Initial commit: Open-IOV vGPU-IO complete stack

- Firmware decompilation: GSP 570.144W (28.5MB ELF, RV64IMAC), 132 decompiled C files
- Kernel modules: nvidia.ko (34MB, OpenRM), open-iov-vfio.ko (2.8MB, VFIO mdev), gsp-vgpu.ko (1.8MB, guest driver)
- Userspace: libiov.so/libvgpu.so (76KB), example_vgpu (28KB), rpc_emulator.c
- PCI config space emulation (256B with PM, MSI, MSI-X, PCIe caps)
- 12 RPC types fully mapped from firmware decompilation
- BAR trap, WPR, scheduling, notification, migration emulation
- Helper scripts: load_vgpu_stack, create_mdev, launch_qemu_vgpu, test_vgpu_stack, setup_guest_driver, unload_vgpu_stack
- Architecture documentation
- Wiki: 10 technical documents
- Git LFS for large binaries (.ko, .bin, .iso, .img, .o, ghidra repos)
- nvidia-open-gpu-kernel-modules as submodule

1 files changed, 95 insertions, 0 deletions

diff --git a/tools/module_ops/load_converged.sh b/tools/module_ops/load_converged.sh
new file mode 100755
index 0000000..77cde09
--- /dev/null
+++ b/tools/module_ops/load_converged.sh
@@ -0,0 +1,95 @@
+#!/bin/bash
+# Load vGPU stack in converged mode
+# Usage: sudo ./load_converged.sh [dry_run]
+
+set -euo pipefail
+
+OPEN_IOV_ROOT="$OPEN_IOV_ROOT"
+NVIDIA_OPEN_KERNEL="$OPEN_IOV_ROOT/nvidia-open-gpu-kernel-modules/kernel-open"
+VGPU_KERNEL="$OPEN_IOV_ROOT/kernel"
+
+DRY_RUN="${1:-false}"
+
+echo "=== Open-IOV Converged Mode Module Load ==="
+echo "Open-IOV Root: $OPEN_IOV_ROOT"
+echo "Dry-run mode: $DRY_RUN"
+echo ""
+
+# Function to check if module is loaded
+is_loaded() {
+    lsmod | grep -q "^$1 "
+}
+
+# Function to load module with error checking
+load_module() {
+    local module_path=$1
+    local module_name=$(basename $module_path .ko)
+    local params="${2:-}"
+    
+    if is_loaded "$module_name"; then
+        echo "Module $module_name already loaded"
+        return 0
+    fi
+    
+    echo "Loading $module_name..."
+    if sudo insmod "$module_path" $params; then
+        echo "  ✓ $module_name loaded successfully"
+        return 0
+    else
+        echo "  ✗ Failed to load $module_name"
+        return 1
+    fi
+}
+
+# 1. Load mdev core
+echo "Step 1: Loading mdev core module..."
+sudo modprobe mdev
+echo "  ✓ mdev loaded"
+echo ""
+
+# 2. Load host nvidia.ko with vGPU support
+echo "Step 2: Loading host nvidia.ko (patched open-gpu-kernel-modules)..."
+load_module "$NVIDIA_OPEN_KERNEL/nvidia.ko" "NVreg_RegisterPCIDriver=1 NVreg_RmMsg=:"
+echo ""
+
+# 3. Load dependent modules
+echo "Step 3: Loading dependent modules..."
+load_module "$NVIDIA_OPEN_KERNEL/nvidia-modeset.ko"
+load_module "$NVIDIA_OPEN_KERNEL/nvidia-uvm.ko"
+load_module "$NVIDIA_OPEN_KERNEL/nvidia-drm.ko"
+load_module "$NVIDIA_OPEN_KERNEL/nvidia-peermem.ko"
+echo ""
+
+# 4. Load nvidia-vgpu-vfio.ko
+echo "Step 4: Loading nvidia-vgpu-vfio.ko..."
+if [ "$DRY_RUN" = "dry_run" ] || [ "$DRY_RUN" = "true" ]; then
+    echo "  Running in DRY-RUN mode first..."
+    load_module "$VGPU_KERNEL/nvidia-vgpu-vfio.ko" "dyndbg='+p' dry_run=1"
+    echo "  Dry-run successful, unloading..."
+    sudo rmmod nvidia-vgpu-vfio
+    echo "  Loading for real..."
+    load_module "$VGPU_KERNEL/nvidia-vgpu-vfio.ko" "dyndbg='+p'"
+else
+    load_module "$VGPU_KERNEL/nvidia-vgpu-vfio.ko" "dyndbg='+p'"
+fi
+echo ""
+
+# 5. Verify
+echo "Step 5: Verifying vGPU registration..."
+sleep 2
+if [ -d "/sys/bus/pci/devices/0000:01:00.0/mdev_supported_types" ]; then
+    echo "  ✓ mdev_supported_types directory exists"
+    ls -la /sys/bus/pci/devices/0000:01:00.0/mdev_supported_types/
+else
+    echo "  ✗ mdev_supported_types NOT found"
+    echo "  Check dmesg for errors: dmesg -T | grep -i vgpu"
+    exit 1
+fi
+
+echo ""
+echo "=== Converged Mode Load Complete ==="
+echo ""
+echo "Next steps:"
+echo "  - Create mdev device: echo <uuid> | sudo tee /sys/bus/pci/devices/0000:01:00.0/mdev_supported_types/nvidia-vgpu-turing/create"
+echo "  - Run QEMU with vGPU passthrough"
+echo "  - Run tests: $OPEN_IOV_ROOT/tools/test/run_integration_tests.sh"
\ No newline at end of file