synchronized 用法
//获得的锁是 当前对象
public synchronized void testSynchronized(){}
//获得的锁是 当前对象的Class对象
public synchronized static void testStaticSynchronized(){}
public final Object lock = new Object();
public void test(){
//获得的锁是 lock对象
synchronized (lock){}
}
synchronized定义在方法体上生成的字节码会多 ACC_SYNCHRONIZED
定义在代码块上字节码会生成 monitorenter,和两个monitorexit，其中第二个monitorexit是防止程序异常可以看到多了athrow字节码 。通过javap -v 命令查看字节码
javap-synchronized.png
不得不说的Java对象头分析(64位机器)
首先一个Java对象大小由三部分组成：对象头，对象体(一般认为对象中的属性大小)，对齐字节。
对象头包括MarkWord（8字节）、Klass Pointer(指向Class对象的指针)
对象如果是数组，对象头额外还会包含4字节的数组长度
下面通过代码演示。代码分析工具使用jol
1.对象中不添加任何属性
public class T{ }
public static void main(String[] args) {
T t = new T();
ClassLayout classLayout = ClassLayout.parseInstance(t); //'org.openjdk.jol', name: 'jol-core', version: '0.9'
System.out.println(classLayout.toPrintable());
}
打印结果
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
可以看到，整个T对象占用16bytes大小，其中对象头（object header）占用12字节
loss due to the next object alignment: 对齐补充，虚拟机要求对象所占用的内存大小必须是8的整数，当不足的时候，对齐补充
2.对象中添加一个属性
public static class T{
long testLong;
}
T类型里面添加一个long属性后，展示效果。同时在TYPE字段也标识了T中的变量属性为long
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (alignment/padding gap)
16 8 long T.testLong 0
Instance size: 24 bytes
Space losses: 4 bytes internal + 0 bytes external = 4 bytes total
可以看到，原本object header占用12字节，long占用8字节。需要对齐补充到8的倍数，所以占用24字节的空间。可以试试int, byte, short 类型自测，也可以证明int占用4字节，short 2字节 ,byte 1字节
3.添加-XX:-UseCompressedOops参数
UseCompressedOops参数用于指针压缩，64位机器上指针占用的大小应该是8字节，但是由于上两次实验得到Klass Pointer 占用4字节。在运行参数中关闭UseCompressedOops，对象头已经更改为占用 16Byte大小。JVM默认开启此参数
T类型里面添加一个byte属性后，展示效果。同时在TYPE字段也标识了T中的变量属性为byte
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 10 35 92 1c (00010000 00110101 10010010 00011100) (479343888)
12 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
16 1 byte T.testLong 0
17 7 (loss due to the next object alignment)
Instance size: 24 bytes
Space losses: 0 bytes internal + 7 bytes external = 7 bytes total
4.对象数组
int[] t = new int[1];
ClassLayout classLayout = ClassLayout.parseInstance(t); //'org.openjdk.jol', name: 'jol-core', version: '0.9'
System.out.println(classLayout.toPrintable());
[I object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 68 0b 98 1b (01101000 00001011 10011000 00011011) (462949224)
12 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
16 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
20 4 (alignment/padding gap)
24 4 int [I.<elements> N/A
28 4 (loss due to the next object alignment)
Instance size: 32 bytes
Space losses: 4 bytes internal + 4 bytes external = 8 bytes total
对象头里多了4字节，表示数组的长度。这也是为什么获取数组长度复杂度是O(1)的原因,是直接从对象头中取出数组的长度
JOL中的对象头如何阅读
0 4 (object header) 05 e8 06 01 (00000101 11101000 00000110 00000001) (17229829)
8 7 6 5
4 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (0)
4 3 2 1
读取方式应该是 1-2-3-4-5-6-7-8
00000000 00000000 00000000 00000001 00000001 00000110 11101000 00000101 从尾往前读
理解了对象的组成结构，接下来着重说明对象头中的MarkWord。在JVM中需要大量存储对象，存储时为了实现一些额外的功能，需要在对象中添加一些标记字段用于增强对象功能，这些标记字段组成了对象头
64classheader.png
着重查看最后3bit的内容,锁的变换对应了对象头的不同标志
演示代码
1.偏向锁
Thread.sleep(5000); //等待jvm开启偏向锁
T t = new T();
