当前位置: 首页>数据库>正文

Android性能优化系列-腾讯matrix-IO监控-IOCanaryPlugin源码分析

作者:秋去无痕

matrix 对io的监控包括四个方面

  1. 监控在主线程执行 IO 操作的问题
  2. 监控缓冲区过小的问题
  3. 监控重复读同一文件
  4. 监控内存泄漏问题

IOCanaryPlugin,内部由IOCanaryCore完成真正的操作。

start方法

根据配置进行hook的安装

//io流hook
if (ioConfig.isDetectFileIOInMainThread() || ioConfig.isDetectFileIOBufferTooSmall() || ioConfig.isDetectFileIORepeatReadSameFile()) {
    IOCanaryJniBridge.install(ioConfig, this);
}
//内存泄漏hook
if (ioConfig.isDetectIOClosableLeak()) {
    this.mCloseGuardHooker = new CloseGuardHooker(this);
    this.mCloseGuardHooker.hook();
}

stop方法

取消hook

if (this.mCloseGuardHooker != null) {
    this.mCloseGuardHooker.unHook();
}

IOCanaryJniBridge.uninstall();

IOCanaryJniBridge.install()

底层hook安装包函几个步骤,加载so,设置hook内容,分别对应了下面几个方法

loadJni

System.loadLibrary("io-canary")

执行了System.loadLibrary("io-canary"),此时会进入io_canary_jni.cc中的JNI_OnLoad方法,在这个方法中有两项关键操作,1.获取到java层的一些信息,2.设置一个回调接口,用于上传监控信息.

InitJniEnv()

static bool InitJniEnv(JavaVM *vm) {
    ....
    jclass temp_cls = env->FindClass("com/tencent/matrix/iocanary/core/IOCanaryJniBridge");
    ....
}

SetIssuedCallback()

iocanary::IOCanary::Get().SetIssuedCallback(OnIssuePublish)

其中OnIssuePublish是在拿到信息之后将信息组装成java层的对象IOIssue,然后放入List中,通过调用java层IOCanaryJniBridge类的onIssuePublish实现信息的抛出。

com/tencent/matrix/iocanary/core/IOIssue
com/tencent/matrix/iocanary/core/IOCanaryJniBridge

enableDetector

通过传入定义好的type类型到底层,实现此类型的io监控,代码如下

iocanary::IOCanary::Get().RegisterDetector(static_cast<DetectorType>(detector_type));

可以看到最终是往detectors_这个vector集合中存入了对应的Detector,每个Detector都FileIODetector的子类。

  • FileIOMainThreadDetector
  • FileIORepeatReadDetector
  • FileIOSmallBufferDetector
void IOCanary::RegisterDetector(DetectorType type) {
    switch (type) {
        case DetectorType::kDetectorMainThreadIO:
            detectors_.push_back(new FileIOMainThreadDetector());
            break;
        case DetectorType::kDetectorSmallBuffer:
            detectors_.push_back(new FileIOSmallBufferDetector());
            break;
        case DetectorType::kDetectorRepeatRead:
            detectors_.push_back(new FileIORepeatReadDetector());
            break;
        default:
            break;
    }
}

setConfig

给对应的io监控设置监控阈值,存入configs_数组,配置和对应的默认值如下,超过阈值则触发监控

  • kMainThreadThreshold = 500 毫秒
  • kSmallBufferThreshold = 4096 kb
  • kRepeatReadThreshold = 20 次

iocanary::IOCanary::Get().SetConfig(static_cast(key), val);

void IOCanaryEnv::SetConfig(IOCanaryConfigKey key, long val) {
    if (key >= IOCanaryConfigKey::kConfigKeysLen) {
        return;
    }

    configs_[key] = val;
}

dohook

dohook是核心方法,前边配置信息准备好后,这里开始进行对应方法的hook。被hook的so文件为

const static char* TARGET_MODULES[] = {
    "libopenjdkjvm.so",
    "libjavacore.so",
    "libopenjdk.so"
};

