Segment routing with IPv6 data plane (SRv6)

Introduction to SRv6

Segment routing steers packets by encoding the packet-processing instructions for each intermediate and destination router directly in the packet header.

The datapath pushes a list of instructions in the form of Segment Identifiers (SIDs) onto the packet, to forward the payload packet directly to the destination using the shortest path or using source routing via one or more transit routers. Each router that terminates a SID in the segment list performs the instructions related to that SID.

Segment Routing standards specify the following two methods for programming the datapath of the encapsulated packet:

  • Segment Routing MPLS (SR-MPLS)

    The SID is encoded in 32 bits and programmed as an MPLS label; this method provides a tunnel to both IPv4 and IPv6 destinations. See Segment routing with MPLS data plane (SR-MPLS).

  • Segment Routing IPv6 (SRv6)

    The SID is encoded in 128 bits and programmed as an IPv6 address; this method provides a tunnel to IPv6 destinations.

SRv6 datapath encapsulation models each SID using a 128-bit address. In shortest path routing, the destination SID is encoded in the Destination Address (DA) field of the outer IPv6 header. In source routing, the SIDs of the nodes the packet must traverse are encoded as a SID list in a Segment Routing Header (SRH), which is a new type of routing header compliant with RFC 8200. The next SID in a segment list to which the packet is to be forwarded is copied from the SRH into the DA field of the outer IPv6 header.

SRv6 provides more than IPv6 transport with shortest path and source-routing capabilities; it provides a framework for programmability of IPv6 networks that takes advantage of the large IPv6 address space.

The following figure shows the SRv6 SRH format and fields (excerpt from RFC 8986).

Figure 1. SRv6 SRH format and fields
Table 1. SRv6 field descriptions
Field name Description

Next Header

Defined in RFC 8200, Section 4.4

Hdr Ext Len

Defined in RFC 8200, Section 4.4

Routing Type

4

Segments Left

Defined in RFC 8200, Section 4.4

Last Entry

Contains the index (zero based), in the Segment List, of the last element of the Segment List

Flags

RFC 8754, Section 8.1 creates an IANA registry for new flags to be defined.

Tag

The Tag field is used to mark a packet as part of a class or group of packets; for example, packets sharing the same set of properties. When the Tag field is not used at the source, it must be set to zero on transmission. When the Tag field is not used during SRH processing, it is ignored. The Tag field is not used when processing the SID, as defined in RFC 8754, Section 4.3.1. It may be used when processing other SIDs that are not defined in this document. The allocation and use of tag is outside the scope of this document.

Segment List[0..n]

128-bit IPv6 addresses representing the nth segment in the segment list. The Segment List is encoded starting from the last segment of the SR policy. That is, the first element of the segment list (Segment List[0]) contains the last segment of the SR policy, the second element contains the penultimate segment of the SR policy, and so on.

TLV

Type Length Value (TLV); see RFC 8754, Section 2.1, and SRv6 SID encoding.

The following flags are defined for the SRv6 SRH Flags field.

Figure 2. 8-bits of flags
Table 2. Flag descriptions
Flag Description

U

Unused and for future use.

Must be 0 on transmission and ignored on receipt.

The following figure shows the SRv6 Segment Identifier (SID) encoding format and fields.

Figure 3. SRv6 SID encoding

The 128-bit address of an SRv6 SID is split into a three-field structure: {LOCATOR:FUNCTION:ARGUMENT}. The size of these fields is configurable.

  • LOCATOR field - encodes the transport or reachability information
  • FUNCTION field - encodes the node SID function (End SID), an adjacency SID function (End.X SID), or a service function that is the equivalent of a service label in SR-MPLS
  • ARGUMENT field - typically encodes a value that identifies the source Ethernet Segment for EVPN services that require multihoming or Etree procedures; can be used to carry limited service or application metadata

The following figures shows the operation of the data and control planes when an IP-VPN route is resolved to an SRv6 tunnel.

Figure 4. SRv6 data and control plane operation

The PE6 egress router advertises the locator route that contains its locator prefix and, optionally, a local node SID (End) in IS-IS. It also advertises the SID of each adjacency (End.X) to its IS-IS neighbors.

The locator prefix provides the route to reach PE6 and is used by other routers to forward an SRv6 packet destined for any SID owned by PE6. Other routers use the End and End.X SIDs to create the repair tunnel for the Remote LFA and TI-LFA backup paths.

In BGP, PE6 advertises a VPN-IPv4 route and includes the End.DT4 service SID, which is equivalent to the SR-MPLS service label in the label per-VRF model. Unlike the SR-MPLS service label, the SRv6 End.DT4 SID contains both the function value that identifies the specific VRF-ID in PE6 and the locator prefix that provides the reachability to router PE6.

The PE1 router resolves the received VPN-IPv4 route by validating the next hop and checking the reachability of the locator prefix of PE6 in the routing table. When PE1 receives an IPv4 packet from a CE node, it pushes an outer IPv6 header that contains the End.DT4 SID in the DA field and looks up the address in the routing table. The packet is then forwarded to one of the next hops of the route of the locator prefix of PE6.

SR OS also supports micro-segment SRv6. The system supports operating in both modes (regular and micro) concurrently on the same platform, but requires the SIDs of each type to come from different SID blocks.

Micro-segment SRv6 provides the same functionality as regular SRv6 but uses 16-bit SIDs while regular SRv6 uses 128-bit SIDs. The 16-bit SIDs are referred to as micro-SIDs.

Micro-SIDs follow the general structure of regular SIDs (LOCATOR:FUNCTION:ARGUMENT), although the following differentiates the two.

In regular SRv6, identifiers are assigned to nodes and form, together with the SID block, the LOCATOR part of any SID. SIDs are assigned per node and associated with a specific behavior (or function), resulting in the LOCATOR:FUNCTION structure. Micro-segment SRv6 introduces a specific function (uN) which acts both as a locator (when combined with the block) and as a function (corresponding to END in regular SRv6) but without using any FUNCTION bits. In other words, the LOCATOR and END constructs of regular SRv6 correspond to a unique construct in micro-segment SRv6: <block><uN>.

Micro-segment SRv6 allows micro-SIDs to be compressed. Compression consists in coalescing several micro-SIDs into a 128-bit structure called a container.

The following figure shows three micro-SIDs, each identifying a different node in a network. The figure shows the result of compressing these three micro-SIDs in a container. Compression is possible because all micro-SIDs belong to the same block.

Figure 5. Compression of three micro-SIDs in a container

A container can be placed in the DA field of an IPv6 header and in a segment of the segment list of an SRH.

Compression allows the router to build longer paths with less overhead than with regular SRv6. However, it leads to specific datapath behaviors. See Datapath support for more information.

Micro-segment SRv6 introduces specific micro-segment SRv6 functions that correspond to functions defined for regular SRv6 (see Micro-segment SRv6 and Overview of the BGP requirements).

In general, any SR OS capability that applies to regular SRv6 also applies to micro-segment SRv6. In the documentation, only the differences are highlighted and commonalities are not repeated using micro-segment SRv6 specific terminology.

Configuring the locator and SIDs

Configuring the SRv6 locator and SIDs

This section describes configuration of the SRv6 locator.

An SRv6 SID is a 128-bit IPv6 address which follows the structure defined by RFC 8986:

SRv6 SID={LOCATOR:FUNCTION:ARGUMENT}.

The user must configure the main SRv6 subnet for this node. This is the locator and is, essentially, an IPv6 subnet (prefix and length) that provides reachability (longest prefix match) to all the SIDs originated by this node. The prefix is encoded in the LOCATOR field of the SRv6 SID and can have a length of either 64 or 96 bits.

The locator is further subdivided into a SID block and a node ID. For example, locator 3FFE:0:0:A1::/64 has a SID block of 3FFE:0 and node ID of 0:A1.

All nodes participating in an SRv6 domain must draw their locator and SIDs from the same SID block. In the previous example, the SID block is subnet 3FFE:0::/32.

Use the following commands to configure the SRv6 locator.

  • MD-CLI

    *[pr:/configure router "Base" segment-routing segment-routing-v6 locator "test"]
    A:admin@test# tree detail
    +-- admin-state <keyword>
    +-- algorithm <number>
    +-- apply-groups <reference>
    +-- apply-groups-exclude <reference>
    +-- block-length <number>
    +-- function-length <number>
    +-- label-block <reference>
    +-- prefix
    |   +-- ip-prefix <global-unicast-ipv6-address-and-prefix>
    +-- static-function
    |   +-- label-block <reference>
    |   +-- max-entries <number>
    +-- termination-fpe <reference>
    +-- origination-fpe <reference>
    +-- source-address <global-unicast-ipv6-address>
    
    
  • classic CLI

    configure+--router
       +--segment-routing
          +--segment-routing-v6
             +--origination-fpe* <fpe>
             +--source-address <ipv6-address>
             +--locator* <locator-name>
                +--admin-state (enable|disable)
                +--termination-fpe <fpe>
                +--algorithm <0, 128-255>
                +--prefix
                   +--ip-prefix <ipv6-address/prefix-length>
                +--block-length <0-96>
                +--function-length <16 | 20-96>
                +--label-block <block-name>
                +--static-function
                          +--max-entries <integer>
                          +--label-block <block-name> 
    

The two reserved label ranges (locator>label-block and locator>static-function>label-block) are mutually exclusive.

If locator>static-function>label-block is configured, then the labels associated with the service functions are drawn as follows:

  • from the label block for static service function

  • from the dynamic label range for dynamic service functions

If locator>label-block is configured, labels associated with service functions are drawn from the label block for both static and dynamic service functions.

Configuring a function length of 16 requires the use of locator>label-block. Function lengths of 16 are not supported on locator>static-function>label-block.

One locator is required for algorithm 0 and one for each IGP flexible algorithm (128-255). The same locator can be shared by multiple IGP instances for the same algorithm number.

The locator prefix,(in this example, 3FFE:0:0:A1::/64), is advertised in the SRv6 Locator TLV in IS-IS in both algorithm 0 and any configured flexible algorithm number, as defined in draft-ietf-lsr-isis-srv6-extensions. It is also advertised as a prefix in IP Reachability TLV (IS-IS TLV 236) in algorithm 0, so the routers that do not support SRv6 can still route the packet to the next hop of the locator and, eventually, to its destination node.

The FUNCTION field is user-configurable. The ARGUMENT field is set to all zeros and is not configurable. The ARGUMENT length must be signaled as zero in the IS-IS End and End.X SIDs and in the BGP service SIDs. Also, the sum of the LOCATOR and FUNCTION lengths must be less than or equal to 128.

Within algorithm 0 and each IGP flexible algorithm, the locator function (FUNCTION field) assigns the values of the End SID and End.X SID which correspond to the node SID and the adjacency or adjacency SET SID, respectively.

The locator function also assigns the value of the service SIDs owned by this node and advertised in the BGP control plane (End.DT4, End.DT6, End.DT46, and End.DX2).

The FUNCTION field can be subdivided into a static and a dynamic subrange. The user can draw from the static subrange to manually assign an SRv6 SID to a node, a local adjacency, or a service. IS-IS and BGP can draw from the dynamic subrange to assign a SID to a local adjacency or a service. The SID of an adjacency to an IS-IS neighbor, over a broadcast interface (LAN End.X), is always dynamically assigned and is not configurable.

The following CLI commands enable the allocation of an SRv6 SID function value. Manual allocation of a static function value is supported for nodes (End SIDs), adjacencies over a P2P interface (End.X SIDs), and service SIDs. Auto-allocation is supported for adjacencies over a P2P interface (End.X SIDs) and service SIDs.

