CON26-J. Do not publish partially initialized objects

Sometimes an object is required to be shared amongst multiple threads. During initialization, the object must remain exclusive to the thread constructing it. However, once the object is initialized, it can be safely published. That is, it can be made visible to other threads. The [Java Memory Model] allows multiple threads to observe the object after its initialization has begun, but before it has concluded. Consequently, it is important to ensure that a partially initialized object is not published.

This guideline is similar to [CON14-J. Do not let the "this" reference escape during object construction]. In this guideline, a reference to a partially initialized member object instance is published before initialization is over, instead of the this reference of the current object.

Noncompliant Code Example

This noncompliant code example initializes a Helper object inside class Foo.

class Foo {
  private Helper helper;

  public Helper getHelper() {
    return helper;
  }

  public void initialize() {
    helper = new Helper(42);
  }
}

public class Helper {
  private int n;

  public Helper(int n) {
    this.n = n;
  }
  // ...
}

Suppose two or more threads can access the field helper using the getHelper() method, and initialize() has not been called yet. Consequently, all threads will see the helper field as uninitialized. Later, if one thread calls initialize(), and another calls getHelper(), the second thread may either see the helper reference as null, observe a fully-initialized Helper object with the n field set to 42, or observe a partially-initialized Helper object with an n that has not been initialized yet, and which contains the default value 0.

In particular, the [Java Memory Model (JMM)] permits compilers to allocate memory for the new Helper object and assign it to the helper field before initializing it. In other words, the compiler can reorder the write to the instance field helper with the write that initializes the object Helper (that is, this.n = n) such that the former occurs first. This introduces a race window during which other threads may see a partially-initialized Helper object instance.

There is another consideration in that if two threads both call initialize(), then two Helper objects will be created, with one eventually being garbage-collected. This is a performance issue, not a correctness issue.

Compliant Solution (synchronized)

The reference of a partially-constructed object can be prevented from being made visible by using method synchronization.

class Foo {
  private Helper helper;

  public synchronized Helper getHelper() {
    return helper;
  }

  public synchronized void initialize() {
    helper = new Helper(42);
  }
}

Synchronizing both methods guarantees that they will never run simultaneously in different threads. If one thread were to call initialize() just before another thread calls getHelper(), the synchronized initialize() method will always finish first. Furthermore, this establishes a [happens-before relationship] between the two threads. This guarantees that the thread calling getHelper() sees the fully initialized Helper object or none at all (helper may contain a null reference). This approach guarantees proper publication for both immutable and mutable members.

Compliant Solution (final field)

If the helper field is declared as final, it is guaranteed to be fully constructed before its reference is made visible.

class Foo {
  private final Helper helper;

  public Helper getHelper() {
    return helper;
  }

  public Foo() {
    helper = new Helper(42);
  }
}

However, the compiler now disallows setting the helper field to a new object using the initialize() method. Instead, a constructor must be used to initialize helper.

According to the Java Language Specification [[JLS 05]], section 17.5.2 "Reading Final Fields During Construction":

A read of a final field of an object within the thread that constructs that object is ordered with respect to the initialization of that field within the constructor by the usual happens-before rules. If the read occurs after the field is set in the constructor, it sees the value the final field is assigned, otherwise it sees the default value.

Consequently, the reference to the helper field should not be published before class Foo's constructor has finished its initialization.

Compliant Solution (thread-safe composition)

Some collection classes provide thread-safety of accesses to contained elements. If the helper field is contained in such a collection, the Helper object is guaranteed to be fully initialized before its reference is made visible. This compliant solution encases the helper field in a Vector.

class Foo {
  private Vector<Helper> helper;

  public Helper getHelper() {
    return helper.elementAt(0);
  }

  public synchronized void initialize() {
    if (helper == null) {
      helper = new Vector<Helper>();
      helper.add(new Helper(42));
    }
  }
}

It synchronizes the initialization and used a null} check to prevent two {{Helper objects from being created and added to the vector (in this case the new instances are not garbage collected, necessitating this step).

Compliant Solution (final and thread-safe composition)

This solution is similar to the previous one, except that the helper field is declared final. Consequently, the helper is required to be created in the constructor. It can be suitably initialized using the initialize() method.

class Foo {
  private final Vector<Helper> helper;

  public Helper getHelper() {
    return helper.elementAt(0);
  }

  Foo() {
    helper = new Vector<Helper>();  
  }
  public synchronized void initialize() {
    if (helper.isEmpty()) {
      helper.add(new Helper(42));
    }
  }
}

Compliant Solution (static initialization)

In this compliant solution, the helper field is initialized in a static block. When initialized statically, an object is guaranteed to be fully initialized before its reference is made visible.

class Foo {
  private static final Helper helper = new Helper(42);

  public static Helper getHelper() {
    return helper;
  } 
}

This requires the helper field to be declared static. It is recommended that the field be declared final as well to sufficiently document the class's immutability property.

