ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE The Netwok Layer IPv4 and IPv6 Part 1 Jean Yves Le Boudec 2014 1 Contents 1. The Two Principles of IP Unicast 2. IPv4 addresses 3. IPv6 addresses 4. Subnets and Masks 5. NATs 6. ARP Textbook Chapter 5: The Network Layer 2
IP Principle #1 = Structured addresses + Longest prefix match Recall goal of Internet Protocol (IP) = interconnect all systems in the world Principle #1: every interface has an IP address IP address is structured to reflect where the system is in the world every packet contains IP address of destination every system has a forwarding table ( = routing table) and performs longest prefix match on address destination 3 Lisa A.H1 Forwarding table to output B.* 2 A.* 0 1 Forwarding table to output A.* 1 B.D.* 2 B.* 3 0 2 1 3 1 0 router R1 to: B.D.H2 router R2 router R3 2 Forwarding table to output A.* 1 B.D.* 1 B.C.* 0 Homer B.C.H2 to output A.* 1 B.* 2 1 router R4 2 Bart B.D.H2 Forwarding table 4
IP Principle #2 = Don t use routers inside a LAN B WiFi base station P W E Ethernet concentrator B E and W P should not go through router W E goes through router Terminology: LAN = subnet IP principle 2 says: between subnets use routers, inside subnet don t 5 We observe a packet from W to P at 1. Which IP destination address do we see? 1. The IP address of P 1 0% 0% 0% 0% 2. The IP address of an Ethernet interface of the Ethernet concentrator 3. There is no destination IP address in the packet since communication is inside the subnet and does not go through a router 4. I don t know The IP address of P dress of an Ethernet... no destination IP a... I don t know 6
The Internet Protocol (IP) Communication between IP hosts requires knowledge of IP addresses An IP address is unique across the whole network (= the world in general) An IP address is the address of an interface There are two versions: IPv4 (current version) and IPv6 (next version) There are two network layers: Internet4 and Internet6 Terminology: packet = IP data unit intermediate system = system that forwards data units to another system; an IP intermediate system is called a router an IP system that does not forward is called a host 7 2. IPv4 addresses IPv4 address Uniquely identifies one interface in the world (in principle) An IPv4 address is 32 bits, usually noted in dotted decimal notation dotted decimal: 4 integers (one integer = 8 bits) example 1: 128.191.151.1 example 2: 129.192.152.2 hexadecimal: 8 hexa digits (one hexa digit = 4 bits) example 1: x80 bf 97 01 example 2: x81 c0 98 02 binary: 32 bits example 1: b1000 0000 1011 1111 1001 0111 0000 0001 example 2: b1000 0001 1100 0000 1001 1000 0000 0010 8
Binary, Decimal and Hexadecimal Given an integer B the basis : any integer can be represented in base B by means of an alphabet of B symbols Usual cases are decimal: 234 binary: 1110 1010 hexadecimal: ea Mapping binary < > hexa is simple: one hexa digit is 4 binary digits e = 1110 a = 1010 ea = b1110 1010 Mapping binary < > decimal is best done by a calculator 1110 1010 = 128 + 64 + 32 + 8 + 2 = 234 Special Cases to remember f = 1111 = 15 ff = 11111111 = 255 9 lrcsuns 128.178.156.24 08:00:20:71:0D:D4 lrcpc3 128.178.156.7 00:00:C0:B8:C2:8D Example Modem + PPP sic500cs 128.178.84.130 128.178.84.133 182.5 in-inr 128.178.156.1 00:00:0C:02:78:36 LRC ed2-in 182.1 DI ETHZ-Backbone 129.132.100.12 129.132.100.27 ezci7-ethz-switch 129.132.35.1 sw-zu-03 Switch 130.59.x.x sw-la-01 128.178.47.3 128.178.84.1 128.178.47.5 ed0-ext EPFL-Backbone ed0-swi 15.13 128.178.100.12 stisun1 15.7 15.221 128.178.100.3 128.178.182.3 in-inj 128.178.79.1 00:00:0C:17:32:96 disun3 128.178.79.9 lrcmac4 08:00:20:20:46:2E 128.178.156.23 Ring SIDI SUN 08:00:07:01:a2:a5 ezci7-ethz-switch ed2-el Komsys 129.132 66.46 LEMA lrcmac4 128.178.29.64 08:00:07:01:a2:a5 10