关于GOT hook,可以查看爱奇艺的开源框架XHook,这里不再描述细节。

被hook的方法如下 open、open64、close、android_fdsan_close_with_tag, 如果so是libjavacore.so,会尝试hook它内部的这几个方法 read、__read_chk、write、__write_chk

open

当一个文件被打开时,回调到设置好的方法ProxyOpen中,在这里会检测是否是主线程操作,如不是则不做处理,如是主线程,则执行DoProxyOpenLogic逻辑。

int ProxyOpen(const char *pathname, int flags, mode_t mode) {
    if(!IsMainThread()) {
        return original_open(pathname, flags, mode);
    }
    int ret = original_open(pathname, flags, mode);
    if (ret != -1) {
        DoProxyOpenLogic(pathname, flags, mode, ret);
    }
    return ret;
}

在DoProxyOpenLogic方法中会获取到当前堆栈信息

static void DoProxyOpenLogic(const char *pathname, int flags, mode_t mode, int ret) {
    ....
    //kJavaBridgeClass = com/tencent/matrix/iocanary/core/IOCanaryJniBridge
    //kMethodIDGetJavaContext = getJavaContext() 得到的是一个JavaContext,
    //是一个内部类,这个类上有一个变量stack,在java Context 创建的时候,
    //就会获取到堆栈信息,保存在stack变量上
    jobject java_context_obj = env->CallStaticObjectMethod(kJavaBridgeClass, kMethodIDGetJavaContext);
    if (NULL == java_context_obj) {
        return;
    }
    //堆栈信息
    jstring j_stack = (jstring) env->GetObjectField(java_context_obj, kFieldIDStack);
    jstring j_thread_name = (jstring) env->GetObjectField(java_context_obj, kFieldIDThreadName);
    //当前线程名
    char* thread_name = jstringToChars(env, j_thread_name);
    char* stack = jstringToChars(env, j_stack);
    JavaContext java_context(GetCurrentThreadId(), thread_name == NULL "" : thread_name, stack == NULL "" : stack);
    ....
    //pathname是被打开的文件名,java_context中包含了堆栈和线程名
    //flags和mode都是系统open方法调用传过来的值,ret是open执行的结果
    //这里进入了IOCanary OnOpen方法
    iocanary::IOCanary::Get().OnOpen(pathname, flags, mode, ret, java_context);
    ....
}

open64

同open()

close

检测是主线程,进入IOCanary OnClose方法

int ProxyClose(int fd) {
    if(!IsMainThread()) {
        return original_close(fd);
    }
    int ret = original_close(fd);
    iocanary::IOCanary::Get().OnClose(fd, ret);
    return ret;
}

android_fdsan_close_with_tag

同close()

read

主要是获取到read消耗的时长,然后携带信息进入IOCanary OnRead

ssize_t ProxyRead(int fd, void *buf, size_t size) {
    if(!IsMainThread()) {
        return original_read(fd, buf, size);
    }
    //获取到当前时间
    int64_t start = GetTickCountMicros();
    //执行原read方法
    size_t ret = original_read(fd, buf, size);
    //记录read时间间隔
    long read_cost_us = GetTickCountMicros() - start;
    //将信息传入IOCanary OnRead
    iocanary::IOCanary::Get().OnRead(fd, buf, size, ret, read_cost_us);
    return ret;
}

__read_chk

同read

write

ssize_t ProxyWrite(int fd, const void *buf, size_t size) {
    if(!IsMainThread()) {
        return original_write(fd, buf, size);
    }
    //获取到当前时间
    int64_t start = GetTickCountMicros();
    //执行write
    size_t ret = original_write(fd, buf, size);
    //记录时间间隔
    long write_cost_us = GetTickCountMicros() - start;
    //将信息传入IOCanary OnRead
    iocanary::IOCanary::Get().OnWrite(fd, buf, size, ret, write_cost_us);
    return ret;
}

__write_chk

同write

IOCanary

从上边open close read write方法的流向可知,最终都还是汇集到了IOCanary这个C++类中,进入对应的方法可知,IOCanary内部又调用了IOInfoCollector这个类。 。

OnOpen

void IOCanary::OnOpen(const char *pathname, int flags, mode_t mode,
                      int open_ret, const JavaContext& java_context) {
    collector_.OnOpen(pathname, flags, mode, open_ret, java_context);
}