According to JSR-133 [[JSR-133 04]], 9.2.3 Static Final Fields:

The rules for class initialization ensure that any thread that reads a static field will be synchronized with the static initialization of that class, which is the only place where static final fields can be set. Thus, no special rules in the JMM are needed for static final fields.

Compliant Solution (initialized flag)

This compliant solution uses a volatile initialized flag, as recommended by [CON28-J. Prevent partially initialized objects from being used]. This ensures that even if the reference to the Helper object instance is published before its initialization is over, the instance will be unusable. This is because every method within Helper checks the flag to determine whether the initialization has finished.

public class Helper {
  private int n;
  private volatile boolean initialized; // defaults to false

  public Helper(int n) {
    this.n = n;
    this.initialized = true;
  }
  
  public void doSomething() {
    if(!initialized) {
      throw new SecurityException("Cannot use partially initialized instance");
    }
    // ... 
  }
  // ...
}

Compliant Solution (public static factory method)

This compliant solution uses a public static factory method in class Helper to create a new instance of Helper. The instance is created using a private constructor. The code for class Foo is the same as the noncompliant code example.

class Foo {
  private Helper helper;

  public Helper getHelper() {
    return helper;
  }

  public void initialize() {
    helper = Helper.getHelper();
  }
}

public class Helper {
  private int n;

  private Helper(int n) {
    this.n = n;
  }
  
  public static Helper getHelper() {
    Helper helper = new Helper(42);
    return helper;
  }
  // ...
}

This approach guarantees that the Helper object will be fully initialized before it is returned to class Foo. It is useful when the nature of the caller of getHelper() is not known.

Compliant Solution (immutable object, volatile reference)

The Java memory model guarantees that any final fields of the object will be fully initialized before a published object becomes visible [[Goetz 06]]. By making n final, we can render the Helper class [immutable]. Furthermore, if the helper field is declared volatile, the Helper's reference is guaranteed to be made visible to any thread that calls getHelper() after it has been fully initialized.

class Foo {
  private volatile Helper helper;

  public Helper getHelper() {
    return helper;
  }

  public void initialize() {
    helper = new Helper(42);
  }
}

// Immutable Helper
public class Helper {
  private final int n;

  public Helper(int n) {
    this.n = n;
  }
  // ...
}

Note that if the state of the Helper object could be changed after its construction, additional locking would be necessary to ensure all threads see the most recent state. See [CON11-J. Do not assume that declaring an object volatile guarantees visibility of its members] for more information. This is typically the case when a mutable, but thread-safe member object is expected to continually publish its most recent state when its reference is declared volatile. The use of volatile guarantees the visibility of only the initial publication and not of subsequent state changes. Consequently, volatile is often not very useful for publishing thread-safe objects, whereas immutable objects can safely use this solution.

This compliant solution requires that helper be both volatile and immutable. If it is not immutable, the code would violate [CON11-J. Do not assume that declaring an object volatile guarantees visibility of its members] and additional synchronization would be required to fix it. And if it were not volatile, it would violate [CON09-J. Ensure visibility of shared references to immutable objects]. See this rule for more information on how to safely publish immutable objects.

Primitive types can also be safely published by publishing an atomic reference to the corresponding boxed type. For instance, an int field can be safely published by publishing an atomic reference to the equivalent Integer using java.util.concurrent.atomic.AtomicReference<Integer>.

Risk Assessment

Failing to synchronize access to shared mutable data can cause different threads to observe different states of the object or a partially initialized object.

Automated Detection

TODO

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

References

[[API 06]]
[[Bloch 01]] Item 48: "Synchronize access to shared mutable data"
[[Goetz 06]] Section 3.5.3 "Safe Publication Idioms"
[[Goetz 07]] Pattern #2: "one-time safe publication"
[[JPL 06]], 14.10.2. Final Fields and Security:
[[Pugh 04]]

[!The CERT Sun Microsystems Secure Coding Standard for Java^button_arrow_left.png!]      [!The CERT Sun Microsystems Secure Coding Standard for Java^button_arrow_up.png!]      [!The CERT Sun Microsystems Secure Coding Standard for Java^button_arrow_right.png!]