Merge pull request #6965 from giuseppe/followup-pr6324

allow switching of port-forward approaches in rootless/using slirp4netns

13 files changed, 214 insertions, 12 deletions

diff --git a/cmd/podman/common/netflags.go b/cmd/podman/common/netflags.go
index 5c83a3c9c..82a82b310 100644
--- a/cmd/podman/common/netflags.go
+++ b/cmd/podman/common/netflags.go
@@ -2,6 +2,7 @@ package common
 
 import (
 	"net"
+	"strings"
 
 	"github.com/containers/libpod/v2/cmd/podman/parse"
 	"github.com/containers/libpod/v2/libpod/define"
@@ -164,11 +165,18 @@ func NetFlagsToNetOptions(cmd *cobra.Command) (*entities.NetOptions, error) {
 			return nil, err
 		}
 
+		parts := strings.SplitN(network, ":", 2)
+
 		ns, cniNets, err := specgen.ParseNetworkNamespace(network)
 		if err != nil {
 			return nil, err
 		}
 
+		if len(parts) > 1 {
+			opts.NetworkOptions = make(map[string][]string)
+			opts.NetworkOptions[parts[0]] = strings.Split(parts[1], ",")
+			cniNets = nil
+		}
 		opts.Network = ns
 		opts.CNINetworks = cniNets
 	}
diff --git a/cmd/podman/common/specgen.go b/cmd/podman/common/specgen.go
index 8a265cedf..08099da4b 100644
--- a/cmd/podman/common/specgen.go
+++ b/cmd/podman/common/specgen.go
@@ -417,6 +417,7 @@ func FillOutSpecGen(s *specgen.SpecGenerator, c *ContainerCLIOpts, args []string
 	s.DNSOptions = c.Net.DNSOptions
 	s.StaticIP = c.Net.StaticIP
 	s.StaticMAC = c.Net.StaticMAC
+	s.NetworkOptions = c.Net.NetworkOptions
 	s.UseImageHosts = c.Net.NoHosts
 
 	s.ImageVolumeMode = c.ImageVolume
diff --git a/docs/source/markdown/podman-create.1.md b/docs/source/markdown/podman-create.1.md
index fd5e14535..4466e6616 100644
--- a/docs/source/markdown/podman-create.1.md
+++ b/docs/source/markdown/podman-create.1.md
@@ -553,7 +553,10 @@ Valid values are:
 - `<network-name>|<network-id>`: connect to a user-defined network, multiple networks should be comma separated
 - `ns:<path>`: path to a network namespace to join
 - `private`: create a new namespace for the container (default)
-- `slirp4netns`: use slirp4netns to create a user network stack.  This is the default for rootless containers
+- `slirp4netns[:OPTIONS,...]`: use slirp4netns to create a user network stack.  This is the default for rootless containers.  It is possible to specify these additional options:
+  **port_handler=rootlesskit**: Use rootlesskit for port forwarding.  Default.
+  **port_handler=slirp4netns**: Use the slirp4netns port forwarding.
+  **allow_host_loopback=true|false**: Allow the slirp4netns to reach the host loopback IP (`10.0.2.2`).  Default to false.
 
 **--network-alias**=*alias*
 
diff --git a/docs/source/markdown/podman-run.1.md b/docs/source/markdown/podman-run.1.md
index 68ffb69e5..de1d8aff6 100644
--- a/docs/source/markdown/podman-run.1.md
+++ b/docs/source/markdown/podman-run.1.md
@@ -561,7 +561,10 @@ Valid _mode_ values are:
 - _network-id_: connect to a user-defined network, multiple networks should be comma separated;
 - **ns:**_path_: path to a network namespace to join;
 - `private`: create a new namespace for the container (default)
-- **slirp4netns**: use **slirp4netns**(1) to create a user network stack.  This is the default for rootless containers.
+- **slirp4netns[:OPTIONS,...]**: use **slirp4netns**(1) to create a user network stack.  This is the default for rootless containers.  It is possible to specify these additional options:
+  **port_handler=rootlesskit**: Use rootlesskit for port forwarding.  Default.
+  **port_handler=slirp4netns**: Use the slirp4netns port forwarding.
+  **allow_host_loopback=true|false**: Allow the slirp4netns to reach the host loopback IP (`10.0.2.2`).  Default to false.
 
