wip

Makefile

@@ -84,6 +84,7 @@ out/delta.tar.gz: bin/init bin/vsockexec bin/cmd/gcs bin/cmd/gcstools bin/cmd/ho
 	cp bin/cmd/gcs rootfs/bin/
 	cp bin/cmd/gcstools rootfs/bin/
 	cp bin/cmd/hooks/wait-paths rootfs/bin/
+	cp /mnt/c/Users/kevpar/src/test/rebuf rootfs/bin/
 	for tool in $(GCS_TOOLS); do ln -s gcstools rootfs/bin/$$tool; done
 	git -C $(SRCROOT) rev-parse HEAD > rootfs/info/gcs.commit && \
 	git -C $(SRCROOT) rev-parse --abbrev-ref HEAD > rootfs/info/gcs.branch && \

Protobuild.toml

@@ -7,7 +7,7 @@ generators = ["go", "go-grpc"]
 
   # defaults are "/usr/local/include" and "/usr/include", which don't exist on Windows.
   # override defaults to supress errors about non-existant directories.
-  after = []
+  # after = []
 
 # This section maps protobuf imports to Go packages.
 [packages]
@@ -21,5 +21,6 @@ prefixes = [
   "github.com/Microsoft/hcsshim/internal/computeagent",
   "github.com/Microsoft/hcsshim/internal/ncproxyttrpc",
   "github.com/Microsoft/hcsshim/internal/vmservice",
+  "github.com/Microsoft/hcsshim/internal/save",
 ]
 generators = ["go", "go-ttrpc"]

cmd/containerd-shim-runhcs-v1/options/runhcs.proto

@@ -107,6 +107,8 @@ message Options {
 
 	// scrub_logs enables removing environment variables and other potentially sensitive information from logs
 	bool scrub_logs = 20;
+
+	string restore_path = 21;
 }
 
 // ProcessDetails contains additional information about a process. This is the additional

cmd/containerd-shim-runhcs-v1/pod.go

@@ -12,6 +12,7 @@ import (
 
 	"github.com/Microsoft/hcsshim/internal/log"
 	"github.com/Microsoft/hcsshim/internal/oci"
+	"github.com/Microsoft/hcsshim/internal/state"
 	"github.com/Microsoft/hcsshim/internal/uvm"
 	"github.com/Microsoft/hcsshim/osversion"
 	"github.com/Microsoft/hcsshim/pkg/annotations"
@@ -66,6 +67,9 @@ type shimPod interface {
 	// return `errdefs.ErrFailedPrecondition`. Deleting the pod's sandbox task
 	// is a no-op.
 	DeleteTask(ctx context.Context, tid string) error
+
+	StartSave(ctx context.Context, path string) error
+	CompleteSave(ctx context.Context, path string) error
 }
 
 func createPod(ctx context.Context, events publisher, req *task.CreateTaskRequest, s *specs.Spec) (_ shimPod, err error) {
@@ -449,3 +453,66 @@ func (p *pod) DeleteTask(ctx context.Context, tid string) error {
 
 	return nil
 }
+
+type podState struct {
+	ID   string
+	Spec *specs.Spec
+}
+
+func (p *pod) StartSave(ctx context.Context, path string) error {
+	if err := os.MkdirAll(path, 0755); err != nil {
+		return err
+	}
+	if p.host == nil {
+		return fmt.Errorf("can only save VM-isolated pods")
+	}
+	if err := p.host.StartSave(ctx, filepath.Join(path, "uvm")); err != nil {
+		return fmt.Errorf("save UVM: %w", err)
+	}
+	if err := p.sandboxTask.Save(ctx, filepath.Join(path, "sandboxTask")); err != nil {
+		return err
+	}
+	if err := state.Write(filepath.Join(path, "state.json"),
+		&podState{
+			ID:   p.id,
+			Spec: p.spec,
+		}); err != nil {
+		return err
+	}
+	return nil
+}
+
+func (p *pod) CompleteSave(ctx context.Context, path string) error {
+	if err := os.MkdirAll(path, 0755); err != nil {
+		return err
+	}
+	if p.host == nil {
+		return fmt.Errorf("can only save VM-isolated pods")
+	}
+	if err := p.host.CompleteSave(ctx, filepath.Join(path, "uvm")); err != nil {
+		return fmt.Errorf("save UVM: %w", err)
+	}
+	return nil
+}
+
+type standbyPod struct {
+	state *podState
+	host  *uvm.UtilityVM
+}
+
+func restorePod(ctx context.Context, path string, netNS string, scratchPath string, events publisher, req *task.CreateTaskRequest) (_ shimPod, err error) {
+	p := &pod{
+		events: events,
+		id:     req.ID,
+	}
+	p.host, err = uvm.RestoreUVM(ctx, filepath.Join(path, "uvm"), netNS, scratchPath, req.ID)
+	if err != nil {
+		return nil, err
+	}
+	state, err := state.Read[podState](filepath.Join(path, "state.json"))
+	if err != nil {
+		return nil, err
+	}
+	p.spec = state.Spec
+	return p, nil
+}

