VPP for Stateless SRv6/GTP-U Translation
- 1. VPP Prototyping for Stateless SRv6/GTP-U Translation 2019.05.28 S.Matsushima, K.Ebisawa
- 2. Objectives • Implements the key SRv6 functions defined by “I-D.ietf-dmm-srv6- mobile-uplane*” • Needs a flexible platform to impalement all possible functions • Try to hack the missing features already known. • Disclose the implementation codes as open source • Provides feedbacks to the I-D through the hack • To Improve the specs. • Identify the hidden missing parts of the spec. *: https://datatracker.ietf.org/doc/draft-ietf-dmm-srv6-mobile-uplane/
- 3. SRv6 Mobile Upane Functions Target Function Description End.MAP Forwards the receiving IPv6 packet and update the IPv6 DA with mapped SID. End.M.GTP6.D Decap the receiving GTP/UDP/IPv6 packet and encap with IPv6 header, or IPv6 header with SRH based on the address/ID mapping rule and binding SR-Policy End.M.GTP6.E Decap the receiving IPv6+SRH packet and encap with IPv6/UDP/GTP-U header based on the address/ID mapping rule. End.M.GTP4.E Decap the receiving IPv6+SRH packet and encap with IPv4/UDP/GTP-U header base on the address/ID mapping rule. T.M.Tmap Decap the receiving GTP/UDP/IPv4 packet and encap with IPv6 header, or IPv6 header with SRH based on the address/ID mapping rule and binding SR-Policy. End.Limit Limit the throughput of the packet flow with mapped SID. **NEW** Encoding GTP-U Messages into SRH. **NEW** Decoding GTP-U Messages from SRH.
- 4. End.MAP in Downlink Internet gNBUE N3 N9 N6 IPv6 Hdr SA = U2::, DA = G::1 IPv6 Hdr SA = Z::, DA = A:: IPv6 Hdr SA = Z::, DA = A:: IPv6 HdrSA = U2::, DA = U1:: IPv6 Hdr SA = Z::, DA = A:: IPv6 HdrSA = Z::, DA = A:: End.DX End.MAP T.Encaps.Red UPF1 U1 UPF2 U2 SMF AMF N2 N11 N4 Payload Payload Payload Payload G::1
- 5. End.M.GTP6.D in Uplink IPv6 Hdr SA = A::, DA = Z:: Payload IPv6 Hdr SA = G1::, DA = X:A::1 UDP - GTP SA = A::, DA = Z:: Payload TEID: 1234 IPv6 Hdr • Without modifying the N2 interface we are steering the UE packets to an SRGW along the designated algorithm path. IPv6 Hdr SA = X::, DA = S1:: SR Hdr (U2::, C1::) SL=2 Payload IPv6 Hdr SA = A::, DA = Z:: IPv6 Hdr SA = X::, DA = C1:: SR Hdr (U2::, C1::) SL=1 Payload IPv6 Hdr SA = A::, DA = Z:: IPv6 Hdr SA = X::, DA = U2:: Payload IPv6 Hdr SA = A::, DA = Z:: Payload IPv6 Hdr SA = A::, DA = Z:: • SRGW advertise SID in LOC(64b of X ) + Algorithm(8b of A and B) format : • X:A::/72 -> Low latency • X:B::/72 -> High bandwidth • SRGW does End.M.GTP6.D function for X:A:: and X:B::. • X:A::1 is a type of End.M.GTP6.D SID that pops the GTP header without lookup and pushes an outer header with SID list • X:A:1 could be a flex-alg SID (end-to-end network slicing) for other SRv6 nodes. • X:A::1 can be shared across UEs SMF InternetSR GW X UPF U2 AMF N2 N9 N6 N4 N11 Old: IPv6 addr “X::1” + TEID (1234) New: IPv6 SID “X:A::1” + TEID (1234) X:A:1 is BSID at X TEID (1234) could be dummy TE C1 NFV S1 N9N9 gNB A G1 UE N3 End.M.GTP.6.D End.DT6End.* End.* End.*: Appropriate SRv6 End function type for the purposes.
- 6. End.M.GTP6.E in Downlink SMF SR GW X AMF IPv6 Hdr SA = Z::, DA = A:: Payload IPv6 Hdr SA = X::, DA = G1:: UDP - GTP SA = Z::, DA = A:: Payload TEID: 3456 IPv6 Hdr • SRGW does End.M.GTP6.E function. • X::2 is an SRv6 End.M.GTP6.E SID. • Removes SRH, adds UDP and GTP with TEID received in SID ARGs • IPv6 DA is the last segment of the SRH • Scales (no state per UE in SR gateway) • gNB could configure an IPv6 addresses per network slice. • G1::1 -> Low latency • G1::2 -> High bandwidth IPv6 Hdr SA= U2::, DA= X::2:3456 SR Hdr ( G1::, X::2:3456, S1::) SL=1 IPv6 Hdr SA = Z::, DA = A:: Payload IPv6 Hdr SA = Z::, DA = A:: Payload • UPF U2 (anchor point) maintains state of A: • FIB lookup: A/128 is matched to <C1::, S1::, X::2:3456, G1::> • SRGW SID X:: need be pre-configured, or through N4 or other means. N2 N4 N11 UPF U2N9 N6N9N9 IPv6 Hdr SA= U2::, DA= S1 SR Hdr ( G1::, X::2:3456, S1::) SL=2 IPv6 Hdr SA = Z::, DA = A:: Payload IPv6 Hdr SA= U2::, DA= C1 SR Hdr ( G1::, X::2:3456, S1::) SL=3 IPv6 Hdr SA = Z::, DA = A:: Payload Remote Endpoint == G1:: + TEID (3456) G1:: is a gNB IPv6 addr TE C1 NFV S1 Internet gNB A G1 UE N3 End.M.GTP.6.E T.Encaps.RedEnd.*End.* End.*: Appropriate SRv6 End function type for the purposes.
