brpc(1)—bthread和brpc

brpc 是百度开源的一个网络框架, 它几乎是开源的C++高性能网络框架的唯一选择。它是国内C++最优秀的开源作品之一，我相信国内大厂内部的闭源RPC网络库也参考过它

brpc的重要特性

bthread，这是几乎唯一的工业级开源协程库（虽然作者不认为它是协程）,在开源界几乎没有替代品
标准的rpc框架和接口, 四参数service接口 (google::protobuf::RpcController* cntl_base, const Request* request, Response* response, google::protobuf::Closure* done) 终结了rpc 各种各样的自定义调用接口问题，以后开发rpc默认使用这种接口，
完备的可观测性；logging和trace, span; bvar 支持的监控指标

bthread

bthread 是有栈协程, 使用用户栈来代表协程对象。它使用队列管理协程等待和切换（类似goroutine），例如加协程锁时，第一个bthread进入执行，后续bthread进入等待队列，当第一个bthread执行完后，唤醒等待队列的一个bthread进入临界区。

bthread主要依靠TaskGroup和TaskControl两个类管理，task其实就是指bthread。

TaskGroup

taskgroup 是 Thread-local group of tasks. 代表一个pthread 上的bthread协程。taskgroup负责执行一个bthread，切换当前线程的bthread等。主要方法有：

start_foreground 立即运行bthread

C++

static int start_foreground(TaskGroup** pg,
                                bthread_t* __restrict tid,
                                const bthread_attr_t* __restrict attr,
                                void * (*fn)(void*),
                                void* __restrict arg);

start_background 将bthread加入到队列_rq，等待调度

C++

    template <bool REMOTE>
    int start_background(bthread_t* __restrict tid,
                         const bthread_attr_t* __restrict attr,
                         void * (*fn)(void*),
                         void* __restrict arg);

sched(TaskGroup** pg); 从pg的队列取出一个bthread，执行。如果要切换bthread，也是调用TaskGroup::sched

TaskGroup有两个任务队列， _rq和_remote_rq。_rq 是当前线程的bthread执行流加的任务, _remote_rq是其他线程向当前线程提交的任务。_rq相比_remote_rq，始终是单线程执行，不需要加锁，性能更高。使用两个队列也是bthread对性能的优化

C++

void TaskGroup::sched(TaskGroup** pg) {
    TaskGroup* g = *pg;
    bthread_t next_tid = 0;
    // Find next task to run, if none, switch to idle thread of the group.
#ifndef BTHREAD_FAIR_WSQ
    const bool popped = g->_rq.pop(&next_tid);
#else
    const bool popped = g->_rq.steal(&next_tid);
#endif
    if (!popped && !g->steal_task(&next_tid)) {
        // Jump to main task if there's no task to run.
        next_tid = g->_main_tid;
    }
    sched_to(pg, next_tid);
}

    bool steal_task(bthread_t* tid) {
        if (_remote_rq.pop(tid)) {
            return true;
        }
#ifndef BTHREAD_DONT_SAVE_PARKING_STATE
        _last_pl_state = _pl->get_state();
#endif
        return _control->steal_task(tid, &_steal_seed, _steal_offset);
    }

void sched_to(TaskGroup** pg, bthread_t next_tid); 立即执行指定tid的bthread任务, 也就是协程切换

C++

inline void TaskGroup::sched_to(TaskGroup** pg, bthread_t next_tid) {
    TaskMeta* next_meta = address_meta(next_tid);  // 直接通过tid 加地址找到taskmeta
    if (next_meta->stack == NULL) {
        ContextualStack* stk = get_stack(next_meta->stack_type(), task_runner);
        if (stk) {
            next_meta->set_stack(stk);
        } else {
            ...
        }
    }
    // Update now_ns only when wait_task did yield.
    sched_to(pg, next_meta);
}

void TaskGroup::sched_to(TaskGroup** pg, TaskMeta* next_meta) {
    TaskGroup* g = *pg;

    // Save errno so that errno is bthread-specific.
    const int saved_errno = errno;
    void* saved_unique_user_ptr = tls_unique_user_ptr;

