rootless: automatically create a systemd scope

when running in rootless mode and using systemd as cgroup manager
create automatically a systemd scope when the user doesn't own the
current cgroup.

This solves a couple of issues:

on cgroup v2 it is necessary that a process before it can moved to a
different cgroup tree must be in a directory owned by the unprivileged
user.  This is not always true, e.g. when creating a session with su
-l.

Closes: https://github.com/containers/libpod/issues/3937

Also, for running systemd in a container it was before necessary to
specify "systemd-run --scope --user podman ...", now this is done
automatically as part of this PR.

Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>

2 files changed, 68 insertions, 0 deletions

diff --git a/pkg/cgroups/cgroups_supported.go b/pkg/cgroups/cgroups_supported.go
index fcd44dfc8..2a36777d4 100644
--- a/pkg/cgroups/cgroups_supported.go
+++ b/pkg/cgroups/cgroups_supported.go
@@ -3,8 +3,15 @@
 package cgroups
 
 import (
+	"bufio"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
 	"sync"
 	"syscall"
+
+	"github.com/pkg/errors"
 )
 
 var (
@@ -25,3 +32,58 @@ func IsCgroup2UnifiedMode() (bool, error) {
 	})
 	return isUnified, isUnifiedErr
 }
+
+// UserOwnsCurrentSystemdCgroup checks whether the current EUID owns the
+// current cgroup.
+func UserOwnsCurrentSystemdCgroup() (bool, error) {
+	uid := os.Geteuid()
+
+	cgroup2, err := IsCgroup2UnifiedMode()
+	if err != nil {
+		return false, err
+	}
+
+	f, err := os.Open("/proc/self/cgroup")
+	if err != nil {
+		return false, errors.Wrapf(err, "open file /proc/self/cgroup")
+	}
+	defer f.Close()
+
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := scanner.Text()
+		parts := strings.SplitN(line, ":", 3)
+
+		if len(parts) < 3 {
+			continue
+		}
+
+		var cgroupPath string
+
+		if cgroup2 {
+			cgroupPath = filepath.Join(cgroupRoot, parts[2])
+		} else {
+			if parts[1] != "name=systemd" {
+				continue
+			}
+			cgroupPath = filepath.Join(cgroupRoot, "systemd", parts[2])
+		}
+
+		st, err := os.Stat(cgroupPath)
+		if err != nil {
+			return false, err
+		}
+		s := st.Sys()
+		if s == nil {
+			return false, fmt.Errorf("error stat cgroup path %s", cgroupPath)
+		}
+
+		if int(s.(*syscall.Stat_t).Uid) != uid {
+			return false, nil
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		return false, errors.Wrapf(err, "parsing file /proc/self/cgroup")
+	}
+	return true, nil
+}
diff --git a/pkg/cgroups/cgroups_unsupported.go b/pkg/cgroups/cgroups_unsupported.go
index 9dc196e42..cd140fbf3 100644
--- a/pkg/cgroups/cgroups_unsupported.go
+++ b/pkg/cgroups/cgroups_unsupported.go
@@ -6,3 +6,9 @@ package cgroups
 func IsCgroup2UnifiedMode() (bool, error) {
 	return false, nil
 }
+
+// UserOwnsCurrentSystemdCgroup checks whether the current EUID owns the
+// current cgroup.
+func UserOwnsCurrentSystemdCgroup() (bool, error) {
+	return false, nil
+}