diff --git a/libcontainer/standard_init_linux.go b/libcontainer/standard_init_linux.go
index 1a9c4979c26..6ad25c9a4df 100644
--- a/libcontainer/standard_init_linux.go
+++ b/libcontainer/standard_init_linux.go
@@ -198,6 +198,13 @@ func (l *linuxStandardInit) Init() error {
 	if err != nil {
 		return err
 	}
+	// exec.LookPath might return no error for an executable residing on a
+	// file system mounted with noexec flag, so perform this extra check
+	// now while we can still return a proper error.
+	if err := system.Eaccess(name); err != nil {
+		return &os.PathError{Op: "exec", Path: name, Err: err}
+	}
+
 	// Set seccomp as close to execve as possible, so as few syscalls take
 	// place afterward (reducing the amount of syscalls that users need to
 	// enable in their seccomp profiles). However, this needs to be done
diff --git a/libcontainer/system/linux.go b/libcontainer/system/linux.go
index e1d6eb18034..039059a444c 100644
--- a/libcontainer/system/linux.go
+++ b/libcontainer/system/linux.go
@@ -31,6 +31,25 @@ func (p ParentDeathSignal) Set() error {
 	return SetParentDeathSignal(uintptr(p))
 }
 
+// Eaccess is similar to unix.Access except for setuid/setgid binaries
+// it checks against the effective (rather than real) uid and gid.
+func Eaccess(path string) error {
+	err := unix.Faccessat2(unix.AT_FDCWD, path, unix.X_OK, unix.AT_EACCESS)
+	if err != unix.ENOSYS && err != unix.EPERM { //nolint:errorlint // unix errors are bare
+		return err
+	}
+
+	// Faccessat2() not available; check if we are a set[ug]id binary.
+	if os.Getuid() == os.Geteuid() && os.Getgid() == os.Getegid() {
+		// For a non-set[ug]id binary, use access(2).
+		return unix.Access(path, unix.X_OK)
+	}
+
+	// For a setuid/setgid binary, there is no fallback way
+	// so assume we can execute the binary.
+	return nil
+}
+
 func Execv(cmd string, args []string, env []string) error {
 	name, err := exec.LookPath(cmd)
 	if err != nil {