System.out.println(ClassLayout.parseInstance(t).toPrintable());
synchronized (t){
System.out.println(ClassLayout.parseInstance(t).toPrintable());
}
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 05 00 00 00 (00000101 00000000 00000000 00000000) (5)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
com.mouxf.jol.JolTestObjectDetail$T object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 05 e8 06 01 (00000101 11101000 00000110 00000001) (17229829)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
可以看到最后三位是 101 代表着偏向锁，至于为什么两次打印都是101而不是一次 001和101？why？
2.轻量级锁
public static void test3() throws Exception{
Thread.sleep(5000);
T t = new T();
print(t);
for (int i = 0; i < 1; i++) {
Thread thread = new Thread(() -> {print(t);});
thread.start();
thread.join();
}
}
public static void print(T t) {
synchronized (t){
System.out.println(Thread.currentThread().getName());
System.out.println(ClassLayout.parseInstance(t).toPrintable());
}
}
main
com.mouxf.jol.JolTestObjectDetail$T object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 05 e8 ff 02 (00000101 11101000 11111111 00000010) (50325509)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
Thread-0
com.mouxf.jol.JolTestObjectDetail$T object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) d8 f1 1c 1f (11011000 11110001 00011100 00011111) (521990616)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
可以发现t中最后3位由 101 -> 000 转变为轻量级锁
3.重量级锁
public static void test4() throws Exception{
T t = new T();
print(t);
for (int i = 0; i < 10; i++) {
Thread thread = new Thread(() -> {print(t);});
thread.start();
}
}
由于打印内容较多，只拿出一部分
com.mouxf.jol.JolTestObjectDetail$T object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) fa c8 46 03 (11111010 11001000 01000110 00000011) (54970618)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 43 c1 00 f8 (01000011 11000001 00000000 11111000) (-134168253)
12 4 (loss due to the next object alignment)
Instance size: 16 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
可以发现t中最后3位由 010 转变为重量级
以上就是演示synchronized的锁升级过程，其实可以看出synchronized锁升级过程是 无锁->偏向锁->轻量级锁->重量级锁
synchronized锁升级过程
-
偏向锁
- 检查对象头标志 01 与 是否偏向 0 （对象头最后 3bit）
- CAS修改MarkWord获取偏向锁，修改成功后将markword中线程ID指向当前线程A
- 当另一个线程B获取对象锁的时候，程序达到安全点后，查看对象头中线程情况。线程A不存活或者线程A不再使用该对象，线程竞争设置偏向锁
- 如果线程A继续持有对象锁，暂停当前线程，撤销偏向锁，升级为轻量级锁
-
轻量级锁
- 当前线程开辟一块记录空间(Lock Recored)，同时复制Markword
- CAS修改MarkWord，修改成功升级为轻量级锁 00
- 线程B CAS修改失败自旋，达到自旋次数后还未获得锁则升级为重量级锁
盗个图，总结得很好(侵删)=
synchronized_lock.png
锁的实现原理
每个对象都存在着一个monitor与之关联，当一个monitor被某个线程持有后，它便处于锁定状态。
monitor是由ObjectMonitor实现的，位于HotSpot虚拟机源码ObjectMonitor.hpp文件，C++实现的。Java 程序调用wait(),notfiy(),notfiyAll()方法底层实现也是ObjectMonitor
ObjectMonitor() {
_header = NULL;
_count = 0; //线程获取锁的次数
_waiters = 0,
_recursions = 0; //锁的重入次数
_object = NULL;
_owner = NULL; //_owner指向持有ObjectMonitor对象的线程
_WaitSet = NULL; //处于wait状态的线程，会被加入到_WaitSet
_WaitSetLock = 0 ; //保护等待队列
_Responsible = NULL ;
_succ = NULL ;
_cxq = NULL ; //阻塞在entry上最近可达的线程列表
FreeNext = NULL ;
_EntryList = NULL ; //处于等待锁block状态的线程，会被加入到该列表
_SpinFreq = 0 ;
_SpinClock = 0 ;
OwnerIsThread = 0 ;
}
相关参数
-XX:BiasedLockingStartupDelay=0 偏向锁在Java6 Java7默认是开启的，但是它在程序启动几秒之后在激活，关闭延迟
-XX:-UseBiasedLocking=false 关闭偏向锁
网友评论