Saturday, October 6, 2018

StackOverFlowError: Causes & Solutions


StackOverFlowError is one of the common confronted JVM error. In this blog post lets learn inner mechanics of thread stacks, reasons that can trigger StackOverFlowError and potential solutions to address this error.

To gain deeper understanding in to StackOverFlowError, let's review this simple program:

<<start:code>>
public class SimpleExample {

      public static void main(String args[]) {
           
            a();
      }    

      public static void a() {

            int x = 0;
            b();
      }

      public static void b() {

            Car y = new Car();           
            c();
      }

      public static void c() {

            float z = 0f;
            System.out.println("Hello");
      }
}
<<end:code>>

This program is very simple with following execution code:

1.       main() method is invoked first
2.       main() method invokes a() method. Inside a() method integer variable ‘x’ is initialized to value 0.
3.       a() method in turn invokes b() method. Inside b() method Car object is constructed and assigned to variable ‘y’.
4.       b() method in turn invokes c() method. Inside c() method float variable ‘z’ is initialized to value 0.

Now let’s review what happens behind the scenes when above simple program is executed. Each thread in the application has its own stack. Each stack has multiple stack frames. Thread adds the methods it’s executing, primitive data types, object pointers, return values to its stack frame in the sequence order in which they are executed.




Fig 1: Thread's Stack frame.

In step #1: main() method is pushed in to the application thread’s stack.

In step #2: a() method is pushed in to application thread’s stack. In a() method, primitive data type ‘int’ is defined with value 0 and assigned to variable x. This information is also pushed in to same stack frame. Note both data i.e. ‘0’ and variable ‘x’ is pushed in to thread’s stack frame.

In step #3: b() method is pushed in to thread’s stack. In b() method, ‘Car’ object is created and assigned to variable ‘y’. Crucial point to note here is ‘Car’ object is created in the heap and not in the thread’s stack. Only Car object’s reference i.e. y is stored in the thread’s stack frame.

In step #4: c() method is pushed in to thread’s stack. In c() method, primitive data type ‘float’ is defined with value 0f and assigned to variable z. This information is also pushed in to same stack frame. Note both data i.e. ‘0f’ and variable ‘z’ is pushed in to thread’s stack frame.
               
Once each method’s execution is completed, then method and the variables/object pointers which are stored in the stack frame are removed as show in Fig 2.




Fig 2: Thread's stack frame after executing methods.
               

What causes StackOverflowError?


As you can see thread’s stack is storing methods it’s executing, primitive datatypes, variables, object pointers and return values. All of these consume memory. If thread’s stack sizes grow beyond the allocated memory limit then StackOverflowError is thrown. Let's look at the below buggy program, which will result in StackOverflowError:

<<start:code>>

public class SOFDemo {

                public static void a() {
                                               
                                // Buggy line. It will cause method a() to be called infinite number of times.
                                a();                        
                }             
               
                public static void main(String args[]) {
                               
                                a();
                }
}
<<end:code>>

In this program main() method invokes a() method. a() method recursively calls itself. This implementation will cause a() method to be invoked infinite number of times. In this circumstance a() method will be added to thread's stack frame infinite number of times. Thus, after few thousand iterations thread’s stack size limit would be exceeded. Once stack size limit is exceeded it will result in
'StackOverflowError':

<<start:code>>
Exception in thread "main" java.lang.StackOverflowError
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
                at com.buggyapp.stackoverflow.SOFDemo.a(SOFDemo.java:7)
<<end:code>>
  


Fig 3: StackOverflowError progression

What are the solutions to StackOverflowError?


There are couple of strategies to address StackOverflowError.

1. Fix the code


Because of a non-terminating recursive call (as shown in above example), threads stack size can grow to a large size. In those circumstance, you must fix the source code which is causing recursive looping. When 'StackOverflowError' is thrown, it will print the stacktrace of the code that it was recursively executing. This code is a good pointer to start debugging and fixing the issue. In the above example it’s ‘a()’  method.

2. Increase Thread Stack Size (-Xss)


  There might be legitimate reason where a threads stack size needs to be increased. May be thread has to execute large number of methods or lot of local variables/created in the methods thread has been executing. In such circumstance you can increase the thread's stack size using the JVM argument: '-Xss'. This argument needs to be passed when you start the application. Example:

<<start:code>>
 -Xss2m
<<end:code>>

 This will set the thread's stack size to 2 mb.

It might bring question, what is the default thread’s stack size? Default thread stack size varies based on your operating system, java version & vendor.

JVM version
Thread stack size
  Sparc 32-bit JVM   
512k
  Sparc 64-bit JVM  
1024k
  x86 Solaris/Linux 32-bit JVM
320K
  x86 Solaris/Linux 64-bit JVM
1024K
  Windows 32-bit JVM
320K
  Windows 64-bit JVM
1024K

 Author bio:

              
Every single day, millions & millions of people in North America—bank, travel, and commerce—use the applications that RamLakshmanan has architected. Ram is an acclaimed speaker in major conferences on scalability, availability, and performance topics. Recently, he has founded a startup, which specializes in troubleshooting performance problems. 

No comments:

Post a Comment