#!/usr/bin/env python3 import socket from socket import inet_pton, inet_ntop import unittest from parameterized import parameterized import scapy.layers.inet6 as inet6 from scapy.layers.inet import UDP, IP from scapy.contrib.mpls import MPLS from scapy.layers.inet6 import ( IPv6, ICMPv6ND_NS, ICMPv6ND_RS, ICMPv6ND_RA, ICMPv6NDOptMTU, ICMPv6NDOptSrcLLAddr, ICMPv6NDOptPrefixInfo, ICMPv6ND_NA, ICMPv6NDOptDstLLAddr, ICMPv6DestUnreach, icmp6types, ICMPv6TimeExceeded, ICMPv6EchoRequest, ICMPv6EchoReply, IPv6ExtHdrHopByHop, ICMPv6MLReport2, ) from scapy.layers.l2 import Ether, Dot1Q, GRE from scapy.packet import Raw from scapy.utils6 import ( in6_getnsma, in6_getnsmac, in6_ptop, in6_islladdr, ) from framework import VppTestCase from asfframework import VppTestRunner, tag_run_solo from util import ppp, ip6_normalize, mk_ll_addr from vpp_papi import VppEnum from vpp_ip import VppIpPuntPolicer, VppIpPuntRedirect, VppIpPathMtu from vpp_ip_route import ( VppIpRoute, VppRoutePath, find_route, VppIpMRoute, VppMRoutePath, VppMplsIpBind, VppMplsRoute, VppMplsTable, VppIpTable, FibPathType, FibPathProto, VppIpInterfaceAddress, find_route_in_dump, find_mroute_in_dump, VppIp6LinkLocalAddress, VppIpRouteV2, ) from vpp_neighbor import find_nbr, VppNeighbor from vpp_ipip_tun_interface import VppIpIpTunInterface from vpp_pg_interface import is_ipv6_misc from vpp_sub_interface import VppSubInterface, VppDot1QSubint from vpp_policer import VppPolicer, PolicerAction from ipaddress import IPv6Network, IPv6Address from vpp_gre_interface import VppGreInterface from vpp_teib import VppTeib from config import config AF_INET6 = socket.AF_INET6 try: text_type = unicode except NameError: text_type = str NUM_PKTS = 67 class TestIPv6ND(VppTestCase): def validate_ra(self, intf, rx, dst_ip=None): if not dst_ip: dst_ip = intf.remote_ip6 # unicasted packets must come to the unicast mac self.assertEqual(rx[Ether].dst, intf.remote_mac) # and from the router's MAC self.assertEqual(rx[Ether].src, intf.local_mac) # the rx'd RA should be addressed to the sender's source self.assertTrue(rx.haslayer(ICMPv6ND_RA)) self.assertEqual(in6_ptop(rx[IPv6].dst), in6_ptop(dst_ip)) # and come from the router's link local self.assertTrue(in6_islladdr(rx[IPv6].src)) self.assertEqual(in6_ptop(rx[IPv6].src), in6_ptop(mk_ll_addr(intf.local_mac))) def validate_na(self, intf, rx, dst_ip=None, tgt_ip=None): if not dst_ip: dst_ip = intf.remote_ip6 if not tgt_ip: dst_ip = intf.local_ip6 # unicasted packets must come to the unicast mac self.assertEqual(rx[Ether].dst, intf.remote_mac) # and from the router's MAC self.assertEqual(rx[Ether].src, intf.local_mac) # the rx'd NA should be addressed to the sender's source self.assertTrue(rx.haslayer(ICMPv6ND_NA)) self.assertEqual(in6_ptop(rx[IPv6].dst), in6_ptop(dst_ip)) # and come from the target address self.assertEqual(in6_ptop(rx[IPv6].src), in6_ptop(tgt_ip)) # Dest link-layer options should have the router's MAC dll = rx[ICMPv6NDOptDstLLAddr] self.assertEqual(dll.lladdr, intf.local_mac) def validate_ns(self, intf, rx, tgt_ip): nsma = in6_getnsma(inet_pton(AF_INET6, tgt_ip)) dst_ip = inet_ntop(AF_INET6, nsma) # NS is broadcast self.assertEqual(rx[Ether].dst, in6_getnsmac(nsma)) # and from the router's MAC self.assertEqual(rx[Ether].src, intf.local_mac) # the rx'd NS should be addressed to an mcast address # derived from the target address self.assertEqual(in6_ptop(rx[IPv6].dst), in6_ptop(dst_ip)) # expect the tgt IP in the NS header ns = rx[ICMPv6ND_NS] self.assertEqual(in6_ptop(ns.tgt), in6_ptop(tgt_ip)) # packet is from the router's local address self.assertEqual(in6_ptop(rx[IPv6].src), intf.local_ip6_ll) # Src link-layer options should have the router's MAC sll = rx[ICMPv6NDOptSrcLLAddr] self.assertEqual(sll.lladdr, intf.local_mac) def send_and_expect_ra( self, intf, pkts, remark, dst_ip=None, filter_out_fn=is_ipv6_misc ): intf.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = intf.get_capture(1, filter_out_fn=filter_out_fn) self.assertEqual(len(rx), 1) rx = rx[0] self.validate_ra(intf, rx, dst_ip) def send_and_expect_na( self, intf, pkts, remark, dst_ip=None, tgt_ip=None, filter_out_fn=is_ipv6_misc ): intf.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = intf.get_capture(1, filter_out_fn=filter_out_fn) self.assertEqual(len(rx), 1) rx = rx[0] self.validate_na(intf, rx, dst_ip, tgt_ip) def send_and_expect_ns( self, tx_intf, rx_intf, pkts, tgt_ip, filter_out_fn=is_ipv6_misc ): self.vapi.cli("clear trace") tx_intf.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = rx_intf.get_capture(1, filter_out_fn=filter_out_fn) self.assertEqual(len(rx), 1) rx = rx[0] self.validate_ns(rx_intf, rx, tgt_ip) def verify_ip(self, rx, smac, dmac, sip, dip): ether = rx[Ether] self.assertEqual(ether.dst, dmac) self.assertEqual(ether.src, smac) ip = rx[IPv6] self.assertEqual(ip.src, sip) self.assertEqual(ip.dst, dip) def get_ip6_nd_rx_requests(self, itf): """Get IP6 ND RX request stats for and interface""" return self.statistics["/net/ip6-nd/rx/requests"][:, itf.sw_if_index].sum() def get_ip6_nd_tx_requests(self, itf): """Get IP6 ND TX request stats for and interface""" return self.statistics["/net/ip6-nd/tx/requests"][:, itf.sw_if_index].sum() def get_ip6_nd_rx_replies(self, itf): """Get IP6 ND RX replies stats for and interface""" return self.statistics["/net/ip6-nd/rx/replies"][:, itf.sw_if_index].sum() def get_ip6_nd_tx_replies(self, itf): """Get IP6 ND TX replies stats for and interface""" return self.statistics["/net/ip6-nd/tx/replies"][:, itf.sw_if_index].sum() @tag_run_solo class TestIPv6(TestIPv6ND): """IPv6 Test Case""" @classmethod def setUpClass(cls): super(TestIPv6, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPv6, cls).tearDownClass() def setUp(self): """ Perform test setup before test case. **Config:** - create 3 pg interfaces - untagged pg0 interface - Dot1Q subinterface on pg1 - Dot1AD subinterface on pg2 - setup interfaces: - put it into UP state - set IPv6 addresses - resolve neighbor address using NDP - configure 200 fib entries :ivar list interfaces: pg interfaces and subinterfaces. :ivar dict flows: IPv4 packet flows in test. *TODO:* Create AD sub interface """ super(TestIPv6, self).setUp() # create 3 pg interfaces self.create_pg_interfaces(range(3)) # create 2 subinterfaces for p1 and pg2 self.sub_interfaces = [ VppDot1QSubint(self, self.pg1, 100), VppDot1QSubint(self, self.pg2, 200), # TODO: VppDot1ADSubint(self, self.pg2, 200, 300, 400) ] # packet flows mapping pg0 -> pg1.sub, pg2.sub, etc. self.flows = dict() self.flows[self.pg0] = [self.pg1.sub_if, self.pg2.sub_if] self.flows[self.pg1.sub_if] = [self.pg0, self.pg2.sub_if] self.flows[self.pg2.sub_if] = [self.pg0, self.pg1.sub_if] # packet sizes self.pg_if_packet_sizes = [64, 1500, 9020] self.interfaces = list(self.pg_interfaces) self.interfaces.extend(self.sub_interfaces) # setup all interfaces for i in self.interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): """Run standard test teardown and log ``show ip6 neighbors``.""" for i in reversed(self.interfaces): i.unconfig_ip6() i.admin_down() for i in self.sub_interfaces: i.remove_vpp_config() super(TestIPv6, self).tearDown() if not self.vpp_dead: self.logger.info(self.vapi.cli("show ip6 neighbors")) # info(self.vapi.cli("show ip6 fib")) # many entries def modify_packet(self, src_if, packet_size, pkt): """Add load, set destination IP and extend packet to required packet size for defined interface. :param VppInterface src_if: Interface to create packet for. :param int packet_size: Required packet size. :param Scapy pkt: Packet to be modified. """ dst_if_idx = int(packet_size / 10 % 2) dst_if = self.flows[src_if][dst_if_idx] info = self.create_packet_info(src_if, dst_if) payload = self.info_to_payload(info) p = pkt / Raw(payload) p[IPv6].dst = dst_if.remote_ip6 info.data = p.copy() if isinstance(src_if, VppSubInterface): p = src_if.add_dot1_layer(p) self.extend_packet(p, packet_size) return p def create_stream(self, src_if): """Create input packet stream for defined interface. :param VppInterface src_if: Interface to create packet stream for. """ hdr_ext = 4 if isinstance(src_if, VppSubInterface) else 0 pkt_tmpl = ( Ether(dst=src_if.local_mac, src=src_if.remote_mac) / IPv6(src=src_if.remote_ip6) / inet6.UDP(sport=1234, dport=1234) ) pkts = [ self.modify_packet(src_if, i, pkt_tmpl) for i in range(self.pg_if_packet_sizes[0], self.pg_if_packet_sizes[1], 10) ] pkts_b = [ self.modify_packet(src_if, i, pkt_tmpl) for i in range( self.pg_if_packet_sizes[1] + hdr_ext, self.pg_if_packet_sizes[2] + hdr_ext, 50, ) ] pkts.extend(pkts_b) return pkts def verify_capture(self, dst_if, capture): """Verify captured input packet stream for defined interface. :param VppInterface dst_if: Interface to verify captured packet stream for. :param list capture: Captured packet stream. """ self.logger.info("Verifying capture on interface %s" % dst_if.name) last_info = dict() for i in self.interfaces: last_info[i.sw_if_index] = None is_sub_if = False dst_sw_if_index = dst_if.sw_if_index if hasattr(dst_if, "parent"): is_sub_if = True for packet in capture: if is_sub_if: # Check VLAN tags and Ethernet header packet = dst_if.remove_dot1_layer(packet) self.assertTrue(Dot1Q not in packet) try: ip = packet[IPv6] udp = packet[inet6.UDP] payload_info = self.payload_to_info(packet[Raw]) packet_index = payload_info.index self.assertEqual(payload_info.dst, dst_sw_if_index) self.logger.debug( "Got packet on port %s: src=%u (id=%u)" % (dst_if.name, payload_info.src, packet_index) ) next_info = self.get_next_packet_info_for_interface2( payload_info.src, dst_sw_if_index, last_info[payload_info.src] ) last_info[payload_info.src] = next_info self.assertTrue(next_info is not None) self.assertEqual(packet_index, next_info.index) saved_packet = next_info.data # Check standard fields self.assertEqual(ip.src, saved_packet[IPv6].src) self.assertEqual(ip.dst, saved_packet[IPv6].dst) self.assertEqual(udp.sport, saved_packet[inet6.UDP].sport) self.assertEqual(udp.dport, saved_packet[inet6.UDP].dport) except: self.logger.error(ppp("Unexpected or invalid packet:", packet)) raise for i in self.interfaces: remaining_packet = self.get_next_packet_info_for_interface2( i.sw_if_index, dst_sw_if_index, last_info[i.sw_if_index] ) self.assertTrue( remaining_packet is None, "Interface %s: Packet expected from interface %s " "didn't arrive" % (dst_if.name, i.name), ) def test_next_header_anomaly(self): """IPv6 next header anomaly test Test scenario: - ipv6 next header field = Fragment Header (44) - next header is ICMPv6 Echo Request - wait for reassembly """ pkt = ( Ether(src=self.pg0.local_mac, dst=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg0.local_ip6, nh=44) / ICMPv6EchoRequest() ) self.pg0.add_stream(pkt) self.pg_start() # wait for reassembly self.sleep(10) def test_fib(self): """IPv6 FIB test Test scenario: - Create IPv6 stream for pg0 interface - Create IPv6 tagged streams for pg1's and pg2's subinterface. - Send and verify received packets on each interface. """ pkts = self.create_stream(self.pg0) self.pg0.add_stream(pkts) for i in self.sub_interfaces: pkts = self.create_stream(i) i.parent.