NDN新增tag的方案

本文档展示如何修改ndn-cxx和NFD的源码，添加一个值基本类型的Tag，下面的过程将展示添加一个值为uint64_t类型的Tag SrcAddress。

添加Tag的含义

在NDNLPv2协议中定义了LpPacket，具体的格式如下：

LpPacket = LP-PACKET-TYPE TLV-LENGTH*LpHeaderField[Fragment]LpHeaderField = SequenceSequence = SEQUENCE-TYPE TLV-LENGTH 8OCTETFragment = FRAGMENT-TYPE TLV-LENGTH 1*OCTET

其中Fragment放置的可以是一个NDN报的分片，或者是一个完整的NDN包（NDN包足够小，不需要分片的情况）；
*LpHeaderField就是一系列的LpPacket的Header，类似一个变长的数组，所以可以自己定义新的LpHeader；
所以添加一个新的Tag实际上就是添加一个新的LpHeader。

修改ndn-cxx

修改ndn-cxx/lp/tlv.hpp，为将要添加的新Tag分配一个新的TLV类型
```
enum {LpPacket = 100,Fragment = 80,Sequence = 81,FragIndex = 82,FragCount = 83,PitToken = 98,Nack = 800,NackReason = 801,NextHopFaceId = 816,IncomingFaceId = 817,CachePolicy = 820,CachePolicyType = 821,CongestionMark = 832,Ack = 836,TxSequence = 840,NonDiscovery = 844,PrefixAnnouncement = 848,/ add by qjm, for add new tagSrcAddress = 852/
};
```
如上所示，我们给新添加的Tag分配了一个TLV type，这个type是唯一标识一个TLV的。并且该值的分配也是需要遵循一定准则的，具体的分配规则可以参考：https://redmine.named-data.net/projects/nfd/wiki/NDNLPv2

Reserved Blocks¶

Two blocks of TLV-TYPEs have been reserved by link protocols:
- [80, 100]: 1-octet encoding
- [800, 1000]: 3-octet encoding
TLV-TYPE numbers for LpHeaderField SHOULD be assigned according to the following rules:
Note: number assignment for a TLV-TYPE nested within a LpHeaderField is not restricted by the above rules.

修改ndn-cxx/lp/tags.hpp，定义一个新的Tag：

/// add by qjm, for add new tag
typedef SimpleTag<uint64_t, 16> SrcAddressTag;
///

修改ndn-cxx/lp/fields.hpp，添加新Tag对应的Field：

// add by qjm, for add new tag
typedef FieldDecl<field_location_tags::Header,uint64_t, tlv::SrcAddress> SrcAddressField;
BOOST_CONCEPT_ASSERT((Field<SrcAddressField>));
///** \brief Set of all field declarations.
*/
typedef boost
::mpl::set<FragmentField,SequenceField,FragIndexField,FragCountField,PitTokenField,NackField,NextHopFaceIdField,IncomingFaceIdField,CachePolicyField,CongestionMarkField,AckField,TxSequenceField,NonDiscoveryField,PrefixAnnouncementField,/// add by qjm, for add new tagSrcAddressField///
> FieldSet;

修改ndn-cxx/face.cpp => extractLpLocalFields：

template<typename NetPkt>
static void
extractLpLocalFields(NetPkt &netPacket, const lp::Packet &lpPacket) {addTagFromField<lp::IncomingFaceIdTag, lp::IncomingFaceIdField>(netPacket, lpPacket);addTagFromField<lp::CongestionMarkTag, lp::CongestionMarkField>(netPacket, lpPacket);// add by qjm, for add new tagaddTagFromField<lp::SrcAddressTag, lp::SrcAddressField>(netPacket, lpPacket);//
}

上面的静态函数是在一个Face从Transport接收到一个Element，进行解析时调用的，功能是 将收到的LpPacket中的LpHeaderField提取，并保存到指定包的Tag列表当中。