    TaskMeta* const cur_meta = g->_cur_meta;
    const int64_t now = butil::cpuwide_time_ns();
    const int64_t elp_ns = now - g->_last_run_ns;
    g->_last_run_ns = now;
    cur_meta->stat.cputime_ns += elp_ns;

    // Switch to the task
    if (__builtin_expect(next_meta != cur_meta, 1)) {
        g->_cur_meta = next_meta;
        // Switch tls_bls
        cur_meta->local_storage = tls_bls;
        tls_bls = next_meta->local_storage;

        // Logging must be done after switching the local storage, since the logging lib
        // use bthread local storage internally, or will cause memory leak.
        if ((cur_meta->attr.flags & BTHREAD_LOG_CONTEXT_SWITCH) ||
            (next_meta->attr.flags & BTHREAD_LOG_CONTEXT_SWITCH)) {
            LOG(INFO) << "Switch bthread: " << cur_meta->tid << " -> "
                      << next_meta->tid;
        }

        if (cur_meta->stack != NULL) {
            if (next_meta->stack != cur_meta->stack) {
                CheckBthreadScheSafety();
                jump_stack(cur_meta->stack, next_meta->stack);
                // probably went to another group, need to assign g again.
                g = BAIDU_GET_VOLATILE_THREAD_LOCAL(tls_task_group);

    ....
}

int usleep(TaskGroup** pg, uint64_t timeout_us); void yield(TaskGroup** pg);

sleep加一个定时任务，然后切换协程

C++

int TaskGroup::usleep(TaskGroup** pg, uint64_t timeout_us) {
    if (0 == timeout_us) {
        yield(pg);
        return 0;
    }
    TaskGroup* g = *pg;
    // We have to schedule timer after we switched to next bthread otherwise
    // the timer may wake up(jump to) current still-running context.
    SleepArgs e = { timeout_us, g->current_tid(), g->current_task(), g };
    g->set_remained(_add_sleep_event, &e);
    sched(pg);
    g = *pg;

void TaskGroup::yield(TaskGroup** pg) {
    TaskGroup* g = *pg;
    ReadyToRunArgs args = {  g->_cur_meta, false };
    g->set_remained(ready_to_run_in_worker, &args);
    sched(pg);
}

TaskControl

TaskControl管理Taskgroup

创建TaskGroup, 调用_add_group(group, tag)

TaskControl使用std::vector<TaggedGroups> _tagged_groups;管理TaskGroup, 增加TaskGroup 使用_tagged_groups[tag][ngroup] = g;

taskControl的每个tag对应若干taskgroup

C++

    // Create a TaskGroup in this control.
    TaskGroup* create_group(bthread_tag_t tag);

TaskGroup* TaskControl::create_group(bthread_tag_t tag) {
    TaskGroup* g = new (std::nothrow) TaskGroup(this);
    if (NULL == g) {
        LOG(FATAL) << "Fail to new TaskGroup";
        return NULL;
    }
    if (g->init(FLAGS_task_group_runqueue_capacity) != 0) {
        LOG(ERROR) << "Fail to init TaskGroup";
        delete g;
        return NULL;
    }
    if (_add_group(g, tag) != 0) {
        delete g;
        return NULL;
    }
    return g;
}

int TaskControl::_add_group(TaskGroup* g, bthread_tag_t tag) {
    std::unique_lock<butil::Mutex> mu(_modify_group_mutex);
    if (_stop) {
        return -1;
    }
    g->set_tag(tag);
    g->set_pl(&_pl[tag][butil::fmix64(pthread_numeric_id()) % PARKING_LOT_NUM]);
    size_t ngroup = _tagged_ngroup[tag].load(butil::memory_order_relaxed);
    if (ngroup < (size_t)BTHREAD_MAX_CONCURRENCY) {
        _tagged_groups[tag][ngroup] = g;
        _tagged_ngroup[tag].store(ngroup + 1, butil::memory_order_release);
    }
    mu.unlock();
    return 0;
}

typedef std::array<TaskGroup*, BTHREAD_MAX_CONCURRENCY> TaggedGroups;
std::vector<butil::atomic<size_t>> _tagged_ngroup;
std::vector<TaggedGroups> _tagged_groups;

bool steal_task(bthread_t* tid, size_t* seed, size_t offset); TaskControl从当前线程的taskgroup拿一个bthread执行。