方法内部逻辑也很清晰,直接将文件名和相关信息组装成info,然后以文件描述符为key,info为value存入了c++的info_map_(一个std::unordered_map)中,信息存起来肯定是要用的,我们后边会看到。文件打开之后,下一步就是或读或写,继续去看read方法。

void IOInfoCollector::OnOpen(const char *pathname, int flags, mode_t mode
        , int open_ret, const JavaContext& java_context) {
    if (open_ret == -1) {
        return;
    }
    //open_ret参数指的是open方法调用后的结果,也就是当前被打开的文件的文件描述符,
    //如果已存在,则返回
    if (info_map_.find(open_ret) != info_map_.end()) {
        return;
    }
    std::shared_ptr<IOInfo> info = std::make_shared<IOInfo>(pathname, java_context);
    info_map_.insert(std::make_pair(open_ret, info));
}

OnRead

void IOCanary::OnRead(int fd, const void *buf, size_t size,
                      ssize_t read_ret, long read_cost) {
    collector_.OnRead(fd, buf, size, read_ret, read_cost);
}

看起来关键内容在CountRWInfo中,从方法名上可以看出,读和写都与此方法有关,所以我们先不看CountRWInfo方法内容,看完write后再去深入CountRWInfo方法。

void IOInfoCollector::OnRead(int fd, const void *buf, size_t size,
                             ssize_t read_ret, long read_cost) {
    if (read_ret == -1 || read_cost < 0) {
        return;
    }
    if (info_map_.find(fd) == info_map_.end()) {
        return;
    }
    CountRWInfo(fd, FileOpType::kRead, size, read_cost);
}

OnWrite

void IOCanary::OnWrite(int fd, const void *buf, size_t size,
                       ssize_t write_ret, long write_cost) {
    collector_.OnWrite(fd, buf, size, write_ret, write_cost);
}

和read一样,进入了CountRWInfo方法

void IOInfoCollector::OnWrite(int fd, const void *buf, size_t size,
                              ssize_t write_ret, long write_cost) {
    if (write_ret == -1 || write_cost < 0) {
        return;
    }
    if (info_map_.find(fd) == info_map_.end()) {
        return;
    }
    CountRWInfo(fd, FileOpType::kWrite, size, write_cost);
}

CountRWInfo

CountRWInfo将每个文件对应的信息封装到IOInfo这个类中,封装的信息包函:

  • 读(写)次数
  • 文件大小
  • 读(写)消耗的时长
  • 单次读(写)最大时长
  • 读(写)间隔小于8000微妙的总时长
  • 缓存区大小
  • 读写类型,读还是写

在一个文件被读写过程中,这个方法会不断的被调用,并更新对应的信息,读写完成之后,得到最终的信息,执行close方法。

void IOInfoCollector::CountRWInfo(int fd, const FileOpType &fileOpType, long op_size, long rw_cost) {
    if (info_map_.find(fd) == info_map_.end()) {
        return;
    }

    const int64_t now = GetSysTimeMicros();
    //读写次数
    info_map_[fd]->op_cnt_ ++;
    //文件大小
    info_map_[fd]->op_size_ += op_size;
    //读写消耗的时长
    info_map_[fd]->rw_cost_us_ += rw_cost;
    //单次读写最大时长
    if (rw_cost > info_map_[fd]->max_once_rw_cost_time_μs_) {
        info_map_[fd]->max_once_rw_cost_time_μs_ = rw_cost;
    }
    //读写间隔小于8000微妙的总时长
    if (info_map_[fd]->last_rw_time_μs_ > 0 && (now - info_map_[fd]->last_rw_time_μs_) < kContinualThreshold) {
        info_map_[fd]->current_continual_rw_time_μs_ += rw_cost;

    } else {
        info_map_[fd]->current_continual_rw_time_μs_ = rw_cost;
    }
    if (info_map_[fd]->current_continual_rw_time_μs_ > info_map_[fd]->max_continual_rw_cost_time_μs_) {
        info_map_[fd]->max_continual_rw_cost_time_μs_ = info_map_[fd]->current_continual_rw_time_μs_;
    }
    info_map_[fd]->last_rw_time_μs_ = now;
    //缓存区大小
    if (info_map_[fd]->buffer_size_ < op_size) {
        info_map_[fd]->buffer_size_ = op_size;
    }
    //读写类型,读还是写
    if (info_map_[fd]->op_type_ == FileOpType::kInit) {
        info_map_[fd]->op_type_ = fileOpType;
    }
}

