create pod on the fly

when a user specifies --pod to podman create|run, we should create that pod
automatically.  the port bindings from the container are then inherited by
the infra container.  this signicantly improves the workflow of running
containers inside pods with podman.  the user is still encouraged to use
podman pod create to have more granular control of the pod create options.

Signed-off-by: baude <bbaude@redhat.com>

1 files changed, 50 insertions, 15 deletions

diff --git a/cmd/podman/create.go b/cmd/podman/create.go
index bcf830c7c..c40650f83 100644
--- a/cmd/podman/create.go
+++ b/cmd/podman/create.go
@@ -11,6 +11,7 @@ import (
 	"syscall"
 
 	"github.com/containers/libpod/cmd/podman/libpodruntime"
+	"github.com/containers/libpod/cmd/podman/shared"
 	"github.com/containers/libpod/libpod"
 	"github.com/containers/libpod/libpod/image"
 	ann "github.com/containers/libpod/pkg/annotations"
@@ -375,8 +376,8 @@ func configureEntrypoint(c *cli.Context, data *inspect.ImageData) []string {
 	return entrypoint
 }
 
-func configurePod(c *cli.Context, runtime *libpod.Runtime, namespaces map[string]string) (map[string]string, error) {
-	pod, err := runtime.LookupPod(c.String("pod"))
+func configurePod(c *cli.Context, runtime *libpod.Runtime, namespaces map[string]string, podName string) (map[string]string, error) {
+	pod, err := runtime.LookupPod(podName)
 	if err != nil {
 		return namespaces, err
 	}
@@ -409,6 +410,7 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 		inputCommand, command                                    []string
 		memoryLimit, memoryReservation, memorySwap, memoryKernel int64
 		blkioWeight                                              uint16
+		namespaces                                               map[string]string
 	)
 	idmappings, err := util.ParseIDMapping(c.StringSlice("uidmap"), c.StringSlice("gidmap"), c.String("subuidname"), c.String("subgidname"))
 	if err != nil {
@@ -492,12 +494,21 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 		return nil, errors.Errorf("--cpu-quota and --cpus cannot be set together")
 	}
 
+	// EXPOSED PORTS
+	var portBindings map[nat.Port][]nat.PortBinding
+	if data != nil {
+		portBindings, err = cc.ExposedPorts(c.StringSlice("expose"), c.StringSlice("publish"), c.Bool("publish-all"), data.ContainerConfig.ExposedPorts)
+		if err != nil {
+			return nil, err
+		}
+	}
+
 	// Kernel Namespaces
 	// TODO Fix handling of namespace from pod
 	// Instead of integrating here, should be done in libpod
 	// However, that also involves setting up security opts
 	// when the pod's namespace is integrated
-	namespaces := map[string]string{
+	namespaces = map[string]string{
 		"pid":  c.String("pid"),
 		"net":  c.String("net"),
 		"ipc":  c.String("ipc"),
@@ -505,8 +516,41 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 		"uts":  c.String("uts"),
 	}
 
+	originalPodName := c.String("pod")
+	podName := strings.Replace(originalPodName, "new:", "", 1)
+	// after we strip out :new, make sure there is something left for a pod name
+	if len(podName) < 1 && c.IsSet("pod") {
+		return nil, errors.Errorf("new pod name must be at least one character")
+	}
 	if c.IsSet("pod") {
-		namespaces, err = configurePod(c, runtime, namespaces)
+		if strings.HasPrefix(originalPodName, "new:") {
+			// pod does not exist; lets make it
+			var podOptions []libpod.PodCreateOption
+			podOptions = append(podOptions, libpod.WithPodName(podName), libpod.WithInfraContainer(), libpod.WithPodCgroups())
+			if len(portBindings) > 0 {
+				ociPortBindings, err := cc.NatToOCIPortBindings(portBindings)
+				if err != nil {
+					return nil, err
+				}
+				podOptions = append(podOptions, libpod.WithInfraContainerPorts(ociPortBindings))
+			}
+
+			podNsOptions, err := shared.GetNamespaceOptions(strings.Split(DefaultKernelNamespaces, ","))
+			if err != nil {
+				return nil, err
+			}
+			podOptions = append(podOptions, podNsOptions...)
+			// make pod
+			pod, err := runtime.NewPod(ctx, podOptions...)
+			if err != nil {
+				return nil, err
+			}
+			logrus.Debugf("pod %s created by new container request", pod.ID())
+
+			// The container now cannot have port bindings; so we reset the map
+			portBindings = make(map[nat.Port][]nat.PortBinding)
+		}
+		namespaces, err = configurePod(c, runtime, namespaces, podName)
 		if err != nil {
 			return nil, err
 		}
@@ -535,7 +579,7 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 	// Make sure if network is set to container namespace, port binding is not also being asked for
 	netMode := ns.NetworkMode(namespaces["net"])
 	if netMode.IsContainer() {
-		if len(c.StringSlice("publish")) > 0 || c.Bool("publish-all") {
+		if len(portBindings) > 0 {
 			return nil, errors.Errorf("cannot set port bindings on an existing container network namespace")
 		}
 	}
@@ -644,15 +688,6 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 		return nil, errors.Errorf("No command specified on command line or as CMD or ENTRYPOINT in this image")
 	}
 
-	// EXPOSED PORTS
-	var portBindings map[nat.Port][]nat.PortBinding
-	if data != nil {
-		portBindings, err = cc.ExposedPorts(c.StringSlice("expose"), c.StringSlice("publish"), c.Bool("publish-all"), data.ContainerConfig.ExposedPorts)
-		if err != nil {
-			return nil, err
-		}
-	}
-
 	// SHM Size
 	shmSize, err := units.FromHumanSize(c.String("shm-size"))
 	if err != nil {
@@ -746,7 +781,7 @@ func parseCreateOpts(ctx context.Context, c *cli.Context, runtime *libpod.Runtim
 		NetMode:        netMode,
 		UtsMode:        utsMode,
 		PidMode:        pidMode,
-		Pod:            c.String("pod"),
+		Pod:            podName,
 		Privileged:     c.Bool("privileged"),
 		Publish:        c.StringSlice("publish"),
 		PublishAll:     c.Bool("publish-all"),