Manager can start CVM with NVIDIA GPU support (#595)

manager/qemu/config.go

@@ -81,6 +81,14 @@ type DiskConfig struct {
 	SCSIID  string `env:"DISK_SCSI_ID" envDefault:"scsi0"`
 }
 
+type GPUConfig struct {
+	EnableGPU    bool
+	GPUBDF       string `env:"GPU_BDF"              envDefault:""`
+	PCIeRootPort string `env:"GPU_PCIE_ROOT_PORT"   envDefault:"pci.1"`
+	PCIeBus      string `env:"GPU_PCIE_BUS"         envDefault:"pcie.0"`
+	FWCfgPciMmio string `env:"GPU_FW_CFG_MMIO_MB"   envDefault:"262144"`
+}
+
 type Config struct {
 	EnableSEVSNP bool
 	EnableTDX    bool
@@ -121,6 +129,9 @@ type Config struct {
 	// vTPM
 	IGVMConfig
 
+	// GPU passthrough
+	GPUConfig
+
 	// display
 	NoGraphic bool   `env:"NO_GRAPHIC" envDefault:"true"`
 	Monitor   string `env:"MONITOR"    envDefault:"pty"`
@@ -227,6 +238,23 @@ func (config Config) ConstructQemuArgs() []string {
 				config.DiskConfig.SCSIID))
 	}
 
+	// GPU passthrough via VFIO
+	if config.GPUConfig.EnableGPU {
+		args = append(args, "-device",
+			fmt.Sprintf("pcie-root-port,id=%s,bus=%s",
+				config.GPUConfig.PCIeRootPort,
+				config.GPUConfig.PCIeBus))
+
+		args = append(args, "-device",
+			fmt.Sprintf("vfio-pci,host=%s,bus=%s",
+				config.GPUConfig.GPUBDF,
+				config.GPUConfig.PCIeRootPort))
+
+		args = append(args, "-fw_cfg",
+			fmt.Sprintf("name=opt/ovmf/X-PciMmio64Mb,string=%s",
+				config.GPUConfig.FWCfgPciMmio))
+	}
+
 	// SEV-SNP
 	if config.EnableSEVSNP {
 		sevSnpType := "sev-snp-guest"
@@ -317,5 +345,13 @@ func NewConfig() (*Config, error) {
 	cfg.EnableSEVSNP = SEVSNPEnabledOnHost()
 	cfg.EnableTDX = TDXEnabledOnHost()
 
+	bdf, detected := GPUPassthroughAvailable()
+	if cfg.GPUConfig.GPUBDF != "" {
+		cfg.GPUConfig.EnableGPU = true
+	} else if detected {
+		cfg.GPUConfig.EnableGPU = true
+		cfg.GPUConfig.GPUBDF = bdf
+	}
+
 	return &cfg, nil
 }

manager/qemu/config_test.go

@@ -151,6 +151,79 @@ func TestConstructQemuArgs(t *testing.T) {
 				"-monitor", "pty",
 			},
 		},
+		{
+			name: "GPU passthrough configuration",
+			config: Config{
+				QemuBinPath: "qemu-system-x86_64",
+				EnableKVM:   true,
+				Machine:     "q35",
+				CPU:         "EPYC",
+				SMPCount:    4,
+				MaxCPUs:     64,
+				MemID:       "ram1",
+				MemoryConfig: MemoryConfig{
+					Size:  "2048M",
+					Slots: 5,
+					Max:   "30G",
+				},
+				OVMFCodeConfig: OVMFCodeConfig{
+					If:       "pflash",
+					Format:   "raw",
+					Unit:     0,
+					File:     "/usr/share/OVMF/OVMF_CODE.fd",
+					ReadOnly: "on",
+				},
+				OVMFVarsConfig: OVMFVarsConfig{
+					If:     "pflash",
+					Format: "raw",
+					Unit:   1,
+					File:   "/usr/share/OVMF/OVMF_VARS.fd",
+				},
+				NetDevConfig: NetDevConfig{
+					ID:            "vmnic",
+					HostFwdAgent:  7020,
+					GuestFwdAgent: 7002,
+				},
+				VirtioNetPciConfig: VirtioNetPciConfig{
+					DisableLegacy: "on",
+					IOMMUPlatform: true,
+					Addr:          "0x2",
+				},
+				KernelConfig: KernelConfig{
+					KernelFile: "img/bzImage",
+					RootFsFile: "img/rootfs.cpio.gz",
+				},
+				GPUConfig: GPUConfig{
+					EnableGPU:    true,
+					GPUBDF:       "0000:02:00.0",
+					PCIeRootPort: "pci.1",
+					PCIeBus:      "pcie.0",
+					FWCfgPciMmio: "262144",
+				},
+				KernelCommandLine: "quiet console=null",
+				NoGraphic:         true,
+				Monitor:           "pty",
+			},
+			expected: []string{
+				"-enable-kvm",
+				"-machine", "q35",
+				"-cpu", "EPYC",
+				"-smp", "4,maxcpus=64",
+				"-m", "2048M,slots=5,maxmem=30G",
+				"-drive", "if=pflash,format=raw,unit=0,file=/usr/share/OVMF/OVMF_CODE.fd,readonly=on",
+				"-drive", "if=pflash,format=raw,unit=1,file=/usr/share/OVMF/OVMF_VARS.fd",
+				"-netdev", "user,id=vmnic,hostfwd=tcp::7020-:7002",
+				"-device", "virtio-net-pci,disable-legacy=on,iommu_platform=true,netdev=vmnic,addr=0x2,romfile=",
+				"-device", "pcie-root-port,id=pci.1,bus=pcie.0",
+				"-device", "vfio-pci,host=0000:02:00.0,bus=pci.1",
+				"-fw_cfg", "name=opt/ovmf/X-PciMmio64Mb,string=262144",
+				"-kernel", "img/bzImage",
+				"-append", "quiet console=null",
+				"-initrd", "img/rootfs.cpio.gz",
+				"-nographic",
+				"-monitor", "pty",
+			},
+		},
 	}
 
 	for _, tt := range tests {

manager/qemu/vm.go

@@ -319,3 +319,36 @@ func GetVirtualSizeGB(path string) (int, error) {
 	gb := (bytes + (1<<30 - 1)) >> 30
 	return int(gb), nil
 }
+
+// GPUPassthroughAvailable scans for NVIDIA GPU devices bound to the vfio-pci
+// driver and returns the BDF of the first one found.
+func GPUPassthroughAvailable() (string, bool) {
+	const vfioPCIPath = "/sys/bus/pci/drivers/vfio-pci"
+	entries, err := os.ReadDir(vfioPCIPath)
+	if err != nil {
+		return "", false
+	}
+
+	for _, entry := range entries {
+		bdf := entry.Name()
+		if !strings.Contains(bdf, ":") {
+			continue
+		}
+
+		vendor, err := os.ReadFile(fmt.Sprintf("/sys/bus/pci/devices/%s/vendor", bdf))
+		if err != nil || strings.TrimSpace(string(vendor)) != "0x10de" {
+			continue
+		}
+
+		class, err := os.ReadFile(fmt.Sprintf("/sys/bus/pci/devices/%s/class", bdf))
+		if err != nil {
+			continue
+		}
+		classStr := strings.TrimSpace(string(class))
+		// 0x0302xx = 3D Controller (e.g. H100), 0x0300xx = VGA Compatible Controller
+		if strings.HasPrefix(classStr, "0x0302") || strings.HasPrefix(classStr, "0x0300") {
+			return bdf, true
+		}
+	}
+	return "", false
+}