Network Prefix Network prefixes are used in routing tables /24 is the prefix length in bits Extract from routing table at sw-la-0 128.178.100.2 Destination Next hop 128.178.29.1 128.178/16 128.178.47.5 0/0 130.59.23.2 Extract from routing table at ed0-swi Destination Next hop 128.178.29/24 128.178.100.2 128.178/16 128.178.100.3 0/0 128.178.47.3 11 Special Addresses 0.0.0.0 absence of address 127.0.0/24 this host (loopback address) for example 127.0.0.1 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16 private networks (e.g in IEW) cannot be used on the public Internet 169.254.0.0/16 link local address (can be used only between systems on same LAN) 224/4 multicast 240/5 reserved 255.255.255.255/32 link local broadcast 12
IPv4 Packet Format Higher layer protocol 1= ICMP, 6 = TCP, 17 = UDP) Header 20 bytes (+ options, if any) payload We will see the functions of the fields other than the addresses in a following module 13 3. IPv6 Addresses The current IP is IPv4. IPv6 is the next version of IP Why a new version? IPv4 address space is too small (32 bits 4 10 addresses ) What does IPv6 do? Redefine packet format with a larger address: 128 bits ( 3 10 addresses) Otherwise essentially the same as IPv4 IPv6 is incompatible with IPv4; routers and hosts must handle both separately A can talk to W, B can talk to W, A and B cannot communicate Web browser W Application IPv6 TCP MAC IPv4 Dual Stack Local router IPv6 MAC IPv4 B HTTP TCP IPv6 A HTTP TCP IPv4 14
Routing Tables at ed0 swi IP address of next hop lrcsuns 128.178.156.24 08:00:20:71:0D:D4 Destination Next hop 2001:620:618:1a4/64 fe80::1%1 2001:620:618/48 fe80::4%2 ::/0 fe80::1%2 128.178.47.3 sw-la-01 2001:620:618:10a::1 fe80::1 %1 stisun1 ed0-ext ed0-swi 128.178.47.3 %2 128.178.15.7 2001:620:618:10b::1 2001:620:618:1a6:1:80b2:f66:1 fe80::1 182.5 ed2-in 128.178.182.3 2001:620:618:1ac::3 2001:620:618:1ac::5 182.1 in-inr in-inj 2001:620:618:1ac::1 128.178.156.1 DI 128.178.79.1 2001:620:618:1ad::1 2001:620:618:1ab::1 00:00:0C:02:78:36 00:00:0C:17:32:96 LRC interface number 128.178.47.3 2001:620:618:10b::4 fe80::4 ed2-el 128.178.29.1 2001:620: 618:1a4::1 lrcpc3 128.178.156.7 00:00:C0:B8:C2:8D lrcmac4 2001:620:618:1ad:0a00:20ff:fe78:30f9 08:00:20:78:30:F9 15 IPv6 addresses are 128 bit long and are written using hexadecimal digits an EPFL public address: 2001:620:618:1a6:0a00:20ff:fe78:30f9 an EPFL private address: fd24:ec43:12ca:1a6:0a00:20ff:fe78:30f9 This is a private address EPFL private 16
Compression Rules for IPv6 Addresses 1 piece = 16 bits = [0 4 ]hexa digits; leading zeroes in one piece are omitted ; prefer lower case pieces separated by : (colon) one IPv6 address uncompressed = 8 pieces :: replaces any number of 0s in more than one piece; appears only once in address uncompressed 2002:0000:0000:0000:0000:ffff:80b2:0c26 2001:0620:0618:01a6:0000:20ff:fe78:30f9 compressed 2002::ffff:80b2:c26 2001:620:618:1a6:0:20ff:fe78:30f9 17 A Few IPv6 Global Unicast Addresses The block 2000/3 (i.e. 2xxx and 3xxx) is allocated for global unicast addresses 2001:620::/32 Switch 2001:620:618::/48 EPFL 2001:620:8::/48 ETHZ 2a02:1200::/27 Swisscom 2001:678::/29 provider independent address 2001::/32 Teredo 2002::/16 6to4 18
Examples of Special Addresses EPFL Private fc00::/7 (i.e. fcxx: and fdxx:) For example fd24:ec43:12ca:1a6:a00: 20ff:fe78:30f9 fe80::/10 ::/128 absence of address ::1/128 this host (loopback address) Unique local addresses = private networks (e.g in IEW) cannot be used on the public Internet link local address (can be used only between systems on same LAN) ff00::/8 multicast ff02::1:ff00:0/104 Solicited node multicast ff02::1/128 ff02::2/128 link local broadcast all link local routers 19 IPv6 Packet Format e.g. Higher layer protocol 1= ICMP, 6 = TCP, 17 = UDP) 16 bytes Header 40 bytes (+ options, if any) We will see the functions of the fields other than the addresses in a following module payload 20
