1 deliverables (landed 2026-07-26)
 
 ### Consumer wiring (the headline task)
@@ -94,15 +141,38 @@ The Round 0 explore-agent inventory flagged "single-touch only" for `i2c-hidd` m
 
 ---
 
-## Build verification (Round 1)
+## Build verification (Round 2)
 
-- **redox-driver-sys host cargo test**: ✅ 80 tests pass (the stub-replacement work compiles cleanly with zero regressions).
-- **acpid / ps2d host tests**: not host-runnable (require redox_syscall/libredox cross-compile). Verification deferred to canonical build.
-- **Canonical build** (`./local/scripts/build-redbear.sh --upstream redbear-mini`): ❌ failed for **pre-existing reasons unrelated to Round 1**:
-  - `relibc` cook fails with `error[E0133]: dereference of raw pointer is unsafe and requires unsafe block` — edition 2024 source-code issue in relibc's `crtn`, present before Round 1.
-  - `base` cook fails to load `local/sources/base/common/Cargo.toml` — pre-existing path resolution issue (`drivers/acpid/Cargo.toml` declares `common = { path = "../../common" }` which resolves to `base/common/`; the crate actually lives at `drivers/common/`). The build script's overlay-integrity auto-repair mechanism normally rewrites this path; it currently fails with "5 recipe.toml files could not be restored".
-  - Neither failure is in code modified by Round 1. Resolving them is build-system / fork-upstream work, not Round 1 scope.
-- **Submodule/base commit**: `45452c5a` pushed to `origin/submodule/base`.
+- **redox-driver-sys host cargo test**: ✅ 80 tests pass (unchanged from Round 1; no redox-driver-sys changes in Round 2).
+- **Canonical build** (`./local/scripts/build-redbear.sh --upstream redbear-mini`): NOT rerun this round per operator directive. Round 2's two build-blocker fixes (relibc unsafe-block + acpid `common` path) are surgical and well-understood — the relibc fix is a syntactic wrap (same machine code), the acpid path fix is one character (`../../common` → `../common`) to match 8 sibling crates. Verification deferred to operator-driven build.
+- **Pre-Round-1 build state**: failed for two reasons now resolved by Round 2:
+  - `relibc` cook fails with `error[E0133]` → ✅ resolved by relibc `b80f8b47`
+  - `base` cook fails with `failed to load manifest for dependency 'common'` → ✅ resolved by base `28e356d5`
+
+## Outstanding audit questions (cannot resolve without further work)
+
+1. Will the LG Gram 16Z90TP's EC require `_REG(EmbeddedControl,1)` before `\_SB.PC00.LPCB.EC0.QR_EC` accepts commands? (needs BM test)
+2. Will `_LID` edge transition work on this specific firmware? (needs BM test)
+3. Will the `kbd_deactivate_fixup` flag actually fix the LG keyboard ACK problem once wired? (needs BM test, was inferred from Linux source)
+4. Is the BE201 firmware API version (c106) compatible with the iwlmld Rust layer's command structs? (needs protocol bring-up)
+5. ~~Will the canonical build pass after Round 2's two blocker fixes?~~ — to be confirmed by operator-driven build.
+
+---
+
+## Round 2 commit map
+
+| Repo | Branch | Commit | Subject |
+|---|---|---|---|
+| RedBear-OS/relibc fork | `submodule/relibc` | `b80f8b47` | `relibc: fix edition-2024 unsafe-op-in-unsafe-fn in epoll::convert_event` |
+| RedBear-OS/base fork | `submodule/base` | `28e356d5` | `acpid: fix common path bug (was depth-3 notation in depth-2 crate)` |
+| RedBear-OS/base fork | `submodule/base` | `887718da` | `acpid: wire lid-open to exit_s2idle (symmetric with lid-closed)` |
+| RedBear-OS (parent) | `0.3.1` | (pending) | `LG Gram Round 2: build-blocker fixes + symmetric lid-switch wiring + docs` |
+
+Parent repo commit includes: updated submodule pointers for `local/sources/relibc` (→ `b80f8b47`) and `local/sources/base` (→ `887718da`), LG-GRAM plan status update, this assessment doc's Round 2 section.
+
+---
+
+## Pre-Round-2 audit findings (preserved for reference)
 
 ## Outstanding audit questions (cannot resolve without further work)
 
@@ -117,7 +187,6 @@ The Round 0 explore-agent inventory flagged "single-touch only" for `i2c-hidd` m
 
 | Repo | Branch | Commit | Subject |
 |---|---|---|---|
-| RedBear-OS (parent) | `0.3.1` | (pending) | `redox-driver-sys: replace stubs + populate PANEL_ORIENTATION_TABLE; LG Gram Round 1 docs` |
+| RedBear-OS (parent) | `0.3.1` | `6d8ef13dc1` | `LG Gram Round 1: redox-driver-sys stub replacements + doc reference fixes` |
 | RedBear-OS/base fork | `submodule/base` | `45452c5a` | `acpid+ps2d: wire SystemQuirkFlags consumers (LG Gram Round 1)` |
-
-Parent repo commit includes: updated submodule pointer for `local/sources/base` (→ `45452c5a`), redox-driver-sys source changes, LG-GRAM-16Z90TP-COMPATIBILITY-PLAN.md status update, this assessment doc, QUIRKS-AUDIT.md update.
+| RedBear-OS/base fork | `submodule/base` | `263a41a9` | `bootstrap: fix broken INITNSMGR-CONCURRENCY-DESIGN.md reference` |