Merge pull request 'feat(gpu): universal NVIDIA lane — detect + install branch at boot (C&C)' (#31) from nvidia-universal-lane into main

build.cfg

@@ -43,6 +43,14 @@ DEFAULT_DESKTOP="xfce"          # operator USB desktop
 # intel | amd | nvidia-590 | nvidia-470 | nvidia-390 | vesa | "" (auto-detect)
 GPU_DRIVER=""
 
+# Universal NVIDIA lane (opt-in). When YES, the image carries an on-image NVIDIA
+# pkg repo for all three branches (390/470/580), does NOT bake a single branch,
+# and clawdie_live_gpu runs in "nvidia-auto": at boot it detects the GPU, installs
+# the matching branch from the on-image repo, and kldloads it before SDDM. On any
+# failure it falls back to the integrated/scfb path (never worse than today).
+# Builds a larger image. Leave NO for single-branch GPU_DRIVER=nvidia-* builds.
+NVIDIA_UNIVERSAL="${NVIDIA_UNIVERSAL:-NO}"
+
 # Deploy target
 TARGET="baremetal"          # vps | baremetal
 

build.sh

@@ -119,6 +119,7 @@ echo "    Clawdie : ${CLAWDIE_REF}"
 echo "    Desktop : ${DEFAULT_DESKTOP}"
 echo "    Pkg     : ${DEFAULT_PKG_BRANCH}"
 echo "    GPU     : ${GPU_DRIVER:-auto-detect}"
+echo "    NVIDIA universal : ${NVIDIA_UNIVERSAL:-NO}"
 echo "    Target  : ${TARGET:-baremetal}"
 echo "    Colibri : ${FEATURE_COLIBRI:-NO}"
 echo "    Clawdie agent : ${FEATURE_CLAWDIE:-NO}"
@@ -906,6 +907,48 @@ install_clawdie_service() {
     fi
 }
 
+# Stage an on-image NVIDIA pkg repo (all branches) so clawdie_live_gpu can
+# `pkg install` the detected branch at boot (NVIDIA_UNIVERSAL lane).
+#
+# FreeBSD-build-host step (authored on Linux; runs + must be validated on
+# FreeBSD). Verify on the build host: (1) the `pkg fetch -o` layout matches what
+# `pkg repo` expects, (2) the dependency closure is complete for offline boot
+# install, (3) image size headroom, (4) `pkg install -r clawdie-nvidia` resolves
+# from file:// at boot. See doc handoff.
+install_nvidia_universal_repo() {
+    [ "${NVIDIA_UNIVERSAL:-NO}" = "YES" ] || {
+        echo "    NVIDIA universal repo staging disabled (NVIDIA_UNIVERSAL=${NVIDIA_UNIVERSAL:-NO})"
+        return 0
+    }
+
+    echo "    Staging universal NVIDIA repo (all branches) into image..."
+    _nv_repo="${MOUNT_POINT}/usr/local/share/clawdie/nvidia-repo"
+    mkdir -p "${_nv_repo}"
+
+    # Fetch all three branches + deps into the on-image repo dir, then build repo
+    # metadata. Uses pkg's own dependency resolution so the closure is complete.
+    if ! env ASSUME_ALWAYS_YES=yes pkg fetch -y -o "${_nv_repo}" -d \
+        nvidia-driver-390 nvidia-driver-470 nvidia-driver-580 nvidia-settings nvidia-xconfig; then
+        echo "ERROR: failed to fetch NVIDIA packages for universal repo"
+        exit 1
+    fi
+    if ! pkg repo "${_nv_repo}"; then
+        echo "ERROR: failed to generate NVIDIA universal repo metadata"
+        exit 1
+    fi
+
+    mkdir -p "${MOUNT_POINT}/usr/local/etc/pkg/repos"
+    cat > "${MOUNT_POINT}/usr/local/etc/pkg/repos/clawdie-nvidia.conf" <<'EOF'
+# On-image NVIDIA package repo for the universal GPU lane. Consumed at boot by
+# clawdie_live_gpu (nvidia-auto mode) via: pkg install -r clawdie-nvidia ...
+clawdie-nvidia: {
+    url: "file:///usr/local/share/clawdie/nvidia-repo",
+    enabled: yes
+}
+EOF
+    echo "    NVIDIA universal repo staged at /usr/local/share/clawdie/nvidia-repo"
+}
+
 install_live_npm_globals() {
     echo "    Installing bundled npm globals into live image..."
 
