Java sun.misc.Unsafe

Java中的AtomicInteger 等Atomic类型是通过sun.misc.Unsafe 来实现的。 Unsafe类包含108个方法,这些方法中绝大部分都是native方法,是对底层硬件、内存等C语言才能进行的操作的封装。Atomic只是使用了其中的CAS功能。

Unsafe提供了许多实例化对象的方法,例如对于仅有私有构造方法的类,或者修改方法的私有变量。这些通过java的反射也可以实现。但是像Unsafe那样在字节级别精确修改Unsafe,反射恐怕做不到。

这里有一篇关于Unsafe的文章,出处: http://mishadoff.com/blog/java-magic-part-4-sun-dot-misc-dot-unsafe/

Java Magic. Part 4: sun.misc.Unsafe

Java is a safe programming language and prevents programmerfrom doing a lot of stupid mistakes, most of which based on memory management.But, there is a way to do such mistakes intentionally, using Unsafe class.

This article is a quick overview of sun.misc.Unsafe public API and fewinteresting cases of its usage.

Unsafe instantiation

Before usage, we need to create instance of Unsafe object.There is no simple way to do it like Unsafe unsafe = new Unsafe(),because Unsafe class has private constructor. It also has staticgetUnsafe() method, but if you naively try to call Unsafe.getUnsafe() you, probably,get SecurityException. Using this method available only from trusted code.

public static Unsafe getUnsafe() {
    Class cc = sun.reflect.Reflection.getCallerClass(2);
    if (cc.getClassLoader() != null)
        throw new SecurityException("Unsafe");
    return theUnsafe;
}

This is how java validates if code is trusted.It is just checking that our code was loaded withprimary classloader.

We can make our code "trusted". Use option bootclasspath when runningyour program and specifypath to system classes plus your one that will use Unsafe.

java -Xbootclasspath:/usr/jdk1.7.0/jre/lib/rt.jar:. com.mishadoff.magic.UnsafeClient

But it's too hard.

Unsafe class contains its instance called theUnsafe, which marked as private.We can steal that variable via java reflection.

Field f = Unsafe.class.getDeclaredField("theUnsafe");
f.setAccessible(true);
Unsafe unsafe = (Unsafe) f.get(null);

Note: Ignore your IDE. For example, eclipse show error "Access restriction..."but if you run code, all works just fine. If the error is annoying, ignore errors onUnsafe usage in:

Preferences -> Java -> Compiler -> Errors/Warnings ->
Deprecated and restricted API -> Forbidden reference -> Warning

Unsafe API

Class sun.misc.Unsafeconsists of 105 methods. There are, actually,few groups of important methods for manipulating with various entities.Here is some of them:

  • Info. Just returns some low-level memory information.

    • addressSize
    • pageSize
  • Objects. Provides methods for object and its fields manipulation.

    • allocateInstance
    • objectFieldOffset
  • Classes. Provides methods for classes and static fields manipulation.

    • staticFieldOffset
    • defineClass
    • defineAnonymousClass
    • ensureClassInitialized
  • Arrays. Arrays manipulation.

    • arrayBaseOffset
    • arrayIndexScale
  • Synchronization. Low level primitives for synchronization.

    • monitorEnter
    • tryMonitorEnter
    • monitorExit
    • compareAndSwapInt
    • putOrderedInt
  • Memory. Direct memory access methods.

    • allocateMemory
    • copyMemory
    • freeMemory
    • getAddress
    • getInt
    • putInt

Interesting use cases

Avoid initialization

allocateInstance method can be useful when you need to skip object initialization phaseor bypass security checks in constructor or you want instance of that classbut don't have any public constructor. Consider following class:

class A {
    private long a; // not initialized value

    public A() {
        this.a = 1; // initialization
    }

    public long a() { return this.a; }
}

Instantiating it using constructor, reflection and unsafe givesdifferent results.

A o1 = new A(); // constructor
o1.a(); // prints 1

A o2 = A.class.newInstance(); // reflection
o2.a(); // prints 1

A o3 = (A) unsafe.allocateInstance(A.class); // unsafe
o3.a(); // prints 0

Just think what happens to all your Singletons.

Memory corruption

This one is usual for every C programmer.By the way, its common technique for security bypass.

Consider some simple class that check access rules:

class Guard {
    private int ACCESS_ALLOWED = 1;

    public boolean giveAccess() {
        return 42 == ACCESS_ALLOWED;
    }
}

The client code is very secure and callsgiveAccess() to check access rules. Unfortunately, for clients,it always returns false. Only privileged users somehow can changevalue of ACCESS_ALLOWED constant and get access.