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() pkts = self.pg0.get_capture() self.verify_capture(self.pg0, pkts) for i in self.sub_interfaces: pkts = i.parent.get_capture() self.verify_capture(i, pkts) def test_ns(self): """IPv6 Neighbour Solicitation Exceptions Test scenario: - Send an NS Sourced from an address not covered by the link sub-net - Send an NS to an mcast address the router has not joined - Send NS for a target address the router does not onn. """ n_rx_req_pg0 = self.get_ip6_nd_rx_requests(self.pg0) n_tx_rep_pg0 = self.get_ip6_nd_tx_replies(self.pg0) # # An NS from a non link source address # nsma = in6_getnsma(inet_pton(AF_INET6, self.pg0.local_ip6)) d = inet_ntop(AF_INET6, nsma) p = ( Ether(dst=in6_getnsmac(nsma)) / IPv6(dst=d, src="2002::2") / ICMPv6ND_NS(tgt=self.pg0.local_ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) pkts = [p] self.send_and_assert_no_replies( self.pg0, pkts, "No response to NS source by address not on sub-net" ) self.assert_equal(self.get_ip6_nd_rx_requests(self.pg0), n_rx_req_pg0 + 1) # # An NS for sent to a solicited mcast group the router is # not a member of FAILS # if 0: nsma = in6_getnsma(inet_pton(AF_INET6, "fd::ffff")) d = inet_ntop(AF_INET6, nsma) p = ( Ether(dst=in6_getnsmac(nsma)) / IPv6(dst=d, src=self.pg0.remote_ip6) / ICMPv6ND_NS(tgt=self.pg0.local_ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) pkts = [p] self.send_and_assert_no_replies( self.pg0, pkts, "No response to NS sent to unjoined mcast address" ) # # An NS whose target address is one the router does not own # nsma = in6_getnsma(inet_pton(AF_INET6, self.pg0.local_ip6)) d = inet_ntop(AF_INET6, nsma) p = ( Ether(dst=in6_getnsmac(nsma)) / IPv6(dst=d, src=self.pg0.remote_ip6) / ICMPv6ND_NS(tgt="fd::ffff") / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) pkts = [p] self.send_and_assert_no_replies( self.pg0, pkts, "No response to NS for unknown target" ) self.assert_equal(self.get_ip6_nd_rx_requests(self.pg0), n_rx_req_pg0 + 2) # # A neighbor entry that has no associated FIB-entry # self.pg0.generate_remote_hosts(4) nd_entry = VppNeighbor( self, self.pg0.sw_if_index, self.pg0.remote_hosts[2].mac, self.pg0.remote_hosts[2].ip6, is_no_fib_entry=1, ) nd_entry.add_vpp_config() # # check we have the neighbor, but no route # self.assertTrue( find_nbr(self, self.pg0.sw_if_index, self.pg0._remote_hosts[2].ip6) ) self.assertFalse(find_route(self, self.pg0._remote_hosts[2].ip6, 128)) # # send an NS from a link local address to the interface's global # address # p = ( Ether(dst=in6_getnsmac(nsma), src=self.pg0.remote_mac) / IPv6(dst=d, src=self.pg0._remote_hosts[2].ip6_ll) / ICMPv6ND_NS(tgt=self.pg0.local_ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) self.send_and_expect_na( self.pg0, p, "NS from link-local", dst_ip=self.pg0._remote_hosts[2].ip6_ll, tgt_ip=self.pg0.local_ip6, ) self.assert_equal(self.get_ip6_nd_rx_requests(self.pg0), n_rx_req_pg0 + 3) self.assert_equal(self.get_ip6_nd_tx_replies(self.pg0), n_tx_rep_pg0 + 1) # # we should have learned an ND entry for the peer's link-local # but not inserted a route to it in the FIB # self.assertTrue( find_nbr(self, self.pg0.sw_if_index, self.pg0._remote_hosts[2].ip6_ll) ) self.assertFalse(find_route(self, self.pg0._remote_hosts[2].ip6_ll, 128)) # # An NS to the router's own Link-local # p = ( Ether(dst=in6_getnsmac(nsma), src=self.pg0.remote_mac) / IPv6(dst=d, src=self.pg0._remote_hosts[3].ip6_ll) / ICMPv6ND_NS(tgt=self.pg0.local_ip6_ll) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) self.send_and_expect_na( self.pg0, p, "NS to/from link-local", dst_ip=self.pg0._remote_hosts[3].ip6_ll, tgt_ip=self.pg0.local_ip6_ll, ) # # do not respond to a NS for the peer's address # p = ( Ether(dst=in6_getnsmac(nsma), src=self.pg0.remote_mac) / IPv6(dst=d, src=self.pg0._remote_hosts[3].ip6_ll) / ICMPv6ND_NS(tgt=self.pg0._remote_hosts[3].ip6_ll) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) self.send_and_assert_no_replies(self.pg0, p) # # we should have learned an ND entry for the peer's link-local # but not inserted a route to it in the FIB # self.assertTrue( find_nbr(self, self.pg0.sw_if_index, self.pg0._remote_hosts[3].ip6_ll) ) self.assertFalse(find_route(self, self.pg0._remote_hosts[3].ip6_ll, 128)) def test_nd_incomplete(self): """IP6-ND Incomplete""" self.pg1.generate_remote_hosts(3) p0 = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg1.remote_hosts[1].ip6) / UDP(sport=1234, dport=1234) / Raw() ) # # a packet to an unresolved destination generates an ND request # n_tx_req_pg1 = self.get_ip6_nd_tx_requests(self.pg1) self.send_and_expect_ns(self.pg0, self.pg1, p0, self.pg1.remote_hosts[1].ip6) self.assert_equal(self.get_ip6_nd_tx_requests(self.pg1), n_tx_req_pg1 + 1) # # a reply to the request # self.assert_equal(self.get_ip6_nd_rx_replies(self.pg1), 0) na = ( Ether(dst=self.pg1.local_mac, src=self.pg1.remote_mac) / IPv6(dst=self.pg1.local_ip6, src=self.pg1.remote_hosts[1].ip6) / ICMPv6ND_NA(tgt=self.pg1.remote_hosts[1].ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg1.remote_hosts[1].mac) ) self.send_and_assert_no_replies(self.pg1, [na]) self.assert_equal(self.get_ip6_nd_rx_replies(self.pg1), 1) def test_ns_duplicates(self): """ND Duplicates""" # # Generate some hosts on the LAN # self.pg1.generate_remote_hosts(3) # # Add host 1 on pg1 and pg2 # ns_pg1 = VppNeighbor( self, self.pg1.sw_if_index, self.pg1.remote_hosts[1].mac, self.pg1.remote_hosts[1].ip6, ) ns_pg1.add_vpp_config() ns_pg2 = VppNeighbor( self, self.pg2.sw_if_index, self.pg2.remote_mac, self.pg1.remote_hosts[1].ip6, ) ns_pg2.add_vpp_config() # # IP packet destined for pg1 remote host arrives on pg1 again. # p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg1.remote_hosts[1].ip6) / inet6.UDP(sport=1234, dport=1234) / Raw() ) self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx1 = self.pg1.get_capture(1) self.verify_ip( rx1[0], self.pg1.local_mac, self.pg1.remote_hosts[1].mac, self.pg0.remote_ip6, self.pg1.remote_hosts[1].ip6, ) # # remove the duplicate on pg1 # packet stream should generate NSs out of pg1 # ns_pg1.remove_vpp_config() self.send_and_expect_ns(self.pg0, self.pg1, p, self.pg1.remote_hosts[1].ip6) # # Add it back # ns_pg1.add_vpp_config() self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx1 = self.pg1.get_capture(1) self.verify_ip( rx1[0], self.pg1.local_mac, self.pg1.remote_hosts[1].mac, self.pg0.remote_ip6, self.pg1.remote_hosts[1].ip6, ) def validate_ra(self, intf, rx, dst_ip=None, src_ip=None, mtu=9000, pi_opt=None): if not dst_ip: dst_ip = intf.remote_ip6 if not src_ip: src_ip = mk_ll_addr(intf.local_mac) # unicasted packets must come to the unicast mac self.assertEqual(rx[Ether].dst, intf.remote_mac) # and from the router's MAC self.assertEqual(rx[Ether].src, intf.local_mac) # the rx'd RA should be addressed to the sender's source self.assertTrue(rx.haslayer(ICMPv6ND_RA)) self.assertEqual(in6_ptop(rx[IPv6].dst), in6_ptop(dst_ip)) # and come from the router's link local self.assertTrue(in6_islladdr(rx[IPv6].src)) self.assertEqual(in6_ptop(rx[IPv6].src), in6_ptop(src_ip)) # it should contain the links MTU ra = rx[ICMPv6ND_RA] self.assertEqual(ra[ICMPv6NDOptMTU].mtu, mtu) # it should contain the source's link layer address option sll = ra[ICMPv6NDOptSrcLLAddr] self.assertEqual(sll.lladdr, intf.local_mac) if not pi_opt: # the RA should not contain prefix information self.assertFalse(ra.haslayer(ICMPv6NDOptPrefixInfo)) else: raos = rx.getlayer(ICMPv6NDOptPrefixInfo, 1) # the options are nested in the scapy packet in way that i cannot # decipher how to decode. this 1st layer of option always returns # nested classes, so a direct obj1=obj2 comparison always fails. # however, the getlayer(.., 2) does give one instance. # so we cheat here and construct a new opt instance for comparison rd = ICMPv6NDOptPrefixInfo( prefixlen=raos.prefixlen, prefix=raos.prefix, L=raos.L, A=raos.A ) if type(pi_opt) is list: for ii in range(len(pi_opt)): self.assertEqual(pi_opt[ii], rd) rd = rx.getlayer(ICMPv6NDOptPrefixInfo, ii + 2) else: self.assertEqual( pi_opt, raos, "Expected: %s, received: %s" % (pi_opt.show(dump=True), raos.show(dump=True)), ) def send_and_expect_ra( self, intf, pkts, remark, dst_ip=None, filter_out_fn=is_ipv6_misc, opt=None, src_ip=None, ): self.vapi.cli("clear trace") intf.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = intf.get_capture(1, filter_out_fn=filter_out_fn) self.assertEqual(len(rx), 1) rx = rx[0] self.validate_ra(intf, rx, dst_ip, src_ip=src_ip, pi_opt=opt) def test_ip6_ra_dump(self): """IPv6 RA dump""" # Dump IPv6 RA for all interfaces ip6_ra_dump = self.vapi.sw_interface_ip6nd_ra_dump(sw_if_index=0xFFFFFFFF) self.assertEqual(len(ip6_ra_dump), len(self.interfaces)) for ip6_ra in ip6_ra_dump: self.assertFalse(ip6_ra.send_radv) self.assertEqual(ip6_ra.n_prefixes, 0) self.assertEqual(len(ip6_ra.prefixes), 0) self.assertEqual(ip6_ra.last_radv_time, 0.0) self.assertEqual(ip6_ra.last_multicast_time, 0.0) self.assertEqual(ip6_ra.next_multicast_time, 0.0) self.assertEqual(ip6_ra.n_advertisements_sent, 0) self.assertEqual(ip6_ra.n_solicitations_rcvd, 0) self.assertEqual(ip6_ra.n_solicitations_dropped, 0) # Enable sending IPv6 RA for an interface self.pg0.ip6_ra_config(no=1, suppress=1) # Add IPv6 RA prefixes for the interface pfx0 = IPv6Network( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), strict=False ) pfx1 = IPv6Network("fafa::/96") self.pg0.ip6_ra_prefix(pfx0, off_link=0, no_autoconfig=0) self.pg0.ip6_ra_prefix(pfx1, off_link=1, no_autoconfig=1) # Wait for multicast IPv6 RA self.sleep(1) # Dump IPv6 RA for the interface ip6_ra_dump = self.vapi.sw_interface_ip6nd_ra_dump( sw_if_index=self.pg0.sw_if_index ) self.assertEqual(len(ip6_ra_dump), 1) ip6_ra = ip6_ra_dump[0] self.assertEqual(ip6_ra.sw_if_index, self.pg0.sw_if_index) self.assertTrue(ip6_ra.send_radv) self.assertEqual(ip6_ra.n_prefixes, 2) self.assertEqual(len(ip6_ra.prefixes), 2) self.assertEqual(ip6_ra.last_radv_time, 0.0) self.assertGreater(ip6_ra.last_multicast_time, 0.0) self.assertGreater(ip6_ra.next_multicast_time, 0.0) self.assertGreater(ip6_ra.n_advertisements_sent, 0) self.assertEqual(ip6_ra.n_solicitations_rcvd, 0) self.assertEqual(ip6_ra.n_solicitations_dropped, 0) self.assertEqual(ip6_ra.prefixes[0].prefix, pfx0) self.assertTrue(ip6_ra.prefixes[0].onlink_flag) self.assertTrue(ip6_ra.prefixes[0].autonomous_flag) self.assertFalse(ip6_ra.prefixes[0].no_advertise) self.assertEqual(ip6_ra.prefixes[1].prefix, pfx1) self.assertFalse(ip6_ra.prefixes[1].onlink_flag) self.assertFalse(ip6_ra.prefixes[1].autonomous_flag) self.assertFalse(ip6_ra.prefixes[1].no_advertise) # Reset sending IPv6 RA for the interface self.pg0.ip6_ra_config(suppress=1) # Remove IPv6 RA prefixes for the interface self.pg0.ip6_ra_prefix(pfx0, is_no=1) self.pg0.ip6_ra_prefix(pfx1, is_no=1) # Dump IPv6 RA for the interface ip6_ra_dump = self.vapi.sw_interface_ip6nd_ra_dump( sw_if_index=self.pg0.sw_if_index ) self.assertEqual(len(ip6_ra_dump), 1) ip6_ra = ip6_ra_dump[0] self.assertEqual(ip6_ra.sw_if_index, self.pg0.sw_if_index) self.assertFalse(ip6_ra.send_radv) self.assertEqual(ip6_ra.n_prefixes, 0) self.assertEqual(len(ip6_ra.prefixes), 0) def test_rs(self): """IPv6 Router Solicitation Exceptions Test scenario: """ self.pg0.ip6_ra_config(no=1, suppress=1) # # Before we begin change the IPv6 RA responses to use the unicast # address - that way we will not confuse them with the periodic # RAs which go to the mcast address # Sit and wait for the first periodic RA. # # TODO # self.pg0.ip6_ra_config(send_unicast=1) # # An RS from a link source address # - expect an RA in return # p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src=self.pg0.remote_ip6) / ICMPv6ND_RS() ) pkts = [p] self.send_and_expect_ra(self.pg0, pkts, "Genuine RS") # # For the next RS sent the RA should be rate limited # self.send_and_assert_no_replies(self.pg0, pkts, "RA rate limited") # # When we reconfigure the IPv6 RA config, # we reset the RA rate limiting, # so we need to do this before each test below so as not to drop # packets for rate limiting reasons. Test this works here. # self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra(self.pg0, pkts, "Rate limit reset RS") # # An RS sent from a non-link local source # self.pg0.ip6_ra_config(send_unicast=1) p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src="2002::ffff") / ICMPv6ND_RS() ) pkts = [p] self.send_and_assert_no_replies(self.pg0, pkts, "RS from non-link source") # # Source an RS from a link local address # self.pg0.ip6_ra_config(send_unicast=1) ll = mk_ll_addr(self.pg0.remote_mac) p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src=ll) / ICMPv6ND_RS() ) pkts = [p] self.send_and_expect_ra(self.pg0, pkts, "RS sourced from link-local", dst_ip=ll) # # Source an RS from a link local address # Ensure suppress also applies to solicited RS # self.pg0.ip6_ra_config(send_unicast=1, suppress=1) ll = mk_ll_addr(self.pg0.remote_mac) p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src=ll) / ICMPv6ND_RS() ) pkts = [p] self.send_and_assert_no_replies(self.pg0, pkts, "Suppressed RS from link-local") # # Send the RS multicast # self.pg0.ip6_ra_config(no=1, suppress=1) # Reset suppress flag to zero self.pg0.ip6_ra_config(send_unicast=1) dmac = in6_getnsmac(inet_pton(AF_INET6, "ff02::2")) ll = mk_ll_addr(self.pg0.remote_mac) p = ( Ether(dst=dmac, src=self.pg0.remote_mac) / IPv6(dst="ff02::2", src=ll) / ICMPv6ND_RS() ) pkts = [p] self.send_and_expect_ra(self.pg0, pkts, "RS sourced from link-local", dst_ip=ll) # # Source from the unspecified address ::. This happens when the RS # is sent before the host has a configured address/sub-net, # i.e. auto-config. Since the sender has no IP address, the reply # comes back mcast - so the capture needs to not filter this. # If we happen to pick up the periodic RA at this point then so be it, # it's not an error. # self.pg0.ip6_ra_config(send_unicast=1) p = ( Ether(dst=dmac, src=self.pg0.remote_mac) / IPv6(dst="ff02::2", src="::") / ICMPv6ND_RS() ) pkts = [p] self.send_and_expect_ra( self.pg0, pkts, "RS sourced from unspecified", dst_ip="ff02::1", filter_out_fn=None, ) # # Configure The RA to announce the links prefix # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len) ) # # RAs should now contain the prefix information option # opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=1, A=1 ) self.pg0.ip6_ra_config(send_unicast=1) ll = mk_ll_addr(self.pg0.remote_mac) p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src=ll) / ICMPv6ND_RS() ) self.send_and_expect_ra(self.pg0, p, "RA with prefix-info", dst_ip=ll, opt=opt) # # Change the prefix info to not off-link # L-flag is clear # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), off_link=1 ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=0, A=1 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix info with L-flag=0", dst_ip=ll, opt=opt ) # # Change the prefix info to not off-link, no-autoconfig # L and A flag are clear in the advert # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), off_link=1, no_autoconfig=1, ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=0, A=0 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix info with A & L-flag=0", dst_ip=ll, opt=opt ) # # Change the flag settings back to the defaults # L and A flag are set in the advert # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len) ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=1, A=1 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra(self.pg0, p, "RA with Prefix info", dst_ip=ll, opt=opt) # # Change the prefix info to not off-link, no-autoconfig # L and A flag are clear in the advert # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), off_link=1, no_autoconfig=1, ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=0, A=0 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix info with A & L-flag=0", dst_ip=ll, opt=opt ) # # Use the reset to defaults option to revert to defaults # L and A flag are clear in the advert # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), use_default=1 ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=1, A=1 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix reverted to defaults", dst_ip=ll, opt=opt ) # # Advertise Another prefix. With no L-flag/A-flag # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg1.local_ip6, self.pg1.local_ip6_prefix_len), off_link=1, no_autoconfig=1, ) opt = [ ICMPv6NDOptPrefixInfo( prefixlen=self.pg0.local_ip6_prefix_len, prefix=self.pg0.local_ip6, L=1, A=1, ), ICMPv6NDOptPrefixInfo( prefixlen=self.pg1.local_ip6_prefix_len, prefix=self.pg1.local_ip6, L=0, A=0, ), ] self.pg0.ip6_ra_config(send_unicast=1) ll = mk_ll_addr(self.pg0.remote_mac) p = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0.local_ip6, src=ll) / ICMPv6ND_RS() ) self.send_and_expect_ra( self.pg0, p, "RA with multiple Prefix infos", dst_ip=ll, opt=opt ) # # Remove the first prefix-info - expect the second is still in the # advert # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg0.local_ip6, self.pg0.local_ip6_prefix_len), is_no=1 ) opt = ICMPv6NDOptPrefixInfo( prefixlen=self.pg1.local_ip6_prefix_len, prefix=self.pg1.local_ip6, L=0, A=0 ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix reverted to defaults", dst_ip=ll, opt=opt ) # # Remove the second prefix-info - expect no prefix-info in the adverts # self.pg0.ip6_ra_prefix( "%s/%s" % (self.pg1.local_ip6, self.pg1.local_ip6_prefix_len), is_no=1 ) # # change the link's link local, so we know that works too. # self.vapi.sw_interface_ip6_set_link_local_address( sw_if_index=self.pg0.sw_if_index, ip="fe80::88" ) self.pg0.ip6_ra_config(send_unicast=1) self.send_and_expect_ra( self.pg0, p, "RA with Prefix reverted to defaults", dst_ip=ll, src_ip="fe80::88", ) # # Reset the periodic advertisements back to default values # self.pg0.ip6_ra_config(suppress=1) self.pg0.ip6_ra_config(no=1, send_unicast=1) def test_mld(self): """MLD Report""" # # test one MLD is sent after applying an IPv6 Address on an interface # self.pg_enable_capture(self.pg_interfaces) self.pg_start() subitf = VppDot1QSubint(self, self.pg1, 99) self.interfaces.append(subitf) self.sub_interfaces.append(subitf) subitf.admin_up() subitf.config_ip6() rxs = self.pg1._get_capture(timeout=4, filter_out_fn=None) # # hunt for the MLD on vlan 99 # for rx in rxs: # make sure ipv6 packets with hop by hop options have # correct checksums self.assert_packet_checksums_valid(rx) if ( rx.haslayer(IPv6ExtHdrHopByHop) and rx.haslayer(Dot1Q) and rx[Dot1Q].vlan == 99 ): mld = rx[ICMPv6MLReport2] self.assertEqual(mld.records_number, 4) class TestIPv6RouteLookup(VppTestCase): """IPv6 Route Lookup Test Case""" routes = [] def route_lookup(self, prefix, exact): return self.vapi.api( self.vapi.papi.ip_route_lookup, { "table_id": 0, "exact": exact, "prefix": prefix, }, ) @classmethod def setUpClass(cls): super(TestIPv6RouteLookup, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPv6RouteLookup, cls).tearDownClass() def setUp(self): super(TestIPv6RouteLookup, self).setUp() drop_nh = VppRoutePath("::1", 0xFFFFFFFF, type=FibPathType.FIB_PATH_TYPE_DROP) # Add 3 routes r = VppIpRoute(self, "2001:1111::", 32, [drop_nh]) r.add_vpp_config() self.routes.append(r) r = VppIpRoute(self, "2001:1111:2222::", 48, [drop_nh]) r.add_vpp_config() self.routes.append(r) r = VppIpRoute(self, "2001:1111:2222::1", 128, [drop_nh]) r.add_vpp_config() self.routes.append(r) def tearDown(self): # Remove the routes we added for r in self.routes: r.remove_vpp_config() super(TestIPv6RouteLookup, self).tearDown() def test_exact_match(self): # Verify we find the host route prefix = "2001:1111:2222::1/128" result = self.route_lookup(prefix, True) assert prefix == str(result.route.prefix) # Verify we find a middle prefix route prefix = "2001:1111:2222::/48" result = self.route_lookup(prefix, True) assert prefix == str(result.route.prefix) # Verify we do not find an available LPM. with self.vapi.assert_negative_api_retval(): self.route_lookup("2001::2/128", True) def test_longest_prefix_match(self): # verify we find lpm lpm_prefix = "2001:1111:2222::/48" result = self.route_lookup("2001:1111:2222::2/128", False) assert lpm_prefix == str(result.route.prefix) # Verify we find the exact when not requested result = self.route_lookup(lpm_prefix, False) assert lpm_prefix == str(result.route.prefix) # Can't seem to delete the default route so no negative LPM test. class TestIPv6IfAddrRoute(VppTestCase): """IPv6 Interface Addr Route Test Case""" @classmethod def setUpClass(cls): super(TestIPv6IfAddrRoute, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPv6IfAddrRoute, cls).tearDownClass() def setUp(self): super(TestIPv6IfAddrRoute, self).setUp() # create 1 pg interface self.create_pg_interfaces(range(1)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIPv6IfAddrRoute, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_ipv6_ifaddrs_same_prefix(self): """IPv6 Interface Addresses Same Prefix test Test scenario: - Verify no route in FIB for prefix 2001:10::/64 - Configure IPv4 address 2001:10::10/64 on an interface - Verify route in FIB for prefix 2001:10::/64 - Configure IPv4 address 2001:10::20/64 on an interface - Delete 2001:10::10/64 from interface - Verify route in FIB for prefix 2001:10::/64 - Delete 2001:10::20/64 from interface - Verify no route in FIB for prefix 2001:10::/64 """ addr1 = "2001:10::10" addr2 = "2001:10::20" if_addr1 = VppIpInterfaceAddress(self, self.pg0, addr1, 64) if_addr2 = VppIpInterfaceAddress(self, self.pg0, addr2, 64) self.assertFalse(if_addr1.query_vpp_config()) self.assertFalse(find_route(self, addr1, 128)) self.assertFalse(find_route(self, addr2, 128)) # configure first address, verify route present if_addr1.add_vpp_config() self.assertTrue(if_addr1.query_vpp_config()) self.assertTrue(find_route(self, addr1, 128)) self.assertFalse(find_route(self, addr2, 128)) # configure second address, delete first, verify route not removed if_addr2.add_vpp_config() if_addr1.remove_vpp_config() self.assertFalse(if_addr1.query_vpp_config()) self.assertTrue(if_addr2.query_vpp_config()) self.assertFalse(find_route(self, addr1, 128)) self.assertTrue(find_route(self, addr2, 128)) # delete second address, verify route removed if_addr2.remove_vpp_config() self.assertFalse(if_addr1.query_vpp_config()) self.assertFalse(find_route(self, addr1, 128)) self.assertFalse(find_route(self, addr2, 128)) def test_ipv6_ifaddr_del(self): """Delete an interface address that does not exist""" loopbacks = self.create_loopback_interfaces(1) lo = self.lo_interfaces[0] lo.config_ip6() lo.admin_up() # # try and remove pg0's subnet from lo # with self.vapi.assert_negative_api_retval(): self.vapi.sw_interface_add_del_address( sw_if_index=lo.sw_if_index, prefix=self.pg0.local_ip6_prefix, is_add=0 ) @unittest.skipIf( "ping" in config.excluded_plugins, "Exclude tests requiring Ping plugin" ) class TestICMPv6Echo(VppTestCase): """ICMPv6 Echo Test Case""" @classmethod def setUpClass(cls): super(TestICMPv6Echo, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestICMPv6Echo, cls).tearDownClass() def setUp(self): super(TestICMPv6Echo, self).setUp() # create 1 pg interface self.create_pg_interfaces(range(1)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp(link_layer=True) i.resolve_ndp() def tearDown(self): super(TestICMPv6Echo, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_icmpv6_echo(self): """VPP replies to ICMPv6 Echo Request Test scenario: - Receive ICMPv6 Echo Request message on pg0 interface. - Check outgoing ICMPv6 Echo Reply message on pg0 interface. """ # test both with global and local ipv6 addresses dsts = (self.pg0.local_ip6, self.pg0.local_ip6_ll) id = 0xB seq = 5 data = b"\x0a" * 18 p = list() for dst in dsts: p.