Android Vold简介(二)

上一篇讲完了Vold的启动，本篇主要来介绍一下SM(StorageManager)和Vold、Kerel和Vold之间如何建立联系，以及信息是如何接收处理的。
 
之前已经提到了，NM(NetlinkManager)启动的时候会执行startListener开始监听Kernel的uevent事件，这里详细来看下是如何接收和处理消息事件的。

int SocketListener::startListener() {
    return startListener(4);
}

int SocketListener::startListener(int backlog) {

    if (!mSocketName && mSock == -1) {
        SLOGE("Failed to start unbound listener");
        errno = EINVAL;
        return -1;
    } else if (mSocketName) {
        if ((mSock = android_get_control_socket(mSocketName)) < 0) {
            SLOGE("Obtaining file descriptor socket '%s' failed: %s",
                 mSocketName, strerror(errno));
            return -1;
        }
        SLOGV("got mSock = %d for %s", mSock, mSocketName);
        fcntl(mSock, F_SETFD, FD_CLOEXEC);
    }

    if (mListen && listen(mSock, backlog) < 0) {
        SLOGE("Unable to listen on socket (%s)", strerror(errno));
        return -1;
    } else if (!mListen)
        mClients[mSock] = new SocketClient(mSock, false, mUseCmdNum);

    if (pipe2(mCtrlPipe, O_CLOEXEC)) {
        SLOGE("pipe failed (%s)", strerror(errno));
        return -1;
    }

    if (pthread_create(&mThread, nullptr, SocketListener::threadStart, this)) {
        SLOGE("pthread_create (%s)", strerror(errno));
        return -1;
    }

    return 0;
}

void *SocketListener::threadStart(void *obj) {
    SocketListener *me = reinterpret_cast<SocketListener *>(obj);

    me->runListener();
    pthread_exit(nullptr);
    return nullptr;
}

NetlinkHandler继承自NetlinkListener，NetlinkListener又是SoctetListener的子类，因此最终还是调用SoctetListener的startListener函数。该函数就是开始监听消息，接着启动一个线程执行runListener函数，循环读取socket消息并发送给各个client端，NetlinkListener作为其中一个，收到Callback后则会去进一步处理。

void SocketListener::runListener() {
    while (true) {
        std::vector<pollfd> fds;

        pthread_mutex_lock(&mClientsLock);
        fds.reserve(2 + mClients.size());
        fds.push_back({.fd = mCtrlPipe[0], .events = POLLIN});
        if (mListen) fds.push_back({.fd = mSock, .events = POLLIN});
        for (auto pair : mClients) {
            // NB: calling out to an other object with mClientsLock held (safe)
            const int fd = pair.second->getSocket();
            if (fd != pair.first) SLOGE("fd mismatch: %d != %d", fd, pair.first);
            fds.push_back({.fd = fd, .events = POLLIN});
        }
        pthread_mutex_unlock(&mClientsLock);

        SLOGV("mListen=%d, mSocketName=%s", mListen, mSocketName);
        int rc = TEMP_FAILURE_RETRY(poll(fds.data(), fds.size(), -1));
        if (rc < 0) {
            SLOGE("poll failed (%s) mListen=%d", strerror(errno), mListen);
            sleep(1);
            continue;
        }

        if (fds[0].revents & (POLLIN | POLLERR)) {
            char c = CtrlPipe_Shutdown;
            TEMP_FAILURE_RETRY(read(mCtrlPipe[0], &c, 1));
            if (c == CtrlPipe_Shutdown) {
                break;
            }
            continue;
        }
        if (mListen && (fds[1].revents & (POLLIN | POLLERR))) {
            int c = TEMP_FAILURE_RETRY(accept4(mSock, nullptr, nullptr, SOCK_CLOEXEC));
            if (c < 0) {
                SLOGE("accept failed (%s)", strerror(errno));
                sleep(1);
                continue;
            }
            pthread_mutex_lock(&mClientsLock);
            mClients[c] = new SocketClient(c, true, mUseCmdNum);
            pthread_mutex_unlock(&mClientsLock);
        }