OnClose

void IOCanary::OnClose(int fd, int close_ret) {
    std::shared_ptr<IOInfo> info = collector_.OnClose(fd, close_ret);
    if (info == nullptr) {
        return;
    }

    OfferFileIOInfo(info);
}

close时记录总时长,文件大小,然后返回,返回后进入OfferFileIOInfo方法

std::shared_ptr<IOInfo> IOInfoCollector::OnClose(int fd, int close_ret) {
    if (info_map_.find(fd) == info_map_.end()) {
        return nullptr;
    }
    //从打开到关闭的总时长
    info_map_[fd]->total_cost_μs_ = GetSysTimeMicros() - info_map_[fd]->start_time_μs_;
    //获取到文件大小
    info_map_[fd]->file_size_ = GetFileSize(info_map_[fd]->path_.c_str());
    std::shared_ptr<IOInfo> info = info_map_[fd];
    //从map中移除
    info_map_.erase(fd);
    //返回信息
    return info;
}

OfferFileIOInfo将info放入队列,并调用notify_one方法通知消费者消费,这里用到了生产消费模式,生产者将生产果实放在队列中,消费者从队列取出进行消费,我们找下消费者在哪。

void IOCanary::OfferFileIOInfo(std::shared_ptr<IOInfo> file_io_info) {
    std::unique_lock<std::mutex> lock(queue_mutex_);
    queue_.push_back(file_io_info);
    queue_cv_.notify_one();
    lock.unlock();
}

可以看到,IOCanary在创建的时候,启动了一个线程

IOCanary::IOCanary() {
    exit_ = false;
    std::thread detect_thread(&IOCanary::Detect, this);
    detect_thread.detach();
}

线程中有一个无限循环,它负责不停的从队列中拿info,如果队列为空则挂起线程等待。

前边我们看到了拿到一条info之后,将info放入到队列中,然后通知消费者消费,此时消费者线程会从TakeFileIOInfo方法中被唤醒,并拿到一条info,交给各个detector去检测。

检测完成之后,满足条件的信息会被放入published_issues中,然后issued_callback_将信息回调出去。前边提到有三个detector接下来具体看下他们的内部逻辑。

void IOCanary::Detect() {
    std::vector<Issue> published_issues;
    std::shared_ptr<IOInfo> file_io_info;
    while (true) {
        published_issues.clear();
        int ret = TakeFileIOInfo(file_io_info);
        if (ret != 0) {
            break;
        }
        for (auto detector : detectors_) {
            detector->Detect(env_, *file_io_info, published_issues);
        }
        if (issued_callback_ && !published_issues.empty()) {
            issued_callback_(published_issues);
        }
        file_io_info = nullptr;
    }
}

FileIOMainThreadDetector

检测主线程io

void FileIOMainThreadDetector::Detect(const IOCanaryEnv &env, const IOInfo &file_io_info,
                                      std::vector<Issue>& issues) {
    //必须是主线程才会执行
    if (GetMainThreadId() == file_io_info.java_context_.thread_id_) {
        int type = 0;
        //单次io时长超过13毫秒,要记录
        //constexpr static const int kPossibleNegativeThreshold = 13*1000;
        if (file_io_info.max_once_rw_cost_time_μs_ > IOCanaryEnv::kPossibleNegativeThreshold) {
            type = 1;
        }
        //最大连续读写时长超过env.GetMainThreadThreshold()=500
        if(file_io_info.max_continual_rw_cost_time_μs_ > env.GetMainThreadThreshold()) {
            type |= 2;
        }
        if (type != 0) {
            Issue issue(kType, file_io_info);
            issue.repeat_read_cnt_ = type; 
            //存入
            PublishIssue(issue, issues);
        }
    }
}

FileIORepeatReadDetector

监听重复读取同一文件


void FileIORepeatReadDetector::Detect(const IOCanaryEnv &env,
                                      const IOInfo &file_io_info,
                                      std::vector<Issue>& issues) {
    const std::string& path = file_io_info.path_;
    if (observing_map_.find(path) == observing_map_.end()) {
        if (file_io_info.max_continual_rw_cost_time_μs_ < env.kPossibleNegativeThreshold) {
            return;
        }

        observing_map_.insert(std::make_pair(path, std::vector<RepeatReadInfo>()));
    }
    std::vector<RepeatReadInfo>& repeat_infos = observing_map_[path];
    if (file_io_info.op_type_ == FileOpType::kWrite) {
        repeat_infos.clear();
        return;
    }

