(gdb) bt
#0 0x00007f475cb540d5 in raise () from /lib/x86_64-linux-gnu/libc.so.6
#1 0x00007f475cb5783b in abort () from /lib/x86_64-linux-gnu/libc.so.6
#2 0x00007f475cb4cd9e in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#3 0x00007f475cb4ce42 in __assert_fail () from /lib/x86_64-linux-gnu/libc.so.6
#4 0x00000000015ba15a in FlowEntry::SetRpfNH (this=0x7f46d40048c0, ft=0x7f4754051550, rt=0x7f46e00194a0)
at controller/src/vnsw/agent/pkt/flow_table.cc:1286
#5 0x00000000015ba9d7 in FlowEntry::InitFwdFlow (this=0x7f46d40048c0, info=0x7f4700f7a8a0, pkt=0x3088a50, ctrl=0x7f4700f7a7e0,
rev_ctrl=0x7f4700f7a820) at controller/src/vnsw/agent/pkt/flow_table.cc:1423
#6 0x00000000015f2259 in PktFlowInfo::Add (this=0x7f4700f7a8a0, pkt=0x3088a50, in=0x7f4700f7a7e0, out=0x7f4700f7a820)
at controller/src/vnsw/agent/pkt/pkt_flow_info.cc:1399
#7 0x00000000015fce77 in FlowHandler::Run (this=0x7f46d401e3c0) at controller/src/vnsw/agent/pkt/flow_handler.cc:80
#8 0x00000000015f8bd5 in Proto::ProcessProto (this=0x7f4754037d40, msg_info=...) at controller/src/vnsw/agent/pkt/proto.cc:44
#9 0x00000000015faf0e in boost::_mfi::mf1<bool, Proto, boost::shared_ptr<PktInfo> >::operator() (this=0x7f4700f7abe8,
p=0x7f4754037d40, a1=...) at build/include/boost/bind/mem_fn_template.hpp:165
#10 0x00000000015fa920 in boost::_bi::list2<boost::_bi::value<Proto*>, boost::arg<1> >::operator()<bool, boost::_mfi::mf1<bool, Proto, boost::shared_ptr<PktInfo> >, boost::_bi::list1<boost::shared_ptr<PktInfo>&> > (this=0x7f4700f7abf8, f=..., a=...)
at build/include/boost/bind/bind.hpp:303
#11 0x00000000015fa308 in boost::_bi::bind_t<bool, boost::_mfi::mf1<bool, Proto, boost::shared_ptr<PktInfo> >, boost::_bi::list2<boost::_bi::value<Proto*>, boost::arg<1> > >::operator()<boost::shared_ptr<PktInfo> > (this=0x7f4700f7abe8, a1=...)
at build/include/boost/bind/bind_template.hpp:32
#12 0x00000000015f9ebf in boost::detail::function::function_obj_invoker1<boost::_bi::bind_t<bool, boost::_mfi::mf1<bool, Proto, boost::shared_ptr<PktInfo> >, boost::_bi::list2<boost::_bi::value<Proto*>, boost::arg<1> > >, bool, boost::shared_ptr<PktInfo> >::invoke (
function_obj_ptr=..., a0=...) at build/include/boost/function/function_template.hpp:132
#13 0x00000000015ed361 in boost::function1<bool, boost::shared_ptr<PktInfo> >::operator() (this=0x7f4700f7abe0, a0=...)
at build/include/boost/function/function_template.hpp:763
#14 0x00000000015fb2f6 in QueueTaskRunner<boost::shared_ptr<PktInfo>, WorkQueue<boost::shared_ptr<PktInfo> > >::RunQueue (
this=0x3094800) at controller/src/base/queue_task.h:81
#15 0x00000000015fb12e in QueueTaskRunner<boost::shared_ptr<PktInfo>, WorkQueue<boost::shared_ptr<PktInfo> > >::Run (this=0x3094800)
at controller/src/base/queue_task.h:64
#16 0x0000000001bfdf21 in TaskImpl::execute (this=0x3081a40) at controller/src/base/task.cc:238
#17 0x00007f475d71eece in ?? () from /usr/lib/libtbb_debug.so.2
#18 0x00007f475d715e0b in ?? () from /usr/lib/libtbb_debug.so.2
#19 0x00007f475d7146f2 in ?? () from /usr/lib/libtbb_debug.so.2
#20 0x00007f475d70f3ce in ?? () from /usr/lib/libtbb_debug.so.2
#21 0x00007f475d70f270 in ?? () from /usr/lib/libtbb_debug.so.2
#22 0x00007f475d948e9a in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0
#23 0x00007f475cc1138d in clone () from /lib/x86_64-linux-gnu/libc.so.6
#24 0x0000000000000000 in ?? ()
The setup used here is,
1. VN is created with subnet 1.1.1.0/24. VXLAN is 2000
2. There are 2 MX in network. Gateway ip 1.1.1.1 is assigned as VRRP-IP on MX
3. The MAC for VRRP IP is 00:00:5E:00:01:01
4. Due to 2 MX, route for 1.1.1.0/24 is ECMP with the 2 MX
5. Baremetal with IP 1.1.1.5 pings Gateway IP 1.1.1.1
6. The ToR does not have FDB for 00:00:5E:00:01:01. As a result, it sends packet to TSN for flooding
VRouter on TSN does following:
1. VRouter receives packets on VXLAN 2000
2. VRouter does route lookup for src-ip (1.1.1.5)
3. Route lookup for src-ip 1.1.1.5 hits ECMP route 1.1.1.0/24. ECMP NH contains the two MX
4. VRouter traps packet for flow-processing
Agent on TSN does following:
1. Agent tries to create L2 Flow
2. Agent tries to set RPF NH for the flow
3. The RPF lookup for 1.1.1.5 hits ECMP Route
Problem:
Agent is doing IP lookup for RPF checks for packet from Baremetal. In L2 Flow setup, agent is not expecting ECMP-NH in RPF processing.
Fix:
When flow is being setup for BMS, agent is not aware of IP for BMS. So, RPF check should be based on Src-MAC instead of Src-IP