        // Add all active clients to the pending list first, so we can release
        // the lock before invoking the callbacks.
        std::vector<SocketClient*> pending;
        pthread_mutex_lock(&mClientsLock);
        const int size = fds.size();
        for (int i = mListen ? 2 : 1; i < size; ++i) {
            const struct pollfd& p = fds[i];
            if (p.revents & (POLLIN | POLLERR)) {
                auto it = mClients.find(p.fd);
                if (it == mClients.end()) {
                    SLOGE("fd vanished: %d", p.fd);
                    continue;
                }
                SocketClient* c = it->second;
                pending.push_back(c);
                c->incRef();
            }
        }
        pthread_mutex_unlock(&mClientsLock);

        for (SocketClient* c : pending) {
            // Process it, if false is returned, remove from the map
            SLOGV("processing fd %d", c->getSocket());
            if (!onDataAvailable(c)) {
                release(c, false);
            }
            c->decRef();
        }
    }
}

详细梳理一下这个过程，mCtrlPipe[0]则作为此处的读端，会从流中读取读取POLLIN对应的事件，每一个client对应的socket文件描述符也都会存储到vector中。接下来就是利用poll函数去轮询vector中的每一个文件描述符。
 
接下来根据revents来决定是否有事件被读到，如果没有事件读到则进行下一轮，有事件被读到就将保存对应文件描述符的client，最后Callback onDataAvailable函数。接着继续看onDataAvailable。

bool NetlinkListener::onDataAvailable(SocketClient *cli)
{
    int socket = cli->getSocket();
    ssize_t count;
    uid_t uid = -1;

    bool require_group = true;
    if (mFormat == NETLINK_FORMAT_BINARY_UNICAST) {
        require_group = false;
    }

    count = TEMP_FAILURE_RETRY(uevent_kernel_recv(socket,
            mBuffer, sizeof(mBuffer), require_group, &uid));
    if (count < 0) {
        SLOGE("recvmsg failed (%s)", strerror(errno));
        return false;
    }

    NetlinkEvent *evt = new NetlinkEvent();
    if (evt->decode(mBuffer, count, mFormat)) {
        onEvent(evt);
    } else if (mFormat != NETLINK_FORMAT_BINARY) {
        // Don't complain if parseBinaryNetlinkMessage returns false. That can
        // just mean that the buffer contained no messages we're interested in.
        SLOGE("Error decoding NetlinkEvent");
    }

    delete evt;
    return true;
}

void NetlinkHandler::onEvent(NetlinkEvent* evt) {
    VolumeManager* vm = VolumeManager::Instance();
    const char* subsys = evt->getSubsystem();

    if (!subsys) {
        LOG(WARNING) << "No subsystem found in netlink event";
        return;
    }

    if (std::string(subsys) == "block") {
        vm->handleBlockEvent(evt);
    }
}

onDataAvailable这边是来真正接收数据的，通过给到的client字段与之建立socket通信，借助uevent_kernel_recv函数读取数据，通过decode解析后调用onEvent分发下去。这时候NetlinkHandler就登场了，获取VM单例对象，调用handleBlockEvent进行事件真正的处理。具体的处理过程这里不再详细介绍。
 
上半部分介绍了Kernel和Vold之间的联系，下面来看下SM和Vold之间的联系。

@Override
public void onStart() {
    mStorageManagerService = new StorageManagerService(getContext());
    publishBinderService("mount", mStorageManagerService);
    mStorageManagerService.start();
}

if (binder != null) {
    mVold = IVold.Stub.asInterface(binder);
    try {
        mVold.setListener(mListener);
    } catch (RemoteException e) {
        mVold = null;
        Slog.w(TAG, "vold listener rejected; trying again", e);
    }
} else {
    Slog.w(TAG, "vold not found; trying again");
}

在SM中，当SystemServer启动SM时候，回调onStart方法，然后会通过binder获得IVold接口，并且将mListener注册过去，这里便是Vold的aidl接口，对应便是VoldNativeService这个接口类。
 
当开机广播BootCompleted发出后，SM则会进行lockuser或者unlockuser的操作，接着会添加内置存储，并且reset Vold，其实对应的是调用VM的reset函数，最后是start user，创建对应文件目录，链接到对应的data存储分区。

private void resetIfBootedAndConnected() {
    if (mBootCompleted && mDaemonConnected) {
        final List<UserInfo> users = mContext.getSystemService(UserManager.class).getUsers();
        killMediaProvider(users);
        final int[] systemUnlockedUsers;
        synchronized (mLock) {
            systemUnlockedUsers = mSystemUnlockedUsers;
            mDisks.clear();
            mVolumes.clear();
            addInternalVolumeLocked();
        }
        try {
            mVold.reset();
            // Tell vold about all existing and started users
            for (UserInfo user : users) {
                mVold.onUserAdded(user.id, user.serialNumber);
            }
            for (int userId : systemUnlockedUsers) {
                    mVold.onUserStarted(userId);
                    mStoraged.onUserStarted(userId);
            }
            mVold.onSecureKeyguardStateChanged(mSecureKeyguardShowing);
            mStorageManagerInternal.onReset(mVold);
        } catch (Exception e) {
            Slog.wtf(TAG, e);
        }
    }
}

