Skip to end of metadata
Go to start of metadata

Each thread in Java is assigned to a thread group upon the thread's creation. These groups are implemented by the java.lang.ThreadGroup class. When the thread group name is not specified explicitly, the main default group is assigned by the Java Virtual Machine (JVM) [Java Tutorials]. The convenience methods of the ThreadGroup class can be used to operate on all threads belonging to a thread group at once. For instance, the ThreadGroup.interrupt() method interrupts all threads in the thread group. Thread groups also help reinforce layered security by confining threads into groups so that they avoid interference with threads in other groups [JavaThreads 2004].

Even though thread groups are useful for keeping threads organized, programmers seldom benefit from their use because many of the methods of the ThreadGroup class (for example, allowThreadSuspension(), resume(), stop(), and suspend()) are deprecated. Furthermore, many nondeprecated methods are obsolete in that they offer little desirable functionality. Ironically, a few ThreadGroup methods are not even thread-safe [Bloch 2001].

Insecure yet nondeprecated methods include

  • ThreadGroup.activeCount()
    According to the Java API [API 2014], the activeCount() method

    returns an estimate of the number of active threads in the current thread's thread group and its subgroups.

    This method is often used as a precursor to thread enumeration. Threads that have never started nevertheless reside in the thread group and are considered to be active. The active count is also affected by the presence of certain system threads [API 2014]. Consequently, the activeCount() method might fail to reflect the actual number of running tasks in the thread group.
  • ThreadGroup.enumerate()
    According to the Java API [API 2014], ThreadGroup class documentation

    [The enumerate() method] copies into the specified array every active thread in this thread group and its subgroups....
    An application might use the activeCount method to get an estimate of how big the array should be; however, if the array is too short to hold all the threads, the extra threads are silently ignored.

Using the ThreadGroup APIs to shut down threads also has pitfalls. Because the stop() method is deprecated, programs require alternative methods to stop threads. According to The Java Programming Language [JPL 2006]:

One way is for the thread initiating the termination to join the other threads and so know when those threads have terminated. However, an application may have to maintain its own list of the threads it creates because simply inspecting the ThreadGroup may return library threads that do not terminate and for which join will not return.

The Executor framework provides a better API for managing a logical grouping of threads and offers secure facilities for handling shutdown and thread exceptions [Bloch 2008]. Consequently, programs must not invoke ThreadGroup methods.

Noncompliant Code Example

This noncompliant code example contains a NetworkHandler class that maintains a controller thread. The controller thread delegates each new request to a worker thread. To demonstrate the race condition in this example, the controller thread serves three requests by starting three threads in succession from its run() method. All threads are defined to belong to the Chief thread group.

final class HandleRequest implements Runnable {
  public void run() {
    // Do something
  }
}

public final class NetworkHandler implements Runnable {
  private static ThreadGroup tg = new ThreadGroup("Chief");

  @Override public void run() {
    new Thread(tg, new HandleRequest(), "thread1").start();
    new Thread(tg, new HandleRequest(), "thread2").start();
    new Thread(tg, new HandleRequest(), "thread3").start();
  }

  public static void printActiveCount(int point) {
    System.out.println("Active Threads in Thread Group " + tg.getName() +
        " at point(" + point + "):" + " " + tg.activeCount());
  }

  public static void printEnumeratedThreads(Thread[] ta, int len) {
    System.out.println("Enumerating all threads...");
    for (int i = 0; i < len; i++) {
      System.out.println("Thread " + i + " = " + ta[i].getName());
    }
  }

  public static void main(String[] args) throws InterruptedException {
    // Start thread controller
    Thread thread = new Thread(tg, new NetworkHandler(), "controller");
    thread.start();

    // Gets the active count (insecure)
    Thread[] ta = new Thread[tg.activeCount()];

    printActiveCount(1); // P1
    // Delay to demonstrate TOCTOU condition (race window)
    Thread.sleep(1000);
    // P2: the thread count changes as new threads are initiated
    printActiveCount(2);  
    // Incorrectly uses the (now stale) thread count obtained at P1
    int n = tg.enumerate(ta);  
    // Silently ignores newly initiated threads 
    printEnumeratedThreads(ta, n); 
                                   // (between P1 and P2)

    // This code destroys the thread group if it does 
    // not have any live threads
    for (Thread thr : ta) {
      thr.interrupt();
      while(thr.isAlive());
    }
    tg.destroy();
  }
}

This implementation contains a time-of-check, time-of-use (TOCTOU) vulnerability because it obtains the count and enumerates the list without ensuring atomicity. If one or more new requests were to occur after the call to activeCount() and before the call to enumerate() in the main() method, the total number of threads in the group would increase, but the enumerated list ta would contain only the initial number, that is, two thread references: main and controller. Consequently, the program would fail to account for the newly started threads in the Chief thread group.