  • MD-CLI

    [pr:/configure router "Base" segment-routing segment-routing-v6]
    A:admin@test# tree detail
    +-- base-routing-instance
    |   +-- apply-groups <reference>
    |   +-- apply-groups-exclude <reference>
    |   +-- locator <reference>
    |       +-- function
    |           +-- end <number>
    |           |   +-- srh-mode <keyword>
    |           +-- end-x-auto-allocate <keyword> protection <keyword>
    |           +-- end-x <number>
    |           |   +-- interface-name <reference>
    |           |   +-- protection <keyword>
    |           |   +-- srh-mode <keyword>
    |           +-- end-dt4
    |           +-- end-dt46
    |           +-- end-dt6
    
    
  • classic CLI

    configure
    +--router
    +---segment-routing
    |   +---segment-routing-v6
    |   |   +--- base-routing-instance
    |   |   |   |   locator <locator-name>
    |   |   |   |   +---function
    |   |   |   |   |   |   end <integer>
    |   |   |   |   |   |   +---srh-mode <psp | usp>   
    |   |   |   |   |   +---end-x-auto-allocate <psp|usp> <protected|unprotected>
    |   |   |   |   |   +---end-x <integer>
    |   |   |   |   |   |   +---interface <name>
    |   |   |   |   |   |   +---srh-mode <psp | usp>   
    |   |   |   |   |   |   +---protection <protected|unprotected>
    |   |   |   |   |   +---end-dt4 [<integer>]
    |   |   |   |   |   +---end-dt6 [<integer>]
    |   |   |   |   |   +---end-dt46 [<integer>]

Configuring the micro-segment locator and SIDs

Use the commands in the following contexts to configure micro-segment locators and SIDS.
configure router segment-routing segment-routing-v6 micro-segment
configure router segment-routing segment-routing-v6 micro-segment-locator
The following commands in micro-segment SRv6 apply to any (micro-segment) locator:
  • block-length

    This command provides the same function as for regular SRv6. The value must be the same on every platform network wide.

  • global-sid-entries
    This command defines the maximum number of unique micro-segment locators that can be configured network wide. The value must be the same on every platform network wide. This command splits in two the total number of values that can be encoded in 16 bits. A 16-bit field allows values in the range 0x0000 to 0xFFFF. Configuring 16 (default value) for the global-sid-entries value splits that range into the following:
    • 0x0000 to 0x3FFF for global identifiers
    • 0x4000 to 0xFFFF for local identifiers
  • sid-length

    This command defines the length of micro-SIDs. The only supported value is 16 (bits).

To configure a block, assign a name to it and configure specific commands associated with the block. Those commands include:
  • label-block

    The reserved label block serves both static and dynamic service functions. The theoretical maximum number of functions in micro-segment SRv6 is 216. A label block that provides more values than the theoretical maximum leads to a waste of labels. Because in practice, the number of local functions is less than (216 - global-sid-entries), a label-block should not need exceed that value.

  • prefix ip-prefix ipv6-prefix/prefix-length

    The IP prefix configures the block as an IPv6 address. The prefix-length value must be equal to the block-length.

  • static-function max-entries

    This command sets the maximum number of static functions the user needs.

To configure a micro-segment locator, use the following configuration knobs that are specific for micro-segment SRv6:
  • block

    This command configures a reference (by name) to a previously configured block.

  • un value

    This command creates a node identifier as an IPv6 address composed of the block part followed by a 16-bit SID. The value of the SID is the nth global SID entry. The value, which must be unique network wide, is configured as part of the locator because it acts as a locator.

The configuration of micro-SIDs follows the same logic as for regular SRv6 SIDs except that:
  • The user configures SIDs that are specific to micro-segment SRv6 (uA, uDT4, uDT6, uDT46).
  • The configured value only needs to be unique on the system and not network wide.
  • The resulting SID value is derived from the local range.
  • Although the uN function is equivalent to the regular SRv6 END function, it is not configurable under the Base instance because it is configured in the micro-segment-locator context.

IS-IS control plane extensions

Use the commands in the following context to enable SRv6 in the IS-IS instance and assign a locator to each algorithm (0 or flexible algorithm).

configure router isis segment-routing-v6 locator

The IS-IS control plane extensions in support of SRv6 are defined in draft-ietf-lsr-isis-srv6-extensions.

The IS-IS control plane advertises the SRv6 capabilities sub-TLV and the SRv6 Locator TLV. The latter includes the End function sub-TLV (equivalent to the prefix SID sub-TLV in SR-MPLS). IS-IS also advertises the End.X function sub-TLV (equivalent to the adjacency SID sub-TLV for a P2P link and a LAN in SR-MPLS) in the Extended IS Reachability TLV (top-level link TLV).

The weight fields in the End.X and LAN End.X sub-TLVs are not filled in on transmit and are ignored on receipt of the link TLV.

The following table describes the supported IS-IS SRv6 TLVs in SR OS.

Table 3. SRv6 IS-IS TLVs
SRv6 TLV/sub-TLV Codepoint IS-IS context TLV Description SR OS support

SRv6 Capabilities sub-TLV

25

Router CAPABILITY TLV (242)

Indicates SRv6 support

Yes

SR-Algorithm sub-TLV

19

Router CAPABILITY TLV (242)

Indicates base algorithm 0 and Flex-Algo 128-255 support

Yes

Maximum Segments Left MSD Type sub-TLV

41

Router CAPABILITY TLV (242)

Indicates how deep a node terminating current segment can process Segments Left field of the SRH to move the next SID to outer IPv6 header DA field

Yes

(Advertised value = 10 SIDs)

Received TLV is displayed in the LSDB but is not used for any purpose

Maximum End Pop MSD Type sub-TLV

42

Router CAPABILITY TLV (242)

Maximum number of SIDs in a SRH when a node removes the SRH (Penultimate Segment Pop (PSP) or Ultimate Segment Pop (USP) modes of SRH)

Yes

(Advertised value = 9 SIDs)

Received TLV is displayed in the LSDB but is not used for any purpose

Maximum H.Encaps MSD type sub-TLV

44

Router CAPABILITY TLV (242)

Indicates the maximum number of SIDs in an SRH that a router can push when forwarding an IP or L2 packet over a SRv6 policy

Yes

(Advertised value = 7 SIDs)

Received TLV is displayed in the LSDB but is not used for any purpose

Maximum End D MSD Type Sub-TLV

45

Router CAPABILITY TLV (242)

The maximum number of SIDs in an SRH when a node removes the SRH and performs the End.DX2/4/6 or End.DT4/6 function (USP mode of SRH)

Yes

(Advertised value = 9 SIDs)

Received TLV is displayed in the LSDB but is not used for any purpose

SRv6 Locator TLV

27

Is a Top-level IS-IS TLV

Advertises the locator prefix configured on this node to terminate SIDs in algorithm 0 and flex-algo 128-255

Yes

SRv6 End SID sub-TLV

5

SRv6 Locator TLV

Advertises the SID for the endpoint or End function (equivalent to the prefix SID sub-TLV in SR-MPLS)

Yes

Prefix Attribute Flags Sub-TLV

4

SRv6 Locator TLV

(Also in IP Reach TLV 236)

Provides attributes of a prefix that is leaked between IS-IS levels

Yes

SRv6 End.X SID sub-TLV

43

Top-level Extended IS reachability TLV (22)

Advertises the SID for the adjacency over a P2P link (equivalent to the adjacency SID sub-TLV for P2P link in SR-MPLS)

Yes

SRv6 LAN End.X SID sub-TLV

44

Top-level Extended IS reachability TLV (22)

Advertises the SID for the adjacency over a LAN (equivalent to the adjacency SID sub-TLV for LAN in SR-MPLS)

Yes

SRv6 SID Structure Sub-Sub-TLV

1

SRv6 End SID Sub-TLV

SRv6 End.X SID Sub-TLV

SRv6 LAN End.X SID Sub-TLV

Provides the length of each field (Block, Locator, Function, and Argument) of the SRv6 SID that it is advertised with

No

SR OS does not advertise this sub-sub-TLV. If received from other vendor's implementation, it is not displayed in the Link-State database and is also not propagated with the locator TLV

When both SR-MPLS and SRv6 are enabled on the same IS-IS instance, an MPLS node SID cannot be configured for a prefix of the locator or an End SID. This is because the SRv6 locator subnet cannot be added to a network interface and an MPLS node SID is configurable against a network interface only.

However, both the SRv6 locator tunnel and the SR-MPLS tunnel are programmed if IS-IS receives a /128 prefix that has both a locator TLV and a prefix SID TLV (with the node flag enabled) from a third-party router implementation. If the prefix SID is for a subnet larger than /128, only the SRv6 locator tunnel is programmed and the SR-MPLS tunnel is not.

Each function encoded in a SRv6 SID has its own endpoint behavior codepoint as listed in SRv6 SID function endpoint behavior codepoints.

Note:

SRv6 standards provide the flexibility to advertise transport and service SIDs in both IS-IS and BGP. In SR OS, the IS-IS control plane only advertises the transport SID functions shown in SRv6 SID function endpoint behavior codepoints and only uses the transport SIDs advertised in IS-IS for building LFA repair tunnels.

SR OS advertises service SID functions in the BGP control plane only as described in BGP service control plane extensions.

For the SRH processing and removal at the SID termination, the following modes of operation are associated with the termination of the End or End.X SID. These modes are sometimes referred to as SID flavors and IS-IS assigns a unique codepoint for each mode of the End or End.X SID of a given adjacency.

  • Basic or unflavored SRH mode

    The router that terminates an End or End.X SID and the Segments-Left field in the received packet is 0 before decrementing, keeps the SRH in the packet, and processes the packet identified by the next-header in the SRH. Typically, the next-header indicates another SRH and the packet is then forwarded based on the lookup of that next-SID; SR OS does not support this mode.

  • Ultimate Segment Pop (USP) SRH mode

    The egress PE terminates the last segment in the outer IPv6 header, removes the SRH and processes the inner service or control plane packet as indicated by the SRH Next-Header field. SR OS supports this mode.

  • Penultimate Segment Pop (PSP) SRH mode

    The router that terminates the End or End.X segment before the last in the segment list, meaning the Segments-Left field before decrementing has a value of 1, removes the SRH on behalf of the egress PE. SR OS supports this mode.

  • PSP&USP mode

    This is a combination of both the USP and PSP modes. The router that terminates the End or End.X segment applies the corresponding behavior for value 0 and 1 of the Segments-Left field. SR OS does not support this mode.

  • Ultimate Segment Decapsulation (USD) mode (PSP&USD, USP&USD, and PSP&USP&USD flavors)

    This is a variant of the USP mode in which the node removes the outer IPv6 header and associated SRH and moves directly to process the inner IPv6 packet as indicated by the SRH Next-Header field. This mode is used to terminate a TI-LFA or a Remote LFA repair tunnel originated with the H.Encaps.Red encapsulation.

    An SRv6 ingress PE or transit P router normally uses the H.Insert.Red behavior to build an LFA (remote LFA or a TI-LFA) repair tunnel. This inserts a dedicated SRH to carry the additional node and adjacency SIDs of the repair tunnel. SR OS supports originating and terminating repair tunnels using H.Insert.Red behavior.

    Some third party implementations use the H.Encaps.Red behavior, which inserts a dedicated outer IPv6 header and SRH to carry the additional SIDs of the LFA repair tunnel. While SR OS does not support originating an LFA repair tunnel with the H.Encaps.Red behavior, it does support terminating the repair tunnel.

    To allow third party implementations to activate the H.Encaps.Red repair tunnel, IS-IS in SR OS can signal support for the USD mode of the outer IPv6 header and SRH processing with End and End.X SIDs in classic SRv6, and with uN and uA in micro-segment SRv6.

    The user can configure the advertisement of the PSP&USD, USP&USD, or PSP&USP&USD flavor of the USD mode for each configured and auto-allocated node SID or adjacency SID. The PSP&USP&USD behavior is similar to the PSP&USD behavior because the datapath in SR OS always performs the USP SRH mode when Segments-Left = 0 on a received SRv6 packet.

Note: On the 7750 SR, when handling a uN+uA, the implementation always applies PSP&USP, regardless of the flavors associated with the uN and the uA.
Table 4. SRv6 SID function endpoint behavior codepoints
SID function endpoint behavior Codepoint (per RFC 8986) SID type: End.SID SID type: End.X SID SID type: LAN End.X SID Advertising protocol Supported

End

(PSP, USP, USD)

1-4, 28-31

Yes

No

No

IS-IS

Yes

IS-IS only

(PSP value = 2, USP value = 3)

(PSP&USD value = 29, USP&USD value = 30)

(PSP&USP&USD value = 31)

End.X

(PSP, USP, USD)

5-8, 32-35

No

Yes

Yes

IS-IS

Yes

IS-IS only

(PSP value = 6, USP value = 7)

(PSP&USD value = 33, USP&USD value = 34)

(PSP&USP&USD value = 35)

End.T

(PSP, USP, USD)

9-12, 36-39

Yes

No

No

IS-IS

No1

End.DX6

16

No

Yes

Yes

BGP or IS-IS

No1

End.DX4

17

No

Yes

Yes

BGP or IS-IS

No1

End.DT6

18

Yes

No

No

BGP or static

Yes1

BGP only

End.DT4

19

Yes

No

No

BGP or static

Yes1

BGP only

End.DT46

20

Yes

No

No

BGP or static

Yes1

BGP only

End.DX2

21

Yes

No

No

BGP or static

Yes1

BGP only

1 IS-IS saves SID sub-TLVs for endpoint behavior values that it does not support, if received from a third-party implementation. However, it only uses End and End.X endpoint behaviors in RLFA and TI-LFA.