In fact, it's not true. Here is the code demostrates it:

Guard guard = new Guard();
guard.giveAccess();   // false, no access

// bypass
Unsafe unsafe = getUnsafe();
Field f = guard.getClass().getDeclaredField("ACCESS_ALLOWED");
unsafe.putInt(guard, unsafe.objectFieldOffset(f), 42); // memory corruption

guard.giveAccess(); // true, access granted

Now all clients will get unlimited access.

Actually, the same functionality can be achieved by reflection.But interesting, that we can modify any object, even onesthat we do not have references to.

For example, there is another Guard object in memorylocated next to current guard object. We can modify its ACCESS_ALLOWED field with the following code

unsafe.putInt(guard, 16 + unsafe.objectFieldOffset(f), 42); // memory corruption

Note, we didn't use any reference to this object.16 is size of Guard object in 32 bit architecture.We can calculate it manually or use sizeOf method, that defined... right now.

sizeOf

Using objectFieldOffset method we can implement C-style sizeof function.This implementation returns shallow size of object:

public static long sizeOf(Object o) {
    Unsafe u = getUnsafe();
    HashSet<Field> fields = new HashSet<Field>();
    Class c = o.getClass();
    while (c != Object.class) {
        for (Field f : c.getDeclaredFields()) {
            if ((f.getModifiers() & Modifier.STATIC) == 0) {
                fields.add(f);
            }
        }
        c = c.getSuperclass();
    }

    // get offset
    long maxSize = 0;
    for (Field f : fields) {
        long offset = u.objectFieldOffset(f);
        if (offset > maxSize) {
            maxSize = offset;
        }
    }

    return ((maxSize/8) + 1) * 8;   // padding
}

Algorithm is the following: go through all non-static fields including allsuperclases, get offset for each field, find maximum and add padding.Probably, I missed something, but idea is clear.

Much simpler sizeOf can be achieved if we just read size value fromthe class struct for this object, which located with offset 12 in JVM 1.7 32 bit.

public static long sizeOf(Object object){
    return getUnsafe().getAddress(
        normalize(getUnsafe().getInt(object, 4L)) + 12L);
}

normalize is a method for casting signed int to unsigned long, forcorrect address usage.

private static long normalize(int value) {
    if(value >= 0) return value;
    return (~0L >>> 32) & value;
}

Awesome, this method returns the same result as our previous sizeoffunction.

In fact, for good, safe and accurate sizeof function better to usejava.lang.instrument package,but it requires specifyng agent option in your JVM.

Shallow copy

Having implementation of calculating shallow object size, we can simplyadd function that copy objects. Standard solution need modify your code with Cloneable,or you can implement custom copy function in your object, but it won't be multipurpose function.

Shallow copy:

static Object shallowCopy(Object obj) {
    long size = sizeOf(obj);
    long start = toAddress(obj);
    long address = getUnsafe().allocateMemory(size);
    getUnsafe().copyMemory(start, address, size);
    return fromAddress(address);
}

toAddress and fromAddress convert object to its address in memory and vice versa.

static long toAddress(Object obj) {
    Object[] array = new Object[] {obj};
    long baseOffset = getUnsafe().arrayBaseOffset(Object[].class);
    return normalize(getUnsafe().getInt(array, baseOffset));
}

static Object fromAddress(long address) {
    Object[] array = new Object[] {null};
    long baseOffset = getUnsafe().arrayBaseOffset(Object[].class);
    getUnsafe().putLong(array, baseOffset, address);
    return array[0];
}

This copy function can be used to copy object of any type, its size will be calculateddynamically. Note that after copying you need to cast object to specific type.

Hide Password

One more interesting usage of direct memory access in Unsafe is removingunwanted objects from memory.

Most of the APIs for retrieving user's password, have signatureas byte[] or char[]. Why arrays?

It is completely for security reason, because we can nullify array elements after we don't need them.If we retrieve password as String it can be saved like an object in memory and nullifying thatstring just perform dereference operation. This object still in memory by the time GC decide to perform cleanup.

This trick creates fake String object with the same size and replace original one in memory:

String password = new String("l00k@myHor$e");
String fake = new String(password.replaceAll(".", "?"));
System.out.println(password); // l00k@myHor$e
System.out.println(fake); // ????????????

getUnsafe().copyMemory(
          fake, 0L, null, toAddress(password), sizeOf(password));

System.out.println(password); // ????????????
System.out.println(fake); // ????????????

Feel safe.