Any subsequent use of the ta array would be insecure. For example, calling the destroy() method to destroy the thread group and its subgroups would not work as expected. The precondition to calling destroy() is that the thread group must be empty with no executing threads. The code attempts to comply with the precondition by interrupting every thread in the thread group. However, the thread group would not be empty when the destroy() method was called, causing a java.lang.IllegalThreadStateException to be thrown.

Compliant Solution

This compliant solution uses a fixed thread pool rather than a ThreadGroup to group its three tasks. The java.util.concurrent.ExecutorService interface provides methods to manage the thread pool. Although the interface lacks methods for finding the number of actively executing threads or for enumerating the threads, the logical grouping can help control the behavior of the group as a whole. For instance, invoking the shutdownPool() method terminates all threads belonging to a particular thread pool.

public final class NetworkHandler {
  private final ExecutorService executor;

  NetworkHandler(int poolSize) {
    this.executor = Executors.newFixedThreadPool(poolSize);
  }

  public void startThreads() {
    for (int i = 0; i < 3; i++) {
      executor.execute(new HandleRequest());
    }
  }

  public void shutdownPool() {
    executor.shutdown();
  }

  public static void main(String[] args)  {
    NetworkHandler nh = new NetworkHandler(3);
    nh.startThreads();
    nh.shutdownPool();
  }
}

Before Java SE 5.0, applications that needed to catch an uncaught exception in a separate thread had to extend the ThreadGroup class because this was the only direct approach to provide the required functionality. Specifically, an application's UncaughtExceptionHandler could only be controlled by subclassing ThreadGroup. In more recent versions of Java, UncaughtExceptionHandler is maintained on a per-thread basis using an interface enclosed by the Thread class. Consequently, the ThreadGroup class provides little unique functionality [Goetz 2006], [Bloch 2008].

Refer to TPS03-J. Ensure that tasks executing in a thread pool do not fail silently for more information on using uncaught exception handlers in thread pools.

Risk Assessment

Use of the ThreadGroup APIs may result in race conditions, memory leaks, and inconsistent object state.

Rule

Severity

Likelihood

Remediation Cost

Priority

Level

THI01-J

Low

Probable

Medium

P4

L3

Automated Detection

ToolVersionCheckerDescription
CodeSonar
5.2p0
PMD.Basic.AvoidThreadGroupAvoid ThreadGroup
Parasoft Jtest
10.3
TRS.AUTGImplemented
SonarQube
6.7
S3014"ThreadGroup" should not be used

Bibliography

[API 2006]

Class Thread
Class ThreadGroup:
  Method activeCount
  Method enumerate

[Bloch 2001]

Item 53, "Avoid Thread Groups"

[Bloch 2008]

Item 73, "Avoid Thread Groups"

[Goetz 2006]

Section 7.3.1, "Uncaught Exception Handlers"

[JavaThreads 2004]

Section 13.1, "ThreadGroups"

[Java Tutorials]

[JPL 2006]

Section 23.3.3, "Shutdown Strategies"

[SDN 2006]

Bug ID 4089701
Bug ID 4229558



6 Comments

  1. I plan to change to a stronger / different title for this rule:

    Do not invoke ThreadGroup methods

    comments?

    1. I think it's reasonable to forbid it. Another choice is something like "Do not use the ThreadGroup API for performing non-trivial (or critical) operations".

      1. that is less analyzable. I'm going to leave it as is unless there are objections.

        1. Maybe a better choice would be "Do not use the ThreadGroup API for Thread control operations".  It can be very useful for diagnostic/info purposes, in which case the weaknesses of the API don't really matter.  The way this rule is currently worded (using the API is forbidden), it doesn't seem to allow for such cases.

  2. From JavaThreads 04:

    The second advantage of thread groups relates to thread security. If you write custom security code for your application, decisions about whether one thread can access and/or modify the state of another thread take into account the thread group to which the threads belong. The Java Plug-in and appletviewer provide such customization so that threads in one applet are prevented from modifying the threads in another applet. To make security decisions in this way, however, requires that you write a custom security manager.

    I didn't want to go into details because I don't want to promote the use of thread groups for enforcing any form of security.

  3. Instead of saying at "Point 1" and at "Point 2" can we place p1, p2, etc. labels in the code so we know precisely what point in the program execution we are referring to?

    Thanks,
    rCs