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst=dst) / ICMPv6EchoRequest(id=id, seq=seq, data=data) ) ) self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rxs = self.pg0.get_capture(len(dsts)) for rx, dst in zip(rxs, dsts): ether = rx[Ether] ipv6 = rx[IPv6] icmpv6 = rx[ICMPv6EchoReply] self.assertEqual(ether.src, self.pg0.local_mac) self.assertEqual(ether.dst, self.pg0.remote_mac) self.assertEqual(ipv6.src, dst) self.assertEqual(ipv6.dst, self.pg0.remote_ip6) self.assertEqual(icmp6types[icmpv6.type], "Echo Reply") self.assertEqual(icmpv6.id, id) self.assertEqual(icmpv6.seq, seq) self.assertEqual(icmpv6.data, data) class TestIPv6RD(TestIPv6ND): """IPv6 Router Discovery Test Case""" @classmethod def setUpClass(cls): super(TestIPv6RD, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPv6RD, cls).tearDownClass() def setUp(self): super(TestIPv6RD, self).setUp() # create 2 pg interfaces self.create_pg_interfaces(range(2)) self.interfaces = list(self.pg_interfaces) # setup all interfaces for i in self.interfaces: i.admin_up() i.config_ip6() def tearDown(self): for i in self.interfaces: i.unconfig_ip6() i.admin_down() super(TestIPv6RD, self).tearDown() def test_rd_send_router_solicitation(self): """Verify router solicitation packets""" count = 2 self.pg_enable_capture(self.pg_interfaces) self.pg_start() self.vapi.ip6nd_send_router_solicitation(self.pg1.sw_if_index, mrc=count) rx_list = self.pg1.get_capture(count, timeout=3) self.assertEqual(len(rx_list), count) for packet in rx_list: self.assertEqual(packet.haslayer(IPv6), 1) self.assertEqual(packet[IPv6].haslayer(ICMPv6ND_RS), 1) dst = ip6_normalize(packet[IPv6].dst) dst2 = ip6_normalize("ff02::2") self.assert_equal(dst, dst2) src = ip6_normalize(packet[IPv6].src) src2 = ip6_normalize(self.pg1.local_ip6_ll) self.assert_equal(src, src2) self.assertTrue(bool(packet[ICMPv6ND_RS].haslayer(ICMPv6NDOptSrcLLAddr))) self.assert_equal(packet[ICMPv6NDOptSrcLLAddr].lladdr, self.pg1.local_mac) def verify_prefix_info(self, reported_prefix, prefix_option): prefix = IPv6Network( text_type( prefix_option.getfieldval("prefix") + "/" + text_type(prefix_option.getfieldval("prefixlen")) ), strict=False, ) self.assert_equal( reported_prefix.prefix.network_address, prefix.network_address ) L = prefix_option.getfieldval("L") A = prefix_option.getfieldval("A") option_flags = (L << 7) | (A << 6) self.assert_equal(reported_prefix.flags, option_flags) self.assert_equal( reported_prefix.valid_time, prefix_option.getfieldval("validlifetime") ) self.assert_equal( reported_prefix.preferred_time, prefix_option.getfieldval("preferredlifetime"), ) def test_rd_receive_router_advertisement(self): """Verify events triggered by received RA packets""" self.vapi.want_ip6_ra_events(enable=1) prefix_info_1 = ICMPv6NDOptPrefixInfo( prefix="1::2", prefixlen=50, validlifetime=200, preferredlifetime=500, L=1, A=1, ) prefix_info_2 = ICMPv6NDOptPrefixInfo( prefix="7::4", prefixlen=20, validlifetime=70, preferredlifetime=1000, L=1, A=0, ) p = ( Ether(dst=self.pg1.local_mac, src=self.pg1.remote_mac) / IPv6(dst=self.pg1.local_ip6_ll, src=mk_ll_addr(self.pg1.remote_mac)) / ICMPv6ND_RA() / prefix_info_1 / prefix_info_2 ) self.pg1.add_stream([p]) self.pg_start() ev = self.vapi.wait_for_event(10, "ip6_ra_event") self.assert_equal(ev.current_hop_limit, 0) self.assert_equal(ev.flags, 8) self.assert_equal(ev.router_lifetime_in_sec, 1800) self.assert_equal(ev.neighbor_reachable_time_in_msec, 0) self.assert_equal( ev.time_in_msec_between_retransmitted_neighbor_solicitations, 0 ) self.assert_equal(ev.n_prefixes, 2) self.verify_prefix_info(ev.prefixes[0], prefix_info_1) self.verify_prefix_info(ev.prefixes[1], prefix_info_2) class TestIPv6RDControlPlane(TestIPv6ND): """IPv6 Router Discovery Control Plane Test Case""" @classmethod def setUpClass(cls): super(TestIPv6RDControlPlane, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPv6RDControlPlane, cls).tearDownClass() def setUp(self): super(TestIPv6RDControlPlane, self).setUp() # create 1 pg interface self.create_pg_interfaces(range(1)) self.interfaces = list(self.pg_interfaces) # setup all interfaces for i in self.interfaces: i.admin_up() i.config_ip6() def tearDown(self): super(TestIPv6RDControlPlane, self).tearDown() @staticmethod def create_ra_packet(pg, routerlifetime=None): src_ip = pg.remote_ip6_ll dst_ip = pg.local_ip6 if routerlifetime is not None: ra = ICMPv6ND_RA(routerlifetime=routerlifetime) else: ra = ICMPv6ND_RA() p = ( Ether(dst=pg.local_mac, src=pg.remote_mac) / IPv6(dst=dst_ip, src=src_ip) / ra ) return p @staticmethod def get_default_routes(fib): list = [] for entry in fib: if entry.route.prefix.prefixlen == 0: for path in entry.route.paths: if path.sw_if_index != 0xFFFFFFFF: defaut_route = {} defaut_route["sw_if_index"] = path.sw_if_index defaut_route["next_hop"] = path.nh.address.ip6 list.append(defaut_route) return list @staticmethod def get_interface_addresses(fib, pg): list = [] for entry in fib: if entry.route.prefix.prefixlen == 128: path = entry.route.paths[0] if path.sw_if_index == pg.sw_if_index: list.append(str(entry.route.prefix.network_address)) return list def wait_for_no_default_route(self, n_tries=50, s_time=1): while n_tries: fib = self.vapi.ip_route_dump(0, True) default_routes = self.get_default_routes(fib) if 0 == len(default_routes): return True n_tries = n_tries - 1 self.sleep(s_time) return False def test_all(self): """Test handling of SLAAC addresses and default routes""" fib = self.vapi.ip_route_dump(0, True) default_routes = self.get_default_routes(fib) initial_addresses = set(self.get_interface_addresses(fib, self.pg0)) self.assertEqual(default_routes, []) router_address = IPv6Address(text_type(self.pg0.remote_ip6_ll)) self.vapi.ip6_nd_address_autoconfig(self.pg0.sw_if_index, 1, 1) self.sleep(0.1) # send RA packet = ( self.create_ra_packet(self.pg0) / ICMPv6NDOptPrefixInfo( prefix="1::", prefixlen=64, validlifetime=2, preferredlifetime=2, L=1, A=1, ) / ICMPv6NDOptPrefixInfo( prefix="7::", prefixlen=20, validlifetime=1500, preferredlifetime=1000, L=1, A=0, ) ) self.pg0.add_stream([packet]) self.pg_start() self.sleep_on_vpp_time(0.1) fib = self.vapi.ip_route_dump(0, True) # check FIB for new address addresses = set(self.get_interface_addresses(fib, self.pg0)) new_addresses = addresses.difference(initial_addresses) self.assertEqual(len(new_addresses), 1) prefix = IPv6Network( text_type("%s/%d" % (list(new_addresses)[0], 20)), strict=False ) self.assertEqual(prefix, IPv6Network(text_type("1::/20"))) # check FIB for new default route default_routes = self.get_default_routes(fib) self.assertEqual(len(default_routes), 1) dr = default_routes[0] self.assertEqual(dr["sw_if_index"], self.pg0.sw_if_index) self.assertEqual(dr["next_hop"], router_address) # send RA to delete default route packet = self.create_ra_packet(self.pg0, routerlifetime=0) self.pg0.add_stream([packet]) self.pg_start() self.sleep_on_vpp_time(0.1) # check that default route is deleted fib = self.vapi.ip_route_dump(0, True) default_routes = self.get_default_routes(fib) self.assertEqual(len(default_routes), 0) self.sleep_on_vpp_time(0.1) # send RA packet = self.create_ra_packet(self.pg0) self.pg0.add_stream([packet]) self.pg_start() self.sleep_on_vpp_time(0.1) # check FIB for new default route fib = self.vapi.ip_route_dump(0, True) default_routes = self.get_default_routes(fib) self.assertEqual(len(default_routes), 1) dr = default_routes[0] self.assertEqual(dr["sw_if_index"], self.pg0.sw_if_index) self.assertEqual(dr["next_hop"], router_address) # send RA, updating router lifetime to 1s packet = self.create_ra_packet(self.pg0, 1) self.pg0.add_stream([packet]) self.pg_start() self.sleep_on_vpp_time(0.1) # check that default route still exists fib = self.vapi.ip_route_dump(0, True) default_routes = self.get_default_routes(fib) self.assertEqual(len(default_routes), 1) dr = default_routes[0] self.assertEqual(dr["sw_if_index"], self.pg0.sw_if_index) self.assertEqual(dr["next_hop"], router_address) self.sleep_on_vpp_time(1) # check that default route is deleted self.assertTrue(self.wait_for_no_default_route()) # check FIB still contains the SLAAC address addresses = set(self.get_interface_addresses(fib, self.pg0)) new_addresses = addresses.difference(initial_addresses) self.assertEqual(len(new_addresses), 1) prefix = IPv6Network( text_type("%s/%d" % (list(new_addresses)[0], 20)), strict=False ) self.assertEqual(prefix, IPv6Network(text_type("1::/20"))) self.sleep_on_vpp_time(1) # check that SLAAC address is deleted fib = self.vapi.ip_route_dump(0, True) addresses = set(self.get_interface_addresses(fib, self.pg0)) new_addresses = addresses.difference(initial_addresses) self.assertEqual(len(new_addresses), 0) class IPv6NDProxyTest(TestIPv6ND): """IPv6 ND ProxyTest Case""" @classmethod def setUpClass(cls): super(IPv6NDProxyTest, cls).setUpClass() @classmethod def tearDownClass(cls): super(IPv6NDProxyTest, cls).tearDownClass() def setUp(self): super(IPv6NDProxyTest, self).setUp() # create 3 pg interfaces self.create_pg_interfaces(range(3)) # pg0 is the master interface, with the configured subnet self.pg0.admin_up() self.pg0.config_ip6() self.pg0.resolve_ndp() self.pg1.ip6_enable() self.pg2.ip6_enable() def tearDown(self): super(IPv6NDProxyTest, self).tearDown() def test_nd_proxy(self): """IPv6 Proxy ND""" # # Generate some hosts in the subnet that we are proxying # self.pg0.generate_remote_hosts(8) nsma = in6_getnsma(inet_pton(AF_INET6, self.pg0.local_ip6)) d = inet_ntop(AF_INET6, nsma) # # Send an NS for one of those remote hosts on one of the proxy links # expect no response since it's from an address that is not # on the link that has the prefix configured # ns_pg1 = ( Ether(dst=in6_getnsmac(nsma), src=self.pg1.remote_mac) / IPv6(dst=d, src=self.pg0._remote_hosts[2].ip6) / ICMPv6ND_NS(tgt=self.pg0.local_ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0._remote_hosts[2].mac) ) self.send_and_assert_no_replies(self.pg1, ns_pg1, "Off link NS") # # Add proxy support for the host # self.vapi.ip6nd_proxy_add_del( is_add=1, ip=inet_pton(AF_INET6, self.pg0._remote_hosts[2].ip6), sw_if_index=self.pg1.sw_if_index, ) # # try that NS again. this time we expect an NA back # self.send_and_expect_na( self.pg1, ns_pg1, "NS to proxy entry", dst_ip=self.pg0._remote_hosts[2].ip6, tgt_ip=self.pg0.local_ip6, ) # # ... and that we have an entry in the ND cache # self.assertTrue( find_nbr(self, self.pg1.sw_if_index, self.pg0._remote_hosts[2].ip6) ) # # ... and we can route traffic to it # t = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(dst=self.pg0._remote_hosts[2].ip6, src=self.pg0.remote_ip6) / inet6.UDP(sport=10000, dport=20000) / Raw(b"\xa5" * 100) ) self.pg0.add_stream(t) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg1.get_capture(1) rx = rx[0] self.assertEqual(rx[Ether].dst, self.pg0._remote_hosts[2].mac) self.assertEqual(rx[Ether].src, self.pg1.local_mac) self.assertEqual(rx[IPv6].src, t[IPv6].src) self.assertEqual(rx[IPv6].dst, t[IPv6].dst) # # Test we proxy for the host on the main interface # ns_pg0 = ( Ether(dst=in6_getnsmac(nsma), src=self.pg0.remote_mac) / IPv6(dst=d, src=self.pg0.remote_ip6) / ICMPv6ND_NS(tgt=self.pg0._remote_hosts[2].ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0.remote_mac) ) self.send_and_expect_na( self.pg0, ns_pg0, "NS to proxy entry on main", tgt_ip=self.pg0._remote_hosts[2].ip6, dst_ip=self.pg0.remote_ip6, ) # # Setup and resolve proxy for another host on another interface # ns_pg2 = ( Ether(dst=in6_getnsmac(nsma), src=self.pg2.remote_mac) / IPv6(dst=d, src=self.pg0._remote_hosts[3].ip6) / ICMPv6ND_NS(tgt=self.pg0.local_ip6) / ICMPv6NDOptSrcLLAddr(lladdr=self.pg0._remote_hosts[2].mac) ) self.vapi.ip6nd_proxy_add_del( is_add=1, ip=inet_pton(AF_INET6, self.pg0._remote_hosts[3].ip6), sw_if_index=self.pg2.sw_if_index, ) self.send_and_expect_na( self.pg2, ns_pg2, "NS to proxy entry other interface", dst_ip=self.pg0._remote_hosts[3].ip6, tgt_ip=self.pg0.local_ip6, ) self.assertTrue( find_nbr(self, self.pg2.sw_if_index, self.pg0._remote_hosts[3].ip6) ) # # hosts can communicate. pg2->pg1 # t2 = ( Ether(dst=self.pg2.local_mac, src=self.pg0.remote_hosts[3].mac) / IPv6(dst=self.pg0._remote_hosts[2].ip6, src=self.pg0._remote_hosts[3].ip6) / inet6.UDP(sport=10000, dport=20000) / Raw(b"\xa5" * 100) ) self.pg2.add_stream(t2) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg1.get_capture(1) rx = rx[0] self.assertEqual(rx[Ether].dst, self.pg0._remote_hosts[2].mac) self.assertEqual(rx[Ether].src, self.pg1.local_mac) self.assertEqual(rx[IPv6].src, t2[IPv6].src) self.assertEqual(rx[IPv6].dst, t2[IPv6].dst) # # remove the proxy configs # self.vapi.ip6nd_proxy_add_del( ip=inet_pton(AF_INET6, self.pg0._remote_hosts[2].ip6), sw_if_index=self.pg1.sw_if_index, is_add=0, ) self.vapi.ip6nd_proxy_add_del( ip=inet_pton(AF_INET6, self.pg0._remote_hosts[3].ip6), sw_if_index=self.pg2.sw_if_index, is_add=0, ) self.assertFalse( find_nbr(self, self.pg2.sw_if_index, self.pg0._remote_hosts[3].ip6) ) self.assertFalse( find_nbr(self, self.pg1.sw_if_index, self.pg0._remote_hosts[2].ip6) ) # # no longer proxy-ing... # self.send_and_assert_no_replies(self.pg0, ns_pg0, "Proxy unconfigured") self.send_and_assert_no_replies(self.pg1, ns_pg1, "Proxy unconfigured") self.send_and_assert_no_replies(self.pg2, ns_pg2, "Proxy unconfigured") # # no longer forwarding. traffic generates NS out of the glean/main # interface # self.pg2.add_stream(t2) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg0.get_capture(1) self.assertTrue(rx[0].haslayer(ICMPv6ND_NS)) class TestIP6Null(VppTestCase): """IPv6 routes via NULL""" @classmethod def setUpClass(cls): super(TestIP6Null, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6Null, cls).tearDownClass() def setUp(self): super(TestIP6Null, self).setUp() # create 2 pg interfaces self.create_pg_interfaces(range(1)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIP6Null, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_ip_null(self): """IP NULL route""" p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst="2001::1") / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) # # A route via IP NULL that will reply with ICMP unreachables # ip_unreach = VppIpRoute( self, "2001::", 64, [ VppRoutePath( "::", 0xFFFFFFFF, type=FibPathType.FIB_PATH_TYPE_ICMP_UNREACH ) ], ) ip_unreach.add_vpp_config() self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg0.get_capture(1) rx = rx[0] icmp = rx[ICMPv6DestUnreach] # 0 = "No route to destination" self.assertEqual(icmp.code, 0) # ICMP is rate limited. pause a bit self.sleep(1) # # A route via IP NULL that will reply with ICMP prohibited # ip_prohibit = VppIpRoute( self, "2001::1", 128, [ VppRoutePath( "::", 0xFFFFFFFF, type=FibPathType.FIB_PATH_TYPE_ICMP_PROHIBIT ) ], ) ip_prohibit.add_vpp_config() self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg0.get_capture(1) rx = rx[0] icmp = rx[ICMPv6DestUnreach] # 1 = "Communication with destination administratively prohibited" self.assertEqual(icmp.code, 1) class TestIP6Disabled(VppTestCase): """IPv6 disabled""" @classmethod def setUpClass(cls): super(TestIP6Disabled, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6Disabled, cls).tearDownClass() def setUp(self): super(TestIP6Disabled, self).setUp() # create 2 pg interfaces self.create_pg_interfaces(range(2)) # PG0 is IP enabled self.pg0.admin_up() self.pg0.config_ip6() self.pg0.resolve_ndp() # PG 1 is not IP enabled self.pg1.admin_up() def tearDown(self): super(TestIP6Disabled, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip4() i.admin_down() def test_ip_disabled(self): """IP Disabled""" MRouteItfFlags = VppEnum.vl_api_mfib_itf_flags_t MRouteEntryFlags = VppEnum.vl_api_mfib_entry_flags_t # # An (S,G). # one accepting interface, pg0, 2 forwarding interfaces # route_ff_01 = VppIpMRoute( self, "::", "ffef::1", 128, MRouteEntryFlags.MFIB_API_ENTRY_FLAG_NONE, [ VppMRoutePath( self.pg1.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_ACCEPT ), VppMRoutePath( self.pg0.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_FORWARD ), ], ) route_ff_01.add_vpp_config() pu = ( Ether(src=self.pg1.remote_mac, dst=self.pg1.local_mac) / IPv6(src="2001::1", dst=self.pg0.remote_ip6) / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) pm = ( Ether(src=self.pg1.remote_mac, dst=self.pg1.local_mac) / IPv6(src="2001::1", dst="ffef::1") / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) # # PG1 does not forward IP traffic # self.send_and_assert_no_replies(self.pg1, pu, "IPv6 disabled") self.send_and_assert_no_replies(self.pg1, pm, "IPv6 disabled") # # IP enable PG1 # self.pg1.config_ip6() # # Now we get packets through # self.pg1.add_stream(pu) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg0.get_capture(1) self.pg1.add_stream(pm) self.pg_enable_capture(self.pg_interfaces) self.pg_start() rx = self.pg0.get_capture(1) # # Disable PG1 # self.pg1.unconfig_ip6() # # PG1 does not forward IP traffic # self.send_and_assert_no_replies(self.pg1, pu, "IPv6 disabled") self.send_and_assert_no_replies(self.pg1, pm, "IPv6 disabled") class TestIP6LoadBalance(VppTestCase): """IPv6 Load-Balancing""" @classmethod def setUpClass(cls): super(TestIP6LoadBalance, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6LoadBalance, cls).tearDownClass() def setUp(self): super(TestIP6LoadBalance, self).setUp() self.create_pg_interfaces(range(5)) mpls_tbl = VppMplsTable(self, 0) mpls_tbl.add_vpp_config() for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() i.enable_mpls() def tearDown(self): for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() i.disable_mpls() super(TestIP6LoadBalance, self).tearDown() def test_ip6_load_balance(self): """IPv6 Load-Balancing""" # # An array of packets that differ only in the destination port # - IP only # - MPLS EOS # - MPLS non-EOS # - MPLS non-EOS with an entropy label # port_ip_pkts = [] port_mpls_pkts = [] port_mpls_neos_pkts = [] port_ent_pkts = [] # # An array of packets that differ only in the source address # src_ip_pkts = [] src_mpls_pkts = [] for ii in range(NUM_PKTS): port_ip_hdr = ( IPv6(dst="3000::1", src="3000:1::1") / inet6.UDP(sport=1234, dport=1234 + ii) / Raw(b"\xa5" * 100) ) port_ip_pkts.append( (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / port_ip_hdr) ) port_mpls_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / MPLS(label=66, ttl=2) / port_ip_hdr ) ) port_mpls_neos_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / MPLS(label=67, ttl=2) / MPLS(label=77, ttl=2) / port_ip_hdr ) ) port_ent_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / MPLS(label=67, ttl=2) / MPLS(label=14, ttl=2) / MPLS(label=999, ttl=2) / port_ip_hdr ) ) src_ip_hdr = ( IPv6(dst="3000::1", src="3000:1::%d" % ii) / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) src_ip_pkts.append( (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / src_ip_hdr) ) src_mpls_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / MPLS(label=66, ttl=2) / src_ip_hdr ) ) # # A route for the IP packets # route_3000_1 = VppIpRoute( self, "3000::1", 128, [ VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index), VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index), ], ) route_3000_1.add_vpp_config() # # a local-label for the EOS packets # binding = VppMplsIpBind(self, 66, "3000::1", 128, is_ip6=1) binding.add_vpp_config() # # An MPLS route for the non-EOS packets # route_67 = VppMplsRoute( self, 67, 0, [ VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index, labels=[67]), VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index, labels=[67]), ], ) route_67.add_vpp_config() # # inject the packet on pg0 - expect load-balancing across the 2 paths # - since the default hash config is to use IP src,dst and port # src,dst # We are not going to ensure equal amounts of packets across each link, # since the hash algorithm is statistical and therefore this can never # be guaranteed. But with 64 different packets we do expect some # balancing. So instead just ensure there is traffic on each link. # rx = self.send_and_expect_load_balancing( self.pg0, port_ip_pkts, [self.pg1, self.pg2] ) n_ip_pg0 = len(rx[0]) self.send_and_expect_load_balancing(self.pg0, src_ip_pkts, [self.pg1, self.pg2]) self.send_and_expect_load_balancing( self.pg0, port_mpls_pkts, [self.pg1, self.pg2] ) self.send_and_expect_load_balancing( self.pg0, src_mpls_pkts, [self.pg1, self.pg2] ) rx = self.send_and_expect_load_balancing( self.pg0, port_mpls_neos_pkts, [self.pg1, self.pg2] ) n_mpls_pg0 = len(rx[0]) # # change the router ID and expect the distribution changes # self.vapi.set_ip_flow_hash_router_id(router_id=0x11111111) rx = self.send_and_expect_load_balancing( self.pg0, port_ip_pkts, [self.pg1, self.pg2] ) self.assertNotEqual(n_ip_pg0, len(rx[0])) rx = self.send_and_expect_load_balancing( self.pg0, src_mpls_pkts, [self.pg1, self.pg2] ) self.assertNotEqual(n_mpls_pg0, len(rx[0])) # # The packets with Entropy label in should not load-balance, # since the Entropy value is fixed. # self.send_and_expect_only(self.pg0, port_ent_pkts, self.pg1) # # change the flow hash config so it's only IP src,dst # - now only the stream with differing source address will # load-balance # self.vapi.set_ip_flow_hash( vrf_id=0, src=1, dst=1, proto=1, sport=0, dport=0, is_ipv6=1 ) self.send_and_expect_load_balancing(self.pg0, src_ip_pkts, [self.pg1, self.pg2]) self.send_and_expect_load_balancing( self.pg0, src_mpls_pkts, [self.pg1, self.pg2] ) self.send_and_expect_only(self.pg0, port_ip_pkts, self.pg2) # # change the flow hash config back to defaults # self.vapi.set_ip_flow_hash( vrf_id=0, src=1, dst=1, sport=1, dport=1, proto=1, is_ipv6=1 ) # # Recursive prefixes # - testing that 2 stages of load-balancing occurs and there is no # polarisation (i.e. only 2 of 4 paths are used) # port_pkts = [] src_pkts = [] for ii in range(257): port_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(dst="4000::1", src="4000:1::1") / inet6.UDP(sport=1234, dport=1234 + ii) / Raw(b"\xa5" * 100) ) ) src_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(dst="4000::1", src="4000:1::%d" % ii) / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) ) route_3000_2 = VppIpRoute( self, "3000::2", 128, [ VppRoutePath(self.pg3.remote_ip6, self.pg3.sw_if_index), VppRoutePath(self.pg4.remote_ip6, self.pg4.sw_if_index), ], ) route_3000_2.add_vpp_config() route_4000_1 = VppIpRoute( self, "4000::1", 128, [VppRoutePath("3000::1", 0xFFFFFFFF), VppRoutePath("3000::2", 0xFFFFFFFF)], ) route_4000_1.add_vpp_config() # # inject the packet on pg0 - expect load-balancing across all 4 paths # self.vapi.cli("clear trace") self.send_and_expect_load_balancing( self.pg0, port_pkts, [self.pg1, self.pg2, self.pg3, self.pg4] ) self.send_and_expect_load_balancing( self.pg0, src_pkts, [self.pg1, self.pg2, self.pg3, self.pg4] ) # # Recursive prefixes # - testing that 2 stages of load-balancing no choices # port_pkts = [] for ii in range(257): port_pkts.append( ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(dst="6000::1", src="6000:1::1") / inet6.UDP(sport=1234, dport=1234 + ii) / Raw(b"\xa5" * 100) ) ) route_5000_2 = VppIpRoute( self, "5000::2", 128, [VppRoutePath(self.pg3.remote_ip6, self.pg3.sw_if_index)], ) route_5000_2.add_vpp_config() route_6000_1 = VppIpRoute( self, "6000::1", 128, [VppRoutePath("5000::2", 0xFFFFFFFF)] ) route_6000_1.add_vpp_config() # # inject the packet on pg0 - expect load-balancing across all 4 paths # self.vapi.cli("clear trace") self.send_and_expect_only(self.pg0, port_pkts, self.pg3) class IP6PuntSetup(object): """Setup for IPv6 Punt Police/Redirect""" def punt_setup(self): self.create_pg_interfaces(range(4)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() self.pkt = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg0.local_ip6) / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) def punt_teardown(self): for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() class TestIP6Punt(IP6PuntSetup, VppTestCase): """IPv6 Punt Police/Redirect""" def setUp(self): super(TestIP6Punt, self).setUp() super(TestIP6Punt, self).punt_setup() def tearDown(self): super(TestIP6Punt, self).punt_teardown() super(TestIP6Punt, self).tearDown() def test_ip_punt(self): """IP6 punt police and redirect""" pkts = self.pkt * 1025 # # Configure a punt redirect via pg1. # nh_addr = self.pg1.remote_ip6 ip_punt_redirect = VppIpPuntRedirect( self, self.pg0.sw_if_index, self.pg1.sw_if_index, nh_addr ) ip_punt_redirect.add_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) # # add a policer # policer = VppPolicer(self, "ip6-punt", 400, 0, 10, 0, rate_type=1) policer.add_vpp_config() ip_punt_policer = VppIpPuntPolicer(self, policer.policer_index, is_ip6=True) ip_punt_policer.add_vpp_config() self.vapi.cli("clear trace") self.pg0.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() # # the number of packet received should be greater than 0, # but not equal to the number sent, since some were policed # rx = self.pg1._get_capture(1) stats = policer.get_stats() # Single rate policer - expect conform, violate but no exceed self.assertGreater(stats["conform_packets"], 0) self.assertEqual(stats["exceed_packets"], 0) self.assertGreater(stats["violate_packets"], 0) self.assertGreater(len(rx), 0) self.assertLess(len(rx), len(pkts)) # # remove the policer. back to full rx # ip_punt_policer.remove_vpp_config() policer.remove_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) # # remove the redirect. expect full drop. # ip_punt_redirect.remove_vpp_config() self.send_and_assert_no_replies(self.pg0, pkts, "IP no punt config") # # Add a redirect that is not input port selective # ip_punt_redirect = VppIpPuntRedirect( self, 0xFFFFFFFF, self.pg1.sw_if_index, nh_addr ) ip_punt_redirect.add_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) ip_punt_redirect.remove_vpp_config() def test_ip_punt_dump(self): """IP6 punt redirect dump""" # # Configure a punt redirects # nh_address = self.pg3.remote_ip6 ipr_03 = VppIpPuntRedirect( self, self.pg0.sw_if_index, self.pg3.sw_if_index, nh_address ) ipr_13 = VppIpPuntRedirect( self, self.pg1.sw_if_index, self.pg3.sw_if_index, nh_address ) ipr_23 = VppIpPuntRedirect( self, self.pg2.sw_if_index, self.pg3.sw_if_index, "0::0" ) ipr_03.add_vpp_config() ipr_13.add_vpp_config() ipr_23.add_vpp_config() # # Dump pg0 punt redirects # self.assertTrue(ipr_03.query_vpp_config()) self.assertTrue(ipr_13.query_vpp_config()) self.assertTrue(ipr_23.query_vpp_config()) # # Dump punt redirects for all interfaces # punts = self.vapi.ip_punt_redirect_dump(0xFFFFFFFF, is_ipv6=1) self.assertEqual(len(punts), 3) for p in punts: self.assertEqual(p.punt.tx_sw_if_index, self.pg3.sw_if_index) self.assertNotEqual(punts[1].punt.nh, self.pg3.remote_ip6) self.assertEqual(str(punts[2].punt.nh), "::") class TestIP6PuntHandoff(IP6PuntSetup, VppTestCase): """IPv6 Punt Police/Redirect""" vpp_worker_count = 2 def setUp(self): super(TestIP6PuntHandoff, self).setUp() super(TestIP6PuntHandoff, self).punt_setup() def tearDown(self): super(TestIP6PuntHandoff, self).punt_teardown() super(TestIP6PuntHandoff, self).tearDown() def test_ip_punt(self): """IP6 punt policer thread handoff""" pkts = self.pkt * NUM_PKTS # # Configure a punt redirect via pg1. # nh_addr = self.pg1.remote_ip6 ip_punt_redirect = VppIpPuntRedirect( self, self.pg0.sw_if_index, self.pg1.sw_if_index, nh_addr ) ip_punt_redirect.add_vpp_config() action_tx = PolicerAction( VppEnum.vl_api_sse2_qos_action_type_t.SSE2_QOS_ACTION_API_TRANSMIT, 0 ) # # This policer drops no packets, we are just # testing that they get to the right thread. # policer = VppPolicer( self, "ip6-punt", 400, 0, 10, 0, 1, 0, 0, False, action_tx, action_tx, action_tx, ) policer.add_vpp_config() ip_punt_policer = VppIpPuntPolicer(self, policer.policer_index, is_ip6=True) ip_punt_policer.add_vpp_config() for worker in [0, 1]: self.send_and_expect(self.pg0, pkts, self.pg1, worker=worker) if worker == 0: self.logger.debug(self.vapi.cli("show trace max 100")) # Combined stats, all threads stats = policer.get_stats() # Single rate policer - expect conform, violate but no exceed self.assertGreater(stats["conform_packets"], 0) self.assertEqual(stats["exceed_packets"], 0) self.assertGreater(stats["violate_packets"], 0) # Worker 0, should have done all the policing stats0 = policer.get_stats(worker=0) self.assertEqual(stats, stats0) # Worker 1, should have handed everything off stats1 = policer.get_stats(worker=1) self.assertEqual(stats1["conform_packets"], 0) self.assertEqual(stats1["exceed_packets"], 0) self.assertEqual(stats1["violate_packets"], 0) # Bind the policer to worker 1 and repeat policer.bind_vpp_config(1, True) for worker in [0, 1]: self.send_and_expect(self.pg0, pkts, self.pg1, worker=worker) self.logger.debug(self.vapi.cli("show trace max 100")) # The 2 workers should now have policed the same amount stats = policer.get_stats() stats0 = policer.get_stats(worker=0) stats1 = policer.get_stats(worker=1) self.assertGreater(stats0["conform_packets"], 0) self.assertEqual(stats0["exceed_packets"], 0) self.assertGreater(stats0["violate_packets"], 0) self.assertGreater(stats1["conform_packets"], 0) self.assertEqual(stats1["exceed_packets"], 0) self.assertGreater(stats1["violate_packets"], 0) self.assertEqual( stats0["conform_packets"] + stats1["conform_packets"], stats["conform_packets"], ) self.assertEqual( stats0["violate_packets"] + stats1["violate_packets"], stats["violate_packets"], ) # Unbind the policer and repeat policer.bind_vpp_config(1, False) for worker in [0, 1]: self.send_and_expect(self.pg0, pkts, self.pg1, worker=worker) self.logger.debug(self.vapi.cli("show trace max 100")) # The policer should auto-bind to worker 0 when packets arrive stats = policer.get_stats() stats0new = policer.get_stats(worker=0) stats1new = policer.get_stats(worker=1) self.assertGreater(stats0new["conform_packets"], stats0["conform_packets"]) self.assertEqual(stats0new["exceed_packets"], 0) self.assertGreater(stats0new["violate_packets"], stats0["violate_packets"]) self.assertEqual(stats1, stats1new) # # Clean up # ip_punt_policer.remove_vpp_config() policer.remove_vpp_config() ip_punt_redirect.remove_vpp_config() class TestIP6Deag(VppTestCase): """IPv6 Deaggregate Routes""" @classmethod def setUpClass(cls): super(TestIP6Deag, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6Deag, cls).tearDownClass() def setUp(self): super(TestIP6Deag, self).setUp() self.create_pg_interfaces(range(3)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIP6Deag, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_ip_deag(self): """IP Deag Routes""" # # Create a table to be used for: # 1 - another destination address lookup # 2 - a source address lookup # table_dst = VppIpTable(self, 1, is_ip6=1) table_src = VppIpTable(self, 2, is_ip6=1) table_dst.add_vpp_config() table_src.add_vpp_config() # # Add a route in the default table to point to a deag/ # second lookup in each of these tables # route_to_dst = VppIpRoute( self, "1::1", 128, [VppRoutePath("::", 0xFFFFFFFF, nh_table_id=1)] ) route_to_src = VppIpRoute( self, "1::2", 128, [ VppRoutePath( "::", 0xFFFFFFFF, nh_table_id=2, type=FibPathType.FIB_PATH_TYPE_SOURCE_LOOKUP, ) ], ) route_to_dst.add_vpp_config() route_to_src.add_vpp_config() # # packets to these destination are dropped, since they'll # hit the respective default routes in the second table # p_dst = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src="5::5", dst="1::1") / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) p_src = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src="2::2", dst="1::2") / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) pkts_dst = p_dst * 257 pkts_src = p_src * 257 self.send_and_assert_no_replies(self.pg0, pkts_dst, "IP in dst table") self.send_and_assert_no_replies(self.pg0, pkts_src, "IP in src table") # # add a route in the dst table to forward via pg1 # route_in_dst = VppIpRoute( self, "1::1", 128, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)], table_id=1, ) route_in_dst.add_vpp_config() self.send_and_expect(self.pg0, pkts_dst, self.pg1) # # add a route in the src table to forward via pg2 # route_in_src = VppIpRoute( self, "2::2", 128, [VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index)], table_id=2, ) route_in_src.add_vpp_config() self.send_and_expect(self.pg0, pkts_src, self.pg2) # # loop in the lookup DP # route_loop = VppIpRoute(self, "3::3", 128, [VppRoutePath("::", 0xFFFFFFFF)]) route_loop.add_vpp_config() p_l = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src="3::4", dst="3::3") / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) self.send_and_assert_no_replies(self.pg0, p_l * 257, "IP lookup loop") class TestIP6Input(VppTestCase): """IPv6 Input Exception Test Cases""" @classmethod def setUpClass(cls): super(TestIP6Input, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6Input, cls).tearDownClass() def setUp(self): super(TestIP6Input, self).setUp() self.create_pg_interfaces(range(2)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIP6Input, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_ip_input_icmp_reply(self): """IP6 Input Exception - Return ICMP (3,0)""" # # hop limit - ICMP replies # p_version = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg1.remote_ip6, hlim=1) / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect_some(self.pg0, p_version * NUM_PKTS, self.pg0) for rx in rxs: icmp = rx[ICMPv6TimeExceeded] # 0: "hop limit exceeded in transit", self.assertEqual((icmp.type, icmp.code), (3, 0)) icmpv6_data = "\x0a" * 18 all_0s = "::" all_1s = "FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF" @parameterized.expand( [ # Name, src, dst, l4proto, msg, timeout ( "src='iface', dst='iface'", None, None, inet6.UDP(sport=1234, dport=1234), "funky version", None, ), ( "src='All 0's', dst='iface'", all_0s, None, ICMPv6EchoRequest(id=0xB, seq=5, data=icmpv6_data), None, 0.1, ), ( "src='iface', dst='All 0's'", None, all_0s, ICMPv6EchoRequest(id=0xB, seq=5, data=icmpv6_data), None, 0.1, ), ( "src='All 1's', dst='iface'", all_1s, None, ICMPv6EchoRequest(id=0xB, seq=5, data=icmpv6_data), None, 0.1, ), ( "src='iface', dst='All 1's'", None, all_1s, ICMPv6EchoRequest(id=0xB, seq=5, data=icmpv6_data), None, 0.1, ), ( "src='All 1's', dst='All 1's'", all_1s, all_1s, ICMPv6EchoRequest(id=0xB, seq=5, data=icmpv6_data), None, 0.1, ), ] ) def test_ip_input_no_replies(self, name, src, dst, l4, msg, timeout): self._testMethodDoc = "IPv6 Input Exception - %s" % name p_version = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6( src=src or self.pg0.remote_ip6, dst=dst or self.pg1.remote_ip6, version=3, ) / l4 / Raw(b"\xa5" * 100) ) self.send_and_assert_no_replies( self.pg0, p_version * NUM_PKTS, remark=msg or "", timeout=timeout ) def test_hop_by_hop(self): """Hop-by-hop header test""" p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg0.local_ip6) / IPv6ExtHdrHopByHop() / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) self.pg0.add_stream(p) self.pg_enable_capture(self.pg_interfaces) self.pg_start() class TestIP6Replace(VppTestCase): """IPv6 Table Replace""" @classmethod def setUpClass(cls): super(TestIP6Replace, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6Replace, cls).tearDownClass() def setUp(self): super(TestIP6Replace, self).setUp() self.create_pg_interfaces(range(4)) table_id = 1 self.tables = [] for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.generate_remote_hosts(2) self.tables.append(VppIpTable(self, table_id, True).add_vpp_config()) table_id += 1 def tearDown(self): super(TestIP6Replace, self).tearDown() for i in self.pg_interfaces: i.admin_down() i.unconfig_ip6() def test_replace(self): """IP Table Replace""" MRouteItfFlags = VppEnum.vl_api_mfib_itf_flags_t MRouteEntryFlags = VppEnum.vl_api_mfib_entry_flags_t N_ROUTES = 20 links = [self.pg0, self.pg1, self.pg2, self.pg3] routes = [[], [], [], []] # the sizes of 'empty' tables for t in self.tables: self.assertEqual(len(t.dump()), 2) self.assertEqual(len(t.mdump()), 5) # load up the tables with some routes for ii, t in enumerate(self.tables): for jj in range(1, N_ROUTES): uni = VppIpRoute( self, "2001::%d" % jj if jj != 0 else "2001::", 128, [ VppRoutePath( links[ii].remote_hosts[0].ip6, links[ii].sw_if_index ), VppRoutePath( links[ii].remote_hosts[1].ip6, links[ii].sw_if_index ), ], table_id=t.table_id, ).add_vpp_config() multi = VppIpMRoute( self, "::", "ff:2001::%d" % jj, 128, MRouteEntryFlags.MFIB_API_ENTRY_FLAG_NONE, [ VppMRoutePath( self.pg0.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_ACCEPT, proto=FibPathProto.FIB_PATH_NH_PROTO_IP6, ), VppMRoutePath( self.pg1.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_FORWARD, proto=FibPathProto.FIB_PATH_NH_PROTO_IP6, ), VppMRoutePath( self.pg2.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_FORWARD, proto=FibPathProto.FIB_PATH_NH_PROTO_IP6, ), VppMRoutePath( self.pg3.sw_if_index, MRouteItfFlags.MFIB_API_ITF_FLAG_FORWARD, proto=FibPathProto.FIB_PATH_NH_PROTO_IP6, ), ], table_id=t.table_id, ).add_vpp_config() routes[ii].append({"uni": uni, "multi": multi}) # # replace the tables a few times # for kk in range(3): # replace each table for t in self.tables: t.replace_begin() # all the routes are still there for ii, t in enumerate(self.tables): dump = t.dump() mdump = t.mdump() for r in routes[ii]: self.assertTrue(find_route_in_dump(dump, r["uni"], t)) self.assertTrue(find_mroute_in_dump(mdump, r["multi"], t)) # redownload the even numbered routes for ii, t in enumerate(self.tables): for jj in range(0, N_ROUTES, 2): routes[ii][jj]["uni"].add_vpp_config() routes[ii][jj]["multi"].add_vpp_config() # signal each table converged for t in self.tables: t.replace_end() # we should find the even routes, but not the odd for ii, t in enumerate(self.tables): dump = t.dump() mdump = t.mdump() for jj in range(0, N_ROUTES, 2): self.assertTrue(find_route_in_dump(dump, routes[ii][jj]["uni"], t)) self.assertTrue( find_mroute_in_dump(mdump, routes[ii][jj]["multi"], t) ) for jj in range(1, N_ROUTES - 1, 2): self.assertFalse(find_route_in_dump(dump, routes[ii][jj]["uni"], t)) self.assertFalse( find_mroute_in_dump(mdump, routes[ii][jj]["multi"], t) ) # reload all the routes for ii, t in enumerate(self.tables): for r in routes[ii]: r["uni"].add_vpp_config() r["multi"].add_vpp_config() # all the routes are still there for ii, t in enumerate(self.tables): dump = t.dump() mdump = t.mdump() for r in routes[ii]: self.assertTrue(find_route_in_dump(dump, r["uni"], t)) self.assertTrue(find_mroute_in_dump(mdump, r["multi"], t)) # # finally flush the tables for good measure # for t in self.tables: t.flush() self.assertEqual(len(t.dump()), 2) self.assertEqual(len(t.mdump()), 5) class TestIP6AddrReplace(VppTestCase): """IPv6 Interface Address Replace""" @classmethod def setUpClass(cls): super(TestIP6AddrReplace, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6AddrReplace, cls).tearDownClass() def setUp(self): super(TestIP6AddrReplace, self).setUp() self.create_pg_interfaces(range(4)) for i in self.pg_interfaces: i.admin_up() def tearDown(self): super(TestIP6AddrReplace, self).tearDown() for i in self.pg_interfaces: i.admin_down() def get_n_pfxs(self, intf): return len(self.vapi.ip_address_dump(intf.sw_if_index, True)) def test_replace(self): """IP interface address replace""" intf_pfxs = [[], [], [], []] # add prefixes to each of the interfaces for i in range(len(self.pg_interfaces)): intf = self.pg_interfaces[i] # 2001:16:x::1/64 addr = "2001:16:%d::1" % intf.sw_if_index a = VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config() intf_pfxs[i].append(a) # 2001:16:x::2/64 - a different address in the same subnet as above addr = "2001:16:%d::2" % intf.sw_if_index a = VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config() intf_pfxs[i].append(a) # 2001:15:x::2/64 - a different address and subnet addr = "2001:15:%d::2" % intf.sw_if_index a = VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config() intf_pfxs[i].append(a) # a dump should n_address in it for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 3) # # remove all the address thru a replace # self.vapi.sw_interface_address_replace_begin() self.vapi.sw_interface_address_replace_end() for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 0) # # add all the interface addresses back # for p in intf_pfxs: for v in p: v.add_vpp_config() for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 3) # # replace again, but this time update/re-add the address on the first # two interfaces # self.vapi.sw_interface_address_replace_begin() for p in intf_pfxs[:2]: for v in p: v.add_vpp_config() self.vapi.sw_interface_address_replace_end() # on the first two the address still exist, # on the other two they do not for intf in self.pg_interfaces[:2]: self.assertEqual(self.get_n_pfxs(intf), 3) for p in intf_pfxs[:2]: for v in p: self.assertTrue(v.query_vpp_config()) for intf in self.pg_interfaces[2:]: self.assertEqual(self.get_n_pfxs(intf), 0) # # add all the interface addresses back on the last two # for p in intf_pfxs[2:]: for v in p: v.add_vpp_config() for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 3) # # replace again, this time add different prefixes on all the interfaces # self.vapi.sw_interface_address_replace_begin() pfxs = [] for intf in self.pg_interfaces: # 2001:18:x::1/64 addr = "2001:18:%d::1" % intf.sw_if_index pfxs.append(VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config()) self.vapi.sw_interface_address_replace_end() # only .18 should exist on each interface for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 1) for pfx in pfxs: self.assertTrue(pfx.query_vpp_config()) # # remove everything # self.vapi.sw_interface_address_replace_begin() self.vapi.sw_interface_address_replace_end() for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 0) # # add prefixes to each interface. post-begin add the prefix from # interface X onto interface Y. this would normally be an error # since it would generate a 'duplicate address' warning. but in # this case, since what is newly downloaded is sane, it's ok # for intf in self.pg_interfaces: # 2001:18:x::1/64 addr = "2001:18:%d::1" % intf.sw_if_index VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config() self.vapi.sw_interface_address_replace_begin() pfxs = [] for intf in self.pg_interfaces: # 2001:18:x::1/64 addr = "2001:18:%d::1" % (intf.sw_if_index + 1) pfxs.append(VppIpInterfaceAddress(self, intf, addr, 64).add_vpp_config()) self.vapi.sw_interface_address_replace_end() self.logger.info(self.vapi.cli("sh int addr")) for intf in self.pg_interfaces: self.assertEqual(self.get_n_pfxs(intf), 1) for pfx in pfxs: self.assertTrue(pfx.query_vpp_config()) @unittest.skipIf( "ping" in config.excluded_plugins, "Exclude tests requiring Ping plugin" ) class TestIP6LinkLocal(VppTestCase): """IPv6 Link Local""" @classmethod def setUpClass(cls): super(TestIP6LinkLocal, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6LinkLocal, cls).tearDownClass() def setUp(self): super(TestIP6LinkLocal, self).setUp() self.create_pg_interfaces(range(2)) for i in self.pg_interfaces: i.admin_up() def tearDown(self): super(TestIP6LinkLocal, self).tearDown() for i in self.pg_interfaces: i.admin_down() def test_ip6_ll(self): """IPv6 Link Local""" # # two APIs to add a link local address. # 1 - just like any other prefix # 2 - with the special set LL API # # # First with the API to set a 'normal' prefix # ll1 = "fe80:1::1" ll2 = "fe80:2::2" ll3 = "fe80:3::3" VppNeighbor( self, self.pg0.sw_if_index, self.pg0.remote_mac, ll2 ).add_vpp_config() VppIpInterfaceAddress(self, self.pg0, ll1, 128).add_vpp_config() # # should be able to ping the ll # p_echo_request_1 = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=ll2, dst=ll1) / ICMPv6EchoRequest() ) self.send_and_expect(self.pg0, [p_echo_request_1], self.pg0) # # change the link-local on pg0 # v_ll3 = VppIpInterfaceAddress(self, self.pg0, ll3, 128).add_vpp_config() p_echo_request_3 = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=ll2, dst=ll3) / ICMPv6EchoRequest() ) self.send_and_expect(self.pg0, [p_echo_request_3], self.pg0) # # set a normal v6 prefix on the link # self.pg0.config_ip6() self.send_and_expect(self.pg0, [p_echo_request_3], self.pg0) # the link-local cannot be removed with self.vapi.assert_negative_api_retval(): v_ll3.remove_vpp_config() # # Use the specific link-local API on pg1 # VppIp6LinkLocalAddress(self, self.pg1, ll1).add_vpp_config() p_echo_request_1.dst = self.pg1.local_mac self.send_and_expect(self.pg1, [p_echo_request_1], self.pg1) VppIp6LinkLocalAddress(self, self.pg1, ll3).add_vpp_config() p_echo_request_3.dst = self.pg1.local_mac self.send_and_expect(self.pg1, [p_echo_request_3], self.pg1) @unittest.skipIf( "gre" in config.excluded_plugins, "Exclude tests requiring GRE plugin" ) def test_ip6_ll_p2p(self): """IPv6 Link Local P2P (GRE)""" self.pg0.config_ip4() self.pg0.resolve_arp() gre_if = VppGreInterface( self, self.pg0.local_ip4, self.pg0.remote_ip4 ).add_vpp_config() gre_if.admin_up() ll1 = "fe80:1::1" ll2 = "fe80:2::2" VppIpInterfaceAddress(self, gre_if, ll1, 128).add_vpp_config() self.logger.info(self.vapi.cli("sh ip6-ll gre0 fe80:2::2")) p_echo_request_1 = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IP(src=self.pg0.remote_ip4, dst=self.pg0.local_ip4) / GRE() / IPv6(src=ll2, dst=ll1) / ICMPv6EchoRequest() ) self.send_and_expect(self.pg0, [p_echo_request_1], self.pg0) self.pg0.unconfig_ip4() gre_if.remove_vpp_config() @unittest.skipIf( "gre" in config.excluded_plugins, "Exclude tests requiring GRE plugin" ) def test_ip6_ll_p2mp(self): """IPv6 Link Local P2MP (GRE)""" self.pg0.config_ip4() self.pg0.resolve_arp() gre_if = VppGreInterface( self, self.pg0.local_ip4, "0.0.0.0", mode=(VppEnum.vl_api_tunnel_mode_t.TUNNEL_API_MODE_MP), ).add_vpp_config() gre_if.admin_up() ll1 = "fe80:1::1" ll2 = "fe80:2::2" VppIpInterfaceAddress(self, gre_if, ll1, 128).add_vpp_config() p_echo_request_1 = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IP(src=self.pg0.remote_ip4, dst=self.pg0.local_ip4) / GRE() / IPv6(src=ll2, dst=ll1) / ICMPv6EchoRequest() ) # no route back at this point self.send_and_assert_no_replies(self.pg0, [p_echo_request_1]) # add teib entry for the peer teib = VppTeib(self, gre_if, ll2, self.pg0.remote_ip4) teib.add_vpp_config() self.logger.info(self.vapi.cli("sh ip6-ll gre0 %s" % ll2)) self.send_and_expect(self.pg0, [p_echo_request_1], self.pg0) # teardown self.pg0.unconfig_ip4() class TestIPv6PathMTU(VppTestCase): """IPv6 Path MTU""" def setUp(self): super(TestIPv6PathMTU, self).setUp() self.create_pg_interfaces(range(2)) # setup all interfaces for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIPv6PathMTU, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_path_mtu_local(self): """Path MTU for attached neighbour""" self.vapi.cli("set log class ip level debug") # # The goal here is not test that fragmentation works correctly, # that's done elsewhere, the intent is to ensure that the Path MTU # settings are honoured. # # # IPv6 will only frag locally generated packets, so use tunnelled # packets post encap # tun = VppIpIpTunInterface( self, self.pg1, self.pg1.local_ip6, self.pg1.remote_ip6 ) tun.add_vpp_config() tun.admin_up() tun.config_ip6() # set the interface MTU to a reasonable value self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2800, 0, 0, 0]) p_6k = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=tun.remote_ip6) / UDP(sport=1234, dport=5678) / Raw(b"0xa" * 2000) ) p_2k = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=tun.remote_ip6) / UDP(sport=1234, dport=5678) / Raw(b"0xa" * 1000) ) p_1k = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=tun.remote_ip6) / UDP(sport=1234, dport=5678) / Raw(b"0xa" * 600) ) nbr = VppNeighbor( self, self.pg1.sw_if_index, self.pg1.remote_mac, self.pg1.remote_ip6 ).add_vpp_config() # this is now the interface MTU frags self.send_and_expect(self.pg0, [p_6k], self.pg1, n_rx=4) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) # drop the path MTU for this neighbour to below the interface MTU # expect more frags pmtu = VppIpPathMtu(self, self.pg1.remote_ip6, 1300).add_vpp_config() # print/format the adj delegate and trackers self.logger.info(self.vapi.cli("sh ip pmtu")) self.logger.info(self.vapi.cli("sh adj 7")) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # increase the path MTU to more than the interface # expect to use the interface MTU pmtu.modify(8192) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) # go back to an MTU from the path pmtu.modify(1300) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # raise the interface's MTU # should still use that of the path self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2000, 0, 0, 0]) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # set path high and interface low pmtu.modify(2000) self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [1300, 0, 0, 0]) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # remove the path MTU self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2800, 0, 0, 0]) pmtu.modify(0) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) def test_path_mtu_remote(self): """Path MTU for remote neighbour""" self.vapi.cli("set log class ip level debug") # # The goal here is not test that fragmentation works correctly, # that's done elsewhere, the intent is to ensure that the Path MTU # settings are honoured. # tun_dst = "2001::1" route = VppIpRoute( self, tun_dst, 64, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)] ).add_vpp_config() # # IPv6 will only frag locally generated packets, so use tunnelled # packets post encap # tun = VppIpIpTunInterface(self, self.pg1, self.pg1.local_ip6, tun_dst) tun.add_vpp_config() tun.admin_up() tun.config_ip6() # set the interface MTU to a reasonable value self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2800, 0, 0, 0]) p_2k = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=tun.remote_ip6) / UDP(sport=1234, dport=5678) / Raw(b"0xa" * 1000) ) p_1k = ( Ether(dst=self.pg0.local_mac, src=self.pg0.remote_mac) / IPv6(src=self.pg0.remote_ip6, dst=tun.remote_ip6) / UDP(sport=1234, dport=5678) / Raw(b"0xa" * 600) ) nbr = VppNeighbor( self, self.pg1.sw_if_index, self.pg1.remote_mac, self.pg1.remote_ip6 ).add_vpp_config() # this is now the interface MTU frags self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) # drop the path MTU for this neighbour to below the interface MTU # expect more frags pmtu = VppIpPathMtu(self, tun_dst, 1300).add_vpp_config() # print/format the fib entry/dpo self.logger.info(self.vapi.cli("sh ip6 fib 