UPDATE: That way is not really safe. For real safety we need to nullifybacked char array via reflection:

Field stringValue = String.class.getDeclaredField("value");
stringValue.setAccessible(true);
char[] mem = (char[]) stringValue.get(password);
for (int i=0; i < mem.length; i++) {
  mem[i] = '?';
}

Thanks to Peter Verhas for pointing out that.

Multiple Inheritance

There is no multiple inheritance in java.

Correct, except we can cast every type to every another one, if we want.

long intClassAddress = normalize(getUnsafe().getInt(new Integer(0), 4L));
long strClassAddress = normalize(getUnsafe().getInt("", 4L));
getUnsafe().putAddress(intClassAddress + 36, strClassAddress);

This snippet adds String class to Integer superclasses, so we can castwithout runtime exception.

(String) (Object) (new Integer(666))

One problem that we must do it with pre-casting to object. To cheat compiler.

Dynamic classes

We can create classes in runtime, for example fromcompiled .class file. To perform that read class contentsto byte array and pass it properly to defineClass method.

byte[] classContents = getClassContent();
Class c = getUnsafe().defineClass(
              null, classContents, 0, classContents.length);
    c.getMethod("a").invoke(c.newInstance(), null); // 1

And reading from file defined as:

private static byte[] getClassContent() throws Exception {
    File f = new File("/home/mishadoff/tmp/A.class");
    FileInputStream input = new FileInputStream(f);
    byte[] content = new byte[(int)f.length()];
    input.read(content);
    input.close();
    return content;
}

This can be useful, when you must create classes dynamically, some proxiesor aspects for existing code.

Throw an Exception

Don't like checked exceptions? Not a problem.

getUnsafe().throwException(new IOException());

This method throws checked exception, but your code not forced to catch or rethrow it.Just like runtime exception.

Fast Serialization

This one is more practical.

Everyone knows that standard java Serializable capabilityto perform serialization is very slow. It also require classto have public non-argument constructor.

Externalizable is better, but it needs to define schema forclass to be serialized.

Popular high-performance libraries, like kryohave dependencies, which can be unacceptable with low-memory requirements.

But full serialization cycle can be easily achieved with unsafe class.

Serialization:

  • Build schema for object using reflection. It can be done once for class.
  • Use Unsafe methods getLong, getInt, getObject, etc. to retrieve actual field values.
  • Add class identifier to have capability restore this object.
  • Write them to the file or any output.

You can also add compression to save space.

Deserialization:

  • Create instance of serialized class. allocateInstance helps, because does not require any constructor.
  • Build schema. The same as 1 step in serialization.
  • Read all fields from file or any input.
  • Use Unsafe methods putLong, putInt, putObject, etc. to fill the object.

Actually, there are much more details in correct inplementation, but intuition is clear.

This serialization will be really fast.

By the way, there are some attempts in kryo to use Unsafe http://code.google.com/p/kryo/issues/detail?id=75

Big Arrays

As you know Integer.MAX_VALUE constant is a max size of java array.Using direct memory allocation we can create arrays with size limited by only heap size.

Here is SuperArray implementation:

class SuperArray {
    private final static int BYTE = 1;

    private long size;
    private long address;

    public SuperArray(long size) {
        this.size = size;
        address = getUnsafe().allocateMemory(size * BYTE);
    }

    public void set(long i, byte value) {
        getUnsafe().putByte(address + i * BYTE, value);
    }

    public int get(long idx) {
        return getUnsafe().getByte(address + idx * BYTE);
    }

    public long size() {
        return size;
    }
}

And sample usage:

long SUPER_SIZE = (long)Integer.MAX_VALUE * 2;
SuperArray array = new SuperArray(SUPER_SIZE);
System.out.println("Array size:" + array.size()); // 4294967294
for (int i = 0; i < 100; i++) {
    array.set((long)Integer.MAX_VALUE + i, (byte)3);
    sum += array.get((long)Integer.MAX_VALUE + i);
}
System.out.println("Sum of 100 elements:" + sum);  // 300

In fact, this technique uses off-heap memory and partiallyavailable in java.nio package.

Memory allocated this way not located in the heap and not under GC management, so take care of itusing Unsafe.freeMemory(). It also does not perform any boundary checks, so anyillegal access may cause JVM crash.

It can be useful for math computations, where code can operate with large arrays of data.Also, it can be interesting for realtime programmers, where GC delays on large arrays canbreak the limits.

Concurrency

And few words about concurrency with Unsafe.compareAndSwap methods are atomic and can be used to implementhigh-performance lock-free data structures.