@@ -1561,6 +1604,15 @@ EOF
             ;;
     esac
 
+    # Universal NVIDIA lane: no branch is baked; clawdie_live_gpu installs the
+    # matching one from the on-image repo at boot. Overrides the mode set above.
+    # kld_list intentionally does NOT preload nvidia (it is not installed yet at
+    # first boot); the rc.d service kldloads it after install.
+    if [ "${NVIDIA_UNIVERSAL:-NO}" = "YES" ]; then
+        set_config_line "${MOUNT_POINT}/etc/rc.conf" 'clawdie_live_gpu_mode="nvidia-auto"'
+        set_config_line "${MOUNT_POINT}/etc/rc.conf" 'clawdie_live_gpu_nvidia_branch=""'
+    fi
+
     mkdir -p "${MOUNT_POINT}/etc/ssh/sshd_config.d"
     install -m 0644 "${LIVE_SESSION_DIR}/sshd-live.conf" \
         "${MOUNT_POINT}/etc/ssh/sshd_config.d/clawdie-live.conf"
@@ -2158,6 +2210,7 @@ install_live_runtime_packages
 configure_live_operator_session
 install_colibri_service
 install_clawdie_service
+install_nvidia_universal_repo
 
 # Copy payload
 # Rebuild payload paths from scratch inside the reusable work image. A failed

doc/NVIDIA-UNIVERSAL-HANDOFF.md

@@ -0,0 +1,74 @@
+# NVIDIA universal lane — Codex build + hardware validation handoff
+
+**From:** Claude (Linux) · **To:** Codex ISO Builder (FreeBSD 15 host)
+**Goal:** one ISO that boots on any NVIDIA machine — detect the GPU pre-SDDM,
+install the matching branch from an on-image repo, kldload it, then SDDM/XFCE.
+
+Authored on Linux (sh -n clean; detector + branch/pkg map unit-tested). The
+package/boot behavior **must be built and proven on FreeBSD + real NVIDIA
+hardware** — it cannot be validated from Linux.
+
+## What landed (behind `NVIDIA_UNIVERSAL`, default NO — normal builds unaffected)
+
+- `build.cfg`: `NVIDIA_UNIVERSAL` flag.
+- `build.sh`:
+  - `install_nvidia_universal_repo()` — `pkg fetch -d` the three families
+    (`nvidia-driver-390/470/580` + settings) into
+    `/usr/local/share/clawdie/nvidia-repo`, `pkg repo` it, write
+    `/usr/local/etc/pkg/repos/clawdie-nvidia.conf` (file:// , enabled).
+  - In universal mode: no branch baked into the root; rc.conf
+    `clawdie_live_gpu_mode="nvidia-auto"`.
+- `live/operator-session/clawdie-live-gpu` (`nvidia-auto` mode): detect device id
+  → branch {390,470,590} → `pkg install -r clawdie-nvidia nvidia-driver-<n>
+nvidia-settings` → kldload via existing select/load. **Best-effort:** any
+  failure falls through to the integrated/scfb path — never worse than today.
+
+## Must verify on the FreeBSD build host
+
+1. **`pkg fetch -o` / `pkg repo` layout** — confirm `pkg fetch -y -o DIR -d ...`
+   places archives where `pkg repo DIR` expects them; adjust if `pkg repo` needs
+   `DIR/All`. Confirm the dependency closure is complete (offline boot install
+   must not reach the network).
+2. **Boot install command** — `pkg install -y -r clawdie-nvidia nvidia-driver-470
+nvidia-settings` resolves purely from the file:// repo with no network.
+3. **Writable root at boot** — the USB-root install must allow `pkg install`
+   before SDDM (it does on a real on-USB FreeBSD; confirm).
+4. **Kernel ABI** — the staged kmod packages must match the image's kernel.
+5. **Image size** — three branches + deps add several hundred MB; confirm
+   headroom against `IMAGE_SIZE`.
+6. **Exactly one branch** installed at boot (no mixing).
+
+## Build commands
+
+```sh
+# colibri artifacts first (do NOT cargo clean after)
+cd /home/clawdie/colibri && git fetch origin && git reset --hard origin/main
+cargo build --workspace --release
+
+# universal image (set COLIBRI_REPO explicitly if building from a tmp worktree)
+cd /home/clawdie/clawdie-iso && git fetch origin && git reset --hard origin/main
+sudo env NVIDIA_UNIVERSAL=YES FEATURE_CLAWDIE=YES \
+  COLIBRI_REPO=/home/clawdie/colibri \
+  CLAWDIE_ARTIFACT_DIR=/home/clawdie/colibri/target/release \
+  ./build.sh --skip-memstick-fetch --live-default-password
+```
+
+## Hardware acceptance
+
+1. Boot on the NVIDIA target.
+2. `cat /var/log/clawdie-live-gpu.log` — expect: detected device id, chosen
+   branch + pkg, "installed nvidia-driver-<n>", selected/loaded modules.
+3. `kldstat | grep nvidia` and SDDM → XFCE comes up accelerated.
+4. If install/load failed, the log shows the fallback and the desktop still
+   comes up on integrated/scfb (no hard failure).
+
+## Deletion criteria
+
+Delete this handoff once a `NVIDIA_UNIVERSAL=YES` image has, on real NVIDIA
+hardware, auto-installed the correct branch and reached XFCE — with the result
+recorded below.
+
+## Results
+
+_(Codex: build host, GPU model + device id, branch chosen, pkg install outcome,
+image size delta, SDDM/XFCE result.)_

