diff options
Diffstat (limited to 'scripts/stack_manager.sh')
| -rwxr-xr-x | scripts/stack_manager.sh | 34 |
1 files changed, 27 insertions, 7 deletions
diff --git a/scripts/stack_manager.sh b/scripts/stack_manager.sh index c4685da..0b62170 100755 --- a/scripts/stack_manager.sh +++ b/scripts/stack_manager.sh @@ -81,7 +81,14 @@ cmd_build_all() { cmd_load_prereqs() { log_info "Loading kernel prerequisites (vfio, mdev)..." - sudo modprobe vfio vfio_iommu_type1 vfio_pci mdev vfio_mdev 2>/dev/null || true + # vfio_mdev does not exist on kernels >= 6.4 (mdev core is standalone mdev.ko); + # modprobe each individually so a missing one never blocks the rest. + local m + for m in vfio vfio_iommu_type1 vfio_pci mdev vfio_mdev; do + if modinfo "$m" &>/dev/null; then + sudo modprobe "$m" 2>/dev/null || true + fi + done lsmod | grep -E "vfio|mdev" } @@ -104,13 +111,26 @@ cmd_load_nvidia() { log_err "Missing $NVIDIA_KO - run 'build-nvidia' first" return 1 fi - if ! lsmod | grep -q "^nvidia "; then - log_info "Loading nvidia.ko (OpenRM)..." - sudo insmod "$NVIDIA_KO" || { log_err "Failed to load nvidia.ko"; return 1; } - log_ok "nvidia.ko loaded" - else - log_ok "nvidia.ko already loaded" + if lsmod | grep -q "^nvidia "; then + # A module named 'nvidia' is already loaded. Verify it is OUR build + # by comparing versions (ours: 570.144 OpenRM w/ vGPU VFIO patches). + local loaded_ver want_ver + loaded_ver=$(cat /sys/module/nvidia/version 2>/dev/null) + want_ver=$(modinfo -F version "$NVIDIA_KO" 2>/dev/null) + if [[ -n "$loaded_ver" && "$loaded_ver" == "$want_ver" ]]; then + log_ok "nvidia.ko (Open-IOV OpenRM $loaded_ver) already loaded" + return 0 + fi + log_err "A foreign nvidia module (version '$loaded_ver') is already loaded;" + log_err "Open-IOV needs its own OpenRM build ('$want_ver'). Cannot load both." + log_err "Unload the stock driver first (requires nothing holding /dev/nvidia*," + log_err "i.e. no GUI session), or reboot - stock nvidia is blacklisted via" + log_err "/etc/modprobe.d/blacklist-nvidia-open-iov.conf." + return 1 fi + log_info "Loading nvidia.ko (OpenRM)..." + sudo insmod "$NVIDIA_KO" || { log_err "Failed to load nvidia.ko"; return 1; } + log_ok "nvidia.ko loaded" } cmd_install_firmware() { |