The dotted decimal notation for 80 1: is 1. 128.193.255.255 2. 228.393.255.255 3. I don t know 0% 0% 0% 128.193.255.255 228.393.255.255 I don t know 21 The hexadecimal notation «2001::bad:babe» denotes a string of 1. 32 bits 2. 44 bits 3. 48 bits 4. 64 bits 5. 128 bits 6. None of the above 7. I don t know 7% 2% 11% 4% 77% 0% 0% 32 bits 44 bits 48 bits 64 bits 128 bits None of the above I don t know 22
4. Subnets and Masks Recall the IP principles: longest prefix match + routers between subnets only An IP system needs to know (in its forwarding table) which addresses are in same LAN as self ( = on link ) ed2 in has an IPv4 packet to destination address 128.178.15.7: packet is sent directly to 128.178.15.7 128.178.156.24 : packet is sent to in inr This is done using the subnet mask dest next-hop interface 128.178.15.0/24 On-link eth0 128.178.10.0/24 On-link eth1 128.178.182.0/24 On-link eth2 128.178.156.0/24 128.178.182.5 eth2 0.0.0.0/0 128.178.10.1 eth1 23 One IP subnet must correspond to one network part An IP address is usually interpreted a network part a host part 128.178.151.24 2001:620:618:1a6:0a00:20ff:fe78:30f9 Network part identifies subnet One subnet = one LAN All hosts in same LAN must have same network part The size of the network part may vary EPFL IPv4 network part is 24 bits ETHZ IPv4 network part 26 bits IPv6 network part is very often 64 bits 24
A system computes its network part from its IP address using the subnet mask, configured with the address = string of bits equal to 1 in network part, to 0 in host part network == IP address && mask At EPFL, IPv4 mask = 255.255.255.0 128.178.15.7 and 128.178.15.221 are on same subnet because 128.178.15.7 && 255.255.255.0 = 128.178.15.221 && 255.255.255.0 =128.178.15.0 IPv6 mask is very often 64 bits i.e. = ffff:ffff:ffff:ffff:: The notation /<length of network part> is also used Subnet Mask We could use the notation /24 instead of 255.255.255.0 25 Same as saying Mask = ffff:ffff:ffff:ffff:: 26
Reserved Addresses with IPv4 0 and «all 1» are often avoided as host part Example: 128.178.156.0 and 128.178.156.255 are avoided (to prevent confusions with broadcast) 27 IPv4 address classes Long ago, IPv4 addresses had a class subnet mask was not necessary This is now obsolete but many people continue to use it. 0 1 2 3 8 16 24 31 class A 0 Net Id Subnet Id Host Id class B 10 Net Id Subnet Id Host Id class C 110 Net Id Host Id class D 1110 Multicast address class E 11110 Reserved Class A B C D E Range 0.0.0.0 to 127.255.255.255 128.0.0.0 to 191.255.255.255 192.0.0.0 to 223.255.255.255 224.0.0.0 to 239.255.255.255 240.0.0.0 to 247.255.255.255 28
bridge?... 187.44.....??...253...1? 192.44.78.254 X? bridge host A 192.44.77.254 192.44.77.2 Can Host A have this address? Masks are all 255.255.255.0 1. Yes 2. No 3. I don t know 0% 0% 0% Yes No know 29 The IPv4 Subnet Mask at ETHZ is 1. 255.255.255.0 2. 255.255.255.1 3. 255.255.255.2 4. 255.255.255.192 5. 255.255.255.198 6. 255.255.255.332 7. ffff:ffff:ffff:ffff:: 8. ffff:ffff:ffff:ffff:c000:: 9. I don t know 14% 4% 11% 57% 7% 4% 4% 0% 0% 255.255.255.0 255.255.255.1 255.255.255.2 255.255.255.192 255.255.255.198 255.255.255.332 ffff:ffff:ffff:ffff:: ffff:ffff:ffff:ffff:c000:: I don t know 30
The IPv6 Subnet Mask at ETHZ is 1. 255.255.255.0 2. 255.255.255.1 3. 255.255.255.2 4. 255.255.255.192 5. 255.255.255.198 6. /48 7. ffff:ffff:ffff:ffff:: 8. ffff:ffff:ffff:ffff:c000:: 9. I don t know 4% 0% 0% 4% 2% 4% 56% 27% 4% 255.255.255.0 255.255.255.1 255.255.255.2 255.255.255.192 255.255.255.198 /48 ffff:ffff:ffff:ffff:: ffff:ffff:ffff:ffff:c000:: I don t know 31 Solution The subnet mask at ETHZ for IPv4 is 26 bits, i.e., in binary 11111111 11111111 11111111 11000000 in dotted decimal, this is 255.255.255.192 For IPv6 it is 64 bits (the most frequent case), i.e. ffff:ffff:ffff:ffff:: 32