 **--network-alias**=*alias*
 
diff --git a/libpod/container.go b/libpod/container.go
index f7abfb005..1ca38ae7e 100644
--- a/libpod/container.go
+++ b/libpod/container.go
@@ -341,6 +341,8 @@ type ContainerConfig struct {
 	Networks []string `json:"networks,omitempty"`
 	// Network mode specified for the default network.
 	NetMode namespaces.NetworkMode `json:"networkMode,omitempty"`
+	// NetworkOptions are additional options for each network
+	NetworkOptions map[string][]string `json:"network_options,omitempty"`
 
 	// Image Config
 
diff --git a/libpod/networking_linux.go b/libpod/networking_linux.go
index 1e79e8732..35e58f916 100644
--- a/libpod/networking_linux.go
+++ b/libpod/networking_linux.go
@@ -173,6 +173,19 @@ type slirpFeatures struct {
 	HasEnableSeccomp       bool
 }
 
+type slirp4netnsCmdArg struct {
+	Proto     string `json:"proto,omitempty"`
+	HostAddr  string `json:"host_addr"`
+	HostPort  int32  `json:"host_port"`
+	GuestAddr string `json:"guest_addr"`
+	GuestPort int32  `json:"guest_port"`
+}
+
+type slirp4netnsCmd struct {
+	Execute string            `json:"execute"`
+	Args    slirp4netnsCmdArg `json:"arguments"`
+}
+
 func checkSlirpFlags(path string) (*slirpFeatures, error) {
 	cmd := exec.Command(path, "--help")
 	out, err := cmd.CombinedOutput()
@@ -210,12 +223,33 @@ func (r *Runtime) setupRootlessNetNS(ctr *Container) error {
 	havePortMapping := len(ctr.Config().PortMappings) > 0
 	logPath := filepath.Join(ctr.runtime.config.Engine.TmpDir, fmt.Sprintf("slirp4netns-%s.log", ctr.config.ID))
 
+	isSlirpHostForward := false
+	disableHostLoopback := true
+	if ctr.config.NetworkOptions != nil {
+		slirpOptions := ctr.config.NetworkOptions["slirp4netns"]
+		for _, o := range slirpOptions {
+			switch o {
+			case "port_handler=slirp4netns":
+				isSlirpHostForward = true
+			case "port_handler=rootlesskit":
+				isSlirpHostForward = false
+			case "allow_host_loopback=true":
+				disableHostLoopback = false
+			case "allow_host_loopback=false":
+				disableHostLoopback = true
+			default:
+				return errors.Errorf("unknown option for slirp4netns: %q", o)
+
+			}
+		}
+	}
+
 	cmdArgs := []string{}
 	slirpFeatures, err := checkSlirpFlags(path)
 	if err != nil {
 		return errors.Wrapf(err, "error checking slirp4netns binary %s: %q", path, err)
 	}
-	if slirpFeatures.HasDisableHostLoopback {
+	if disableHostLoopback && slirpFeatures.HasDisableHostLoopback {
 		cmdArgs = append(cmdArgs, "--disable-host-loopback")
 	}
 	if slirpFeatures.HasMTU {
@@ -228,6 +262,12 @@ func (r *Runtime) setupRootlessNetNS(ctr *Container) error {
 		cmdArgs = append(cmdArgs, "--enable-seccomp")
 	}
 
+	var apiSocket string
+	if havePortMapping && isSlirpHostForward {
+		apiSocket = filepath.Join(ctr.runtime.config.Engine.TmpDir, fmt.Sprintf("%s.net", ctr.config.ID))
+		cmdArgs = append(cmdArgs, "--api-socket", apiSocket)
+	}
+
 	// the slirp4netns arguments being passed are describes as follows:
 	// from the slirp4netns documentation: https://github.com/rootless-containers/slirp4netns
 	// -c, --configure Brings up the tap interface
@@ -291,7 +331,11 @@ func (r *Runtime) setupRootlessNetNS(ctr *Container) error {
 	}
 
 	if havePortMapping {
-		return r.setupRootlessPortMapping(ctr, netnsPath)
+		if isSlirpHostForward {
+			return r.setupRootlessPortMappingViaSlirp(ctr, cmd, apiSocket)
+		} else {
+			return r.setupRootlessPortMappingViaRLK(ctr, netnsPath)
+		}
 	}
 	return nil
 }
@@ -342,7 +386,7 @@ func waitForSync(syncR *os.File, cmd *exec.Cmd, logFile io.ReadSeeker, timeout t
 	return nil
 }
 