如果插入的Tag要在Interest包中生效，则需要修改ndn-cxx/impl/face-impl.hpp => asyncExpressInterest：

void
asyncExpressInterest(RecordId id, shared_ptr<const Interest> interest,const DataCallback &afterSatisfied,const NackCallback &afterNacked,const TimeoutCallback &afterTimeout) {NDN_LOG_DEBUG("<I " << *interest);this->ensureConnected(true);const Interest &interest2 = *interest;auto &entry = m_pendingInterestTable.put(id, std::move(interest), afterSatisfied, afterNacked,afterTimeout, ref(m_scheduler));lp::Packet lpPacket;addFieldFromTag<lp::NextHopFaceIdField, lp::NextHopFaceIdTag>(lpPacket, interest2);addFieldFromTag<lp::CongestionMarkField, lp::CongestionMarkTag>(lpPacket, interest2);// add by qjm, for add new tagaddFieldFromTag<lp::SrcAddressField, lp::SrcAddressTag>(lpPacket, interest2);//entry.recordForwarding();m_face.m_transport->send(finishEncoding(std::move(lpPacket), interest2.wireEncode(),'I', interest2.getName()));dispatchInterest(entry, interest2);
}

上面新添的代码的功能是将一个要发送的包（Interest、Data等继承自TagHost）中的Tag提取处理出来，并作为一个LpHeaderField写入新构造的lpPacket当中。

如果插入的Tag要在Data包中生效，则需要修改ndn-cxx/impl/face-impl.hpp => asyncPutData：

void
asyncPutData(const Data &data) {NDN_LOG_DEBUG("<D " << data.getName());bool shouldSendToForwarder = satisfyPendingInterests(data);if (!shouldSendToForwarder) {return;}this->ensureConnected(true);lp::Packet lpPacket;addFieldFromTag<lp::CachePolicyField, lp::CachePolicyTag>(lpPacket, data);addFieldFromTag<lp::CongestionMarkField, lp::CongestionMarkTag>(lpPacket, data);/// add by qjm, for add new tagaddFieldFromTag<lp::SrcAddressField, lp::SrcAddressTag>(lpPacket, data);//m_face.m_transport->send(finishEncoding(std::move(lpPacket), data.wireEncode(),'D', data.getName()));
}

上面新添的代码的功能是将一个要发送的包（Interest、Data等继承自TagHost）中的Tag提取处理出来，并作为一个LpHeaderField写入新构造的lpPacket当中。

修改NFD

NFD中需要修改daemon/face/generic-link-service.cpp中的两处代码：

GenericLinkService::decodeInterest
```
void
GenericLinkService::decodeInterest(const Block &netPkt, const lp::Packet &firstPkt,const EndpointId &endpointId) {...../// add by qjm, for add srcAddress Tagif (firstPkt.has<lp::SrcAddressField>()) {interest->setTag(make_shared<lp::SrcAddressTag>(firstPkt.get<lp::SrcAddressField>()));}///this->receiveInterest(*interest, endpointId);
}
```
上面的代码在GenericLinkService解析Interest的时候，提取对应LpPacket中的Field，写入Interest的tags列表当中。

ps: 由于此处演示的是添加一个Tag，只对于Interest生效，所以需要修改这个函数，如果要对Data包或者Nack等包生效，类比上述代码，对GenericLinkService::decodeData，GenericLinkService::decodeNack之类的进行修改即可

GenericLinkService::encodeLpFields

void
GenericLinkService::encodeLpFields(const ndn::PacketBase &netPkt, lp::Packet &lpPacket) {...../// add by qjm, for add srcAddress Tagauto srcAddressTag = netPkt.getTag<lp::SrcAddressTag>();if (srcAddressTag != nullptr) {lpPacket.add<lp::SrcAddressField>(*srcAddressTag);}///auto pitToken = netPkt.getTag<lp::PitToken>();if (pitToken != nullptr) {lpPacket.add<lp::PitTokenField>(*pitToken);}
}

上面新添的代码会在发送一个包的时候调用，目的是将Tag提取出来，座位一个LpHeaderField 添加到LpPacket中。

添加成功后的效果