多线程是Java工程师进阶所必须掌握的一项技能，也是面试中绕不过的一个环节，而死锁又是多线程同步失败的经典案例，对于复杂的系统，死锁是很难通过代码层面来做静态检测和排查的，所以有的面试官会从反向出发，让你手写一个死锁程序。
先来看一个网络上常见的死锁程序（可能存在问题）：
public class DeadLockTest {
private static Object lock1 = new Object();
private static Object lock2 = new Object();
public static void main(String[] args) {
new Thread(() -> {
synchronized (lock1) {
System.out.println("thread1 acquired lock1");
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println("thread1 try to acquire lock2");
synchronized (lock2) {
System.out.println("thread1 acquired lock2");
}
}
}, "t1").start();
new Thread(() -> {
synchronized (lock2) {
System.out.println("thread2 acquired lock2");
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println("thread2 try to acquire lock1");
synchronized (lock1) {
System.out.println("thread2 acquired lock1");
}
}
}, "t2").start();
// 检测死锁
checkDeadLock();
System.out.println("main thread end");
}
public static void checkDeadLock() {
ThreadMXBean mxBean = ManagementFactory.getThreadMXBean();
ScheduledExecutorService scheduled = Executors.newScheduledThreadPool(1);
// 初始等待5秒，每隔10秒检测一次
scheduled.scheduleAtFixedRate(()->{
long[] threadIds = mxBean.findDeadlockedThreads();
if (threadIds != null) {
System.out.println("检测到死锁线程：");
ThreadInfo[] threadInfos = mxBean.getThreadInfo(threadIds);
for (ThreadInfo info : threadInfos) {
System.out.println(info.getThreadId() + "：" + info.getThreadName());
}
}
}, 5L, 10L, TimeUnit.SECONDS);
}
}
上面这段程序在99.99%的情况下都会发生死锁，但是从理论的角度来讲，死锁并不是100%会发生的，比如：线程t1先启动并获取了锁lock1，在休眠的这1s的过程中，JVM并未发生线程调度（实际上基本不可能），t2未得到执行也未获取到锁lock2，这时候t1休眠结束继续执行并获取了锁lock2，那么这种情况下就不会发生死锁了。
如何写一个100%会发生死锁的程序呢？直接上代码：
public class DeadLockTest {
private static Object lock1 = new Object();
private static Object lock2 = new Object();
// 这里的flag需要用volatile修饰，以保证线程间的可见性
private static volatile boolean flag1 = false;
private static volatile boolean flag2 = false;
public static void main(String[] args) {
new Thread(() -> {
synchronized (lock1) {
flag1 = true;
System.out.println("thread1 acquired lock1");
while (!flag2) {
// 无限循环，等待thread2获取到lock2后再继续往下执行（相比使用Thread.sleep(1000)在理论上是100%会出现死锁）
Thread.yield();
}
System.out.println("thread1 try to acquire lock2");
synchronized (lock2) {
System.out.println("thread1 acquired lock2");
}
}
}, "t1").start();
new Thread(() -> {
synchronized (lock2) {
flag2 = true;
System.out.println("thread2 acquired lock2");
while (!flag1) {
Thread.yield();
}
System.out.println("thread2 try to acquire lock1");
synchronized (lock1) {
System.out.println("thread2 acquired lock1");
}
}
}, "t2").start();
// 检测死锁
checkDeadLock();
System.out.println("main thread end");
}
public static void checkDeadLock() {
ThreadMXBean mxBean = ManagementFactory.getThreadMXBean();
ScheduledExecutorService scheduled = Executors.newScheduledThreadPool(1);
// 初始等待5秒，每隔10秒检测一次
scheduled.scheduleAtFixedRate(() -> {
long[] threadIds = mxBean.findDeadlockedThreads();
if (threadIds != null) {
System.out.println("检测到死锁线程：");
ThreadInfo[] threadInfos = mxBean.getThreadInfo(threadIds);
for (ThreadInfo info : threadInfos) {
System.out.println(info.getThreadId() + "：" + info.getThreadName());
}
}
}, 5L, 10L, TimeUnit.SECONDS);
}
}