Pod Security Option support

Added support for pod security options. These are applied to infra and passed down to the
containers as added (unless overridden).

Modified the inheritance process from infra, creating a new function Inherit() which reads the config, and marshals the compatible options into an intermediate struct `InfraInherit`
This is then unmarshaled into a container config and all of this is added to the CtrCreateOptions. Removes the need (mostly) for special additons which complicate the Container_create
code and pod creation.

resolves #12173

Signed-off-by: cdoern <cdoern@redhat.com>

2 files changed, 98 insertions, 50 deletions

diff --git a/pkg/specgen/generate/container_create.go b/pkg/specgen/generate/container_create.go
index 7ab9d1b29..7d792b3b1 100644
--- a/pkg/specgen/generate/container_create.go
+++ b/pkg/specgen/generate/container_create.go
@@ -2,13 +2,14 @@ package generate
 
 import (
 	"context"
-	"fmt"
+	"encoding/json"
 	"path/filepath"
 	"strings"
 
 	cdi "github.com/container-orchestrated-devices/container-device-interface/pkg"
 	"github.com/containers/common/libimage"
 	"github.com/containers/podman/v3/libpod"
+	"github.com/containers/podman/v3/libpod/define"
 	"github.com/containers/podman/v3/pkg/namespaces"
 	"github.com/containers/podman/v3/pkg/specgen"
 	"github.com/containers/podman/v3/pkg/util"
@@ -29,43 +30,30 @@ func MakeContainer(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGener
 
 	// If joining a pod, retrieve the pod for use, and its infra container
 	var pod *libpod.Pod
-	var infraConfig *libpod.ContainerConfig
+	var infra *libpod.Container
 	if s.Pod != "" {
 		pod, err = rt.LookupPod(s.Pod)
 		if err != nil {
 			return nil, nil, nil, errors.Wrapf(err, "error retrieving pod %s", s.Pod)
 		}
 		if pod.HasInfraContainer() {
-			infra, err := pod.InfraContainer()
+			infra, err = pod.InfraContainer()
 			if err != nil {
 				return nil, nil, nil, err
 			}
-			infraConfig = infra.Config()
 		}
 	}
 
-	if infraConfig != nil && (len(infraConfig.NamedVolumes) > 0 || len(infraConfig.UserVolumes) > 0 || len(infraConfig.ImageVolumes) > 0 || len(infraConfig.OverlayVolumes) > 0) {
-		s.VolumesFrom = append(s.VolumesFrom, infraConfig.ID)
-	}
-
-	if infraConfig != nil && len(infraConfig.Spec.Linux.Devices) > 0 {
-		s.DevicesFrom = append(s.DevicesFrom, infraConfig.ID)
-	}
-	if infraConfig != nil && infraConfig.Spec.Linux.Resources != nil && infraConfig.Spec.Linux.Resources.BlockIO != nil && len(infraConfig.Spec.Linux.Resources.BlockIO.ThrottleReadBpsDevice) > 0 {
-		tempDev := make(map[string]spec.LinuxThrottleDevice)
-		for _, val := range infraConfig.Spec.Linux.Resources.BlockIO.ThrottleReadBpsDevice {
-			nodes, err := util.FindDeviceNodes()
-			if err != nil {
-				return nil, nil, nil, err
-			}
-			key := fmt.Sprintf("%d:%d", val.Major, val.Minor)
-			tempDev[nodes[key]] = spec.LinuxThrottleDevice{Rate: uint64(val.Rate)}
-		}
-		for i, dev := range s.ThrottleReadBpsDevice {
-			tempDev[i] = dev
+	options := []libpod.CtrCreateOption{}
+	compatibleOptions := &libpod.InfraInherit{}
+	var infraSpec *spec.Spec
+	if infra != nil {
+		options, infraSpec, compatibleOptions, err = Inherit(*infra)
+		if err != nil {
+			return nil, nil, nil, err
 		}
-		s.ThrottleReadBpsDevice = tempDev
 	}
+
 	if err := FinishThrottleDevices(s); err != nil {
 		return nil, nil, nil, err
 	}
@@ -119,8 +107,6 @@ func MakeContainer(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGener
 		s.CgroupNS = defaultNS
 	}
 
-	options := []libpod.CtrCreateOption{}
-
 	if s.ContainerCreateCommand != nil {
 		options = append(options, libpod.WithCreateCommand(s.ContainerCreateCommand))
 	}
@@ -165,7 +151,8 @@ func MakeContainer(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGener
 		return nil, nil, nil, err
 	}
 
-	opts, err := createContainerOptions(ctx, rt, s, pod, finalVolumes, finalOverlays, imageData, command)
+	infraVolumes := (len(compatibleOptions.InfraVolumes) > 0 || len(compatibleOptions.InfraUserVolumes) > 0 || len(compatibleOptions.InfraImageVolumes) > 0)
+	opts, err := createContainerOptions(ctx, rt, s, pod, finalVolumes, finalOverlays, imageData, command, infraVolumes, *compatibleOptions)
 	if err != nil {
 		return nil, nil, nil, err
 	}
@@ -178,27 +165,29 @@ func MakeContainer(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGener
 		logrus.Debugf("setting container name %s", s.Name)
 		options = append(options, libpod.WithName(s.Name))
 	}
-	if len(s.DevicesFrom) > 0 {
-		for _, dev := range s.DevicesFrom {
-			ctr, err := rt.GetContainer(dev)
-			if err != nil {
-				return nil, nil, nil, err
-			}
-			devices := ctr.DeviceHostSrc()
-			s.Devices = append(s.Devices, devices...)
-		}
-	}
 	if len(s.Devices) > 0 {
-		opts = extractCDIDevices(s)
+		opts = ExtractCDIDevices(s)
 		options = append(options, opts...)
 	}
-	runtimeSpec, err := SpecGenToOCI(ctx, s, rt, rtc, newImage, finalMounts, pod, command)
+	runtimeSpec, err := SpecGenToOCI(ctx, s, rt, rtc, newImage, finalMounts, pod, command, compatibleOptions)
 	if err != nil {
 		return nil, nil, nil, err
 	}
 	if len(s.HostDeviceList) > 0 {
 		options = append(options, libpod.WithHostDevice(s.HostDeviceList))
 	}
+	if infraSpec != nil && infraSpec.Linux != nil { // if we are inheriting Linux info from a pod...
+		// Pass Security annotations
+		if len(infraSpec.Annotations[define.InspectAnnotationLabel]) > 0 && len(runtimeSpec.Annotations[define.InspectAnnotationLabel]) == 0 {
+			runtimeSpec.Annotations[define.InspectAnnotationLabel] = infraSpec.Annotations[define.InspectAnnotationLabel]
+		}
+		if len(infraSpec.Annotations[define.InspectAnnotationSeccomp]) > 0 && len(runtimeSpec.Annotations[define.InspectAnnotationSeccomp]) == 0 {
+			runtimeSpec.Annotations[define.InspectAnnotationSeccomp] = infraSpec.Annotations[define.InspectAnnotationSeccomp]
+		}
+		if len(infraSpec.Annotations[define.InspectAnnotationApparmor]) > 0 && len(runtimeSpec.Annotations[define.InspectAnnotationApparmor]) == 0 {
+			runtimeSpec.Annotations[define.InspectAnnotationApparmor] = infraSpec.Annotations[define.InspectAnnotationApparmor]
+		}
+	}
 	return runtimeSpec, s, options, err
 }
 func ExecuteCreate(ctx context.Context, rt *libpod.Runtime, runtimeSpec *spec.Spec, s *specgen.SpecGenerator, infra bool, options ...libpod.CtrCreateOption) (*libpod.Container, error) {
@@ -210,7 +199,7 @@ func ExecuteCreate(ctx context.Context, rt *libpod.Runtime, runtimeSpec *spec.Sp
 	return ctr, rt.PrepareVolumeOnCreateContainer(ctx, ctr)
 }
 
-func extractCDIDevices(s *specgen.SpecGenerator) []libpod.CtrCreateOption {
+func ExtractCDIDevices(s *specgen.SpecGenerator) []libpod.CtrCreateOption {
 	devs := make([]spec.LinuxDevice, 0, len(s.Devices))
 	var cdiDevs []string
 	var options []libpod.CtrCreateOption
@@ -224,19 +213,16 @@ func extractCDIDevices(s *specgen.SpecGenerator) []libpod.CtrCreateOption {
 			cdiDevs = append(cdiDevs, device.Path)
 			continue
 		}
-
 		devs = append(devs, device)
 	}
-
 	s.Devices = devs
 	if len(cdiDevs) > 0 {
 		options = append(options, libpod.WithCDI(cdiDevs))
 	}
-
 	return options
 }
 
-func createContainerOptions(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGenerator, pod *libpod.Pod, volumes []*specgen.NamedVolume, overlays []*specgen.OverlayVolume, imageData *libimage.ImageData, command []string) ([]libpod.CtrCreateOption, error) {
+func createContainerOptions(ctx context.Context, rt *libpod.Runtime, s *specgen.SpecGenerator, pod *libpod.Pod, volumes []*specgen.NamedVolume, overlays []*specgen.OverlayVolume, imageData *libimage.ImageData, command []string, infraVolumes bool, compatibleOptions libpod.InfraInherit) ([]libpod.CtrCreateOption, error) {
 	var options []libpod.CtrCreateOption
 	var err error
 
@@ -317,7 +303,10 @@ func createContainerOptions(ctx context.Context, rt *libpod.Runtime, s *specgen.
 	for _, imageVolume := range s.ImageVolumes {
 		destinations = append(destinations, imageVolume.Destination)
 	}
-	options = append(options, libpod.WithUserVolumes(destinations))
+
+	if len(destinations) > 0 || !infraVolumes {
+		options = append(options, libpod.WithUserVolumes(destinations))
+	}
 
 	if len(volumes) != 0 {
 		var vols []*libpod.ContainerNamedVolume
@@ -405,7 +394,7 @@ func createContainerOptions(ctx context.Context, rt *libpod.Runtime, s *specgen.
 	if len(s.SelinuxOpts) > 0 {
 		options = append(options, libpod.WithSecLabels(s.SelinuxOpts))
 	} else {
-		if pod != nil {
+		if pod != nil && len(compatibleOptions.InfraLabels) == 0 {
 			// duplicate the security options from the pod
 			processLabel, err := pod.ProcessLabel()
 			if err != nil {
@@ -498,3 +487,33 @@ func createContainerOptions(ctx context.Context, rt *libpod.Runtime, s *specgen.
 
 	return options, nil
 }
+
+func Inherit(infra libpod.Container) (opts []libpod.CtrCreateOption, infraS *spec.Spec, compat *libpod.InfraInherit, err error) {
+	options := []libpod.CtrCreateOption{}
+	compatibleOptions := &libpod.InfraInherit{}
+	infraConf := infra.Config()
+	infraSpec := infraConf.Spec
+
+	config, err := json.Marshal(infraConf)
+	if err != nil {
+		return nil, nil, nil, err
+	}
+	err = json.Unmarshal(config, compatibleOptions)
+	if err != nil {
+		return nil, nil, nil, err
+	}
+	if infraSpec.Linux != nil && infraSpec.Linux.Resources != nil {
+		resources, err := json.Marshal(infraSpec.Linux.Resources)
+		if err != nil {
+			return nil, nil, nil, err
+		}
+		err = json.Unmarshal(resources, &compatibleOptions.InfraResources)
+		if err != nil {
+			return nil, nil, nil, err
+		}
+	}
+	if compatibleOptions != nil {
+		options = append(options, libpod.WithInfraConfig(*compatibleOptions))
+	}
+	return options, infraSpec, compatibleOptions, nil
+}
diff --git a/pkg/specgen/generate/oci.go b/pkg/specgen/generate/oci.go
index efac53104..ee3a990fc 100644
--- a/pkg/specgen/generate/oci.go
+++ b/pkg/specgen/generate/oci.go
@@ -2,6 +2,7 @@ package generate
 
 import (
 	"context"
+	"encoding/json"
 	"path"
 	"strings"
 
@@ -174,7 +175,7 @@ func getCGroupPermissons(unmask []string) string {
 }
 
 // SpecGenToOCI returns the base configuration for the container.
-func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runtime, rtc *config.Config, newImage *libimage.Image, mounts []spec.Mount, pod *libpod.Pod, finalCmd []string) (*spec.Spec, error) {
+func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runtime, rtc *config.Config, newImage *libimage.Image, mounts []spec.Mount, pod *libpod.Pod, finalCmd []string, compatibleOptions *libpod.InfraInherit) (*spec.Spec, error) {
 	cgroupPerm := getCGroupPermissons(s.Unmask)
 
 	g, err := generate.New("linux")
@@ -299,9 +300,32 @@ func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runt
 		g.AddAnnotation(key, val)
 	}
 
-	g.Config.Linux.Resources = s.ResourceLimits
+	if compatibleOptions.InfraResources == nil && s.ResourceLimits != nil {
+		g.Config.Linux.Resources = s.ResourceLimits
+	} else if s.ResourceLimits != nil { // if we have predefined resource limits we need to make sure we keep the infra and container limits
+		originalResources, err := json.Marshal(s.ResourceLimits)
+		if err != nil {
+			return nil, err
+		}
+		infraResources, err := json.Marshal(compatibleOptions.InfraResources)
+		if err != nil {
+			return nil, err
+		}
+		err = json.Unmarshal(infraResources, s.ResourceLimits) // put infra's resource limits in the container
+		if err != nil {
+			return nil, err
+		}
+		err = json.Unmarshal(originalResources, s.ResourceLimits) // make sure we did not override anything
+		if err != nil {
+			return nil, err
+		}
+		g.Config.Linux.Resources = s.ResourceLimits
+	} else {
+		g.Config.Linux.Resources = compatibleOptions.InfraResources
+	}
 	// Devices
 
+	var userDevices []spec.LinuxDevice
 	if s.Privileged {
 		// If privileged, we need to add all the host devices to the
 		// spec.  We do not add the user provided ones because we are
@@ -316,14 +340,19 @@ func SpecGenToOCI(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.Runt
 				return nil, err
 			}
 		}
+		if len(compatibleOptions.InfraDevices) > 0 && len(s.Devices) == 0 {
+			userDevices = compatibleOptions.InfraDevices
+		} else {
+			userDevices = s.Devices
+		}
 		// add default devices specified by caller
-		for _, device := range s.Devices {
+		for _, device := range userDevices {
 			if err = DevicesFromPath(&g, device.Path); err != nil {
 				return nil, err
 			}
 		}
 	}
-	s.HostDeviceList = s.Devices
+	s.HostDeviceList = userDevices
 
 	// set the devices cgroup when not running in a user namespace
 	if !inUserNS && !s.Privileged {