C++

bool TaskControl::steal_task(bthread_t* tid, size_t* seed, size_t offset) {
    // 当前线程的taskgroup, tls_task_group。
    // tls_task_group 定义为每个线程的内部变量
    auto tag = tls_task_group->tag();

    const size_t ngroup = tag_ngroup(tag).load(butil::memory_order_acquire/*1*/);
    if (0 == ngroup) {
        return false;
    }

    // NOTE: Don't return inside `for' iteration since we need to update |seed|
    bool stolen = false;
    size_t s = *seed;
    auto& groups = tag_group(tag);
    for (size_t i = 0; i < ngroup; ++i, s += offset) {
        TaskGroup* g = groups[s % ngroup];
        // g is possibly NULL because of concurrent _destroy_group
        if (g) {
            if (g->_rq.steal(tid)) {
                stolen = true;
                break;
            }
            if (g->_remote_rq.pop(tid)) {
                stolen = true;
                break;
            }
        }
    }
    *seed = s;
    return stolen;
}

TaskControl::signal_task(int num_task, bthread_tag_t tag) 唤醒tag对应的taskgroup

值得注意的是, signal 借助ParkingLot 类，通过futex_wait_private和futex_wake_private封装了一种信号量

C++

void TaskControl::signal_task(int num_task, bthread_tag_t tag) {
    if (num_task <= 0) {
        return;
    }
    if (num_task > 2) {
        num_task = 2;
    }
    auto& pl = tag_pl(tag);
    int start_index = butil::fmix64(pthread_numeric_id()) % PARKING_LOT_NUM;
    num_task -= pl[start_index].signal(1);
    if (num_task > 0) {
        for (int i = 1; i < PARKING_LOT_NUM && num_task > 0; ++i) {
            if (++start_index >= PARKING_LOT_NUM) {
                start_index = 0;
            }
            num_task -= pl[start_index].signal(1);
        }
    }
}

TaskGroup* choose_one_group(bthread_tag_t tag); 随机选择一个tag对应的TaskGroup

C++

TaskGroup* TaskControl::choose_one_group(bthread_tag_t tag) {
    CHECK(tag >= BTHREAD_TAG_DEFAULT && tag < FLAGS_task_group_ntags);
    auto& groups = tag_group(tag);
    const auto ngroup = tag_ngroup(tag).load(butil::memory_order_acquire);
    if (ngroup != 0) {
        return groups[butil::fast_rand_less_than(ngroup)];
    }
    CHECK(false) << "Impossible: ngroup is 0";
    return NULL;
}

bthread协程的实现原理

bthread是有栈协程，在创建协程时，会在线程空间创建协程栈，并装载协程函数；运行协程是运行协程栈的函数，类似线程栈的函数入栈出栈；切换协程是保存协程上下文到协程栈；退出协程则会销毁协程栈。

协程实现包括bthread_make_fcontext 创建协程栈、bthread_jump_fcontext 切换协程栈函数，在context.cpp文件里。

协程实现原理还可以参考：https://sf-zhou.github.io/brpc/brpc_01_bthread.html

总结

bthread 提供了工业级的协程实现，让brpc的函数可以在bthread中执行。bthread提供sleep, yield, join等常用调用。

taskcontrol是管理和调度bthread的类, 还提供steal_task方法从某个线程中拿取bthread，这个接口可以用来优化线程load的均衡。

用户可以通过以下步骤管理协程

taskcontrol.init(int nconcurrency) 创建指定数量的线程
taskcontrol.create_group(tag) 创建taskgroup, taskgroup会由固定的一个线程负责，返回taskgroup
taskgroup.start_background, 在taskgroup启动一个bthread任务并执行, 可以实现《在指定线程创建协程》这个灵活操作

brpc

brpc 是一个网络库，下层接socket、event和网络包，上层接http/proto网络协议和service。同时提供name service根据域名获取下游机器组和lb 负载均衡选择下游机器。

channel 和客户端

channel通道，是网络中相当常见的概念。在brpc中，专门指客户端发起的通道。A Channel represents a communication line to one server or multiple servers which can be used to call that Server's services. Servers may be running on another machines. Normally, you should not call a Channel directly, but instead construct a stub Service wrapping it.