    RepeatReadInfo repeat_read_info(file_io_info.path_, file_io_info.java_context_.stack_, file_io_info.java_context_.thread_id_,
                                  file_io_info.op_size_, file_io_info.file_size_);
    if (repeat_infos.size() == 0) {
        repeat_infos.push_back(repeat_read_info);
        return;
    }
    if((GetTickCount() - repeat_infos[repeat_infos.size() - 1].op_timems) > 17) {   //17ms todo astrozhou add to params
        repeat_infos.clear();
    }
    bool found = false;
    int repeatCnt;
    for (auto& info : repeat_infos) {
        if (info == repeat_read_info) {
            found = true;

            info.IncRepeatReadCount();

            repeatCnt = info.GetRepeatReadCount();
            break;
        }
    }
    if (!found) {
        repeat_infos.push_back(repeat_read_info);
        return;
    }
    if (repeatCnt >= env.GetRepeatReadThreshold()) {
        Issue issue(kType, file_io_info);
        issue.repeat_read_cnt_ = repeatCnt;
        issue.stack = repeat_read_info.GetStack();
        PublishIssue(issue, issues);
    }
}

FileIOSmallBufferDetector

监听缓存区过小

void FileIOSmallBufferDetector::Detect(const IOCanaryEnv &env, const IOInfo &file_io_info,
                                       std::vector<Issue>& issues) {
    if (file_io_info.op_cnt_ > env.kSmallBufferOpTimesThreshold && (file_io_info.op_size_ / file_io_info.op_cnt_) < env.GetSmallBufferThreshold()
            && file_io_info.max_continual_rw_cost_time_μs_ >= env.kPossibleNegativeThreshold) {
        PublishIssue(Issue(kType, file_io_info), issues);
    }
}

OnIssuePublish

所有信息都拿到之后就开始要回调了,也就回到了我们最开始开到的 iocanary::IOCanary::Get().SetIssuedCallback(OnIssuePublish)

void OnIssuePublish(const std::vector<Issue>& published_issues) {
    ....
    //这里new了一个Java层的List
    jobject j_issues = env->NewObject(kListClass, kMethodIDListConstruct);
    //遍历所有的info,拿到信息,每一条信息创建一个Java层的IOIssue对象,封装到这个对象中

    for (const auto& issue : published_issues) {
        jint type = issue.type_;
        jstring path = env->NewStringUTF(issue.file_io_info_.path_.c_str());
        jlong file_size = issue.file_io_info_.file_size_;
        jint op_cnt = issue.file_io_info_.op_cnt_;
        jlong buffer_size = issue.file_io_info_.buffer_size_;
        jlong op_cost_time = issue.file_io_info_.rw_cost_us_/1000;
        jint op_type = issue.file_io_info_.op_type_;
        jlong op_size = issue.file_io_info_.op_size_;
        jstring thread_name = env->NewStringUTF(issue.file_io_info_.java_context_.thread_name_.c_str());
        jstring stack = env->NewStringUTF(issue.stack.c_str());
        jint repeat_read_cnt = issue.repeat_read_cnt_;

        jobject issue_obj = env->NewObject(kIssueClass, kMethodIDIssueConstruct, type, path, file_size, op_cnt, buffer_size,
                                           op_cost_time, op_type, op_size, thread_name, stack, repeat_read_cnt);
        //讲IOIssue对象add到List中
        env->CallBooleanMethod(j_issues, kMethodIDListAdd, issue_obj);
        ....
    }
    //回调到Java层的IOCanaryJniBridge类中的静态方法onIssuePublish中
    env->CallStaticVoidMethod(kJavaBridgeClass, kMethodIDOnIssuePublish, j_issues);
    ....
}

后边在Java层onIssuePublish中就开始拼接信息转为json打印到控制台或上传服务器,流程至此就算结束了。

总结

IOCanaryPlugin通过hook底层io方法open、read、write、close来实现对io操作的拦截,于是所有的io操作都会被监控到,这样就可以在每一个io操作的过程中记录操作的信息,并分析io操作是否超过设定阈值,如满足条件则进行上报。


https://www.xamrdz.com/database/6bd1994671.html

相关文章: