组播发送的流程

组播都是udp报文发送，发送用的地方在选路

ns3有组播的例子，加默认路由，静态组播路由，这几个例子我没看懂。FIXME

这几个例子与我想要的区别是，我要针对quagga加增内容，改造，而不能像ns3的组播路由那样自由发挥FIXME

现状：采用dce直接路由组播的结果是：

assert failed. cond="oif", msg="Try to send on link-local multicast address, and no interface index is given!", file=../model/ipv4-dce-routing.cc, line=409
terminate called without an active exception

需要变更成组播顺利发送

另一个想法：如果直接使用linux栈（日本人做的那个ns-next-sim协议栈），可能是一切工作正常的。那个栈的劣势是和ns3其他部分有些脱离。FIXME

回来处理那个assert

Breakpoint 4, ns3::Ipv4DceRouting::LookupStatic (this=0x6c1d40, dest=..., oif=...) at ../model/ipv4-dce-routing.cc:409
409          NS_ASSERT_MSG (oif, "Try to send on link-local multicast address, and no interface index is given!");
(gdb) bt
#0  ns3::Ipv4DceRouting::LookupStatic (this=0x6c1d40, dest=..., oif=...) at ../model/ipv4-dce-routing.cc:409
#1  0x00007ffff7ac912b in ns3::Ipv4DceRouting::RouteOutput (this=0x6c1d40, p=..., header=..., oif=..., sockerr=@0x7fffffffcfec: 32767) at ../model/ipv4-dce-routing.cc:664
#2  0x00007ffff6d1170a in ns3::UdpSocketImpl::DoSendTo (this=0x7fffdc006200, p=..., dest=..., port=646) at ../src/internet/model/udp-socket-impl.cc:621
#3  0x00007ffff6d140ca in ns3::UdpSocketImpl::SendTo (this=0x7fffdc006200, p=..., flags=64, address=...) at ../src/internet/model/udp-socket-impl.cc:807
#4  0x00007ffff7a44df4 in ns3::UnixDatagramSocketFd::MainSendTo (this=0x7fffdc007ed0, r=0x7ffff0a29e6c, p=..., f=64, ad=...) at ../model/unix-datagram-socket-fd.cc:510
#5  0x00007ffff7a45787 in ns3::EventImpl* ns3::MakeEvent, unsigned int, ns3::Address), ns3::UnixDatagramSocketFd*, int*, ns3::Ptr, int, ns3::Address>(void (ns3::UnixDatagramSocketFd::*)(int*, ns3::Ptr, unsigned int, ns3::Address), ns3::UnixDatagramSocketFd*, int*, ns3::Ptr, int, ns3::Address)::EventMemberImpl4::Notify() (this=0x7fffdc024bf0) at /home/z/new/dce/build/include/ns3.23/ns3/make-event.h:452
#6  0x00007ffff733b5e7 in ns3::EventImpl::Invoke (this=0x7fffdc024bf0) at ../src/core/model/event-impl.cc:51
#7  0x00007ffff7a8a5ba in ns3::TaskManager::Schedule (this=0x67dc00) at ../model/task-manager.cc:441
#8  0x00007ffff7a8c5ad in ns3::EventImpl* ns3::MakeEvent(void (ns3::TaskManager::*)(), ns3::TaskManager*)::EventMemberImpl0::Notify() (this=0x64f120) at /home/z/new/dce/build/include/ns3.23/ns3/make-event.h:323
#9  0x00007ffff733b5e7 in ns3::EventImpl::Invoke (this=0x64f120) at ../src/core/model/event-impl.cc:51
#10 0x00007ffff73402ac in ns3::DefaultSimulatorImpl::ProcessOneEvent (this=0x6b1870) at ../src/core/model/default-simulator-impl.cc:147
#11 0x00007ffff7340638 in ns3::DefaultSimulatorImpl::Run (this=0x6b1870) at ../src/core/model/default-simulator-impl.cc:200
#12 0x00007ffff733c378 in ns3::Simulator::Run () at ../src/core/model/simulator.cc:204
#13 0x000000000040cd90 in main (argc=1, argv=0x7fffffffdc08) at ../myscripts/ns-3-dce-quagga/example/dce-quagga-mpls.cc:219
(gdb)

发送是由ldp hello触发的，

查看

#2  0x00007ffff6d1170a in ns3::UdpSocketImpl::DoSendTo (this=0x7fffdc006200, p=..., dest=..., port=646) at ../src/internet/model/udp-socket-impl.cc:621

发现m_boundnetdevice是0，表示并没有发生过bind操作。

查这篇文章，看到client和server都是有bind的。

http://www.cnblogs.com/hnrainll/archive/2011/11/07/2238671.html

查quagga，发现hello有绑定的

mpls_socket_bind

mpls_socket_udp_sendto

如此找bind，linux kernel的处理代码是inet_bind，对于组播来说，这个绑定也许真没用FIXME

～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～～