channel 主要是两个接口

channel::Init 创建shannel，发起链接

int Init(const char* server_addr_and_port, const ChannelOptions* options); 初始化channel，也就是创建发起一个连接

C++

int Channel::Init(const char* server_addr, int port,
                  const ChannelOptions* options) {
    GlobalInitializeOrDie();
    butil::EndPoint point;
    const AdaptiveProtocolType& ptype = (options ? options->protocol : _options.protocol);
    const Protocol* protocol = FindProtocol(ptype);
    if (protocol == NULL || !protocol->support_client()) {
        LOG(ERROR) << "Channel does not support the protocol";
        return -1;
    }
    if (protocol->parse_server_address != NULL) {
        if (!protocol->parse_server_address(&point, server_addr)) {
            LOG(ERROR) << "Fail to parse address=`" << server_addr << '\'';
            return -1;
        }
        point.port = port;
    } else {
        if (str2endpoint(server_addr, port, &point) != 0 &&
            hostname2endpoint(server_addr, port, &point) != 0) {
            return -1;
        }
    }
    return InitSingle(point, server_addr, options, port);
}

Channel::InitSingle 函数

a. GlobalInitializeOrDie(); b. InitChannelOptions(options) c. CreateSocketSSLContext(_options, &ssl_ctx) d. SocketMapInsert(SocketMapKey(server_addr_and_port, sig),

显然核心函数在SocketMapInsert, SocketMap::Insert(const SocketMapKey& key, SocketId* id,

a. SingleConnection* sc = _map.seek(key); 如果链接已经有了，返回 b. _options.socket_creator->CreateSocket(opt, &tmp_id) 创建链接 c. _map[key] = new_sc; 链接加入到map

Socket::OnCreated

a. 配置成员变量，如_remote_side，_local_side 等 b. DoConnect(options.connect_abstime, NULL, NULL); 创建和发起socket链接

C++

int Socket::DoConnect(const timespec* abstime,
                      int (*on_connect)(int, int, void*), void* data) {
    if (_conn) {
        return _conn->Connect(this, abstime, on_connect, data);
    } else {
	    // 会重新创建socket，然后::connect
        return Connect(abstime, on_connect, data);
    }
}

stub.method 通过channel 向服务端发送请求

客户端调用

C++

    // 初始化channel
    brpc::Channel channel;

    // Initialize the channel, NULL means using default options.
    brpc::ChannelOptions options;
    options.protocol = FLAGS_protocol;
    options.connection_type = FLAGS_connection_type;
    options.timeout_ms = FLAGS_timeout_ms/*milliseconds*/;
    options.max_retry = FLAGS_max_retry;
    if (channel.Init(FLAGS_server.c_str(), FLAGS_load_balancer.c_str(), &options) != 0) {
        LOG(ERROR) << "Fail to initialize channel";
        return -1;
    }

    example::EchoService_Stub stub(&channel);
	
	example::EchoRequest request;
	example::EchoResponse response;
	brpc::Controller cntl;

	request.set_message(g_request);
	cntl.set_log_id(log_id++);  // set by user

	cntl.request_attachment().append(g_attachment);

	stub.Echo(&cntl, &request, &response, NULL);

// 通过channel->CallMethod 发送数据
void EchoService_Stub::Echo(::PROTOBUF_NAMESPACE_ID::RpcController* controller,
                              const ::example::EchoRequest* request,
                              ::example::EchoResponse* response,
                              ::google::protobuf::Closure* done) {
  channel_->CallMethod(descriptor()->method(0),
                       controller, request, response, done);
}

