podman cp: support copying on tmpfs mounts

Traditionally, the path resolution for containers has been resolved on
the *host*; relative to the container's mount point or relative to
specified bind mounts or volumes.

While this works nicely for non-running containers, it poses a problem
for running ones.  In that case, certain kinds of mounts (e.g., tmpfs)
will not resolve correctly.  A tmpfs is held in memory and hence cannot
be resolved relatively to the container's mount point.  A copy operation
will succeed but the data will not show up inside the container.

To support these kinds of mounts, we need to join the *running*
container's mount namespace (and PID namespace) when copying.

Note that this change implies moving the copy and stat logic into
`libpod` since we need to keep the container locked to avoid race
conditions.  The immediate benefit is that all logic is now inside
`libpod`; the code isn't scattered anymore.

Further note that Docker does not support copying to tmpfs mounts.

Tests have been extended to cover *both* path resolutions for running
and created containers.  New tests have been added to exercise the
tmpfs-mount case.

For the record: Some tests could be improved by using `start -a` instead
of a start-exec sequence.  Unfortunately, `start -a` is flaky in the CI
which forced me to use the more expensive start-exec option.

Signed-off-by: Valentin Rothberg <rothberg@redhat.com>

libpod/container_copy_linux.go

@@ -0,0 +1,266 @@
+// +build linux
+
+package libpod
+
+import (
+	"context"
+	"io"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strings"
+
+	buildahCopiah "github.com/containers/buildah/copier"
+	"github.com/containers/buildah/pkg/chrootuser"
+	"github.com/containers/buildah/util"
+	"github.com/containers/podman/v3/libpod/define"
+	"github.com/containers/storage"
+	"github.com/containers/storage/pkg/idtools"
+	"github.com/docker/docker/pkg/archive"
+	"github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+	"golang.org/x/sys/unix"
+)
+
+func (c *Container) copyFromArchive(ctx context.Context, path string, reader io.Reader) (func() error, error) {
+	var (
+		mountPoint   string
+		resolvedRoot string
+		resolvedPath string
+		unmount      func()
+		err          error
+	)
+
+	// Make sure that "/" copies the *contents* of the mount point and not
+	// the directory.
+	if path == "/" {
+		path = "/."
+	}
+
+	// Optimization: only mount if the container is not already.
+	if c.state.Mounted {
+		mountPoint = c.state.Mountpoint
+		unmount = func() {}
+	} else {
+		// NOTE: make sure to unmount in error paths.
+		mountPoint, err = c.mount()
+		if err != nil {
+			return nil, err
+		}
+		unmount = func() { c.unmount(false) }
+	}
+
+	if c.state.State == define.ContainerStateRunning {
+		resolvedRoot = "/"
+		resolvedPath = c.pathAbs(path)
+	} else {
+		resolvedRoot, resolvedPath, err = c.resolvePath(mountPoint, path)
+		if err != nil {
+			unmount()
+			return nil, err
+		}
+	}
+
+	decompressed, err := archive.DecompressStream(reader)
+	if err != nil {
+		unmount()
+		return nil, err
+	}
+
+	idMappings, idPair, err := getIDMappingsAndPair(c, mountPoint)
+	if err != nil {
+		decompressed.Close()
+		unmount()
+		return nil, err
+	}
+
+	logrus.Debugf("Container copy *to* %q (resolved: %q) on container %q (ID: %s)", path, resolvedPath, c.Name(), c.ID())
+
+	return func() error {
+		defer unmount()
+		defer decompressed.Close()
+		putOptions := buildahCopiah.PutOptions{
+			UIDMap:     idMappings.UIDMap,
+			GIDMap:     idMappings.GIDMap,
+			ChownDirs:  idPair,
+			ChownFiles: idPair,
+		}
+
+		return c.joinMountAndExec(ctx,
+			func() error {
+				return buildahCopiah.Put(resolvedRoot, resolvedPath, putOptions, decompressed)
+			},
+		)
+	}, nil
+}
+
+func (c *Container) copyToArchive(ctx context.Context, path string, writer io.Writer) (func() error, error) {
+	var (
+		mountPoint string
+		unmount    func()
+		err        error
+	)
+
+	// Optimization: only mount if the container is not already.
+	if c.state.Mounted {
+		mountPoint = c.state.Mountpoint
+		unmount = func() {}
+	} else {
+		// NOTE: make sure to unmount in error paths.
+		mountPoint, err = c.mount()
+		if err != nil {
+			return nil, err
+		}
+		unmount = func() { c.unmount(false) }