2001::1")) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # increase the path MTU to more than the interface # expect to use the interface MTU pmtu.modify(8192) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) # go back to an MTU from the path pmtu.modify(1300) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # raise the interface's MTU # should still use that of the path self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2000, 0, 0, 0]) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # turn the tun_dst into an attached neighbour route.modify([VppRoutePath("::", self.pg1.sw_if_index)]) nbr2 = VppNeighbor( self, self.pg1.sw_if_index, self.pg1.remote_mac, tun_dst ).add_vpp_config() self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # add back to not attached nbr2.remove_vpp_config() route.modify([VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)]) # set path high and interface low pmtu.modify(2000) self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [1300, 0, 0, 0]) self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=3) self.send_and_expect(self.pg0, [p_1k], self.pg1, n_rx=2) # remove the path MTU self.vapi.sw_interface_set_mtu(self.pg1.sw_if_index, [2800, 0, 0, 0]) pmtu.remove_vpp_config() self.send_and_expect(self.pg0, [p_2k], self.pg1, n_rx=2) self.send_and_expect(self.pg0, [p_1k], self.pg1) class TestIPFibSource(VppTestCase): """IPv6 Table FibSource""" @classmethod def setUpClass(cls): super(TestIPFibSource, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPFibSource, cls).tearDownClass() def setUp(self): super(TestIPFibSource, self).setUp() self.create_pg_interfaces(range(2)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_arp() i.generate_remote_hosts(2) i.configure_ipv6_neighbors() def tearDown(self): super(TestIPFibSource, self).tearDown() for i in self.pg_interfaces: i.admin_down() i.unconfig_ip4() def test_fib_source(self): """IP Table FibSource""" routes = self.vapi.ip_route_v2_dump(0, True) # 2 interfaces (4 routes) + 2 specials + 4 neighbours = 10 routes self.assertEqual(len(routes), 10) # dump all the sources in the FIB sources = self.vapi.fib_source_dump() for source in sources: if source.src.name == "API": api_source = source.src if source.src.name == "interface": intf_source = source.src if source.src.name == "adjacency": adj_source = source.src if source.src.name == "special": special_source = source.src if source.src.name == "default-route": dr_source = source.src # dump the individual route types routes = self.vapi.ip_route_v2_dump(0, True, src=adj_source.id) self.assertEqual(len(routes), 4) routes = self.vapi.ip_route_v2_dump(0, True, src=intf_source.id) self.assertEqual(len(routes), 4) routes = self.vapi.ip_route_v2_dump(0, True, src=special_source.id) self.assertEqual(len(routes), 1) routes = self.vapi.ip_route_v2_dump(0, True, src=dr_source.id) self.assertEqual(len(routes), 1) # add a new soure that'a better than the API self.vapi.fib_source_add( src={"name": "bgp", "priority": api_source.priority - 1} ) # dump all the sources to check our new one is there sources = self.vapi.fib_source_dump() for source in sources: if source.src.name == "bgp": bgp_source = source.src self.assertTrue(bgp_source) self.assertEqual(bgp_source.priority, api_source.priority - 1) # add a route with the default API source r1 = VppIpRouteV2( self, "2001::1", 128, [VppRoutePath(self.pg0.remote_ip6, self.pg0.sw_if_index)], ).add_vpp_config() r2 = VppIpRouteV2( self, "2001::1", 128, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)], src=bgp_source.id, ).add_vpp_config() # ensure the BGP source takes priority p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst="2001::1") / inet6.UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) self.send_and_expect(self.pg0, [p], self.pg1) r2.remove_vpp_config() r1.remove_vpp_config() self.assertFalse(find_route(self, "2001::1", 128)) class TestIPxAF(VppTestCase): """IP cross AF""" @classmethod def setUpClass(cls): super(TestIPxAF, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIPxAF, cls).tearDownClass() def setUp(self): super(TestIPxAF, self).setUp() self.create_pg_interfaces(range(2)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.config_ip4() i.resolve_arp() i.resolve_ndp() def tearDown(self): super(TestIPxAF, self).tearDown() for i in self.pg_interfaces: i.admin_down() i.unconfig_ip4() i.unconfig_ip6() def test_x_af(self): """Cross AF routing""" N_PKTS = 63 # a v4 route via a v6 attached next-hop VppIpRoute( self, "1.1.1.1", 32, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)], ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IP(src=self.pg0.remote_ip4, dst="1.1.1.1") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) for rx in rxs: self.assertEqual(rx[IP].dst, "1.1.1.1") # a v6 route via a v4 attached next-hop VppIpRoute( self, "2001::1", 128, [VppRoutePath(self.pg1.remote_ip4, self.pg1.sw_if_index)], ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst="2001::1") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) for rx in rxs: self.assertEqual(rx[IPv6].dst, "2001::1") # a recursive v4 route via a v6 next-hop (from above) VppIpRoute( self, "2.2.2.2", 32, [VppRoutePath("2001::1", 0xFFFFFFFF)] ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IP(src=self.pg0.remote_ip4, dst="2.2.2.2") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) # a recursive v4 route via a v6 next-hop VppIpRoute( self, "2.2.2.3", 32, [VppRoutePath(self.pg1.remote_ip6, 0xFFFFFFFF)] ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IP(src=self.pg0.remote_ip4, dst="2.2.2.3") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) # a recursive v6 route via a v4 next-hop VppIpRoute( self, "3001::1", 128, [VppRoutePath(self.pg1.remote_ip4, 0xFFFFFFFF)] ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst="3001::1") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) for rx in rxs: self.assertEqual(rx[IPv6].dst, "3001::1") VppIpRoute( self, "3001::2", 128, [VppRoutePath("1.1.1.1", 0xFFFFFFFF)] ).add_vpp_config() p = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst="3001::2") / UDP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) rxs = self.send_and_expect(self.pg0, p * N_PKTS, self.pg1) for rx in rxs: self.assertEqual(rx[IPv6].dst, "3001::2") class TestIPv6Punt(VppTestCase): """IPv6 Punt Police/Redirect""" def setUp(self): super(TestIPv6Punt, self).setUp() self.create_pg_interfaces(range(4)) for i in self.pg_interfaces: i.admin_up() i.config_ip6() i.resolve_ndp() def tearDown(self): super(TestIPv6Punt, self).tearDown() for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() def test_ip6_punt(self): """IPv6 punt police and redirect""" # use UDP packet that have a port we need to explicitly # register to get punted. pt_l4 = VppEnum.vl_api_punt_type_t.PUNT_API_TYPE_L4 af_ip6 = VppEnum.vl_api_address_family_t.ADDRESS_IP6 udp_proto = VppEnum.vl_api_ip_proto_t.IP_API_PROTO_UDP punt_udp = { "type": pt_l4, "punt": { "l4": { "af": af_ip6, "protocol": udp_proto, "port": 7654, } }, } self.vapi.set_punt(is_add=1, punt=punt_udp) pkts = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src=self.pg0.remote_ip6, dst=self.pg0.local_ip6) / UDP(sport=1234, dport=7654) / Raw(b"\xa5" * 100) ) * 1025 # # Configure a punt redirect via pg1. # nh_addr = self.pg1.remote_ip6 ip_punt_redirect = VppIpPuntRedirect( self, self.pg0.sw_if_index, self.pg1.sw_if_index, nh_addr ) ip_punt_redirect.add_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) # # add a policer # policer = VppPolicer(self, "ip6-punt", 400, 0, 10, 0, rate_type=1) policer.add_vpp_config() ip_punt_policer = VppIpPuntPolicer(self, policer.policer_index, is_ip6=True) ip_punt_policer.add_vpp_config() self.vapi.cli("clear trace") self.pg0.add_stream(pkts) self.pg_enable_capture(self.pg_interfaces) self.pg_start() # # the number of packet received should be greater than 0, # but not equal to the number sent, since some were policed # rx = self.pg1._get_capture(1) stats = policer.get_stats() # Single rate policer - expect conform, violate but no exceed self.assertGreater(stats["conform_packets"], 0) self.assertEqual(stats["exceed_packets"], 0) self.assertGreater(stats["violate_packets"], 0) self.assertGreater(len(rx), 0) self.assertLess(len(rx), len(pkts)) # # remove the policer. back to full rx # ip_punt_policer.remove_vpp_config() policer.remove_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) # # remove the redirect. expect full drop. # ip_punt_redirect.remove_vpp_config() self.send_and_assert_no_replies(self.pg0, pkts, "IP no punt config") # # Add a redirect that is not input port selective # ip_punt_redirect = VppIpPuntRedirect( self, 0xFFFFFFFF, self.pg1.sw_if_index, nh_addr ) ip_punt_redirect.add_vpp_config() self.send_and_expect(self.pg0, pkts, self.pg1) ip_punt_redirect.remove_vpp_config() def test_ip6_punt_dump(self): """IPv6 punt redirect dump""" # # Configure a punt redirects # nh_address = self.pg3.remote_ip6 ipr_03 = VppIpPuntRedirect( self, self.pg0.sw_if_index, self.pg3.sw_if_index, nh_address ) ipr_13 = VppIpPuntRedirect( self, self.pg1.sw_if_index, self.pg3.sw_if_index, nh_address ) ipr_23 = VppIpPuntRedirect( self, self.pg2.sw_if_index, self.pg3.sw_if_index, "::" ) ipr_03.add_vpp_config() ipr_13.add_vpp_config() ipr_23.add_vpp_config() # # Dump pg0 punt redirects # self.assertTrue(ipr_03.query_vpp_config()) self.assertTrue(ipr_13.query_vpp_config()) self.assertTrue(ipr_23.query_vpp_config()) # # Dump punt redirects for all interfaces # punts = self.vapi.ip_punt_redirect_dump(sw_if_index=0xFFFFFFFF, is_ipv6=True) self.assertEqual(len(punts), 3) for p in punts: self.assertEqual(p.punt.tx_sw_if_index, self.pg3.sw_if_index) self.assertNotEqual(punts[1].punt.nh, self.pg3.remote_ip6) self.assertEqual(str(punts[2].punt.nh), "::") class TestIP6InterfaceRx(VppTestCase): """IPv6 Interface Receive""" @classmethod def setUpClass(cls): super(TestIP6InterfaceRx, cls).setUpClass() @classmethod def tearDownClass(cls): super(TestIP6InterfaceRx, cls).tearDownClass() def setUp(self): super(TestIP6InterfaceRx, self).setUp() self.create_pg_interfaces(range(3)) table_id = 0 for i in self.pg_interfaces: i.admin_up() if table_id != 0: table = VppIpTable(self, table_id, is_ip6=1) table.add_vpp_config() i.set_table_ip6(table_id) i.config_ip6() i.resolve_ndp() table_id += 1 def tearDown(self): for i in self.pg_interfaces: i.unconfig_ip6() i.admin_down() i.set_table_ip6(0) super(TestIP6InterfaceRx, self).tearDown() def test_interface_rx(self): """IPv6 Interface Receive""" # # add a route in the default table to receive ... # route_to_dst = VppIpRoute( self, "1::", 122, [ VppRoutePath( "::", self.pg1.sw_if_index, type=FibPathType.FIB_PATH_TYPE_INTERFACE_RX, ) ], ) route_to_dst.add_vpp_config() # # packets to these destination are dropped, since they'll # hit the respective default routes in table 1 # p_dst = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src="5::5", dst="1::1") / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) pkts_dst = p_dst * 10 self.send_and_assert_no_replies(self.pg0, pkts_dst, "IP in table 1") # # add a route in the dst table to forward via pg1 # route_in_dst = VppIpRoute( self, "1::1", 128, [VppRoutePath(self.pg1.remote_ip6, self.pg1.sw_if_index)], table_id=1, ) route_in_dst.add_vpp_config() self.send_and_expect(self.pg0, pkts_dst, self.pg1) # # add a route in the default table to receive ... # route_to_dst = VppIpRoute( self, "1::", 122, [ VppRoutePath( "::", self.pg2.sw_if_index, type=FibPathType.FIB_PATH_TYPE_INTERFACE_RX, ) ], table_id=1, ) route_to_dst.add_vpp_config() # # packets to these destination are dropped, since they'll # hit the respective default routes in table 2 # p_dst = ( Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) / IPv6(src="6::6", dst="1::2") / inet6.TCP(sport=1234, dport=1234) / Raw(b"\xa5" * 100) ) pkts_dst = p_dst * 10 self.send_and_assert_no_replies(self.pg0, pkts_dst, "IP in table 2") # # add a route in the table 2 to forward via pg2 # route_in_dst = VppIpRoute( self, "1::2", 128, [VppRoutePath(self.pg2.remote_ip6, self.pg2.sw_if_index)], table_id=2, ) route_in_dst.add_vpp_config() self.send_and_expect(self.pg0, pkts_dst, self.pg2) if __name__ == "__main__": unittest.main(testRunner=VppTestRunner)