Channel::CallMethod

C++

void Channel::CallMethod(const google::protobuf::MethodDescriptor* method,
                         google::protobuf::RpcController* controller_base,
                         const google::protobuf::Message* request,
                         google::protobuf::Message* response,
                         google::protobuf::Closure* done) {
    const int64_t start_send_real_us = butil::gettimeofday_us();
    Controller* cntl = static_cast<Controller*>(controller_base);
    cntl->OnRPCBegin(start_send_real_us);
    ...
    cntl->_response = response;
    cntl->_done = done;
    cntl->_pack_request = _pack_request;
    cntl->_method = method;
    cntl->_auth = _options.auth;

    if (SingleServer()) {
        cntl->_single_server_id = _server_id;
        cntl->_remote_side = _server_address;
    }

    // Share the lb with controller.
    cntl->_lb = _lb;

    _serialize_request(&cntl->_request_buf, cntl, request);

    cntl->IssueRPC(start_send_real_us);
    if (done == NULL) {
        // done如果是nullptr，同步等待
        Join(correlation_id);
        if (cntl->_span) {
            cntl->SubmitSpan();
        }
        cntl->OnRPCEnd(butil::gettimeofday_us());
    }
}

cntl->IssueRPC

C++

void Controller::IssueRPC(int64_t start_realtime_us) {
    _current_call.begin_time_us = start_realtime_us;

    // Pick a target server for sending RPC
    _current_call.need_feedback = false;
    _current_call.enable_circuit_breaker = has_enabled_circuit_breaker();
    SocketUniquePtr tmp_sock;
    if (SingleServer()) {
        const int rc = Socket::Address(_single_server_id, &tmp_sock);
        
        _current_call.peer_id = _single_server_id;
    } else {
        // 使用lb选择下游服务器
        LoadBalancer::SelectIn sel_in =
            { start_realtime_us, true,
              has_request_code(), _request_code, _accessed };
        LoadBalancer::SelectOut sel_out(&tmp_sock);
        const int rc = _lb->SelectServer(sel_in, &sel_out);
        if (rc != 0) {
            std::ostringstream os;
            DescribeOptions opt;
            opt.verbose = false;
            _lb->Describe(os, opt);
            SetFailed(rc, "Fail to select server from %s", os.str().c_str());
            return HandleSendFailed();
        }
        _current_call.need_feedback = sel_out.need_feedback;
        _current_call.peer_id = tmp_sock->id();

        _remote_side = tmp_sock->remote_side();
    }

    // Make request
    butil::IOBuf packet;
    SocketMessage* user_packet = NULL;
    _pack_request(&packet, &user_packet, cid.value, _method, this,
                  _request_buf, using_auth);
    // TODO: PackRequest may accept SocketMessagePtr<>?
    SocketMessagePtr<> user_packet_guard(user_packet);
    Socket::WriteOptions wopt;
    wopt.id_wait = cid;
    wopt.abstime = pabstime;
    wopt.pipelined_count = _pipelined_count;
    wopt.auth_flags = _auth_flags;
    wopt.ignore_eovercrowded = has_flag(FLAGS_IGNORE_EOVERCROWDED);
    wopt.write_in_background = write_to_socket_in_background();
    int rc;
    size_t packet_size = 0;
    // 写数据到socket
    if (user_packet_guard) {
        if (span) {
            packet_size = user_packet_guard->EstimatedByteSize();
        }
        rc = _current_call.sending_sock->Write(user_packet_guard, &wopt);
    } else {
        packet_size = packet.size();
        rc = _current_call.sending_sock->Write(&packet, &wopt);
    }

    if (using_auth) {
        _current_call.sending_sock->SetAuthentication(rc);
    }
}

Acceptor 和EventDispatcher

brpc 的rpc 客户端是通过channel 创建链接和向链接发送请求。类似channel的两个特点, 服务端的基础就是从socket accpet链接和接收数据。完成这个的分别是acceptor和EventDispatcher

Acceptor 接受链接

Acceptor 是server的成员变量，在server.Start()里调用Acceptor::StartAccept

C++

    Acceptor* _am;
    Acceptor* _internal_am;