+	}
+
+	statInfo, resolvedRoot, resolvedPath, err := c.stat(ctx, mountPoint, path)
+	if err != nil {
+		unmount()
+		return nil, err
+	}
+
+	idMappings, idPair, err := getIDMappingsAndPair(c, mountPoint)
+	if err != nil {
+		unmount()
+		return nil, err
+	}
+
+	logrus.Debugf("Container copy *from* %q (resolved: %q) on container %q (ID: %s)", path, resolvedPath, c.Name(), c.ID())
+
+	return func() error {
+		defer unmount()
+		getOptions := buildahCopiah.GetOptions{
+			// Unless the specified points to ".", we want to copy the base directory.
+			KeepDirectoryNames: statInfo.IsDir && filepath.Base(path) != ".",
+			UIDMap:             idMappings.UIDMap,
+			GIDMap:             idMappings.GIDMap,
+			ChownDirs:          idPair,
+			ChownFiles:         idPair,
+			Excludes:           []string{"dev", "proc", "sys"},
+		}
+		return c.joinMountAndExec(ctx,
+			func() error {
+				return buildahCopiah.Get(resolvedRoot, "", getOptions, []string{resolvedPath}, writer)
+			},
+		)
+	}, nil
+}
+
+// getIDMappingsAndPair returns the ID mappings for the container and the host
+// ID pair.
+func getIDMappingsAndPair(container *Container, containerMount string) (*storage.IDMappingOptions, *idtools.IDPair, error) {
+	user, err := getContainerUser(container, containerMount)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	idMappingOpts, err := container.IDMappings()
+	if err != nil {
+		return nil, nil, err
+	}
+
+	hostUID, hostGID, err := util.GetHostIDs(idtoolsToRuntimeSpec(idMappingOpts.UIDMap), idtoolsToRuntimeSpec(idMappingOpts.GIDMap), user.UID, user.GID)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	idPair := idtools.IDPair{UID: int(hostUID), GID: int(hostGID)}
+	return &idMappingOpts, &idPair, nil
+}
+
+// getContainerUser returns the specs.User of the container.
+func getContainerUser(container *Container, mountPoint string) (specs.User, error) {
+	userspec := container.Config().User
+
+	uid, gid, _, err := chrootuser.GetUser(mountPoint, userspec)
+	u := specs.User{
+		UID:      uid,
+		GID:      gid,
+		Username: userspec,
+	}
+
+	if !strings.Contains(userspec, ":") {
+		groups, err2 := chrootuser.GetAdditionalGroupsForUser(mountPoint, uint64(u.UID))
+		if err2 != nil {
+			if errors.Cause(err2) != chrootuser.ErrNoSuchUser && err == nil {
+				err = err2
+			}
+		} else {
+			u.AdditionalGids = groups
+		}
+	}
+
+	return u, err
+}
+
+// idtoolsToRuntimeSpec converts idtools ID mapping to the one of the runtime spec.
+func idtoolsToRuntimeSpec(idMaps []idtools.IDMap) (convertedIDMap []specs.LinuxIDMapping) {
+	for _, idmap := range idMaps {
+		tempIDMap := specs.LinuxIDMapping{
+			ContainerID: uint32(idmap.ContainerID),
+			HostID:      uint32(idmap.HostID),
+			Size:        uint32(idmap.Size),
+		}
+		convertedIDMap = append(convertedIDMap, tempIDMap)
+	}
+	return convertedIDMap
+}
+
+// joinMountAndExec executes the specified function `f` inside the container's
+// mount and PID namespace.  That allows for having the exact view on the
+// container's file system.
+//
+// Note, if the container is not running `f()` will be executed as is.
+func (c *Container) joinMountAndExec(ctx context.Context, f func() error) error {
+	if c.state.State != define.ContainerStateRunning {
+		return f()
+	}
+
+	// Container's running, so we need to execute `f()` inside its mount NS.
+	errChan := make(chan error)
+	go func() {
+		runtime.LockOSThread()
+
+		// Join the mount and PID NS of the container.
+		getFD := func(ns LinuxNS) (*os.File, error) {
+			nsPath, err := c.namespacePath(ns)
+			if err != nil {
+				return nil, err
+			}
+			return os.Open(nsPath)
+		}
+
+		mountFD, err := getFD(MountNS)
+		if err != nil {
+			errChan <- err
+			return
+		}
+		defer mountFD.Close()
+
+		pidFD, err := getFD(PIDNS)
+		if err != nil {
+			errChan <- err
+			return
+		}
+		defer pidFD.Close()
+		if err := unix.Unshare(unix.CLONE_NEWNS); err != nil {
+			errChan <- err
+			return
+		}
+		if err := unix.Setns(int(pidFD.Fd()), unix.CLONE_NEWPID); err != nil {
+			errChan <- err
+			return
+		}
+
+		if err := unix.Setns(int(mountFD.Fd()), unix.CLONE_NEWNS); err != nil {
+			errChan <- err
+			return
+		}
+
+		// Last but not least, execute the workload.
+		errChan <- f()
+	}()
+	return <-errChan
+}