What is the subnet broadcast address for subnet 129.132.100.0/26? 1. 129.132.100.0 2. 129.132.100.15 3. 129.132.100.63 4. 129.132.100.192 5. 129.132.100.255 6. I don t know 0% 0% 0% 0% 0% 0% 129.132.100.0 129.132.100.15 129.132.100.63 129.132.100.192 129.132.100.255 I don t know 33 Solution The subnet address is 26 bit long and is 129.132.100.0. The subnet broadcast address is obtained by setting the host part (i.e. the last 6 bits) to all 1 s. The last byte is 0 0000 0000 replace the last 6 bits of the last byte by 1 0011 1111 = 63 the last byte of the subnet broadcast address is 63 The subnet broadcast address is 129.132.100.63 34
5. NATs: Why invented? (Network Address Translation boxes) Goal: re use same IP address for several devices / use private addresses This is a special type of «middle box», that is violating the TCP/IP architecture Used in residential networks («ADSL Modem») Used in companies to save IP addresses 35 How does Network Address Translation Work? NAT box modifies IP address and port numbers (port numbers are in TCP and UDP headers see transport protocol module) Maps (IP address, protocol type, port number) Exact matching from NAT Table To: 192.168.10.11 UDP : 1029 To: 130.104.228.200 UDP : 3441 To: 192.168.10.10 UDP : 1029 To: 130.104.228.200 UDP : 3442 LAN IPv 4 NAT box LAN Internet 192.168.10.11 udp 1029 130.104.228.200 udp 3441 192.168.10.11 udp 1029 130.104.228.200 udp 3442 NAT table 36
Creating a NAT table entry: on the fly From: 192.168.10.11 TCP : 1765 To: 201.19.32.45 Port 80 From: 130.104.228.200 TCP : 2343 To: 201.19.32.45 Port 80 LAN IPv 4 NAT box LAN Internet 192.168.10.10 udp 1029 130.104.228.200 udp 3441 192.168.10.11 udp 1029 130.104.228.200 udp 3442 192.168.10.11 tcp 1765 130.104.228.200 tcp 2343 NAT table 201.19.32.45 37 Why some applications don t work with NATs S NAT Assume A behind a NAT and S in the internet Communication between A and S must be initiated by A «punch a hole through the NAT» If A and S are both behind a NAT (e.g. with voice over IP), we have a bootstrap problem A does not know its IP address as seen by S Solving this requires a third party this is what made Skype s fortune Cone or restricted NATs: third party used only to discover translated address Symmetric NAT: third party relays all traffic 38
Types of NATs Cone: translated (address, port) = f (internal (address, port)) ; remains valid as long as refreshed Symmetric: translated (address, port) = f(internal (address, port), correspondent (address, port)) Restricted: translated (address, port) = f (internal (address, port)) ; but valid only for specific correspondent address traffic from an unknown correspondent address is silently discarded This is only a rough classification; many things may happen in practice; NATs are a hack! 39 NAT44 and NAT66 NATs are motivated primarily by shortage of IPv4 addresses NAT44 maps IPv4 IPv4 addresses Widespread Many believe that there is no need for NAT66 since there are as many IPv6 addresses as one may ever need NAT66 are not widespread But NAT66 may be needed for other reasons use private addresses eg IEWv6 to EPFL v6 RFC 6296 specifies NAT66 40
When a NAT has a packet to forward and an association exists in the NAT table A. The NAT looks for a longest prefix match B. The NAT looks for an exact match C. None of the above D. I don t know The NAT looks for a longest. The NAT looks for an exact. 0% 0% 0% 0% None of the above I don t know 41 From WAN to LAN, the NAT may modify A. The source port B. The destination port C. None of the above D. I don t know The source port 0% 0% 0% 0% The destination port None of the above I don t know 42
Solution To: 130.104.228.200 TCP : 2343 To: 192.168.10.11 TCP : 1765 From LAN to WAN, the NAT modifies source port and source address From WAN to LAN, the NAT modifies destination port and destination address 43