- 7. End.M.GTP4.E in Downlink SMF SR GW X AMF N2 N11 UPF U2N9 N6 N4 N9N9 IPv6 Hdr SA = Z::, DA = A:: Payload IPv4 Hdr SA = SRGW, DA = G1 UDP - GTP SA = Z::, DA = A:: Payload TEID: 3456 IPv6 Hdr IPv6 Hdr SA= U2::, DA= X::G1:SRGW:3456 SR Hdr (X::G1:SRGW:3456, S1::)SL=0 IPv6 Hdr SA = Z::, DA = A:: IPv6 Hdr SA = Z::, DA = A:: Payload IPv6 Hdr SA= U2::, DA= S1 SR Hdr (X::G1:SRGW:3456, S1::) SL=1 IPv6 Hdr SA = Z::, DA = A:: Payload IPv6 Hdr SA= U2::, DA= C1 SR Hdr (X::G1:SRGW:3456, S1::) SL=2 IPv6 Hdr SA = Z::, DA = A:: Payload • X::G1:SRGW:3456is an SRv6 End.M.GTP4.E function • Removes SRH, adds IPv4/UDP and GTP • IPv4 SA, DA and GTP TEID are part of the SID’s arguments • Scales (no state per UE in SR gateway) Payload IPv4 addr of gNB + TEID (3456) TE C1 NFV S1 Internet gNB A G1 UE N3 End.M.GTP.4.E T.Encaps.RedEnd.* End.* End.*: Appropriate SRv6 End function type for the purposes.
- 8. SMF SR GW X AMF N2 N11 T.M.Tmap in Uplink UPF U2N9 N6 N4 N9N9 IPv6 Hdr SA = A::, DA = Z:: Payload IPv4 Hdr SA = G1, DA = SRGW UDP - GTP SA = A::, DA = Z:: Payload TEID: 1234 IPv6 Hdr IPv6 Hdr SA = X::, DA = S1:: SR Hdr (U2::, C1::) SL=2 Payload IPv6 Hdr SA = A::, DA = Z:: IPv6 Hdr SA = X::, DA = C1:: SR Hdr (U2::, C1::) SL=1 Payload IPv6 Hdr SA = A::, DA = Z:: IPv6 Hdr SA = X::, DA = U2:: Payload IPv6 Hdr SA = A::, DA = Z:: Payload IPv6 Hdr SA = A::, DA = Z:: • SRGW does T.M.Tmap function. • SRGW is a UPF Uplink Classifier. Coexist with current UPFs. • Steers traffic into SR policy. IPv4 addr of SRGW + TEID (1234) TE C1 NFV S1 Internet gNB A G1 UE N3 T.M.Tmap End.DTEnd.* End.* End.*: Appropriate SRv6 End function type for the purposes.
- 9. SRv6 Mobile User Plane in IETF104 Hackathon
- 10. Feedback from Hackathon • Two target functions has been implemented for VPP and P4 Switch. • New mapping rule of GTP-U<->SRv6 has been studied • It is possible to support both ‘Args.Mob.Session’ and ‘Drop-in’ scenario. • The codes are now open-sourced: • VPP and P4 Target Function Description End.MAP Forwards the receiving IPv6 packet and update the IPv6 DA with mapped SID. End.M.GTP6.D Decap the receiving GTP/UDP/IPv6 packet and encap with IPv6 header, or IPv6 header with SRH based on the address/ID mapping rule and binding SR-Policy End.M.GTP6.E Decap the receiving IPv6+SRH packet and encap with IPv6/UDP/GTP-U header based on the address/ID mapping rule. End.M.GTP4.E (SRv6 -> GTP-U) Decap the receiving IPv6+SRH packet and encap with IPv4/UDP/GTP-U header base on the address/ID mapping rule. T.M.Tmap (GTP-U -> SRv6) Decap the receiving GTP/UDP/IPv4 packet and encap with IPv6 header, or IPv6 header with SRH based on the address/ID mapping rule and binding SR-Policy. End.Limit Limit the throughput of the packet flow with mapped SID. **NEW** Translate GTP-U Echo Request to ICMP Echo Request and vice versa **NEW** Translate GTP-U Echo Reply to ICMP Echo Request and vice versa
- 11. Hackathon Topology Hack target VPP1 VPP4 D4:: IPv4 over GTP/UDP/IPv4 IPv4 over SRH/IPv6 VPP2 D2:: IPv4 over IPv6 IPv4 over GTP/UDP/IPv4 T.M.Tmap End.M.GTP4.E VPP3 D3:: scapy scapy
- 12. VPP Command Lines for SRv6 Mobile UPlane Function Command Line End.M.GTP4.E “sr localsid (del) address <IPv6 address of SRGW>/<mask> behavior end.m.gtp4.e” T.M.Tmap “sr policy add bsid <bsid> next A1:: next B1:: next C1:: (weight 5) (fib-table 3) gtp4_removal sr-prefix <ipv6_prefix_interworking>/<mask> local-prefix <src_prefix>/<mask>” “sr steer l3 <IPv4 dst addr of GTP-U>/<mask> via bsid <bsid>”
- 13. ToDo: • Contribute stateless GTP-U/SRv6 translation to FD.io • Need ”T.M.tmap” and “End.M.GTP4.E” refactoring to be integrated into VPP distribution. • Support IPv6 over GTP-U case with “T.M.Tmap” and “End.M.GTP4.E”. • Implement rest of the functions: • End.M.GTP6.E/D • End.MAP • Encode/decode GTP-U Messages to/from SRH
- 14. SRv6 Mobile Hackathon team in IETF104@Prague