vz: implement save/restore

- Only `aarch64` is supported due to lack of support in `Code-Hex/vz/v3`.
- Attempt to restore if the `vz-machine-state` file is found in the instance directory. If this fails, proceed with a cold boot.
- On stop, always attempt to save to the `vz-machine-state` file; if this fails, perform a shutdown.
- Guest OS and Guest Agent behavior are not considered here. Handling may be needed for saving while port forwarding is active.

Signed-off-by: Norio Nomura <[email protected]>

Author: norio-nomura

pkg/store/filenames/filenames.go

@@ -56,6 +56,7 @@ const (
 	HostAgentStderrLog   = "ha.stderr.log"
 	VzIdentifier         = "vz-identifier"
 	VzEfi                = "vz-efi"           // efi variable store
+	VzMachineState       = "vz-machine-state" // machine state file
 	QemuEfiCodeFD        = "qemu-efi-code.fd" // efi code; not always created
 	AnsibleInventoryYAML = "ansible-inventory.yaml"
 

pkg/vz/save_restore.go

@@ -0,0 +1,18 @@
+//go:build darwin && !arm64 && !no_vz
+
+package vz
+
+import (
+	"fmt"
+	"runtime"
+
+	"github.com/Code-Hex/vz/v3"
+)
+
+func saveVM(vm *vz.VirtualMachine, machineStatePath string) error {
+	return fmt.Errorf("save is not supported on the vz driver for the architecture %s", runtime.GOARCH)
+}
+
+func restoreVM(vm *vz.VirtualMachine, machineStatePath string) error {
+	return fmt.Errorf("restore is not supported on the vz driver for the architecture %s", runtime.GOARCH)
+}

pkg/vz/save_restore_arm64.go

@@ -0,0 +1,76 @@
+//go:build darwin && arm64 && !no_vz
+
+package vz
+
+import (
+	"errors"
+	"fmt"
+	"os"
+
+	"github.com/Code-Hex/vz/v3"
+	"github.com/sirupsen/logrus"
+)
+
+func saveVM(vm *vz.VirtualMachine, machineStatePath string) error {
+	if !vm.CanPause() {
+		return fmt.Errorf("can't pause the VZ machine")
+	}
+
+	// Remove the old machine state file if it exists,
+	// because saving the machine state will fail if the file already exists.
+	if err := os.Remove(machineStatePath); err != nil && !errors.Is(err, os.ErrNotExist) {
+		logrus.WithError(err).Errorf("Failed to remove the old VZ machine state file %q", machineStatePath)
+		return err
+	}
+
+	logrus.Info("Pausing VZ")
+	if err := vm.Pause(); err != nil {
+		return err
+	}
+
+	// If we can't stop the machine after pausing, saving the machine state will be useless.
+	// So we should check this before saving the machine state.
+	if !vm.CanStop() {
+		return fmt.Errorf("can't stop the VZ machine after pausing")
+	}
+
+	logrus.Info("Saving VZ machine state for resuming later")
+	if err := vm.SaveMachineStateToPath(machineStatePath); err != nil {
+		// If we fail to save the machine state, we should resume the machine to call RequestStop() later
+		logrus.WithError(err).Errorf("Failed to save the machine state to %q", machineStatePath)
+		if resumeError := vm.Resume(); resumeError != nil {
+			return resumeError
+		}
+		return err
+	}
+
+	logrus.Info("Stopping VZ")
+	if err := vm.Stop(); err != nil {
+		// If we fail to stop the machine, we should resume the machine to call RequestStop() later
+		logrus.WithError(err).Error("Failed to stop the VZ machine")
+		if resumeError := vm.Resume(); resumeError != nil {
+			return resumeError
+		}
+		return err
+	}
+	return nil
+}
+
+func restoreVM(vm *vz.VirtualMachine, machineStatePath string) error {
+	if _, err := os.Stat(machineStatePath); err != nil {
+		return err
+	}
+	logrus.Info("Saved VZ machine state found, resuming VZ")
+	if err := vm.RestoreMachineStateFromURL(machineStatePath); err != nil {
+		return err
+	}
+	if err := vm.Resume(); err != nil {
+		return err
+	}
+	// Remove the machine state file after resuming the machine
+	if err := os.Remove(machineStatePath); err != nil {
+		// We should log the error but continue the process, because the machine state is already restored
+		logrus.WithError(err).Errorf("Failed to remove the VZ machine state file %q", machineStatePath)
+	}
+	return nil
+}

pkg/vz/vm_darwin.go

@@ -58,10 +58,17 @@ func startVM(ctx context.Context, driver *driver.BaseDriver) (*virtualMachineWra
 	if err != nil {
 		return nil, nil, err
 	}
-
-	err = machine.Start()
-	if err != nil {
-		return nil, nil, err
+	machineStatePath := filepath.Join(driver.Instance.Dir, filenames.VzMachineState)
+	if err := restoreVM(machine, machineStatePath); err != nil {
+		if errors.Is(err, os.ErrNotExist) {
+			logrus.Info("Saved VZ machine state not found, starting VZ")
+		} else {
+			logrus.WithError(err).Warn("Failed to restore VZ. Falling back to starting")
+		}
+		err = machine.Start()
+		if err != nil {
+			return nil, nil, err
+		}
 	}
 
 	wrapper := &virtualMachineWrapper{VirtualMachine: machine, stopped: false}

pkg/vz/vz_driver_darwin.go

@@ -17,6 +17,7 @@ import (
 	"github.com/lima-vm/lima/pkg/driver"
 	"github.com/lima-vm/lima/pkg/limayaml"
 	"github.com/lima-vm/lima/pkg/reflectutil"
+	"github.com/lima-vm/lima/pkg/store/filenames"
 )
 
 var knownYamlProperties = []string{
@@ -192,6 +193,13 @@ func (l *LimaVzDriver) RunGUI() error {
 }
 
 func (l *LimaVzDriver) Stop(_ context.Context) error {
+	machineStatePath := filepath.Join(l.Instance.Dir, filenames.VzMachineState)
+	if err := saveVM(l.machine.VirtualMachine, machineStatePath); err != nil {
+		logrus.WithError(err).Warn("Failed to save VZ. Falling back to shutdown")
+	} else {
+		return nil
+	}
+
 	logrus.Info("Shutting down VZ")
 	canStop := l.machine.CanRequestStop()