BGP advertises the supported endpoint behaviors End.DT4, End.DT6, End.DT46, and End.DX2 and accepts any behavior codepoint with a supported NLRI type.

Micro-segment SRv6

Use commands in the following context to enable micro-segment SRv6 in the IS-IS instance and to assign a micro-segment locator to each algorithm (zero or flexible algorithm).
configure router isis segment-routing-v6 micro-segment-locator

The following table lists the TLVs that micro-segment SRv6 supports in addition to the TLVs described in SRv6 IS-IS TLVs.

Table 5. SRv6 IS-IS TLVs additionally supported by micro-segment SRv6
SRv6 TLV/sub-TLV Codepoint IS-IS context TLV Description SR OS support

SRv6 SID Structure sub-sub-TLV

1

SRv6 uN SID Sub-TLV,

SRv6 uA SID Sub-TLV,

SRv6 LAN uA SID Sub-TLV

Provides the length of each field (block, locator, function, and argument) of the SRv6 uSID it is advertised with

Yes

The following table lists the endpoint behavior codepoint for each SID function encoded in a micro-segment SRv6 SID.

Table 6. SRv6 SID function endpoint behavior codepoints
SID function endpoint behavior Codepoint SID type: End.SID SID type: End.X SID SID type: LAN End.X SID Advertising protocol Supported

uN

42-50

Yes

No

No

IS-IS

Yes 2

IS-IS only

(PSP value = 44, USP value = 45)

(PSP&USD value = 48, USP&USD value = 49)

(PSP&USP&USD value = 50)

uA

51-59

No

Yes

Yes

IS-IS

Yes 2

IS-IS only

(PSP value = 53, USP value = 54)

(PSP&USD value = 57, USP&USD value = 58)

(PSP&USP&USD value = 59)

1 IS-IS saves SID sub-TLVs for endpoint behavior values that it does not support, if received from a third-party implementation. However, it only uses End and End.X endpoint behaviors in RLFA and TI-LFA.

BGP advertises the supported endpoint behaviors End.DT4, End.DT6, End.DT46, and End.DX2 and accepts any behavior codepoint with a supported NLRI type.

SRv6 support in IS-IS multitopology

This feature extends the support of SRv6 to multitopology IPv6 (MT2) in IS-IS.

Feature configuration

The user first configures a locator for use in IS-IS IPv6 unicast multitopology. Use the following CLI syntax to configure a locator for use in IS-IS IPv6 unicast multitopology:
  • classic CLI
    configure
    +--router
       +--isis <isis-instance>
          +--segment-routing-v6
            +---no adj-sid-hold    
            |   adj-sid-hold <seconds>
            +---locator <name>
            |   no locator <name>
            |   +---level {1|2}    
            |   |   +---metric <metric>
            |   |   |   no metric
            |   +---level-capability {level-1|level-2|level-1/2}
            |   |   no level-capability
            |   +---multi-topology [mt0] [mt2]
            |   |   no multi-topology    
            |   +---no tag
            |   |   tag <tag>
            +---shutdown
            |   no shutdown
    
  • MD-CLI
    configure
    +--router
       +--isis <isis-instance>
          +--segment-routing-v6
             +-- adj-sid-hold <seconds>
             +--admin-state (enable|disable)
             +--locator <locator-name> 
                +--level-capability <level-1|level-2|level-1/2>
                +--level <1|2>
                   +--metric <metric>
                +--tag <tag>
            +---multi-topology [mt0] [mt2]
    

The user can enable one or more SRv6 local locators in a specific IS-IS instance. Each locator can be enabled in a single topology of an IS-IS instance, topology 0 (MT0) or topology 2 (MT2). By default, a locator name added to an IS-IS instance is enabled in MT0. A local locator can be used in multiple IS-IS instances and can be assigned to at most one IPv6 topology independently within each IS-IS instance.

Note: To enable processing of local and remote IPv6 prefixes and SRv6 locators in MT0 and MT2, use the admin-state enable MD-CLI command in the configure router isis segment-routing-v6 context. The user must also enable IPv6 routing in MT0 (using the isis ipv6-routing native command), in MT2 (using the isis multi-topology ipv6-unicast command), or in both to enable SRv6 forwarding in the appropriate topologies.

IS-IS control plane changes

When a local locator is enabled in the MT2 IPv6 unicast topology of an IS-IS instance, IS-IS advertises the following routes:

  • The prefix in a top level Multitopology Reachable IPv6 Prefixes TLV type 237 with the MT-ID field set to 2.
  • A top level SRv6 Locator TLV type 27 that contains the locator prefix as well as the End SID sub-TLVs associated with this local locator. The locator TLV is advertised with the MT-ID field set to 2.
  • The End.X or LAN End.X sub-TLV in the top level multitopology Extended IS Reachability TLV (222).
  • The TE Application Specific Link Attributes sub-TLV in the top level multitopology Extended IS Reachability TLV (222), which supports the flex-algo feature.
  • The Application-Specific Shared Risk Link Group (SRLG) TLV (238).

The following table summarizes the new or modified TLVs in support of SRv6 in IS-IS MT2.

Table 7. SRv6 IS-IS MT2 TLVs
Multitopology TLV Codepoint IS-IS context TLV Description SR support
Multitopology IPv6 Reach TLV 237 Is a top-level IS-IS TLV

The main prefix TLV

MT-ID field set to 2

Yes

SRv6 Locator TLV 27 Is a top-level IS-IS TLV

Indicates that the locator, configured on this node, is used to terminate SIDs in algorithm 0 and flex-algo 128-255

MT-ID field set to 2

Yes
SRv6 End SID Sub-TLV 5 SRv6 Locator TLV Advertises the SID for the endpoint or End function (equivalent to the prefix SID in SR-MPLS) Yes
Prefix Attribute Flags Sub-TLV 4

SRv6 Locator TLV

(also in IPv6 Reach TLV 237)

Indicates that the prefix of the locator is anycast Yes
SRv6 End.X SID Sub-TLV 43 Top-level Multitopology Extended IS Reachability TLV (222) Advertises the SID for the adjacency or End.X function over a P2P link (equivalent to the adjacency SID sub-TLV for P2P link in SR-MPLS) Yes
SRv6 LAN End.X SID sub-TLV 44 Top-level Multitopology Extended IS Reachability TLV (222) Advertises the SID for the adjacency or End.X function over a LAN (equivalent to the adjacency SID sub-TLV for LAN in SR-MPLS) Yes
SRv6 SID Structure Sub-Sub-TLV 1

SRv6 End SID Sub-TLV,

SRv6 End.X SID Sub-TLV,

SRv6 LAN End.X SID Sub-TLV

Provides the length of each field (Block, Locator, Function, and Argument) of the SRv6 SID that it is advertised with

No

SR OS does not advertise this sub-sub-TLV. If received from other vendor's implementation, it is not displayed in Link-State database and is also not propagated with the locator TLV.

Application Specific Link Attributes (ASLA) sub-TLV 16 Top-level Multitopology Extended IS reachability TLV (222) Advertises the link attributes for the flex-algo application Yes
Application-Specific Shared Risk Link Group (SRLG) TLV 238 Is a top-level IS-IS TLV Advertises the link SRLG attribute for the flex-algo application

No

The SRLG constraint is not supported in IS-IS MT0 or MT2. IS-IS does not advertise this TLV and does not use it for any purpose if received from the network.

In prior releases of SR OS, only local locator routes and local prefix routes were redistributed with an export policy between topologies MT0 and MT2 of the same IS-IS instance or different IS-IS instances. These local locators were redistributed as a prefix TLVs and not as a locator TLVs. In addition, the tag of the local locator was not redistributed. Routes of remote locators and remote prefixes were not redistributed.

To extend SRv6 at a boundary of two IS-IS instances operating IPv6 in different topologies, the SRv6 support in MT2 feature enhances this behavior as follows:
  • It allows the concurrent enabling of a local locator in multiple IS-IS instances and in either MT0 or MT2 topology of each of these IS-IS instances.
  • It allows the redistribution with an export policy of remote locator routes between MT0 and MT2 topologies of different IS-IS instances. The locator routes are redistributed as locator TLVs. In algorithm 0, a prefix TLV is advertised in addition to the locator TLV.

Redistribution between MT0 and MT2 topologies of the same IS-IS instance is not allowed for either remote locator or remote prefix routes.

Locator and SID resolution

The MT2 local locator, remote locator, SID resolution, and programming into the tunnel table and datapath tunnel follow the same rules as those for enabling SRv6 in the single topology (MT0).

If a remote IPv6 prefix route is received in both MT0 and MT2, the route with the lowest cost (IGP metric) is selected. If both routes have the same metric, the MT0 route is selected. The selected route is programmed in the route table and FIB. If that prefix also advertised a locator TLV, the corresponding SRv6 route is updated in the route table and FIB to point to the SRv6 tunnel which is programmed in the tunnel table.

Note:

The preference of the MT0 route over the MT2 route is solely based on comparing the cost of each route. So, a remote IPv6 prefix route without a locator TLV can win over a remote IPv6 route with a locator TLV. The programmed route in the route table is a regular IS-IS route and does not have an SRv6 tunnel associated with it.

The same selection rule applies to a locator TLV advertised with a flex-algo number in both MT0 and MT2. The selection is based on comparing the cost of the routes using the metric of that specific algorithm (IGP, TE, or latency metric). The selected SRv6 route is added to the tunnel table.

SRv6 locator summarization with IS-IS

If an IS-IS domain exists out of multiple areas, the operator must redistribute SRv6 locators between areas for inter-area SRv6-based transport services. Each SRv6 locator is associated with an applied topology algorithm as follows:
  • algorithm 0 for the base SPF topology
  • algorithm in the range of 128 to 255 for flexible algorithms

Scaling is impacted when all existing SRv6 locators are redistributed between all existing areas. SRv6 locator summarization with IS-IS reduces the sizes of the LSDB and the IPv6 routing table and increases network stability.

SRv6 locators can be summarized when they are redistributed from one area into another area. This helps to reduce the number of SRv6 locators existing in each area. A summary SRv6 algorithm-aware locator address is configured with the following syntax:

summary-address {ip-prefix/ip-prefix-length | ip-prefix netmask} [level] [tag tag] [algorithm <algo-id>]
no summary-address {ip-prefix/ip-prefix-length | ip-prefix netmask} 
The configuration parameters are as follows:
  • ip-prefix/ip-prefix-length | ip-prefix netmask contains the summary locator address of the route

  • level configures the level where the summary route is applied

  • tag assigns a route tag to the summary address

  • algo-id identifies the flex-algorithm topology for the summary address

    The following considerations apply:
    • When an algorithm-aware summary address is generated, all matching algorithm-aware member prefixes are suppressed.
    • When the algorithm is not explicitly configured, the summary address is only generated for algorithm 0 in an IP Reachability TLV, and all more specific prefixes are suppressed from both the IP reach TLV and the SRv6 locator TLV for algorithm 0.
    • When the algorithm is configured in the range 128 to 255 or 0, the summary address is generated only for member prefixes belonging to the configured algorithm (including explicit algorithm 0 configurations). Only member prefixes within the configured algorithm are summarized and suppressed, while member prefixes that belong to a different algorithm are not summarized and not suppressed.
The following apply when the summary address is configured and the route table has matching member prefixes:
  • For algorithm 0, when algorithm 0 is not explicitly configured, the summary is inserted as an IS-IS IP Reachability TLV and all more specific prefixes from both IP Reachability TLV and SRv6 locator TLV are suppressed.
  • For algorithm 0, when algorithm 0 is explicitly configured, the summary is inserted as an IS-IS IP Reachability TLV and as an SRv6 locator TLV, and all more specific prefixes from both the IP Reachability TLV and the SRv6 locator TLV are suppressed.
  • For algorithms in the range of 128 to 255, the summary is inserted as an SRv6 locator TLV and all more specific locator prefixes in the SRv6 locator TLV are suppressed.
  • The summary address SRv6 locator does not contain any END SIDs of member SRv6 locator prefixes in the SRv6 locator TLV.
The following apply when using administrative tags for SRv6 locator summaries:
  • When SRv6 locators are summarized from one IS-IS level into another IS-IS level, special care must be taken to avoid re-redistributing them back into the original IS-IS level and potentially causing routing loops. Routing filters must be used to prevent such routing loops.
  • Existing filters can use 32-bit administrative tags to match upon routes and avoid routing loops. This route tag can be set using the summary-address command when originating an algorithm-aware SRv6 summary locator.
  • A routing policy with a match tag supports matching to both classic IPv6 prefix tags and SRv6 locator tags.