binder::Status VoldNativeService::reset() {
    ENFORCE_UID(AID_SYSTEM);
    ACQUIRE_LOCK;

    return translate(VolumeManager::Instance()->reset());
}

这里来看一下reset的过程，因为前面VM启动过程中，mInternalEmulated->create()会调用getListener()获取监听方法，但是彼时SystemServer还没有启动，SM还没有注册下来，所以并没有通知上去。
 
此处SM启动结束后执行reset，就会重新执行mInternalEmulated->create();过程，这样就会通知到SM，触发InternalEmulated去mount。

int VolumeManager::reset() {
    // Tear down all existing disks/volumes and start from a blank slate so
    // newly connected framework hears all events.
    if (mInternalEmulated != nullptr) {
        mInternalEmulated->destroy();
        mInternalEmulated->create();
    }
    for (const auto& disk : mDisks) {
        disk->destroy();
        disk->create();
    }
    updateVirtualDisk();
    mAddedUsers.clear();
    mStartedUsers.clear();
    return 0;
}

触发mount后最终会执行EmulatedVolume类的doMount()函数，进而拉起sdcard进程去完成真正的mount过程。

status_t EmulatedVolume::doMount() {
    // We could have migrated storage to an adopted private volume, so always
    // call primary storage "emulated" to avoid media rescans.
    std::string label = mLabel;
    if (getMountFlags() & MountFlags::kPrimary) {
        label = "emulated";
    }

    mFuseDefault = StringPrintf("/mnt/runtime/default/%s", label.c_str());
    mFuseRead = StringPrintf("/mnt/runtime/read/%s", label.c_str());
    mFuseWrite = StringPrintf("/mnt/runtime/write/%s", label.c_str());
    mFuseFull = StringPrintf("/mnt/runtime/full/%s", label.c_str());

    setInternalPath(mRawPath);
    setPath(StringPrintf("/storage/%s", label.c_str()));

    if (fs_prepare_dir(mFuseDefault.c_str(), 0700, AID_ROOT, AID_ROOT) ||
        fs_prepare_dir(mFuseRead.c_str(), 0700, AID_ROOT, AID_ROOT) ||
        fs_prepare_dir(mFuseWrite.c_str(), 0700, AID_ROOT, AID_ROOT) ||
        fs_prepare_dir(mFuseFull.c_str(), 0700, AID_ROOT, AID_ROOT)) {
        PLOG(ERROR) << getId() << " failed to create mount points";
        return -errno;
    }

    dev_t before = GetDevice(mFuseFull);

    if (!(mFusePid = fork())) {
        // clang-format off
        if (execl(kFusePath, kFusePath,
                "-u", "1023", // AID_MEDIA_RW
                "-g", "1023", // AID_MEDIA_RW
                "-m",
                "-w",
                "-G",
                "-i",
                "-o",
                mRawPath.c_str(),
                label.c_str(),
                NULL)) {
            // clang-format on
            PLOG(ERROR) << "Failed to exec";
        }

        LOG(ERROR) << "FUSE exiting";
        _exit(1);
    }

    if (mFusePid == -1) {
        PLOG(ERROR) << getId() << " failed to fork";
        return -errno;
    }

    nsecs_t start = systemTime(SYSTEM_TIME_BOOTTIME);
    while (before == GetDevice(mFuseFull)) {
        LOG(DEBUG) << "Waiting for FUSE to spin up...";
        usleep(50000);  // 50ms

        nsecs_t now = systemTime(SYSTEM_TIME_BOOTTIME);
        if (nanoseconds_to_milliseconds(now - start) > 5000) {
            LOG(WARNING) << "Timed out while waiting for FUSE to spin up";
            return -ETIMEDOUT;
        }
    }
    /* sdcardfs will have exited already. FUSE will still be running */
    TEMP_FAILURE_RETRY(waitpid(mFusePid, nullptr, 0));
    mFusePid = 0;

    return OK;
}

上面便是Kernel和Vold以及SM和Vold之间的联系，以及他们之间是如何进行交互。前者是通过监听Kernel uevent事件来完成上报，后者是通过binder建立通信，通过binder由SM向Vold发送命令，并且在Vold注册listener完成消息上报。