NS2包结构解析

      最近在做ns2的“反移植”工作，深入研究了一下NS2中包的结构，其定义主要在packet.h/cc中实现的，但是有许多代码是为了与TCL接口而设计的。其定义如下：

class Packet : public Event {
private:

 unsigned char* bits_; // headerbits
 AppData*data_;  // variable size bufferfor 'data'
....

};

 

     不得不说，上面两个字段域是Packet最重要的动动，其中bits_存储包头结构，而data_存储用户自定义的数据。但是，NS2其实是一个大而全的仿真平台，它在仿真时其实是将所有或者大多数数据包放在一个包里面，即使当前用不到。比如我们建立一个仿真环境，用来测试两个节点的通过TCP建立连接，然后发送数据，理论上这个仿真中只涉及到TCP/IP的基本存储功能，即应该包括hdr_mac, hdr_ll, hdr_ip,hdr_cmn, hdr_tcp，但是实际得到的包应该如下：

 

hdr_ip

hdr_ll

hdr_cmn

hdr_tcp

hdr_XCP

........

data_

那如何访问数据包中相应的子包呢，可以查看packet.h开始，发现定义了许多宏，如：

#defineHDR_CMN(p)     (hdr_cmn::access(p))
#defineHDR_ARP(p)     (hdr_arp::access(p))
#defineHDR_MAC(p)     (hdr_mac::access(p))
#define HDR_MAC802_11(p) ((hdr_mac802_11*)hdr_mac::access(p))
#define HDR_MAC_TDMA(p) ((hdr_mac_tdma *)hdr_mac::access(p))
#defineHDR_SMAC(p)    ((hdr_smac *)hdr_mac::access(p))
#defineHDR_LL(p)      (hdr_ll::access(p))
#defineHDR_HDLC(p)    ((hdr_hdlc *)hdr_ll::access(p))
#defineHDR_IP(p)      (hdr_ip::access(p))
#defineHDR_RTP(p)     (hdr_rtp::access(p))
#defineHDR_TCP(p)     (hdr_tcp::access(p))
#defineHDR_SCTP(p)    (hdr_sctp::access(p))
#defineHDR_SR(p)      (hdr_sr::access(p))
#defineHDR_TFRC(p)    (hdr_tfrc::access(p))
#defineHDR_TORA(p)    (hdr_tora::access(p))

就是用来访问相应的字段，比如HDR_MAC(p)，即可以取得对应的域。而具体如何取则由hdr_***::access来实现。access是每一种包头都包含的，比如hdr_cmn:

  inline static hdr_cmn* access(const Packet* p){
    return (hdr_cmn*)p->access(offset_);
  }

再看看包Packet的access函数，如下：

 inline unsigned char* access(int off) const{
      if (off <0)     abort();
      return(&bits_[off]);
 }

到此，基本明白了，offset其实是每个包相对于packet开始的偏移值，通过这个偏移值即可确定其在整个包中的地址。

 

现在问题是，在哪个地方设置这个offset值，搜索了所有的C++代码也没看到对每个包的offset的初始化。经过仔细比较，发现在tcl/lib/ns-lib.tcl中的Simulator初始化init过程中有一个调用：

  $self create_packetformat

接着在ns-packet.tcl中找到了这个函数：

Simulator instproc create_packetformat { } {
 PacketHeaderManager instvar tab_
 set pm [new PacketHeaderManager]
 foreach cl [PacketHeader info subclass] {
  if [info exists tab_($cl)]{
   set off [$pmallochdr $cl]
   $cl offset$off
  }
 }
 $self set packetManager_ $pm
}