cmd/containerd-shim-runhcs-v1/pod_test.go

@@ -86,6 +86,14 @@ func (tsp *testShimPod) DeleteTask(ctx context.Context, tid string) error {
 	return nil
 }
 
+func (tsp *testShimPod) StartSave(ctx context.Context, path string) error {
+	return fmt.Errorf("not implemented")
+}
+
+func (tsp *testShimPod) CompleteSave(ctx context.Context, path string) error {
+	return fmt.Errorf("not implemented")
+}
+
 // Pod tests
 
 func setupTestPodWithFakes(t *testing.T) (*pod, *testShimTask) {

cmd/containerd-shim-runhcs-v1/saverestore.go

@@ -0,0 +1,52 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"os"
+)
+
+func (s *service) startSave(ctx context.Context, path string) error {
+	if err := os.MkdirAll(path, 0755); err != nil {
+		return err
+	}
+	v := s.taskOrPod.Load()
+	if v == nil {
+		return fmt.Errorf("invalid state: no pod")
+	}
+	p, ok := v.(shimPod)
+	if !ok {
+		return fmt.Errorf("only works with pod, not standalone task")
+	}
+	if err := p.StartSave(ctx, path); err != nil {
+		return err
+	}
+	return nil
+}
+
+func (s *service) completeSave(ctx context.Context, path string) error {
+	if err := os.MkdirAll(path, 0755); err != nil {
+		return err
+	}
+	v := s.taskOrPod.Load()
+	if v == nil {
+		return fmt.Errorf("invalid state: no pod")
+	}
+	p, ok := v.(shimPod)
+	if !ok {
+		return fmt.Errorf("only works with pod, not standalone task")
+	}
+	if err := p.CompleteSave(ctx, path); err != nil {
+		return err
+	}
+	return nil
+}
+
+func (s *service) restore(ctx context.Context, path string) error {
+	// p, err := restorePod(ctx, filepath.Join(path, "pod"))
+	// if err != nil {
+	// 	return err
+	// }
+	// s.standbyPod = p
+	return nil
+}

cmd/containerd-shim-runhcs-v1/serve.go

@@ -26,6 +26,7 @@ import (
 	runhcsopts "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/options"
 	"github.com/Microsoft/hcsshim/internal/extendedtask"
 	hcslog "github.com/Microsoft/hcsshim/internal/log"
+	"github.com/Microsoft/hcsshim/internal/save"
 	"github.com/Microsoft/hcsshim/internal/shimdiag"
 	"github.com/Microsoft/hcsshim/pkg/octtrpc"
 )
@@ -193,6 +194,12 @@ var serveCommand = cli.Command{
 			return fmt.Errorf("failed to create new service: %w", err)
 		}
 
+		if p := newShimOpts.RestorePath; p != "" {
+			if err := svc.restore(context.Background(), p); err != nil {
+				return fmt.Errorf("restore shim from state %q: %w", p, err)
+			}
+		}
+
 		s, err := ttrpc.NewServer(ttrpc.WithUnaryServerInterceptor(octtrpc.ServerInterceptor()))
 		if err != nil {
 			return err
@@ -201,6 +208,7 @@ var serveCommand = cli.Command{
 		task.RegisterTaskService(s, svc)
 		shimdiag.RegisterShimDiagService(s, svc)
 		extendedtask.RegisterExtendedTaskService(s, svc)
+		save.RegisterSaveService(s, svc)
 
 		sl, err := winio.ListenPipe(socket, nil)
 		if err != nil {