HttpClient超时机制(安全问题处理:访问超大文件控制)

最新推荐文章于 2024-05-16 21:11:56 发布

iteye_7245

最新推荐文章于 2024-05-16 21:11:56 发布

阅读量442

点赞数

分类专栏： java 文章标签： JBoss Socket thread JDK

java 专栏收录该内容

66 篇文章 0 订阅

订阅专栏

背景

最近一直在做项目，其中的一个功能点，主要是访问外部网站并获取页面的字符串，具体的网站url完全是由用户输入，所以存在一定的安全隐患。

从测试来看，如果给定的一部电影的url地址，链接会一直不能被关闭，直到数据流被读完，如果来个几十次这样的请求，应用估计也差不多崩溃了

说明：项目中使用的HttpClient版本是3.0.1

测试

一般的HttpClient使用例子：

MultiThreadedHttpConnectionManager manager = new MultiThreadedHttpConnectionManager();
        HttpClient client = new HttpClient(manager);
        client.setConnectionTimeout(30000);
        client.setTimeout(30000);

        GetMethod get = new GetMethod("http://download.jboss.org/jbossas/7.0/jboss-7.0.0.Alpha1/jboss-7.0.0.Alpha1.zip");
        try {
            client.executeMethod(get);  //发起请求
            String result = get.getResponseBodyAsString(); //获取数据
        } catch (Exception e) {
        } finally {
            get.releaseConnection(); //释放链接
        }

这里我给出的一个url是近20MB的一个下载资源，很快发现线程要等个很久。咋办，得加个timeout超时机制。

"main" prio=10 tid=0x0899e800 nid=0x4010 runnable [0xb7618000..0xb761a1c8]
   java.lang.Thread.State: RUNNABLE
	at java.net.SocketInputStream.socketRead0(Native Method)
	at java.net.SocketInputStream.read(SocketInputStream.java:129)
	at java.io.BufferedInputStream.fill(BufferedInputStream.java:218)
	at java.io.BufferedInputStream.read1(BufferedInputStream.java:258)
	at java.io.BufferedInputStream.read(BufferedInputStream.java:317)
	- locked <0xb23a4c30> (a java.io.BufferedInputStream)
	at org.apache.commons.httpclient.ContentLengthInputStream.read(ContentLengthInputStream.java:156)
	at org.apache.commons.httpclient.ContentLengthInputStream.read(ContentLengthInputStream.java:170)
	at org.apache.commons.httpclient.ChunkedInputStream.exhaustInputStream(ChunkedInputStream.java:338)
	at org.apache.commons.httpclient.ContentLengthInputStream.close(ContentLengthInputStream.java:104)
	at java.io.FilterInputStream.close(FilterInputStream.java:155)
	at org.apache.commons.httpclient.AutoCloseInputStream.notifyWatcher(AutoCloseInputStream.java:179)
	at org.apache.commons.httpclient.AutoCloseInputStream.close(AutoCloseInputStream.java:143)
	at org.apache.commons.httpclient.HttpMethodBase.releaseConnection(HttpMethodBase.java:1341)

分析

目前httpClient3.1只支持3种timeout的设置：

connectionTimeout : socket建立链接的超时时间，Httpclient包中通过一个异步线程去创建socket链接，对应的超时控制。
timeoutInMilliseconds : socket read数据的超时时间， socket.setSoTimeout(timeout);
httpConnectionTimeout : 如果那个的是MultiThreadedHttpConnectionManager，对应的是从连接池获取链接的超时时间。

分析一下问题，我们需要的是一个HttpClient整个链接读取的一个超时时间，包括请求发起，Http Head解析，response流读取的一系列时间的总和。

目标很明确，对应的修正后的测试代码：

final MultiThreadedHttpConnectionManager manager = new MultiThreadedHttpConnectionManager();
        final HttpClient client = new HttpClient(manager);
        client.setConnectionTimeout(30000);
        client.setTimeout(30000);
        final GetMethod get = new GetMethod(
                                            "http://download.jboss.org/jbossas/7.0/jboss-7.0.0.Alpha1/jboss-7.0.0.Alpha1.zip");