Configuring IS-IS Flex-Algorithm for SRv6

SRv6 introduces flexible algorithms to the IPv6 data plane.

A router is provisioned with topology- or algorithm-specific locators for each topology or algorithm pair supported by that node. Each locator is a covering prefix for all SIDs provisioned on that router that have the matching topology or algorithm. Locators associated with flexible algorithms are not advertised in a Prefix Reachability TLV (236 or 237). However, locators associated with algorithm 0 are advertised in a Prefix Reachability TLV (236 or 237), which allows legacy routers that do not support SRv6 to install a forwarding entry for algorithm 0 SRv6 traffic.

Each SRv6 locator is associated with an algorithm (either algorithm 0 or a flexible algorithm in the range of 128 to 255) and each algorithm represents a topologically-constrained forwarding construct. The M-flag within the flexible algorithm prefix metric sub-TLV is not applicable to prefixes advertised as SRv6 locators. The metric field in the locator TLV is used regardless of the M-flag in the FAD advertisement.

A router configured to participate in a flexible algorithm must use the selected FAD to compute the corresponding routing table. The available options are as follows:
  • Algorithm 0 (legacy routing table entries) is constructed from information advertised as a traditional IP Reachability TLV or as an SRv6 locator TLV (27). When IP reach TLV and SRv6 locator TLV contain conflicting information, then the IP Reachability TLV information is used.
  • Algorithms ranging from 128 to 255 (Flex-Algorithm routing table entries) are constructed from information advertised and constructed from locators found in the SRv6 locator TLV (27).
For route leaking of flexible algorithm-aware SRv6 locators between IS-IS areas, the following rules apply when a topology TLV (IP Reachability TLV or SRv6 locator TLV) is leaked, including leaked locators and end SIDs:
  • For algorithm 0, this SRv6 locator route is programmed as regular IS-IS route. If an IS-IS route is readvertised and also has an SRv6 locator TLV, it is readvertised as a regular IP Reachability TLV and SRv6 locator TLV.
  • For algorithms ranging from 128 to 255, if locator leaking is enabled, the original SRv6 locator TLV is readvertised as a SRv6 locator TLV into the other area.
  • The default locator leaking behavior between levels is as follows:
    • For Level 1 to Level 2, leaking is enabled by default.
    • For Level 2 to Level 1, leaking is disabled by default.
    • Changing the default leaking behavior requires an export policy where the prefix-list keyword behavior is configured to match upon prefixes or locators found in the routing table regardless of the associated algorithm. The prefix-list keyword allows combined support for algorithm 0 locators (regular prefixes) and algorithm locators ranging from 128 to 255 (Flex-Algorithm prefixes).
      configure
      +---router
          +---policy-options
              +---[no] policy-statement <name>
                  +---from
                      +---[no] level [1|2]
                      +---[no] prefix-list
                      +---[no] protocol
                      +---[no] tag
                  +---to
      	                +---<snip>

If a locator is associated with a flexible algorithm and the LFA is enabled, then LFA paths to the locator prefix must be calculated using the flexible algorithm in the corresponding topology to guarantee that they follow the same constraints as the calculation of the primary paths. LFA paths must only use SRv6 SIDs advertised specifically for the flexible algorithm. The LFA configuration is inherited from algorithm 0. The anycast behavior of SRv6 flexible algorithms is inherited from the standard algorithm 0 (standard SPF) SRv6 configuration.

The IS-IS neighbor advertisements are topology-specific and not algorithm-specific. Therefore, the SRv6 End.X SIDs inherit topology from the associated neighbor advertisement, but the algorithm is specified in the individual SID. All End.X SIDs are a subnet of a locator with matching topology and algorithm which is advertised by the same node in an SRv6 locator TLV. The End.X SIDs that do not meet this requirement are ignored. All End.X SIDs must find a supernet by the subnet of a locator with the matching algorithm which is advertised by the same router in an SRv6 locator TLV. The End.X SIDs that do not meet this requirement are ignored.

IS-IS protocol limitations affect enabling SRv6 flexible algorithms on a broadcast network. On a broadcast network, the LAN End.X SIDs of all neighbors for all participating flexible algorithms need to be advertised in a single LSP fragment because each IS-IS TE-NBR with all its TLV blocks must be advertised in one IS-IS LSP fragment. The amount of information inserted by segment routing for SRv6 into the LSP fragment depends upon the number of the flexible algorithms used, the number of static or auto-end.X configured per locator, and if both SRv6 and SR-MPLS are deployed.

BGP service control plane extensions

This section provides an overview of the BGP service control plane extensions.

Overview of the BGP requirements

The BGP service control plane required extensions are specified in RFC 9252, BGP Overlay Services Based on Segment Routing over IPv6 (SRv6). BGP requires some changes in the IPv6, VPN-IPv4, VPN-IPv6, and EVPN family routes so that the egress PE can signal the following End programming behaviors to the ingress PE:

  • Layer 2 SRv6 service SIDs

    End.DX2
    Layer 2 decapsulation and cross-connect to an Epipe egress SAP, signaled by AD per-EVI routes
  • Layer 3 SRv6 service SIDs

    End.DT4
    a VPRN (or GRT) route-table lookup, signaled by VPN-IPv4 or EVPN in Interface-less (EVPN-IFL) IPv4 prefix routes (also by IPv4)
    End.DT6
    a VPRN (or GRT) route-table IPv6 lookup, signaled by VPN-IPv6 or EVPN-IFL IPv6 prefix routes (also by IPv6)
    End.DT46
    a VPRN route-table lookup for IPv4 or IPv6 prefixes, signaled by VPN-IPv4 or VPN-IPv6 or EVPN-IFL IPv4 or IPv6 prefix routes

The following figure shows an example for the End.DX2 behavior for EVPN-VPWS services.

Figure 6. End.DX2 behavior for EVPN-VPWS

The ingress and egress PEs behave as follows:

  • The egress PE (PE6) advertises an A-D per EVI route with the SRv6 Service SID that identifies the End.DX2 behavior. The service SID includes the configured locator in the Epipe (A6::), as well as the allocated function (E9), which identifies the Epipe at the egress PE.

  • The ingress PE (PE1) imports the A-D per EVI route and creates an EVPN destination in the corresponding Epipe to A6::E9.

  • When PE1 receives frames at the access SAP, it encapsulates the frames into an SRv6 packet using the configured IP SA. The IP DA is the EVPN destination SID.

  • Shortest path forwarding is considered in the example shown in End.DX2 behavior for EVPN-VPWS, and therefore the EVPN destination SID is encoded in the IP DA. If TI-LFA is required, PE1 modifies the encapsulation to include an SRH and additional SIDs.

  • When the SRv6 packet arrives at PE6, the SID encoded in the IP DA identifies the packet for termination on PE6 and the Epipe for decapsulation and forwarding.

Similar procedures are followed for the other required services.

The following table lists the functions that micro-segment SRv6 implements to support the requirements.

Table 8. Micro-segment SRv6 functions
SID function endpoint behavior Codepoint SID type: End.SID SID type: End.X SID SID type: LAN End.X SID Advertising protocol Supported

uDT6

62

Yes

No

No

BGP or static

Yes3

uDT4

63

Yes

No

No

BGP or static

Yes 3

uDT46

64

Yes

No

No

BGP or static

Yes 3

uDX2

65

Yes

No

No

BGP or static

Yes 3

uDT2U

67

Yes

No

No

BGP or static

Yes 3

uDT2M

68

Yes

No

No

BGP or static

Yes 3

1 BGP advertises the supported endpoint behaviors and accepts any behavior codepoint with a supported NLRI type.

BGP extensions

The following BGP extensions are supported, as per RFC 9252:

  • SRv6 Service TLV:

    • SRv6 Service TLV encoded in the BGP Prefix-SID attribute

    • SRv6 SID Information Sub-TLV (SRv6 Service Sub-TLV type 1) encoded in the SRv6 Service TLV

    • SRv6 SID Structure Sub-Sub-TLV (SRv6 Service Data Sub-Sub-TLV type 1)

  • Transposition of 16 or 20 bits of the FUNCTION to the Label field of the NLRI

  • Arg.FE2 arguments used for split-horizon filtering in EVPN multihoming, advertised in the EVPN AD per-ES routes with 16 bits of the argument transposed to the label field of the ESI Label extended community.

The BGP extensions are applied to the following routes by setting the behavior field in the SRv6 Services TLV, as per RFC 8986:

  • VPN-IPv4

  • VPN-IPv6

  • IPv4

  • IPv6

  • EVPN AD per-EVI

  • EVPN AD per-ES

  • EVPN MAC/IP

  • EVPN Inclusive Multicast Ethernet Tag

Advertising SRv6 service TLVs

The EVPN, VPN-IPv4, VPN-IPv6, IPv4, and IPv6 routes for the SRv6-enabled services are advertised along with the SRv6 Service TLV. The TLV format is described in RFC 9252 and shown in the following figure.

Figure 7. SRv6 service TLV format

The SRv6 Service TLV encoded in the BGP Prefix-SID attribute can have two different types:

  • Type 5 is used for Layer 3 service SIDs or the SIDs signaled for VPRN services with VPN-IP or EVPN-IFL routes. Layer 3 service SIDs are also supported for the base router along with IPv4 or IPv6 routes.

  • Type 6 is used for Layer 2 service SIDs or the SIDs signaled for Epipe or VPLS services. Type 6 is supported along with AD per-EVI and per-ES routes for Epipes.

The SRv6 Service TLV may contain an unordered list of sub-TLVs, but currently the SRv6 Service TLV is advertised with only one sub-TLV: the SRv6 SID Info sub-TLV (type 1). This sub-TLV encodes the following information:

  • For the SID value, the entire 128-bit SID is allocated to the service, including the locator configured for the service and the allocated FUNCTION (which can be dynamically allocated or statically configured on the service). The ARGUMENT is always 0.

  • SID flags are all zero.

  • Endpoint behavior encodes the behavior as in RFC 8986, in decimal values. The following values are relevant for SR OS:
    • 18 – End.DT6

    • 19 – End.DT4

    • 20 – End.DT46

    • 21 – End.DX2

    • 24 – End.DT2m (only for AD per-ES routes)

  • One SID Structure Sub-Sub-TLV (Service Data Sub-Sub-TLV type 1)

The SID Structure Sub-Sub-TLV is always included in routes with label fields and always uses the following values when advertised:

  • Locator Block Length - encodes the length of the block configured in the locator for the service

  • Locator Node Length - the length of the node configured in the locator for the service

  • Function Length - configurable in the range 20 to 96

  • Argument Length - 0 (default) or 16 (configured)

  • Transposition Length (TL) - 20 for EVPN and VPN-IP routes with full SIDs and 16 for micro segments and 0 for IP routes in the base router

  • Transposition Offset (TO)

    • for EVPN and VPN-IP routes:

      • If the Function Length equals 20 or 16, the Transposition Offset (TO) value equals the prefix length configured in the locator

      • If the Function Length is greater than 20, the TO value equals:

        (prefix length configured in the locator) + (FunctLength 20)

    • for IP routes in base router, TO value is always 0

  • For IP routes in the base router, the TO value is always 0.

Transposition procedures when advertising service routes

The purpose of the SID Structure Sub-Sub-TLV is twofold:

  • Advertise the structure of the SRv6 SID used in the service, including the length of the Locator block, node, function and argument.

  • Support transposition procedures for efficient service route packing. The FUNCTION is transposed into the label field in the route’s NLRI. Because the rest of the SID is common for routes of the same type in the service, this transposition operation supports efficient packing of routes into the same BGP update.

The following figure shows how the FUNCTION part of the SID is transposed. This example illustrates how transposition works for EVPN-VPWS, and it would be similar for VPN-IP routes.

Figure 8. Transposition of the FUNCTION into the NLRI

In the preceding figure, PE6 is configured with an Epipe that uses a configured locator with LB length = 40 bits and LN length = 24 bits. The Function length is set at 24, and 20 bits are always transposed into the NLRI (non-configurable). In the preceding figure, the following rules apply:

  • On reception, the router can build any SID out of the received route, irrespective of transposition, as long as the lengths are correctly encoded.

  • On transmission, the system performs a transposition for VPN-IP and EVPN service routes as follows:

    • If Function Length is greater than 20 in the Locator configuration, the function bits are put at the right-most bits of the L bits. For example, if LB LEN is 40 bits and the LN Len is 24 bits:

    • If Function Length = 20, the entire function is transposed into the label field, and the following is signaled in the route:

      Length [LBL, LNL, FL, AL] : [40, 24, 20, 0]

      TL:20, TO:64

      *A:PE-4>config>router>segment-routing>srv6>locator# info 
      ----------------------------------------------
                              shutdown
                              block-length 40
                              termination-fpe 1
                              prefix
                                  ip-prefix cafe:1:0:4::/64
                              exit
                              static-function
                                  max-entries 10
                              exit
      ----------------------------------------------
      *A:PE-4>config>router>segment-routing>srv6>locator# no shutdown 
      
      3 2021/01/19 08:21:16.827 UTC MINOR: DEBUG #2001 Base Peer 1: 2001:db8::3
      "Peer 1: 2001:db8::3: UPDATE
      Peer 1: 2001:db8::3 - Send BGP UPDATE:
          Withdrawn Length = 0
          Total Path Attr Length = 102
          Flag: 0x90 Type: 14 Len: 33 Multiprotocol Reachable NLRI:
              Address Family VPN_IPV4
              NextHop len 12 NextHop 192.0.2.4
              10.0.0.3/32 RD 192.0.2.4:20 Label 524254
          Flag: 0x40 Type: 1 Len: 1 Origin: 0
          Flag: 0x40 Type: 2 Len: 0 AS Path:
          Flag: 0x40 Type: 5 Len: 4 Local Preference: 100
          Flag: 0xc0 Type: 16 Len: 8 Extended Community:
              target:64500:20
          Flag: 0xc0 Type: 40 Len: 37 Prefix-SID-attr:
             SRv6 Services TLV (37 bytes):-
               Type: SRV6 L3 Service TLV (5)
               Length: 34 bytes, Reserved: 0x0
               SRv6 Service Information
               Service Information sub-TLV Type 1
                  Type: 1 Len: 30 Rsvd1: 0x0
                  SRv6 SID: cafe:1:0:4::
                  SID Flags: 0x0 Endpoint Behavior: 0x14 Rsvd2: 0x0
                  SRv6 SID Sub-Sub-TLV
                     Type: 1 Len: 6
                     BL:40 NL:24 FL:20 AL0 TL:20 TO:64
      
    • If Function Length = 32, part of the function is transposed into the label field, and the following is signaled in the route:

      Length [LBL, LNL, FL, AL] : [40, 24, 20, 0]

      TL:20, TO:76

      *A:PE-4>config>router>segment-routing>srv6>locator# function-length 32 
      *A:PE-4>config>router>segment-routing>srv6>locator# info 
      ----------------------------------------------
                              shutdown
                              block-length 40
                              function-length 32
                              termination-fpe 1
                              prefix
                                  ip-prefix cafe:1:0:4::/64
                              exit
                              static-function
                                  max-entries 10
                              exit
      ----------------------------------------------
      *A:PE-4>config>router>segment-routing>srv6>locator# no shutdown 
      
      8 2021/01/19 08:27:09.318 UTC MINOR: DEBUG #2001 Base Peer 1: 2001:db8::3
      "Peer 1: 2001:db8::3: UPDATE
      Peer 1: 2001:db8::3 - Send BGP UPDATE:
          Withdrawn Length = 0
          Total Path Attr Length = 102
          Flag: 0x90 Type: 14 Len: 33 Multiprotocol Reachable NLRI:
              Address Family VPN_IPV4
              NextHop len 12 NextHop 192.0.2.4
              10.0.0.3/32 RD 192.0.2.4:20 Label 524254
          Flag: 0x40 Type: 1 Len: 1 Origin: 0
          Flag: 0x40 Type: 2 Len: 0 AS Path:
          Flag: 0x40 Type: 5 Len: 4 Local Preference: 100
          Flag: 0xc0 Type: 16 Len: 8 Extended Community:
              target:64500:20
          Flag: 0xc0 Type: 40 Len: 37 Prefix-SID-attr:
             SRv6 Services TLV (37 bytes):-
               Type: SRV6 L3 Service TLV (5)
               Length: 34 bytes, Reserved: 0x0
               SRv6 Service Information
               Service Information sub-TLV Type 1
                  Type: 1 Len: 30 Rsvd1: 0x0
                  SRv6 SID: cafe:1:0:4::
                  SID Flags: 0x0 Endpoint Behavior: 0x14 Rsvd2: 0x0
                  SRv6 SID Sub-Sub-TLV
                     Type: 1 Len: 6
                     BL:40 NL:24 FL:32 AL0 TL:20 TO:76
      
  • The label field of the NLRI (VPN-IP and EVPN routes) encodes the FUNCTION that is dynamically or statically allocated for the service.

With the transposition procedure, multiple NLRIs with the same common SRv6 SID (minus the function) can be packed into the same BGP update, as is done for regular VPN-IP or EVPN route for MPLS tunnels. The packing benefit is illustrated in the following figure.

Figure 9. Transposition and route packing
Note: The transposition procedures do not apply to service SIDs in the base router, advertised via IPv4 and IPv6 families.
For micro-segment SRv6, the following applies:
  • The Transposition Length (TL) is always 16 for EVPN and VPN-IP routes.
  • For IP routes in the base router, the TL is always 0.
  • Because FL is fixed to 16. the Transposition Offset (TO) = LBL + LNL.

Supported service routes for SRv6

The service routes that are supported for SRv6 are:

  • VPRN services, configured for SRv6:

    • VPN-IPv4 routes

    • VPN-IPv6 routes

  • Base router

    • IPv6 routes, if configured for SRv6 in IPv6 family

    • IPv4 routes, if configured for SRv6 in IPv4 family

  • Epipe services, configured for SRv6:

    • EVPN AD per-EVI routes

    • EVPN AD per-ES routes, for Epipes that make use of Ethernet Segments

  • VPLS services configured for SRv6:
    • EVPN AD per-EVI routes
    • EVPN AD per-ES routes, including the advertisement of arguments
    • EVPN MAC/IP Advertisement routes
    • EVPN Inclusive Multicast Ethernet Tag routes

BGP next hop for SRv6 service routes

As specified in RFC 9252, BGP Overlay Services Based on Segment Routing over IPv6 (SRv6), the egress PE may set the next hop to any of its IPv6 addresses. When the IPv6 address value is not covered by the SRv6 locator from which the SRv6 Service SID is allocated, the ingress PE performs reachability checks for the SRv6 Service SID in addition to the BGP next-hop reachability procedures.

Next hop and locator resolution considerations include the following:

  • On reception of a BGP SRv6 service route, both the locator and the next hop are resolved independently in the route table.
  • For base instance routes (not service routes), the following command triggers the independent resolution of the next hop and the locator reachability (and collates their states that drive route programming):
    • MD-CLI
      configure router bgp segment-routing-v6 family ignore-received-srv6-tlvs false
    • classic CLI
      configure router bgp segment-routing-v6 family no ignore-received-srv6-tlvs

    The locator state is considered only if the preceding command is configured.

  • Whether the router does next-hop-self or next-hop-unchanged, there is no affect on the RIB-IN processing and reachability because the next-hop behavior is a RIB-OUT parameter.
  • If a received route has a resolved next hop but unresolved locator, the show router bgp routes command shows "valid/best" but not "used" in the route flags. The following command displays the locator and the resolution of the locator.
    show router bgp next-hop 192.0.2.3 detail vpn-ipv4
===============================================================================
 BGP Router ID:192.0.2.4        AS:64500       Local AS:64500      
===============================================================================

===============================================================================
BGP VPN Next Hop
===============================================================================
-------------------------------------------------------------------------------
VPN Next Hop          : 192.0.2.3
Autobind              : gre 
Labels                : --
Admin-tag-policy      : --
Strict-tunnel-tagging : N
Color                 : --
Locator               : cafe:1:0:3::/64
-------------------------------------------------------------------------------
Resolving Prefix : 192.0.2.3/32
Preference       : 18                   Metric           : 10
Reference Count  : 6                    Owner            : GRE
Fib Programmed   : Y                    
Resolved Next Hop: 192.168.34.1
Egress Label     : n/a                  TunnelId         : 4294967293
Locator State    : Resolved             
-------------------------------------------------------------------------------
Next Hops : 1
===============================================================================

The ingress PE can use the MT-ISIS Unreachable Prefix Announcement (UPA) feature for expedited BGP next-hop reachability checks when SRv6 route summarization is used toward the BGP next hops. Use the following command to activate UPA for IPv4 or IPv6 over SRv6.

configure router bgp segment-routing-v6 family upa-trigger next-hop frr

Use the following command to activate UPA for BGP-IPVPN.

configure service vprn bgp-ipvpn segment-routing-v6 upa-trigger next-hop frr

When a user applies the preceding commands, the ingress SRv6 BGP router can react with Fast Reroute (FRR) when it detects a matching UPA.

When BGP SRv6 UPA-based SRv6 FRR is activated, BGP starts listening to entries added to the UPA routing table to trigger BGP next-hop resolution when a longest prefix matching UPA is detected.

The BGP FRR processing sequence supports UPA-based next-hop unreachability checks. When BGP detects that a next hop disappears due to a matching UPA entry, BGP sends a trigger to the data plane to switch to the backup path, however:

  • BGP does not withdraw the Network Layer Reachability Information (NLRI) from the BGP table
  • BGP does not begin a new best-path calculation
  • BGP does not send a withdraw request to its neighbors

Instead, BGP waits for the BGP control plane to converge as normal to detect through the traditional control plane that the NLRI is no longer valid. This BGP sequence results in BGP behaving with greater grace after UPA failures. This process allows UPA to serve as a stimulus for swift transitioning from the primary to the backup path, while the control plane remains unaffected. Consequently, BGP requires time to converge and withdraw the NLRI at the legacy speed. The effects of a UPA are contained, and this subdues the reactive nature of BGP. This is advantageous, especially during periods of network instability and the resultant routing fluctuations.

Route policy support for matching and modifying BGP SRv6 service routes

Route matching criteria and route modifying actions that are specific to BGP routes carrying SRv6 TLVs can be configured for route policies.

SR OS supports the following route matching criteria:

  • SRv6 TLV presence as match criterion

    Use the following command in the from clause of a policy-statement entry to configure whether a BGP route with associated SRv6 TLVs matches the entry:

    Note: Named entries are only supported in the MD-CLI.
    • MD-CLI

      configure policy-options policy-statement entry from srv6-tlv
      configure policy-options policy-statement named-entry from srv6-tlv
    • classic CLI

      configure router policy-options policy-statement entry from srv6-tlv

    This match criterion is supported in BGP import policies, BGP export policies, and VRF or VSI import policies.

  • SRv6 SID prefix as match criterion

    Use the following command in the from clause of a policy-statement entry to configure a SID or micro-segment (uSID) value in the SRv6 TLV (including the locator and function) to be used as match criterion for a BGP route:

    • MD-CLI

      configure policy-options policy-statement entry from srv6-sid-prefix
      configure policy-options policy-statement named-entry from srv6-sid-prefix
    • classic CLI

      configure router policy-options policy-statement entry from srv6-sid-prefix

    This match criterion uses longest prefix match logic and is supported in BGP import policies, BGP export policies, and VRF or VSI import policies.

SR OS supports the following route modifying actions:

  • SRv6 locator modification

    Use the following command in the action clause of a policy-statement entry to modify the SRv6 TLV in a BGP route to use a different locator than the default or the locator added using the command options in the configure router bgp segment-routing-v6 family add-srv6-tlvs context:

    • MD-CLI

      configure policy-options policy-statement entry action srv6-locator
      configure policy-options policy-statement named-entry action srv6-locator
    • classic CLI

      configure router policy-options policy-statement entry action srv6-locator

    This route modifying action is supported in BGP export policies and SRv6-specific VRF export policies.

  • SRv6 uSID locator modification

    Use the following command in the action clause of a policy-statement entry to modify the SRv6 TLV in a BGP route to use a different uSID locator than the default or the uSID locator added using the command options in the configure router bgp segment-routing-v6 family add-srv6-tlvs context:

    • MD-CLI

      configure policy-options policy-statement entry action srv6-micro-segment-locator
      configure policy-options policy-statement named-entry action srv6-micro-segment-locator
    • classic CLI

      configure router policy-options policy-statement entry action srv6-micro-segment-locator

    This route modifying action is supported in BGP export policies and SRv6-specific VRF export policies.

Route table, FIB table, and tunnel table support

The following tables and information are needed to process a SRv6 packet at service origination, service termination, and transit router roles.

Route table and FIB

SRv6 locator and SID resolution is performed in the RTM and forwarding of all SRv6 packets is performed in the FIB.

The TTM is used to save details of the SRv6 tunnel but is not used directly to forward user or CPM originated packets.

The RTM and FIB are programmed with the routes of the local and remote locators, the local End.X SIDs, and the local End SIDs.

When a policy is applied to export SRv6 routes from the route table to another IS-IS instance, only the IP Reachability TLV and the locator TLV, along with the End SID sub-TLVs, are advertised by the receiving IS-IS instance. Local End, End.X, and LAN End.X routes are not exported nor advertised as separate routes.

  • remote locator (route owner = SRv6 IS-IS)

    All routers in the SRv6 domain populate a resolved remote locator prefix received in the SRv6 Locator TLV in the route table and the FIB.

    A SRv6 packet is always forwarded out in the datapath using the FIB.

    For algorithm 0, the same prefix is advertised with the IP reach prefix TLV and the SRv6 Locator TLV. A single route entry is programmed in the route table and the FIB.

    The prefix of an IGP flexible algorithm locator TLV is never advertised with an IP reach prefix TLV. Therefore, the route of the locator TLV is programmed in the route table and the FIB.

    • remote locator with up to 64 ECMP next hops

      IS-IS models a remote locator prefix with two or more ECMP next hops as an IGP route with tunneled next hops using a protected NHLFE with hardware PG-ID per tunneled next hop.

      This implementation provides uniform failover in ECMP. IS-IS allocates a hardware PG-ID to each next hop it establishes an adjacency with. That PG-ID is then used when programming SRv6 routes of a remote locator and of a local adjacency that resolve to this next hop.

      The route table manager programs the route into the FIB. IS-IS creates an SRv6 tunnel for the locator prefix. The tunnel is added to the tunnel table. The IS-IS route entries in the route table and the FIB point to the tunnel ID of this tunnel in the tunnel table.

      Weighted ECMP, when enabled on the interfaces of this IS-IS instance, is supported when forwarding packets over the locator next hops.

    • remote locator with primary or backup next hops

      To provide uniform failover, IS-IS models a remote locator prefix with a primary next hop or a primary and LFA backup next-hop pair as an IGP route with a tunneled next hop using a protected NHLFE with a hardware PG-ID.

      The route table manager programs the route into the FIB. IS-IS creates a SRv6 tunnel for the locator prefix. The tunnel is added to the tunnel table. The IS-IS route entries in the route table and the FIB point to the tunnel ID of this tunnel in tunnel table.

  • local locator (route owner = SRv6)

    All routers in the SRv6 domain populate a route entry in the route table and the FIB to terminate packets destined for the local locator. This is modeled like any other local route but with the SRv6-specific route owner.

  • local adjacency SID (route owner = IS-IS)

    All routers in the SRv6 domain populate a route entry in the route table and FIB for each local End.X and LAN End.X adjacency SID with primary and backup next hops.

    The route table, and FIB entries are modeled like a remote locator prefix with primary and backup next hops.

  • local End SID (route owner = SRv6)

    All routers in the SRv6 domain populate a route entry in the route table and the FIB to terminate packets destined for each local End SID. This is modeled like any other local route but with the SRv6-specific route owner.

With micro-segment SRv6, entries pertaining to local services populate the RTM and FIB. In the FIB, those entries are aggregated. At most, 15 entries are needed to cover the whole service addressing space offered by micro-segment SRv6. The route owner is SRv6.

TTM

The tunnel table is not used directly in the SRv6 locator resolution, in SID resolution, or in packet forwarding. All resolution is performed in the RTM and forwarding is performed in the FIB.

Each resolved remote locator or local adjacency creates an entry in TTM (ROUTE_OWNER_SRV6_ISIS). The entries for remote locators and local adjacencies in the RTM and FIB point to the tunnel ID of this tunnel. The TTM entry is not used directly for forwarding SRv6 packets. However, the route resolution behavior for SRv6 services can be modified to preferentially resolve in TTMv6. See SRv6 policy support for Layer 2 and Layer 3 services for more information.

Users of route table SRv6 routes

The SRv6 locator and adjacency routes in route table can be used to forward the following user and CPM originated packets:

  • user packets

  • ICMPv6 echo request and echo reply packets as described in 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide

  • UDP traceroute packets as described in 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide

Forwarding and terminating of any other CPM originated packets are not supported. Specifically, received management protocol and control plane protocol packets that are encapsulated in SRv6 are dropped.

If the user configures the address of a BGP neighbor, an LDP peer, or an RSVP-TE LSP destination to match a locator prefix or a SID, packets are forwarded over the SRv6 tunnel but are dropped at the destination router.

Datapath support

This section describes packet processing in the datapath on ingress PE, egress PE, and transit P router roles.

SR OS supports both regular SRv6 and micro-segment SRv6. Both modes can operate concurrently on the same platform, but it requires that the SIDs of each type come from different SID blocks.

Service origination and termination roles

The SRv6 processing is performed inline in the IP datapath when forwarding service packets over a shortest path SRv6 tunnel at service origination or when terminating an SRv6 packet with the service SID in the DA field of the outer IPv6 header.

SRv6 processing is performed in a specialized SRv6 Forwarding Path Extension (FPE) when forwarding service packets over an SRv6 policy or when terminating an SRv6 packet with a local node SID in the DA field and the service SID in the SRH.

Note: The SRv6 FPE also processes service packets forwarded over an SRv6 shortest path tunnel when the service SID in the DA field matches a non-locator IPv6 route in the FIB. A non-locator IPv6 route is an IS-IS route of the locator prefix advertised without a locator TLV, a static route, an OSPF route, or a BGP route of the locator prefix.

The SRv6 origination and termination cannot share the same FPE. A single FPE can be configured for SRv6 origination. One or more FPEs can be configured for SRv6 termination.

Caution: Regardless of whether packets require SRv6 FPE processing, the user must configure the SRv6 origination FPE at the SRv6 ingress PE and one or more SRv6 termination FPEs at the SRv6 egress PE. On a per-packet basis, the datapath makes a determination if SRv6 FPE processing is required. Transit SRv6 routers do not need SRv6 FPEs except at an SRv6 policy BSID expansion node. In the latter, the user must configure an SRv6 origination FPE. See Segment routing policies with an IPv6 data plane for more information about the SRv6 policy feature.

The following figure shows a packet walk-through including the service origination and termination without SRv6 FPE.

Figure 10. Packet walk-through showing service origination and termination without SRv6 FPE

The following figure shows a packet walk-through including the service origination and termination with SRv6 FPE.

Figure 11. Packet walk-through showing service origination and termination with SRv6 FPE

Micro-segment SRv6

The datapath behavior for micro-segment SRv6 slightly differs from regular SRv6. The general behavior of performing two lookups applies but the FIB entries are not the same.

After the parallel lookups, the same procedures as for regular SRv6 apply with the following exception in the behavior in step 1.b. The notion of matching only on a locator does not exist in micro-segment SRv6 because locators also realize an END/uN function. Therefore, a match on a locator is a match on a uN (shift or END). The packet is not sent to the FPE if the first match is on a local locator/uN, but rather if the first match is on a local locator/uN and the second match is on a service SID.

At the ingress PE

The following occurs at the ingress PE:

  • Packet forwarded to a shortest path SRv6 tunnel

    The datapath pushes the SRv6 encapsulation header on the received Layer 2 or Layer 3 service packet.

    • The Outer DA field is set to service SID (End.DT4, End.DT6, End.DT46, or End.DX2 SID).
    • The Outer next-header field is set to IPv4, IPv6, or Ethernet.
    • The hop-limit field in the outer IPv6 header of the SRv6 tunnel is set to 255 for all transit IPv4, IPv6, and Ethernet packets encapsulated into SRv6. The hop-limit for OAM packets originated by the CPM on the router is set according to the specific OAM probe. See the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide for more information.

    The service ingress datapath forwards the packet to one of the SRv6 tunnel candidate egress network IP interfaces based on a hash of the inner packet headers. See Using flow label in load-balancing of IPv6 and SRv6 encapsulated packets for more information about the spraying of service packets over SRv6 tunnel candidate next hops.

    The following diagram displays details of the datapath processing of a service packet that is originated on a VPRN and forwarded over an SRv6 shortest path tunnel.

    Figure 12. Walk-through SRv6 datapath without SRv6 FPE at service origination node
  • Packet forwarded to a SRv6 policy

    The SRv6 FPE egress datapath receives the Layer 2 or Layer 3 service packet and pushes the SRv6 encapsulation header:

    • The Outer DA field is set to the top SID of the SRv6 policy.
    • The Outer next-header field is set to SRH.
    • The Service SID (End.DT4, End.DT6, End.DT46, or End.DX2 SID) is encoded in the SRH.
    • The SRH next-header field is set to IPv4, IPv6, or Ethernet.
    • The hop-limit field in the outer IPv6 header of the SRv6 tunnel is set to 255 for all transit IPv4, IPv6, and Ethernet packets encapsulated into SRv6. The hop limit for OAM packets originated by the CPM on the router is set according to the specific OAM probe. See the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide for more information.

    The SRv6 FPE ingress datapath does the lookup on the outer DA field and forwards the packet to one of the candidate egress network IP interfaces based on the flow label and the SA or DA or both fields of the outer IPv6 packet header. See Using flow label in load-balancing of IPv6 and SRv6 encapsulated packets for more information about the spraying of SRv6 packets.

    The following diagram displays details of the datapath processing of a service packet that is originated on a VPRN and forwarded over an SRv6 policy.

Figure 13. Walk-through SRv6 datapath with SRv6 FPE at service origination node

At the egress PE

  1. The procedure in this step is common to the transit router role and the service termination router role.

    On the ingress IP network interface, the SRv6 feature concurrently performs two IPv6 address lookups on a received IPv6 packet:
    • A first (longest prefix match) lookup checks if the address in the outer header DA field matches either an SRv6 local locator subnet, a local service SID, a local End function or a local End.X function. This first lookup is for the current SID.
    • A second lookup is performed on the next SID in the SRH (when the IPv6 packet has an SRH). The SRv6 feature reads next SID using the index value after decrementing the Segments-Left field.

    The subsequent processing depends on the outcome of the first lookup:

    1. If the match is on a local service SID:

      • If the payload type is IPv4, IPv6, or Ethernet, the packet is forwarded to the service function processing; see step 2 for more details.

        The payload type refers to the value of the last next-header field in the processing chain of the packet. This could be the next-header field of the outer IPv6 packet, if there is no SRH. This could also be the next-header field of an expired SRH (Segments-Left=0) for which the last SID matches the service SID.

      • If the payload type indicates any other protocol, including ICMPv6 (ICMP ping packet) and UDP (potential traceroute message when hop-limit field has a value of 1), the packet is redirected to the CPM for more processing. Protocol matching ICMPv6 ping and UDP traceroute have their packets processed as described in the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide. Other protocol packets are dropped.

        The CPM generates a specific ICMPv6 message to the address in the SA field of the processed or dropped packet depending on the protocol type and the result of the match of the address in the DA field of the packet. These ICMPv6 reply messages are summarized in ICMPv6 reply messages to extracted SRv6 packets.
    2. If the match is on a local locator only:

      • If the payload type is IPv4, IPv6, or Ethernet, the packet is forwarded to the SRv6 FPE for potential service function processing; see step 2 for more details.

      • If the payload type indicates any other protocol, including ICMPv6 (ICMP ping packet) and UDP (potential traceroute message when hop-limit field has a value of 1), the packet is redirected to the CPM for more processing. Protocol matching ICMPv6 ping and UDP traceroute have their packets processed as described in the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide. Other protocol packets are dropped.

        The CPM generates a specific ICMPv6 message to the address in the SA field of the processed or dropped packet depending on the protocol type and the result of the match of the address in the DA field of the packet. These ICMPv6 reply messages are summarized in ICMPv6 reply messages to extracted SRv6 packets.

    3. If the match is on a specific local End function and the next SID lookup is not a local locator, the packet is processed as per the transit router role for these functions as detailed in Transit router role with or without segment termination.

    4. If the match is on a specific local End function and the next SID resulted in a match on a local locator, the packet is processed as described in step 1.b with the next-header field used in the processing is that of the SRH.

    5. If the match is on a specific local End.X function, regardless of the next SID match outcome the packet is processed in accordance with the transit router role for these functions; see Transit router role with or without segment termination for more information.

    6. If the match is on a regular IPv6 route or there is no match, the packet is forwarded or dropped. For forwarded packets, the destination address could match the locator prefix or a regular IPv6 prefix of a remote node.

  2. The procedure in this step is specific to the service termination router role.

    1. When the first lookup match is on a local service SID (service SID encoded in outer DA field), service packet processing is performed inline in the network interface ingress datapath.

      • The datapath performs the detailed processing of the specific SID function as per https://tools.ietf.org/html/rfc8986. It then removes the SRv6 encapsulation headers, including SRH if any.

      • It decrements and propagates, into the IPv4 TTL field or IPv6 hop-limit field of the forwarded inner packet, the minimum of the incoming outer header hop-limit and inner header hop-limit (or TTL) values.

      • It performs a lookup of the service SID and forwards the packet to the service context for further processing.

      The following figure displays a walk-through of the SRv6 datapath without SRv6 FPE at the service termination node.

      Figure 14. Walk-through SRv6 datapath without SRv6 FPE at service termination node
    2. When the first lookup match is on a local locator entry of the FIB, or on a local End function and second lookup match is on a local locator, the service SID must be in the SRH and service packet processing is performed by the SRv6 termination FPE.

      • The egress SRv6 FPE datapath receives the SRv6 encapsulated packet and performs the detailed processing of the specific SID function as per https://tools.ietf.org/html/rfc8986. It then removes the SRv6 encapsulation headers, including SRH if any, and inserts a service label, with the value derived from the FUNCTION field value, into the inner service packet.

        • The egress SRv6 FPE decrements and propagates, into the IPv4 TTL field or IPv6 hop-limit field of the forwarded packet, the minimum of the incoming outer header hop-limit and inner header hop-limit (or TTL) values.

        • The ingress SRv6 FPE does an ILM lookup on the service label and forwards the packet to the service context for further processing.

The following figure shows a walk-through of the SRv6 datapath with SRv6 FPE at the service termination node.

Figure 15. Walk-through SRv6 datapath with SRv6 FPE at service termination node

Transit router role with or without segment termination

The transit router role does not require the use of an SRv6 FPE except at SRv6 policy BSID expansion node. See Segment routing policies with an IPv6 data plane for more details on the SRv6 policy feature.

The following steps summarize the packet processing for the transit router role. For more information about the specific processing of the SID function see https://tools.ietf.org/html/rfc8986.

  1. The procedure in this step is common to the transit router role and the service termination router role.

    On the ingress IP network interface, the SRv6 feature concurrently performs two IPv6 address lookups on a received IPv6 packet:
    • A first (longest prefix match) lookup checks if the address in the outer header DA field matches either an SRv6 local locator subnet, a local service SID, a local End function or a local End.X function. This first lookup is for the current SID.
    • A second lookup is performed on the next SID in the SRH (when the IPv6 packet has an SRH). The SRv6 feature reads next SID using the index value after decrementing the Segments-Left field.

    The subsequent processing depends on the outcome of the first lookup:

    1. If the match is on a local service SID:

      • If the payload type is IPv4, IPv6, or Ethernet, the packet is forwarded to the service function processing; see step 2 in section At the egress PE for more details.

        The payload type refers to the value of the last next-header field in the processing chain of the packet. This could be the next-header field of the outer IPv6 packet, if there is no SRH. This could also be the next-header field of an expired SRH (Segments-Left=0) for which the last SID matches the service SID.

      • If the payload type indicates any other protocol, including ICMPv6 (ICMP ping packet) and UDP (potential traceroute message when hop-limit field has a value of 1), the packet is redirected to the CPM for more processing. Protocol matching ICMPv6 ping and UDP traceroute have their packets processed as described in the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide. Other protocol packets are dropped.

        The CPM generates a specific ICMPv6 message to the address in the SA field of the processed or dropped packet depending on the protocol type and the result of the match of the address in the DA field of the packet. These ICMPv6 reply messages are summarized in the table that follows.
        Table 9. ICMPv6 reply messages to extracted SRv6 packets
        Protocol Destination IP address match result ICMPv6 reply

        (Type/code)

        ICMP echo request/reply

        (See 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide for ICMPv6 Ping support in SRv6)

        locator prefix [function | any arg]

        echo reply/ping successful

        UDP / TCP

        (See 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide for UDP Traceroute support in SRv6)

        locator prefix [ function | any arg ]

        dest unreachable, port unreachable

        Any other protocol

        locator prefix [ function | any arg ]

        ICMP Parameter Problem/SR Upper-layer Header Error

        All protocols including above

        locator prefix | unsupported function

        dest unreachable, communication prohibited

    2. If the match is on a local locator only:

      • If the payload type is IPv4, IPv6, or Ethernet, the packet is forwarded to the SRv6 FPE for potential service function processing; see step 2 in section At the egress PE for more information.

      • If the payload type indicates any other protocol, including ICMPv6 (ICMP ping packet) and UDP (potential traceroute message when hop-limit field has a value of 1), the packet is redirected to the CPM for more processing. Protocol matching ICMPv6 ping and UDP traceroute have their packets processed as described in the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide. Other protocol packets are dropped.

        The CPM generates a specific ICMPv6 message to the address in the SA field of the processed or dropped packet depending on the protocol type and the result of the match of the address in the DA field of the packet. These ICMPv6 reply messages are summarized in ICMPv6 reply messages to extracted SRv6 packets.

    3. If the match is on a specific local End function and the next SID lookup is not a local locator, the packet is processed as per the transit router role for these functions as detailed in step 2 below.

    4. If the match is on a specific local End function and the next SID resulted in a match on a local locator, the packet is processed as described in the preceding step 1.b with the next-header field used in the processing is that of the SRH.

    5. If the match is on a specific local End.X function, regardless of the next SID match outcome, the packet is processed in accordance with the transit router role for these functions; see 2 below.

    6. If the match is on a regular IPv6 route or there is no match, the packet is forwarded or dropped. For forwarded packets, the destination address could match the locator prefix or a regular IPv6 prefix of a remote node.

  2. The procedure in this step is specific to the transit router role.

    If the match is on a local End or End.X SID, the SID termination processing is performed on the packet.

    1. If the End or End.X SID is the last SID in the packet encapsulation, meaning there is no SRH or there are only expired SRHs (Segments-Left =0), the packet is sent to the CPM for further processing.

      Note: The CPM processes ICMPv6 ping packets and UDP traceroute packets but drops any other protocol type. See the 7450 ESS, 7750 SR, 7950 XRS, and VSR OAM and Diagnostics Guide for more information about the processing of ICPMv6 echo request and reply packets and UDP traceroute packets.
    2. If the next-header in the IPv6 header is an SRH, the Segments-Left field is zero, and the next-header in the SRH is another SRH, the current SRH is removed and the remaining steps are applied on the next SRH.

    3. If the Segments-Left field is 1 and the SRH mode of the terminated SID is PSP, the SRH is removed. Otherwise, the Segments-Left field is decremented and used to read and copy the next SID into the DA field of the outer IPv6 header.

    4. Decrement the incoming outer IPv6 header hop-limit and write it into the outer IPv6 header hop-limit field of the outgoing packet.

    5. If the first SID lookup of the current SID in the FIB matched an End function, use the outcome of the second SID lookup of the next SID to forward the packet to the next hop of the next SID (in the DA field of the outer IPv6 header).

    6. If the first SID lookup of the current SID in the FIB matched an End.X function, override the outcome of the second SID lookup of the next SID with the set of next hops of the adjacency and forward the packet.

    7. If both the current and the next SIDs match a local End or End.X SID, the packet is forwarded as indicated in Forwarding behavior for back-to-back local SIDs.

      Table 10. Forwarding behavior for back-to-back local SIDs
      Current SID match Next SID match Forwarding action

      End

      End

      Packet is extracted to the CPM, which drops it if the next SID in not the last SID. An ICMPv6 packet (type: dest unreachable, code: communication prohibited) is sent to the address in the SA field

      If the next SID is the last SID, the bounced packet is processed as per ICMPv6 reply messages to extracted SRv6 packets.

      End

      End.X

      Packet is forwarded over the adjacency of the next SID to the downstream neighbor, which forwards it back to this node for the next SID processing. The bounced packet is forwarded again over the same adjacency.

      If the next SID is the last SID, the bounced packet is processed as per ICMPv6 reply messages to extracted SRv6 packets.

      End.X

      End

      Packet is forwarded over the adjacency of the current SID to the downstream neighbor, which forwards it back to the current node for the next SID processing

      If the next SID is the last SID, the bounced packet is processed as per ICMPv6 reply messages to extracted SRv6 packets.

      End.X

      End.X

      Packet is forwarded over the adjacency of the current SID to the downstream neighbor, which forwards it back to the current node for the next SID processing

      If the next SID is the last SID, the packet is processed as per ICMPv6 reply messages to extracted SRv6 packets.

Transit router role in micro-segment SRv6

The datapath behavior in transit SRv6-enabled routers distinguishes the following use cases:
  • The DA field of the IPv6 header contains a single SID (shortest path case).
  • Multiple SIDs are used in the DA field and eventually in the SRH (LFA repair tunnel or SRv6 policy).

The DA field of the IPv6 header contains a single SID is of the form 2001:db8:00d1:d547::

where
  • <2001:0db8> is the SID block
  • <00d1> is the identifier associated with a specific node
  • <d547> is an identifier for a specific local service assigned by the node
<2001:0db8><00d1> acts both as a locator for the specific node and as a uN function. A packet with this address in the DA field of the IPv6 header is routed toward the specific node with each node along the path matching on 2001:0db8:00d1::/48 and forwarding to the predetermined next hop.

Multiple SIDs in the DA field and eventually in the SRH illustrate the following micro-segment SRv6 behavior.

Suppose that a source node wants to send traffic for a specific service to the destination node D via the nodes B and F. For that example, consider the following:
  • The micro-segment ID block is 2001:0db8::/32.
  • The node identifiers (uN) are 0x00b1 for node B, 0x00f1 for node F, and 0x00d1 for node D.
  • Node D has advertised 2001:0db8:00d1:d457:: for the specific service.
The source node does the following:
  • It constructs the following container: 2001:0db8:00b1:00f1:00d1:d457::
  • It places this container in the DA field of the IPv6 header.
  • It forwards the packet to the next-hop determined by an LPM on that address.
On receiving the packet, nodes B and F perform a shift operation (bound to the uN).

Transit routers in micro-segment SRv6 illustrates the preceding example.

Node B matches on 2001:0db8:00b1::/48 (as it has that entry in its FIB) and performs the operation associated with that micro-segment ID:
  • It removes the 16-bit from the address.
  • It shifts the right part toward the MSB.
  • It adds a 16-bit block of zeros, the so called End of Container (EoC), as the LSB.
Node B forwards the packet to the next-hop based on a lookup of the resulting DA field.

Node F does the same (based on matching on its own identifier).

Node D does the same but processes the packet in the service corresponding to the identifier.

Figure 16. Transit routers in micro-segment SRv6

It can be that the path is longer than the number of micro-SIDs that can be compressed in 128 bits. In that case, an SRH is used to convey the rest of the path. Each container in the SRH must be of the same form (a block part followed by a sequence of micro-SIDs). At some node along the path, because of the shift operations, the container in the IPv6 DA expires. At that node, micro-segment SRv6 implements a regular SRv6 END function (or END.X function if the match is on a uA).

For example, if node B receives a packet with 2001:0db8:00b1:: in the DA field, it detects that its own identifier is followed by an EoC. Node B copies the next segment or container from the SRH in the DA field.

Using flow label in load-balancing of IPv6 and SRv6 encapsulated packets

When a service is bound to a SRv6 shortest path tunnel, the ingress service SAP or interface sprays the packets over the ECMP next hops of the SRv6 tunnel and LAG links of the outgoing network interfaces.

The default hash calculation on the ingress service SAP or interface is based on the existing hash procedures of an IPv4, IPv6, or an Ethernet packet. For IPv6 service packets, an option is provided to include the packet’s Flow Label field, when not zero, and to hash on the triplet {SA, DA, Flow Label}. The flow-label-load-balancing command is used to enable this behavior on an access or network interface.

