Holding locks while performing time-consuming or blocking operations can severely degrade system performance and can result in starvation. Furthermore, deadlock can result if interdependent threads block indefinitely. Blocking operations include network, file, and console I/O (for example, Console.readLine()
) and object serialization. Deferring a thread indefinitely also constitutes a blocking operation. Consequently, programs must not perform blocking operations while holding a lock.
When the Java Virtual Machine (JVM) interacts with a file system that operates over an unreliable network, file I/O might incur a large performance penalty. In such cases, avoid file I/O over the network while holding a lock. File operations (such as logging) that could block while waiting for the output stream lock or for I/O to complete could be performed in a dedicated thread to speed up task processing. Logging requests can be added to a queue, assuming that the queue's put()
operation incurs little overhead as compared to file I/O [Goetz 2006].
Noncompliant Code Example (Deferring a Thread)
This noncompliant code example defines a utility method that accepts a time
argument:
public synchronized void doSomething(long time) throws InterruptedException { // ... Thread.sleep(time); }
Because the method is synchronized, when the thread is suspended, other threads cannot use the synchronized methods of the class. The current object's monitor continues to be held because the Thread.sleep()
method lacks synchronization semantics.
Compliant Solution (Intrinsic Lock)
This compliant solution defines the doSomething()
method with a timeout
parameter rather than the time
value. Using Object.wait()
instead of Thread.sleep()
allows setting a timeout period during which a notification may awaken the thread.
public synchronized void doSomething(long timeout) throws InterruptedException { // ... while (<condition does not hold>) { wait(timeout); // Immediately releases the current monitor } }
The current object's monitor is immediately released upon entering the wait state. When the timeout period elapses, the thread resumes execution after reacquiring the current object's monitor.
According to the Java API Class Object
documentation [API 2014]
Note that the
wait
method, as it places the current thread into the wait set for this object, unlocks only this object; any other objects on which the current thread may be synchronized remain locked while the thread waits.This method should only be called by a thread that is the owner of this object's monitor.
Programs must ensure that threads that hold locks on other objects release those locks appropriately before entering the wait state. Additional guidance on waiting and notification is available in THI03-J. Always invoke wait() and await() methods inside a loop and THI02-J. Notify all waiting threads rather than a single thread.
Noncompliant Code Example (Network I/O)
This noncompliant code example defines a sendPage()
method that sends a Page
object from a server to a client. The method is synchronized to protect the pageBuff
array when multiple threads request concurrent access.
// Class Page is defined separately. // It stores and returns the Page name via getName() Page[] pageBuff = new Page[MAX_PAGE_SIZE]; public synchronized boolean sendPage(Socket socket, String pageName) throws IOException { // Get the output stream to write the Page to ObjectOutputStream out = new ObjectOutputStream(socket.getOutputStream()); // Find the Page requested by the client // (this operation requires synchronization) Page targetPage = null; for (Page p : pageBuff) { if (p.getName().compareTo(pageName) == 0) { targetPage = p; } } // Requested Page does not exist if (targetPage == null) { return false; } // Send the Page to the client // (does not require any synchronization) out.writeObject(targetPage); out.flush(); out.close(); return true; }
Calling writeObject()
within the synchronized sendPage()
method can result in delays and deadlock-like conditions in high-latency networks or when network connections are inherently lossy.
Compliant Solution
This compliant solution separates the process into a sequence of steps:
- Perform actions on data structures requiring synchronization.
- Create copies of the objects to be sent.
- Perform network calls in a separate unsynchronized method.
In this compliant solution, the unsynchronized sendPage()
method calls the synchronized getPage()
method to retrieve the requested Page
in the pageBuff
array. After the Page
is retrieved, sendPage()
calls the unsynchronized deliverPage()
method to deliver the Page
to the client.
// No synchronization public boolean sendPage(Socket socket, String pageName) { Page targetPage = getPage(pageName); if (targetPage == null){ return false; } return deliverPage(socket, targetPage); } // Requires synchronization private synchronized Page getPage(String pageName) { Page targetPage = null; for (Page p : pageBuff) { if (p.getName().equals(pageName)) { targetPage = p; } } return targetPage; } // Return false if an error occurs, true if successful public boolean deliverPage(Socket socket, Page page) { ObjectOutputStream out = null; boolean result = true; try { // Get the output stream to write the Page to out = new ObjectOutputStream(socket.getOutputStream()); // Send the page to the client out.writeObject(page);out.flush(); } catch (IOException io) { result = false; } finally { if (out != null) { try { out.close(); } catch (IOException e) { result = false; } } } return result; }
Exceptions
LCK09-J-EX0: Classes that provide an appropriate termination mechanism to callers are permitted to violate this rule (see THI04-J. Ensure that threads performing blocking operations can be terminated).
LCK09-J-EX1: Methods that require multiple locks may hold several locks while waiting for the remaining locks to become available. This constitutes a valid exception, although the programmer must follow other applicable rules, especially LCK07-J. Avoid deadlock by requesting and releasing locks in the same order to avoid deadlock .
Risk Assessment
Blocking or lengthy operations performed within synchronized regions could result in a deadlocked or unresponsive system.
Rule | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
LCK09-J | Low | Probable | High | P2 | L3 |
Automated Detection
Some static analysis tools are capable of detecting violations of this rule.
Tool | Version | Checker | Description |
---|---|---|---|
CodeSonar | 8.1p0 | JAVA.CONCURRENCY.STARVE.BLOCKING | Blocking in Critical Section (Java) |
Parasoft Jtest | 2024.1 | CERT.LCK09.TSHL CERT.LCK09.TSHL2 | Do not use blocking methods while holding a lock Do not call 'Thread.sleep()' while holding a lock since doing so can cause poor performance and deadlocks |
PVS-Studio | 7.33 | V6095 | |
ThreadSafe | 1.3 | CCE_LK_LOCKED_BLOCKING_CALLS | Implemented |
SonarQube | 9.9 | S2276 | Implemented |
4 Comments
Dhruv Mohindra
I wrote in CON08-J. Do not call alien methods that synchronize on the same objects as any callers in the execution chain "It is intuitive that a lock will be held for as long as the synchronized block is executing. This means that any time-intensive operations are unsuitable within this block.". Given that the situation described there is a tad more complex, I think this is a much needed guideline. You might also want to cross reference that recommendation in addition to following David's advice.
David Svoboda
Dhruv Mohindra
The guideline has
wait()
in the CS. We could of course suggestCondition.await(long time, Timeunit unit)
or evenExecutorService.invokeAll(Arrays.asList(new Task()), long timeout, TimeUnit unit);
(the Task can then sleep in itsrun()
if it implementsRunnable
) from the j.u.c utilities.James Ahlborn
While this rule is certainly valid, i think it fails to distinguish between the problem situation "holding an unrelated lock while blocking" and a valid situation "holding a related lock while blocking". for instance, one could implement a cache which lazy loads resources which are expensive to create (i.e. they involve i/o or other blocking operations). each resource could have it's own lock such that when the resource is actually being loaded, the resource's lock would be held. simple example:
in this example, i would assert that it is perfectly valid to hold the Wrapper instance lock while a potentially blocking operation is executing.