        Thread t = new Thread(new Runnable() {

            @Override
            public void run() {
                try {
                    client.executeMethod(get);
                    String result = get.getResponseBodyAsString();
                } catch (Exception e) {
                    // ignore
                }
            }
        }, "Timeout guard");
        t.setDaemon(true);
        t.start();
        try {
            t.join(5000l);  //等待5s后结束
        } catch (InterruptedException e) {
            System.out.println("out finally start");
            ((MultiThreadedHttpConnectionManager) client.getHttpConnectionManager()).shutdown();
            System.out.println("out finally end");
        }
        if (t.isAlive()) {
            System.out.println("out finally start");
            ((MultiThreadedHttpConnectionManager) client.getHttpConnectionManager()).shutdown();
            System.out.println("out finally end");
            t.interrupt();
            // throw new TimeoutException();
        }
        System.out.println("done");

这里通过Thread.join方法，设置了超时时间为5000 ms，这是比较早的用法。如果熟悉cocurrent包的，可以直接使用Future和ThreadPoolExecutor进行异步处理，缓存对应的Thread。

ExecutorService service = Executors.newCachedThreadPool();
        Future future = service.submit(new Callable<String>() {

            @Override
            public String call() throws Exception {

                try {
                    client.executeMethod(get);
                    return get.getResponseBodyAsString();
                } catch (Exception e) {
                    e.printStackTrace();
                } finally {
                    System.out.println("future finally start");
                    ((MultiThreadedHttpConnectionManager) client.getHttpConnectionManager()).shutdown();
                    System.out.println("future finally end");
                }

                return "";
            }

        });

        try {
            future.get(5000, TimeUnit.MILLISECONDS);
        } catch (Exception e) {
            System.out.println("out finally");
            e.printStackTrace();
            ((MultiThreadedHttpConnectionManager) client.getHttpConnectionManager()).shutdown();
            System.out.println("out finally end");
        }

        service.shutdown();

说明：这里为什么释放链接未采用get.releaseConnection()

看下release的实现：

public void releaseConnection() {

        if (responseStream != null) {
            try {
                // FYI - this may indirectly invoke responseBodyConsumed.
                responseStream.close(); // 会先关闭流
            } catch (IOException e) {
                // the connection may not have been released, let's make sure
                ensureConnectionRelease();
            }
        } else {
            // Make sure the connection has been released. If the response 
            // stream has not been set, this is the only way to release the 
            // connection. 
            ensureConnectionRelease();
        }
    }

这里会先关闭responseStream流，这就是问题点。
对应的responseStream是在方法：readResponseBody(HttpConnection conn)。一般的html页面返回的是一个ContentLengthInputStream对象

ContentLengthInputStream在调用close方法时会用ChunkedInputStream.exhaustInputStream读完所有流数据

public void close() throws IOException {
        if (!closed) {
            try {
                ChunkedInputStream.exhaustInputStream(this);
            } finally {
                // close after above so that we don't throw an exception trying
                // to read after closed!
                closed = true;
            }
        }
    }

ChunkedInputStream.exhaustInputStream代码

static void exhaustInputStream(InputStream inStream) throws IOException {
        // read and discard the remainder of the message
        byte buffer[] = new byte[1024];
        while (inStream.read(buffer) >= 0) {
            ; 
        }
    }

说明：

因为非sleep和park的方法，不会响应InterruptedException事件，所以普通future超时发起的Thread.interrpt()并没有效果。
默认的SimpleHttpConnectionManager不支持这样的操作，所以选择MultiThreadedHttpConnectionManager.shutdown()方法，强制关闭底层HttpConnection的sock的输入输出流。

总结

理解一下HttpClient这样设计的理由： socket重用，keepAlive协议的支持等，保证上一次数据不会对新的请求有影响。
Thread.interrpt()处理，只会在Thread处于sleep或者wait状态才会被唤醒(api的描述)。而且该方法的调用并不自动产生InterruptedException异常，一般是需要自己判断Thread.isInterrupted()，然后throw异常。我们目前使用的一些jdk cocurrent类比如future.cancel也是类似处理。