Android数据流量的统计流程与分析方法

1.数据流量相关文件

android_net_TrafficStats.cpp (amss\linux\android\frameworks\base\core\jni)
JNI层接口，获取IFace状态，统计数据包

TrafficStats.java (amss\linux\android\frameworks\base\core\java\android\net)
java层接口，读取数据包大小

DcTracker.java (amss\linux\android\frameworks\opt\telephony\src\java\com\android\internal\telephony\dataconnection)
负责数据流量跟踪，开了一个子线程定时更新数据包大小，并发送状态给phoneNotifier.

GsmCdmaPhone.java (amss\linux\android\frameworks\opt\telephony\src\java\com\android\internal\telephony)
向UI提供流量状态的接口

DefaultPhoneNotifier.java (amss\linux\android\frameworks\opt\telephony\src\java\com\android\internal\telephony
执行流量通知，将状态通过TelephonyRegistry发给监听者

TelephonyRegistry.java (amss\linux\android\frameworks\base\services\core\java\com\android\server)
向监听者发送流量更新提示

PhoneStateListener.java (amss\linux\android\frameworks\base\telephony\java\android\telephony)
MobileSignalController.java (amss\linux\android\frameworks\base\packages\systemui\src\com\android\systemui\statusbar
私有类MobilePhoneStateListener继承自PhoneStateListener，通知RSSI更新流量状态。

2. Log分析与工具

分析需抓到下面的log

（高通平台）QXDM log PDCP; （MTK平台）ELT log
logcat radio log system log;
TCP dump

TCP dump抓取方法：
adb root
adb shell tcpdump -i any -s 0 -w /data/tcpdump.pcap
adb pull /data/tcpdump.pcap

相关分析工具：

QXDM、QCAT、WIRESHARK

3.重要代码

Step 1. 统计数据包

见下面的parseIfaceStats方法实现。

这里分别累计了rxBytes，rxPackets，txBytes，txPackets，tcpRxPackets，tcpTxPackets。
根据上层的读取接口可以看出，只有tcpRxPackets和tcpTxPackets被看作有效的下行和上行数据。
static int parseIfaceStats(const char* iface, struct Stats* stats) {
FILE *fp = fopen(QTAGUID_IFACE_STATS, “r”);
if (fp == NULL) {
return -1;
}

char buffer[384];
char cur_iface[32];
bool foundTcp = false;
uint64_t rxBytes, rxPackets, txBytes, txPackets, tcpRxPackets, tcpTxPackets;

while (fgets(buffer, sizeof(buffer), fp) != NULL) {
    int matched = sscanf(buffer, "%31s %" SCNu64 " %" SCNu64 " %" SCNu64
            " %" SCNu64 " " "%*u %" SCNu64 " %*u %*u %*u %*u "
            "%*u %" SCNu64 " %*u %*u %*u %*u", cur_iface, &rxBytes,
            &rxPackets, &txBytes, &txPackets, &tcpRxPackets, &tcpTxPackets);
    if (matched >= 5) {
        if (matched == 7) {
            foundTcp = true;
        }
        if (!iface || !strcmp(iface, cur_iface)) {
            stats->rxBytes += rxBytes;
            stats->rxPackets += rxPackets;
            stats->txBytes += txBytes;
            stats->txPackets += txPackets;
            if (matched == 7) {
                stats->tcpRxPackets += tcpRxPackets;
                stats->tcpTxPackets += tcpTxPackets;
            }
        }
    }
}

if (!foundTcp) {
    stats->tcpRxPackets = UNKNOWN;
    stats->tcpTxPackets = UNKNOWN;
}

if (fclose(fp) != 0) {
    return -1;
}
return 0;

}

step 2 .检测是否有上下行数据的逻辑

先看下这个枚举，定义了流量的几种状态

public enum Activity {
        NONE,
        DATAIN,  //下行
        DATAOUT, //上行
        DATAINANDOUT, //同时上行和下行
        DORMANT
    }

updateDataActivity该方法从3.1中获取到数据包大小，并与上一次检测的结果比较，进而判断是否有流量。
private void updateDataActivity() {
long sent, received;

    DctConstants.Activity newActivity;

    TxRxSum preTxRxSum = new TxRxSum(mTxPkts, mRxPkts);
    TxRxSum curTxRxSum = new TxRxSum();
    curTxRxSum.updateTxRxSum();
    mTxPkts = curTxRxSum.txPkts;
    mRxPkts = curTxRxSum.rxPkts;

    if (VDBG) {
        log("updateDataActivity: curTxRxSum=" + curTxRxSum + " preTxRxSum=" + preTxRxSum);
    }

    if (mNetStatPollEnabled && (preTxRxSum.txPkts > 0 || preTxRxSum.rxPkts > 0)) {
        sent = mTxPkts - preTxRxSum.txPkts;
        received = mRxPkts - preTxRxSum.rxPkts;

        if (VDBG)
            log("updateDataActivity: sent=" + sent + " received=" + received);
        if (sent > 0 && received > 0) {
            newActivity = DctConstants.Activity.DATAINANDOUT;
        } else if (sent > 0 && received == 0) {
            newActivity = DctConstants.Activity.DATAOUT;
        } else if (sent == 0 && received > 0) {
            newActivity = DctConstants.Activity.DATAIN;
        } else {
            newActivity = (mActivity == DctConstants.Activity.DORMANT) ?
                    mActivity : DctConstants.Activity.NONE;
        }

        if (mActivity != newActivity && mIsScreenOn) {
            if (VDBG)
                log("updateDataActivity: newActivity=" + newActivity);
            mActivity = newActivity;
            mPhone.notifyDataActivity();
        }
    }

}

step 3. 通知监听者流量状态更新

public void notifyDataActivityForSubscriber(int subId, int state) {
        if (!checkNotifyPermission("notifyDataActivity()" )) {
            return;
        }
        synchronized (mRecords) {
            int phoneId = SubscriptionManager.getPhoneId(subId);
            if (validatePhoneId(phoneId)) {
                mDataActivity[phoneId] = state;
                for (Record r : mRecords) {
                    // Notify by correct subId.
                    if (r.matchPhoneStateListenerEvent(PhoneStateListener.LISTEN_DATA_ACTIVITY) &&
                            idMatch(r.subId, subId, phoneId)) {
                        try {
                            r.callback.onDataActivity(state);
                        } catch (RemoteException ex) {
                            mRemoveList.add(r.binder);
                        }
                    }
                }
            }
            handleRemoveListLocked();
        }
}

Step 4.通知RSSI状态更新

void setActivity(int activity) {
        mCurrentState.activityIn = activity == TelephonyManager.DATA_ACTIVITY_INOUT
                || activity == TelephonyManager.DATA_ACTIVITY_IN;
        mCurrentState.activityOut = activity == TelephonyManager.DATA_ACTIVITY_INOUT
                || activity == TelephonyManager.DATA_ACTIVITY_OUT;
        if (mConfig.readIconsFromXml) {
            mCurrentState.dataActivity = activity;
        }
        notifyListenersIfNecessary();
}

关系图
待续。

本文为博主原创，可尽情点赞、留言与讨论。