cgroup: plug leaks of /sys/fs/cgroup handle

We auto-close this file descriptor in the final exec step, but it's
probably a good idea to not possibly leak the file descriptor to "runc
init" (we've had issues like this in the past) especially since it is a
directory handle from the host mount namespace.

In practice, on runc 1.1 this does leak to "runc init" but on main the
handle has a low enough file descriptor that it gets clobbered by the
ForkExec of "runc init".

OPEN_TREE_CLONE would let us protect this handle even further, but the
performance impact of creating an anonymous mount namespace is probably
not worth it.

Also, switch to using an *os.File for the handle so if it goes out of
scope during setup (i.e. an error occurs during setup) it will get
cleaned up by the GC.

Fixes: GHSA-xr7r-f8xq-vfvv CVE-2024-21626
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>

libcontainer/cgroups/file.go

@@ -66,16 +66,16 @@ var (
 	// TestMode is set to true by unit tests that need "fake" cgroupfs.
 	TestMode bool
 
-	cgroupFd     int = -1
-	prepOnce     sync.Once
-	prepErr      error
-	resolveFlags uint64
+	cgroupRootHandle *os.File
+	prepOnce         sync.Once
+	prepErr          error
+	resolveFlags     uint64
 )
 
 func prepareOpenat2() error {
 	prepOnce.Do(func() {
 		fd, err := unix.Openat2(-1, cgroupfsDir, &unix.OpenHow{
-			Flags: unix.O_DIRECTORY | unix.O_PATH,
+			Flags: unix.O_DIRECTORY | unix.O_PATH | unix.O_CLOEXEC,
 		})
 		if err != nil {
 			prepErr = &os.PathError{Op: "openat2", Path: cgroupfsDir, Err: err}
@@ -86,15 +86,16 @@ func prepareOpenat2() error {
 			}
 			return
 		}
+		file := os.NewFile(uintptr(fd), cgroupfsDir)
+
 		var st unix.Statfs_t
-		if err = unix.Fstatfs(fd, &st); err != nil {
+		if err := unix.Fstatfs(int(file.Fd()), &st); err != nil {
 			prepErr = &os.PathError{Op: "statfs", Path: cgroupfsDir, Err: err}
 			logrus.Warnf("falling back to securejoin: %s", prepErr)
 			return
 		}
 
-		cgroupFd = fd
-
+		cgroupRootHandle = file
 		resolveFlags = unix.RESOLVE_BENEATH | unix.RESOLVE_NO_MAGICLINKS
 		if st.Type == unix.CGROUP2_SUPER_MAGIC {
 			// cgroupv2 has a single mountpoint and no "cpu,cpuacct" symlinks
@@ -121,29 +122,29 @@ func openFile(dir, file string, flags int) (*os.File, error) {
 		return openFallback(path, flags, mode)
 	}
 
-	fd, err := unix.Openat2(cgroupFd, relPath,
+	fd, err := unix.Openat2(int(cgroupRootHandle.Fd()), relPath,
 		&unix.OpenHow{
 			Resolve: resolveFlags,
 			Flags:   uint64(flags) | unix.O_CLOEXEC,
 			Mode:    uint64(mode),
 		})
 	if err != nil {
 		err = &os.PathError{Op: "openat2", Path: path, Err: err}
-		// Check if cgroupFd is still opened to cgroupfsDir
+		// Check if cgroupRootHandle is still opened to cgroupfsDir
 		// (happens when this package is incorrectly used
 		// across the chroot/pivot_root/mntns boundary, or
 		// when /sys/fs/cgroup is remounted).
 		//
 		// TODO: if such usage will ever be common, amend this
-		// to reopen cgroupFd and retry openat2.
-		fdPath, closer := utils.ProcThreadSelf("fd/" + strconv.Itoa(cgroupFd))
+		// to reopen cgroupRootHandle and retry openat2.
+		fdPath, closer := utils.ProcThreadSelf("fd/" + strconv.Itoa(int(cgroupRootHandle.Fd())))
 		defer closer()
 		fdDest, _ := os.Readlink(fdPath)
 		if fdDest != cgroupfsDir {
-			// Wrap the error so it is clear that cgroupFd
+			// Wrap the error so it is clear that cgroupRootHandle
 			// is opened to an unexpected/wrong directory.
-			err = fmt.Errorf("cgroupFd %d unexpectedly opened to %s != %s: %w",
-				cgroupFd, fdDest, cgroupfsDir, err)
+			err = fmt.Errorf("cgroupRootHandle %d unexpectedly opened to %s != %s: %w",
+				cgroupRootHandle.Fd(), fdDest, cgroupfsDir, err)
 		}
 		return nil, err
 	}