    // listen socket
    _listen_addr = endpoint;
    for (int port = port_range.min_port; port <= port_range.max_port; ++port) {
        _listen_addr.port = port;
        butil::fd_guard sockfd(tcp_listen(_listen_addr));
        if (_listen_addr.port == 0) {
            // port=0 makes kernel dynamically select a port from
            // https://en.wikipedia.org/wiki/Ephemeral_port
            _listen_addr.port = get_port_from_fd(sockfd);    
    ...
    // Pass ownership of `sockfd' to `_am'
    if (_am->StartAccept(sockfd, _options.idle_timeout_sec,
                            _default_ssl_ctx,
                            _options.force_ssl) != 0) {
        LOG(ERROR) << "Fail to start acceptor";
        return -1;
    }

Acceptor::StartAccept 会把 sockfd 注册可读事件给EventDispatcher，回调函数是OnNewConnections，也就是创建新链接。

显然EventDispatcher 是一个epoll模型的程序, 不停的触发网络事件，并发送给对应注册事件的线程。

C++

void Acceptor::OnNewConnections(Socket* acception) {
    int progress = Socket::PROGRESS_INIT;
    do {
        OnNewConnectionsUntilEAGAIN(acception);
        if (acception->Failed()) {
            return;
        }
    } while (acception->MoreReadEvents(&progress));
}

void Acceptor::OnNewConnectionsUntilEAGAIN(Socket* acception) {
    while (1) {
        struct sockaddr_storage in_addr;
        bzero(&in_addr, sizeof(in_addr));
        socklen_t in_len = sizeof(in_addr);
        butil::fd_guard in_fd(accept(acception->fd(), (sockaddr*)&in_addr, &in_len));
        if (in_fd < 0) {
            continue;
        }

        Acceptor* am = dynamic_cast<Acceptor*>(acception->user());
        if (NULL == am) {
            return;
        }
        
        SocketId socket_id;
        SocketOptions options;
        options.keytable_pool = am->_keytable_pool;
        options.fd = in_fd;
        butil::sockaddr2endpoint(&in_addr, in_len, &options.remote_side);
        options.user = acception->user();
        options.force_ssl = am->_force_ssl;
        options.initial_ssl_ctx = am->_ssl_ctx;
            options.on_edge_triggered_events = InputMessenger::OnNewMessages;
        }
        options.use_rdma = am->_use_rdma;
        options.bthread_tag = am->_bthread_tag;
        if (Socket::Create(options, &socket_id) != 0) {
            LOG(ERROR) << "Fail to create Socket";
            continue;
        }

        SocketUniquePtr sock;
        if (Socket::AddressFailedAsWell(socket_id, &sock) >= 0) {
            bool is_running = true;
            {
                BAIDU_SCOPED_LOCK(am->_map_mutex);
                is_running = (am->status() == RUNNING);
                am->_socket_map.insert(socket_id, ConnectStatistics());
            }
}

EventDispatcher

EventDispatcher 可以配置，一般每个tag tasakgroup配置一个dispather，也就是一个线程一个, 以bthread的形式运行

循环执行epoll_wait -> CallInputEventCallback。 epoll_wait不会阻塞整个线程? epoll_wait时线程阻塞，但有事件到来继续执行时，该线程会创建多个bthread处理。这里需要详细分析brpc 线程切换的行为

C++

void InitializeGlobalDispatchers() {
    g_edisp = new EventDispatcher[FLAGS_task_group_ntags * FLAGS_event_dispatcher_num];
    for (int i = 0; i < FLAGS_task_group_ntags; ++i) {
        for (int j = 0; j < FLAGS_event_dispatcher_num; ++j) {
            bthread_attr_t attr =
                FLAGS_usercode_in_pthread ? BTHREAD_ATTR_PTHREAD : BTHREAD_ATTR_NORMAL;
            attr.tag = (BTHREAD_TAG_DEFAULT + i) % FLAGS_task_group_ntags;
            // dispatcher.Start
            CHECK_EQ(0, g_edisp[i * FLAGS_event_dispatcher_num + j].Start(&attr));
        }
    }

int EventDispatcher::Start(const bthread_attr_t* consumer_thread_attr) {
    if (_event_dispatcher_fd < 0) {
        LOG(FATAL) << "epoll was not created";
        return -1;
    }
    if (_tid != 0) {
        LOG(FATAL) << "Already started this dispatcher(" << this 
                   << ") in bthread=" << _tid;
        return -1;
    }
    // 
    int rc = bthread_start_background(&_tid, &epoll_thread_attr, RunThis, this);
    if (rc) {
        LOG(FATAL) << "Fail to create epoll thread: " << berror(rc);
        return -1;
    }
    return 0;
}