For example, consider the problem to increment value in the shared objectusing lot of threads.

First we define simple interface Counter:

interface Counter {
    void increment();
    long getCounter();
}

Then we define worker thread CounterClient, that uses Counter:

class CounterClient implements Runnable {
    private Counter c;
    private int num;

    public CounterClient(Counter c, int num) {
        this.c = c;
        this.num = num;
    }

    @Override
    public void run() {
        for (int i = 0; i < num; i++) {
            c.increment();
        }
    }
}

And this is testing code:

int NUM_OF_THREADS = 1000;
int NUM_OF_INCREMENTS = 100000;
ExecutorService service = Executors.newFixedThreadPool(NUM_OF_THREADS);
Counter counter = ... // creating instance of specific counter
long before = System.currentTimeMillis();
for (int i = 0; i < NUM_OF_THREADS; i++) {
    service.submit(new CounterClient(counter, NUM_OF_INCREMENTS));
}
service.shutdown();
service.awaitTermination(1, TimeUnit.MINUTES);
long after = System.currentTimeMillis();
System.out.println("Counter result: " + c.getCounter());
System.out.println("Time passed in ms:" + (after - before));

First implementation is not-synchronized counter:

class StupidCounter implements Counter {
    private long counter = 0;

    @Override
    public void increment() {
        counter++;
    }

    @Override
    public long getCounter() {
        return counter;
    }
}

Output:

Counter result: 99542945
Time passed in ms: 679

Working fast, but no threads management at all, so result is inaccurate.Second attempt, add easiest java-way synchronization:

class SyncCounter implements Counter {
    private long counter = 0;

    @Override
    public synchronized void increment() {
        counter++;
    }

    @Override
    public long getCounter() {
        return counter;
    }
}

Output:

Counter result: 100000000
Time passed in ms: 10136

Radical synchronization always work. But timings is awful.Let's try ReentrantReadWriteLock:

class LockCounter implements Counter {
    private long counter = 0;
    private WriteLock lock = new ReentrantReadWriteLock().writeLock();

    @Override
    public void increment() {
        lock.lock();
        counter++;
        lock.unlock();
    }

    @Override
    public long getCounter() {
        return counter;
    }
}

Output:

Counter result: 100000000
Time passed in ms: 8065

Still correct, and timings are better. What about atomics?

class AtomicCounter implements Counter {
    AtomicLong counter = new AtomicLong(0);

    @Override
    public void increment() {
        counter.incrementAndGet();
    }

    @Override
    public long getCounter() {
        return counter.get();
    }
}

Output:

Counter result: 100000000
Time passed in ms: 6552

AtomicCounter is even better. Finally, try Unsafeprimitive compareAndSwapLong to see if it is really privilegy to use it.

class CASCounter implements Counter {
    private volatile long counter = 0;
    private Unsafe unsafe;
    private long offset;

    public CASCounter() throws Exception {
        unsafe = getUnsafe();
        offset = unsafe.objectFieldOffset(CASCounter.class.getDeclaredField("counter"));
    }

    @Override
    public void increment() {
        long before = counter;
        while (!unsafe.compareAndSwapLong(this, offset, before, before + 1)) {
            before = counter;
        }
    }

    @Override
    public long getCounter() {
        return counter;
    }

Output:

Counter result: 100000000
Time passed in ms: 6454

Hmm, seems equal to atomics. Maybe atomics use Unsafe? (YES)

In fact this example is easy enough, but it shows some power of Unsafe.

As I said, CAS primitive can be used to implement lock-free data structures.The intuition behind this is simple:

  • Have some state
  • Create a copy of it
  • Modify it
  • Perform CAS
  • Repeat if it fails

Actually, in real it is more hard than you can imagine. There are a lot of problems likeABA Problem, instructions reordering, etc.

If you really interested, you can refer to the awesome presentation about lock-free HashMap

UPDATE: Added volatile keyword to counter variable to avoid risk of infinite loop.

Kudos to Nitsan Wakart

Bonus

Documentation for park method from Unsafe class containslongest English sentence I've ever seen:

Block current thread, returning when a balancingunpark occurs, or a balancing unpark hasalready occurred, or the thread is interrupted, or, if notabsolute and time is not zero, the given time nanoseconds haveelapsed, or if absolute, the given deadline in millisecondssince Epoch has passed, or spuriously (i.e., returning for no"reason"). Note: This operation is in the Unsafe class onlybecause unpark is, so it would be strange to place itelsewhere.

Conclusion

Although, Unsafe has a bunch of useful applications, never use it.


评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值