-func (r *Runtime) setupRootlessPortMapping(ctr *Container, netnsPath string) error {
+func (r *Runtime) setupRootlessPortMappingViaRLK(ctr *Container, netnsPath string) error {
 	syncR, syncW, err := os.Pipe()
 	if err != nil {
 		return errors.Wrapf(err, "failed to open pipe")
@@ -419,6 +463,90 @@ func (r *Runtime) setupRootlessPortMapping(ctr *Container, netnsPath string) err
 	return nil
 }
 
+func (r *Runtime) setupRootlessPortMappingViaSlirp(ctr *Container, cmd *exec.Cmd, apiSocket string) (err error) {
+	const pidWaitTimeout = 60 * time.Second
+	chWait := make(chan error)
+	go func() {
+		interval := 25 * time.Millisecond
+		for i := time.Duration(0); i < pidWaitTimeout; i += interval {
+			// Check if the process is still running.
+			var status syscall.WaitStatus
+			pid, err := syscall.Wait4(cmd.Process.Pid, &status, syscall.WNOHANG, nil)
+			if err != nil {
+				break
+			}
+			if pid != cmd.Process.Pid {
+				continue
+			}
+			if status.Exited() || status.Signaled() {
+				chWait <- fmt.Errorf("slirp4netns exited with status %d", status.ExitStatus())
+			}
+			time.Sleep(interval)
+		}
+	}()
+	defer close(chWait)
+
+	// wait that API socket file appears before trying to use it.
+	if _, err := WaitForFile(apiSocket, chWait, pidWaitTimeout); err != nil {
+		return errors.Wrapf(err, "waiting for slirp4nets to create the api socket file %s", apiSocket)
+	}
+
+	// for each port we want to add we need to open a connection to the slirp4netns control socket
+	// and send the add_hostfwd command.
+	for _, i := range ctr.config.PortMappings {
+		conn, err := net.Dial("unix", apiSocket)
+		if err != nil {
+			return errors.Wrapf(err, "cannot open connection to %s", apiSocket)
+		}
+		defer func() {
+			if err := conn.Close(); err != nil {
+				logrus.Errorf("unable to close connection: %q", err)
+			}
+		}()
+		hostIP := i.HostIP
+		if hostIP == "" {
+			hostIP = "0.0.0.0"
+		}
+		apiCmd := slirp4netnsCmd{
+			Execute: "add_hostfwd",
+			Args: slirp4netnsCmdArg{
+				Proto:     i.Protocol,
+				HostAddr:  hostIP,
+				HostPort:  i.HostPort,
+				GuestPort: i.ContainerPort,
+			},
+		}
+		// create the JSON payload and send it.  Mark the end of request shutting down writes
+		// to the socket, as requested by slirp4netns.
+		data, err := json.Marshal(&apiCmd)
+		if err != nil {
+			return errors.Wrapf(err, "cannot marshal JSON for slirp4netns")
+		}
+		if _, err := conn.Write([]byte(fmt.Sprintf("%s\n", data))); err != nil {
+			return errors.Wrapf(err, "cannot write to control socket %s", apiSocket)
+		}
+		if err := conn.(*net.UnixConn).CloseWrite(); err != nil {
+			return errors.Wrapf(err, "cannot shutdown the socket %s", apiSocket)
+		}
+		buf := make([]byte, 2048)
+		readLength, err := conn.Read(buf)
+		if err != nil {
+			return errors.Wrapf(err, "cannot read from control socket %s", apiSocket)
+		}
+		// if there is no 'error' key in the received JSON data, then the operation was
+		// successful.
+		var y map[string]interface{}
+		if err := json.Unmarshal(buf[0:readLength], &y); err != nil {
+			return errors.Wrapf(err, "error parsing error status from slirp4netns")
+		}
+		if e, found := y["error"]; found {
+			return errors.Errorf("error from slirp4netns while setting up port redirection: %v", e)
+		}
+	}
+	logrus.Debug("slirp4netns port-forwarding setup via add_hostfwd is ready")
+	return nil
+}
+
 // Configure the network namespace using the container process
 func (r *Runtime) setupNetNS(ctr *Container) error {
 	nsProcess := fmt.Sprintf("/proc/%d/ns/net", ctr.state.PID)
diff --git a/libpod/options.go b/libpod/options.go
index b3c11ebc1..32748a3c1 100644
--- a/libpod/options.go
+++ b/libpod/options.go
@@ -1003,6 +1003,19 @@ func WithStaticIP(ip net.IP) CtrCreateOption {
 	}
 }
 
+// WithNetworkOptions sets additional options for the networks.
+func WithNetworkOptions(options map[string][]string) CtrCreateOption {
+	return func(ctr *Container) error {
+		if ctr.valid {
+			return define.ErrCtrFinalized
+		}
+
+		ctr.config.NetworkOptions = options
+
+		return nil
+	}
+}
+
 // WithStaticMAC indicates that the container should request a static MAC from
 // the CNI plugins.
 // It cannot be set unless WithNetNS has already been passed.
diff --git a/pkg/domain/entities/types.go b/pkg/domain/entities/types.go
index 7e910ff61..b313e5f8b 100644
--- a/pkg/domain/entities/types.go
+++ b/pkg/domain/entities/types.go
@@ -42,6 +42,8 @@ type NetOptions struct {
 	PublishPorts       []specgen.PortMapping
 	StaticIP           *net.IP
 	StaticMAC          *net.HardwareAddr
+	// NetworkOptions are additional options for each network
+	NetworkOptions map[string][]string
 }
 
 // All CLI inspect commands and inspect sub-commands use the same options
diff --git a/pkg/namespaces/namespaces.go b/pkg/namespaces/namespaces.go
index 2ffbde977..7831af8f9 100644
--- a/pkg/namespaces/namespaces.go
+++ b/pkg/namespaces/namespaces.go
@@ -385,7 +385,7 @@ func (n NetworkMode) IsBridge() bool {
 
 // IsSlirp4netns indicates if we are running a rootless network stack
 func (n NetworkMode) IsSlirp4netns() bool {
-	return n == slirpType
+	return n == slirpType || strings.HasPrefix(string(n), slirpType+":")
 }
 
 // IsNS indicates a network namespace passed in by path (ns:<path>)
diff --git a/pkg/specgen/generate/namespaces.go b/pkg/specgen/generate/namespaces.go
index 09d6ba445..a19009bc2 100644
--- a/pkg/specgen/generate/namespaces.go
+++ b/pkg/specgen/generate/namespaces.go
@@ -2,6 +2,7 @@ package generate
 
 import (
 	"context"
+	"fmt"
 	"os"
 	"strings"
 
@@ -226,7 +227,11 @@ func namespaceOptions(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.
 		if err != nil {
 			return nil, err
 		}
-		toReturn = append(toReturn, libpod.WithNetNS(portMappings, postConfigureNetNS, "slirp4netns", nil))
+		val := "slirp4netns"
+		if s.NetNS.Value != "" {
+			val = fmt.Sprintf("slirp4netns:%s", s.NetNS.Value)
+		}
+		toReturn = append(toReturn, libpod.WithNetNS(portMappings, postConfigureNetNS, val, nil))
 	case specgen.Bridge:
 		portMappings, err := createPortMappings(ctx, s, img)
 		if err != nil {
@@ -261,6 +266,9 @@ func namespaceOptions(ctx context.Context, s *specgen.SpecGenerator, rt *libpod.
 	if s.StaticMAC != nil {
 		toReturn = append(toReturn, libpod.WithStaticMAC(*s.StaticMAC))
 	}
+	if s.NetworkOptions != nil {
+		toReturn = append(toReturn, libpod.WithNetworkOptions(s.NetworkOptions))
+	}
 
 	return toReturn, nil
 }
@@ -465,7 +473,7 @@ func GetNamespaceOptions(ns []string) ([]libpod.PodCreateOption, error) {
 		case "pid":
 			options = append(options, libpod.WithPodPID())
 		case "user":
-			return erroredOptions, errors.Errorf("User sharing functionality not supported on pod level")
+			continue
 		case "ipc":
 			options = append(options, libpod.WithPodIPC())
 		case "uts":
diff --git a/pkg/specgen/namespaces.go b/pkg/specgen/namespaces.go
index 5f56b242b..9bf2c5d05 100644
--- a/pkg/specgen/namespaces.go
+++ b/pkg/specgen/namespaces.go
@@ -108,7 +108,9 @@ func validateNetNS(n *Namespace) error {
 		return nil
 	}
 	switch n.NSMode {
-	case "", Default, Host, Path, FromContainer, FromPod, Private, NoNetwork, Bridge, Slirp:
+	case Slirp:
+		break
+	case "", Default, Host, Path, FromContainer, FromPod, Private, NoNetwork, Bridge:
 		break
 	default:
 		return errors.Errorf("invalid network %q", n.NSMode)
@@ -119,8 +121,8 @@ func validateNetNS(n *Namespace) error {
 		if len(n.Value) < 1 {
 			return errors.Errorf("namespace mode %s requires a value", n.NSMode)
 		}
-	} else {
-		// All others must NOT set a string value
+	} else if n.NSMode != Slirp {
+		// All others except must NOT set a string value
 		if len(n.Value) > 0 {
 			return errors.Errorf("namespace value %s cannot be provided with namespace mode %s", n.Value, n.NSMode)
 		}
@@ -250,7 +252,7 @@ func ParseNetworkNamespace(ns string) (Namespace, []string, error) {
 	var cniNetworks []string
 	// Net defaults to Slirp on rootless
 	switch {
-	case ns == "slirp4netns":
+	case ns == "slirp4netns", strings.HasPrefix(ns, "slirp4netns:"):
 		toReturn.NSMode = Slirp
 	case ns == "pod":
 		toReturn.NSMode = FromPod
diff --git a/pkg/specgen/specgen.go b/pkg/specgen/specgen.go
index 17583d82a..a346a9742 100644
--- a/pkg/specgen/specgen.go
+++ b/pkg/specgen/specgen.go
@@ -379,6 +379,9 @@ type ContainerNetworkConfig struct {
 	// Conflicts with UseImageHosts.
 	// Optional.
 	HostAdd []string `json:"hostadd,omitempty"`
+	// NetworkOptions are additional options for each network
+	// Optional.
+	NetworkOptions map[string][]string `json:"network_options,omitempty"`
 }
 
 // ContainerResourceConfig contains information on container resource limits.
diff --git a/test/e2e/run_networking_test.go b/test/e2e/run_networking_test.go
index 467d0c5ef..317f760db 100644
--- a/test/e2e/run_networking_test.go
+++ b/test/e2e/run_networking_test.go
@@ -235,6 +235,35 @@ var _ = Describe("Podman run networking", func() {
 		Expect(ncBusy).To(ExitWithError())
 	})
 
+	It("podman run network expose host port 8081 to container port 8000 using rootlesskit port handler", func() {
+		session := podmanTest.Podman([]string{"run", "--network", "slirp4netns:port_handler=rootlesskit", "-dt", "-p", "8081:8000", ALPINE, "/bin/sh"})
+		session.Wait(30)
+		Expect(session.ExitCode()).To(Equal(0))
+
+		ncBusy := SystemExec("nc", []string{"-l", "-p", "8081"})
+		Expect(ncBusy).To(ExitWithError())
+	})
+
+	It("podman run network expose host port 8082 to container port 8000 using slirp4netns port handler", func() {
+		session := podmanTest.Podman([]string{"run", "--network", "slirp4netns:port_handler=slirp4netns", "-dt", "-p", "8082:8000", ALPINE, "/bin/sh"})
+		session.Wait(30)
+		Expect(session.ExitCode()).To(Equal(0))
+		ncBusy := SystemExec("nc", []string{"-l", "-p", "8082"})
+		Expect(ncBusy).To(ExitWithError())
+	})
+
+	It("podman run network expose host port 8080 to container port 8000 using invalid port handler", func() {
+		session := podmanTest.Podman([]string{"run", "--network", "slirp4netns:port_handler=invalid", "-dt", "-p", "8080:8000", ALPINE, "/bin/sh"})
+		session.Wait(30)
+		Expect(session.ExitCode()).To(Not(Equal(0)))
+	})
+
+	It("podman run slirp4netns network with host loopback", func() {
+		session := podmanTest.Podman([]string{"run", "--network", "slirp4netns:allow_host_loopback=true", ALPINE, "ping", "-c1", "10.0.2.2"})
+		session.Wait(30)
+		Expect(session.ExitCode()).To(Equal(0))
+	})
+
 	It("podman run network expose ports in image metadata", func() {
 		session := podmanTest.Podman([]string{"create", "-dt", "-P", nginx})
 		session.Wait(90)