// bthread运行EventDispatcher::Run
void EventDispatcher::Run() {
    while (!_stop) {
        epoll_event e[32];
        const int n = epoll_wait(_event_dispatcher_fd, e, ARRAY_SIZE(e), -1);
        if (_stop) {
            break;
        }
        if (n < 0) {
            if (EINTR == errno) {
                // We've checked _stop, no wake-up will be missed.
                continue;
            }
            PLOG(FATAL) << "Fail to epoll_wait epfd=" << _event_dispatcher_fd;
            break;
        }
        for (int i = 0; i < n; ++i) {
            if (e[i].events & (EPOLLIN | EPOLLERR | EPOLLHUP)
                ) {
                // We don't care about the return value.
                CallInputEventCallback(e[i].data.u64, e[i].events, _thread_attr);
            }
        }
        for (int i = 0; i < n; ++i) {
            if (e[i].events & (EPOLLOUT | EPOLLERR | EPOLLHUP)) {
                // We don't care about the return value.
                CallOutputEventCallback(e[i].data.u64, e[i].events, _thread_attr);
            }
        }
    }
}

protocol service

上层应用层协议

EventDispatcher的CallOutputEventCallback 首先调用InputMessenger::OnNewMessages 从socket读数据，然后调用InputMessenger::CutInputMessage 确认协议类型。

确定协议类型的办法

遍历_handlers
执行 _handlers[cur_index].parse(&m->_read_buf, m, read_eof, _handlers[cur_index].arg);
如果成功，找到Index；如果失败，继续下一轮尝试这个一个个的解析尝试的办法，肯定会影响性能把

handle类型就是协议类型，确认了协议类型就调用对应的handle解析协议了

C++

    Protocol http_protocol = { ParseHttpMessage,
                               SerializeHttpRequest, PackHttpRequest,
                               ProcessHttpRequest, ProcessHttpResponse,
                               VerifyHttpRequest, ParseHttpServerAddress,
                               GetHttpMethodName,
                               CONNECTION_TYPE_POOLED_AND_SHORT,
                               "http" };
    if (RegisterProtocol(PROTOCOL_HTTP, http_protocol) != 0) {
        exit(1);
    }

以protocol rpc service为例，数据从socket读取到msg后，存放在InputMessageBase

C++

// 任何网络包都可以解析成header+data两部分
struct BAIDU_CACHELINE_ALIGNMENT MostCommonMessage : public InputMessageBase {
    butil::IOBuf meta;
    butil::IOBuf payload;
    PipelinedInfo pi;

    inline static MostCommonMessage* Get() {
        return butil::get_object<MostCommonMessage>();
    }

ParsePbFromIOBuf(&meta, msg->meta)
将msg->payload解析到messages->Request()msg->payload.cutn(&req_buf, body_without_attachment_size); ParseFromCompressedData(req_buf, messages->Request(), req_cmp_type)
std::unique_ptr<Controller> cntl(new (std::nothrow) Controller);
设置google::protobuf::Closure* done = ::brpc::NewCallback< int64_t, Controller*, RpcPBMessages*, const Server*, MethodStatus*, int64_t>(&SendRpcResponse,
获得service const Server::MethodProperty* mp = server_accessor.FindMethodPropertyByFullName( svc_name, request_meta.method_name()); svc = mp->service; 拿到service对象
svc->CallMethod(method, cntl.release(), messages->Request(), messages->Response(), done); CallMethod具体的实现在pb.h中
执行完service函数, 最后SendRpcResponse(meta.correlation_id(), cntl.release(), messages, server, method_status, msg->received_us());

TODO

brpc的线程模型和切换

brpc的使用和性能资源分析

brpc和butil实现上的优秀、值得借鉴的地方