live/operator-session/clawdie-live-gpu

@@ -79,6 +79,57 @@ clawdie_live_gpu_nvidia_branch_for_device()
     fi
 }
 
+# Map a branch label {390,470,590} to the on-image pkg name. The "590" lane is
+# the current driver, which is the nvidia-driver-580 package.
+clawdie_live_gpu_nvidia_pkg_for_branch()
+{
+    case "$1" in
+        390) echo "nvidia-driver-390" ;;
+        470) echo "nvidia-driver-470" ;;
+        590) echo "nvidia-driver-580" ;;
+        *)   return 1 ;;
+    esac
+}
+
+# nvidia-auto mode: detect the GPU, install the matching branch from the on-image
+# clawdie-nvidia repo, and leave /boot/modules/nvidia.ko in place for select to
+# load. Fully best-effort: any failure just logs and returns, so the caller falls
+# back to the integrated/scfb path. Never blocks boot.
+clawdie_live_gpu_install_nvidia()
+{
+    _blocks="$1"
+
+    # Already installed (e.g. a previous boot) — nothing to do.
+    if [ -e /boot/modules/nvidia.ko ]; then
+        clawdie_live_gpu_log "nvidia.ko already present; skipping auto-install"
+        return 0
+    fi
+
+    _device_id=$(clawdie_live_gpu_nvidia_device_id "$_blocks" || true)
+    _branch=$(clawdie_live_gpu_nvidia_branch_for_device "$_device_id")
+    _pkg=$(clawdie_live_gpu_nvidia_pkg_for_branch "$_branch") || {
+        clawdie_live_gpu_log "WARN: no package mapping for branch ${_branch}; skipping NVIDIA auto-install"
+        return 1
+    }
+    clawdie_live_gpu_log "nvidia-auto: device ${_device_id:-unknown} -> branch ${_branch} (${_pkg})"
+
+    if ! command -v pkg >/dev/null 2>&1; then
+        clawdie_live_gpu_log "WARN: pkg not available; cannot auto-install NVIDIA"
+        return 1
+    fi
+
+    # Install only from the on-image repo so this works offline and never reaches
+    # the network during boot.
+    if env ASSUME_ALWAYS_YES=yes pkg install -y -r clawdie-nvidia "${_pkg}" nvidia-settings \
+        >> "$LOG_FILE" 2>&1; then
+        clawdie_live_gpu_log "nvidia-auto: installed ${_pkg}"
+        return 0
+    fi
+
+    clawdie_live_gpu_log "WARN: nvidia-auto install of ${_pkg} failed; falling back to integrated/scfb"
+    return 1
+}
+
 clawdie_live_gpu_has_pci_vendor()
 {
     _blocks="$1"
@@ -190,6 +241,14 @@ clawdie_live_gpu_start()
     fi
 
     clawdie_live_gpu_log "detected display devices: $(echo "$_blocks" | grep -i '^vgapci' | tr '\n' ' ')"
+
+    # Universal lane: install the matching NVIDIA branch before module selection.
+    # Best-effort — on failure select_modules just falls through to integrated.
+    if [ "${clawdie_live_gpu_mode:-auto}" = "nvidia-auto" ] &&
+        clawdie_live_gpu_has_pci_vendor "$_blocks" "0x10de"; then
+        clawdie_live_gpu_install_nvidia "$_blocks" || true
+    fi
+
     _modules=$(clawdie_live_gpu_select_modules "$_blocks")
     if [ -z "$_modules" ]; then
         clawdie_live_gpu_log "no live KMS module selected; using scfb/vesa fallback"