 应该是在这个地方实现了offset的初始化，换句话说，是在这个地方指定每个包的offset。设置断点并打印，证实了我的猜想：

PacketHeader/PBC has offset 0 
PacketHeader/LRWPAN has offset 8 
PacketHeader/XCP has offset 216 
PacketHeader/PGM has offset 272 
PacketHeader/PGM_SPM has offset 288 
PacketHeader/PGM_NAK has offset 296 
PacketHeader/Pushback has offset 312 
PacketHeader/NV has offset 320 
PacketHeader/LDP has offset 328 
PacketHeader/MPLS has offset 368 
PacketHeader/rtProtoLS has offset 392 
PacketHeader/Ping has offset 400 
PacketHeader/TFRC has offset 424 
PacketHeader/TFRC_ACK has offset 480 
PacketHeader/Diffusion has offset 544 
PacketHeader/RAP has offset 736 
PacketHeader/AOMDV has offset 760 
PacketHeader/AODV has offset 1568
PacketHeader/SR has offset 2376 
PacketHeader/TORA has offset 3088 
PacketHeader/IMEP has offset 3120 
PacketHeader/ARP has offset 3632 
PacketHeader/MIP has offset 3664 
PacketHeader/IPinIP has offset 3696 
PacketHeader/LL has offset 3704 
PacketHeader/Mac has offset 3736 
PacketHeader/Encap has offset 3776 
PacketHeader/HttpInval has offset 3784 
PacketHeader/SRMEXT has offset 3792 
PacketHeader/SRM has offset 3800 
PacketHeader/aSRM has offset 3816 
PacketHeader/mcastCtrl has offset 3824 
PacketHeader/CtrMcast has offset 3848 
PacketHeader/rtProtoDV has offset 3864 
PacketHeader/GAF has offset 3872 
PacketHeader/Snoop has offset 3880 
PacketHeader/SCTP has offset 3904 
PacketHeader/TCPA has offset 3912
PacketHeader/TCP has offset 3928 
PacketHeader/IVS has offset 4008 
PacketHeader/RTP has offset 4040 
PacketHeader/Message has offset 4056 
PacketHeader/Resv has offset 4120 
PacketHeader/QS has offset 4136 
PacketHeader/UMP has offset 4152 
PacketHeader/Src_rt has offset 4168 
PacketHeader/IP has offset 4248 
PacketHeader/Common has offset 4280 
PacketHeader/Flags has offset 4384

实际上，每个包的offset应该为sizeof(hdr_***)。为了验证，写了一个程序测试。比如hdr_ip,

typedef int32_t nsaddr_t; 
typedef int32_t nsmask_t;

struct ns_addr_t {
 int32_t addr_;
 int32_t port_;
};

struct hdr_ip{
 
 ns_addr_t src_;
 ns_addr_t dst_;
 int  ttl_;

 
  u_int16_t sport_;
  u_int16_t dport_;
 
 
 int  fid_; 
 int  prio_;

 static int offset_;
 inline static int& offset() {return offset_; }

 
 ns_addr_t& src() { return (src_);}
 nsaddr_t& saddr() { return(src_.addr_); }
       int32_t& sport() { return src_.port_;}

 ns_addr_t& dst() { return (dst_);}
 nsaddr_t& daddr() { return(dst_.addr_); }
       int32_t& dport() { return dst_.port_;}
 int& ttl() { return (ttl_);}
 
 int& flowid() { return (fid_);}
 int& prio() { return (prio_);}
};

int main(intargc, char* argv[])
{

    printf("ip:%d\n", sizeof(hdr_ip));
   printf("tora: %d\n", sizeof(hdr_tora));

   getch();
    return 0;

}

程序的输出验证我的猜想。

packet.h的下面定义则实现了两者间的绑定：

class PacketHeaderClass : public TclClass {
protected:
 

PacketHeaderClass(const char* classname, inthdrsize);
  virtual int method(int argc, const char*const*argv);
  void field_offset(const char* fieldname, intoffset);
 inline void bind_offset(int* off) {offset_ = off; }
 inline void offset(int* off) {offset_=off;}
  inthdrlen_;    // # of bytes for thisheader
  int*offset_;    // offset for thisheader

public:

  virtual void bind();
  virtual void export_offsets();
  TclObject* create(int argc, const char*const*argv);

};