The ingress service SAP or interface copies the output of the hash on the inner packet headers into the flow label field of the outer IPv6 header that it pushes on the SRv6 encapsulated packet. This is regardless of whether the flow label is used or not in the computation of the hash on the service packet

When a service is bound to an SRv6 policy, the service packets are first forwarded to the egress network interface of the SRv6 origination FPE to build and push the SRv6 encapsulation. The packets are then handed in to the ingress network interface of the SRv6 origination FPE which sprays the packets over the ECMP next hops of the SRv6 policy and LAG links of the outgoing network interfaces.

The SRv6 origination FPE egress network interface copies the output of the hash on the inner packet headers into the flow label field of the outer IPv6 header that it pushes on the SRv6 encapsulated packet. This is regardless of whether the flow label is used or not in the computation of the hash on the service packet.

The SRv6 origination FPE ingress network interface does not require the flow-label-load-balancing command to be enabled. All SRv6 packets are automatically sprayed to the ECMP next hops of the SRv6 tunnel and LAG links of the outgoing network interfaces using a hash on the triplet {SA, DA, Flow Label} in the SRv6 packet’s outer IPv6 header.

On a transit router, the hashing of SRv6 encapsulated packets can also use the Flow Label field in the outer IPv6 header to provide more entropy to the load-balancing process of SRv6 packets. The flow-label-load-balancing command can be configured on a network interface to hash on the triplet {SA, DA, Flow Label}. By default, a transit router only hashes on the tuple {SA, DA} in the header of a received IPv6 packet with a non-zero flow label field, including when the packet is SRv6. The description of the flow-label-load-balancing command and the detailed behavior of the hash feature based on the IPv6 packet flow label field and its general application to access and network interfaces is described in section IPv6 Flow Label Load Balancing of the 7450 ESS, 7750 SR, 7950 XRS, and VSR Interface Configuration Guide.

Interaction with other datapath features

The following describes the interaction of the SRv6 feature with other datapath features:

  • When SRv6 is enabled on the base router, datapath enables forwarding and receiving SRv6 encapsulated packets on all network interfaces of the router. The datapath, however, drops an SRv6 encapsulated packet if received from or needs forwarding to an access interface (IES, VPRN). A service packet received on either a network interface or an access interface can be encapsulated into a SRv6 packet and forwarded to a network interface.

  • SRv6 feature performs concurrently a couple of IPv6 address lookups on a packet received with an SRH. The first lookup is for the current SID in the DA field in the header of the received SRv6 packet and the second is for the next SID in the SRH.

    The datapath repurposes the IPv6 unicast RPF (IPv6 uRPF) check for the next SID lookup, which means the IPv6 uRPF feature cannot be performed on all IPv6 packets received on that interface. CLI enforces this interaction and so SRv6 cannot be enabled in the base router context (config>router>segment-routing>segment-routing-v6) if the IPv6 uPRF check (config>router>if>ipv6>urpf) is enabled on one or more network interfaces.

    Conversely, the IPv6 uRPF check cannot be enabled on a network interface if SRv6 is enabled in the base router context.

    Note: SRv6 does not impact uRPF checks of IPv4 packets received on a network interface.
  • When SRv6 is enabled in one or more IGP instances, a transit router cannot check the SA field in the outer IPv6 header of a SRv6 encapsulated packet received on a network interface and that also has an SRH header. Normally an IPv6 packet which uses 0::0 or a link-local address format should be dropped. All other IPv6 packets, including a SRv6 encapsulated packet that does not have an SRH header, are checked for these two situations.

  • Policy Based Routing (PBR) is allowed on flows of packets of an SRv6 tunnel. In other words, the user can apply an ACL filter on a network interface which matches on the outer SA and DA fields of the SRv6 packet and execute an action such as a redirection.

    The operator must ensure that SRv6 matching packets are directed to a router that can process and forward the SRv6 packets.
    Note: Redirecting an SRv6 packet even to an SRv6-capable router is not recommended because the processing of the SID list in the SRH can create loops for any of the SIDs in the outer DA and SRH.

LFA support

Note: This section and the following IS-IS procedures and Datapath procedures sections are applicable to regular and micro-segment SRv6, unless stated otherwise.

LFA, remote LFA, and TI-LFA are supported in the following router roles:

  • service originating role

  • transit role with segment termination

  • transit role without segment termination

The backup is computed and programmed for each remote SRv6 node SID, service SID (for example, End.DT4, End.DT6, End.DT46, and End.DX2), and for each local adjacency or LAN adjacency SID.

A base LFA backup path or a TI-LFA backup path that uses a direct IP next hop (not a repair tunnel), requires configuring a next hop that is different from the primary path and does not modify the SID list pushed on the primary path.

When the RLFA or TI-LFA backup path uses a repair tunnel (source routed or not), the additional SIDs of the repair tunnel must be inserted into the packet when the backup is activated. This requires the insertion of an LFA dedicated SRH into the packet. The SRv6 behavior is referred to as H.Insert.Red and is described in draft-filsfils-spring-srv6-net-pgm-insertion-04. The application of this behavior to the LFA repair tunnel is described in draft-voyer-6man-extension-header-insertion-10.

The following figure illustrates the packet encoding for the primary and remote LFA backup path using the H.Insert.Red behavior and pushing a dedicated reduced SRH for the repair tunnel. The figure uses regular SIDs but is equally applicable to micro SIDs.

Figure 17. LFA repair tunnel packet encoding

The LFA backup path for the prefix of a remote node SID and of a remote service SID is programmed in the route table and tunnel table with the entry of the prefix of the locator of that SID. This practice is required because forwarding to these SIDs requires looking up the remote locator they were derived from.

The LFA backup for a local adjacency SID or a local LAN adjacency SID is programmed in route table with the specific entry corresponding to this SID.

IS-IS procedures

The base LFA, remote LFA, and TI-LFA features operate with SRv6 tunnels in the same way as with SR-MPLS tunnels.

Use the commands in the following context to enable the base LFA:
  • MD-CLI
    configure router isis loopfree-alternate
  • classic CLI
    configure router isis loopfree-alternates
Use the commands in the following contexts to enable remote LFA and TI-LFA:
  • MD-CLI
    configure router isis loopfree-alternate remote-lfa
    configure router isis loopfree-alternate ti-lfa
  • classic CLI
    configure router isis loopfree-alternates remote-lfa
    configure router isis loopfree-alternates ti-lfa

The max-srv6-frr-sids command option configures the maximum number of SRv6 SIDs allowed in the segment list of a TI-LFA repair tunnel. A higher max-srv6-frr-sids value results in better coverage by TI-LFA, at the expense of increased packet encapsulation overhead. The SR OS implementation can insert up to 1 additional regular SID or up to 3 additional micro SIDs in the segment list of a TI-LFA repair tunnel.

The max-srv6-frr-sids parameter is used by TI-LFA to limit the search for the P-Q set in the post-convergence path. Users can configure the following values:

  • 0

    The IGP LFA SPF restricts the search to TI-LFA backup next hop, which does not require a repair tunnel. This means that the P and Q nodes are the same and match a neighbor of the LFA SPF computing router.

  • 1 (default)

    This option corresponds to a repair tunnel to the node SID of a PQ node (P=Q) or to an adjacency SID between adjacent P and Q nodes. With regular 128-bit SIDs, an SRv6 adjacency is globally routable, therefore there is no need to insert the node SID of the P node; the adjacency SID between the P and Q nodes is sufficient.

    With micro-segment SRv6, an adjacency SID is globally routable only if it is combined with the corresponding node SID (<uN,uA>). In that case and in all other cases, the combination counts as a single SID toward the limit configured by max-srv6-frr-sids.

  • 2 to 3

    When protecting a node SID, the IGP LFA SPF widens the search to include a repair tunnel to a P-Q set in which P and Q nodes are interconnected with two to three adjacencies. IGP can also protect an adjacency using the backup path of the neighbor’s node SID. In this case, the neighbor’s node SID is added to the SIDs of the adjacencies between the P and Q nodes, for a total of two to three SIDs.

    With micro-segment SRv6, the addition of the neighbor's node SID can exceed the max-srv6-frr-sids value as long as it fits in the same LFA SRH container.

The following is a description of the SRv6 IS-IS tunnel backup path computation. A compression of the SIDs is applied to minimize the SID list of the computed backup path.

The backup path of a remote locator is as follows:

  • TI-LFA

    IS-IS adds the segment list of node and adjacency SIDs of the P-Q set to the repair tunnel. The datapath encodes the top SID of this segment list into the DA field of the outer IPv6 header and moves the received SID value in the DA field and the remaining SIDs of the repair tunnel segment list into the LFA SRH. The protected locator prefix backup next hop in the route table points to the tunnel ID of the locator prefix of the top SID of the P-Q set segment list.

  • base LFA

    If an alternate equal-cost parallel link to the LFA neighbor exists, IS-IS programs it as a regular IP next hop of the protected locator.

    If the alternate parallel link to the LFA neighbor is not equal-cost, IS-IS programs a null SID repair tunnel. The datapath does not push an LFA SRH in this case. The protected locator prefix backup next hop in the route table points to the tunnel ID of the local adjacency over the outgoing link to the LFA neighbor.

  • remote LFA

    IS-IS adds the node SID of the PQ node to the repair tunnel. The datapath encodes this SID into the DA field of the outer IPv6 header and moves the received SID value of the DA field into the LFA SRH. The protected locator prefix backup next hop in the route table points to the tunnel ID of the locator prefix of the PQ node.

    Note: If the P node is a neighbor of the computing node, the SID list is empty and the backup next hop of the protected locator prefix in the route table points to the tunnel ID of the local adjacency over the outgoing link to the LFA neighbor.

The backup path of a local adjacency is as follows:

  • TI-LFA

    IS-IS computes the repair tunnel by using the segment list of the P-Q set plus a node SID of the neighbor node on the other side of the protected adjacency. If the Q node is the same as the neighbor node on the other side of the protected adjacency, the repair tunnel is programmed after compressing the SID list to the adjacency SIDs of the links between the P node and the neighbor node. The datapath encodes the top SID of this segment list into the DA field of the outer IPv6 header and moves the received SID value in the DA field and the remaining SIDs of the repair tunnel segment list into the LFA SRH.

  • base LFA

    IS-IS attempts first to program a null SID repair tunnel by using the adjacency of an alternate parallel link to the neighbor on the other side of the protected adjacency. The datapath does not push an LFA SRH in this case. The protected locator prefix backup next hop in the route table points to the tunnel ID of the local adjacency over the outgoing link to the LFA neighbor.

    If the first option fails, IS-IS programs a repair tunnel by using the node SID of the neighbor on the other side of the protected adjacency. The datapath encodes that node SID of the neighbor into the DA field of the outer IPv6 header and moves the received value of the DA field into the LFA SRH.

  • remote LFA

    IS-IS computes the repair tunnel by using the node SID of the PQ node plus a node SID of the neighbor node on the other side of the protected adjacency. If the PQ node is one hop from the node on the other side of the protected adjacency, the repair tunnel SID list is compressed to the adjacency SID of the link between the PQ node and the neighbor node. The datapath encodes the top SID of this segment list into the DA field of the outer IPv6 header and moves the received SID value in the DA field and the remaining SIDs of the repair tunnel segment list into the LFA SRH.

Note: IS-IS prefers a PSP over a USP SID when selecting the PQ node SID or the P-Q set adjacency SID of a remote LFA or TI-LFA repair tunnel. In general, if a third-party implementation signals other SRH modes, IS-IS selects the SID in the ascending order of the SRH mode codepoint for the SID. For example, node SIDs in the regular SRv6 have the following order:
  1. End 0x01

  2. End_PSP 0x02

  3. End_USP 0x03

  4. End_PSP_USP 0x04

  5. End_PSP_USD 0x1D

  6. End_USP_USD 0x1E

  7. End_PSP_USP_USD 0x1F

When two or more node SIDs of the same SRH mode exist, IS-IS selects the SID with the lowest function value.

When two or more adjacency SIDs of the same SRH mode exist, IS-IS prefers the SID with protection enabled and selects the SID with lowest function value from the protected or unprotected SID subset.

The TI-LFA algorithm prefers a micro-segment locator over a regular locator in case both exist. As such, a micro-segment path may protect a regular SRv6 path.

Datapath procedures

The datapath procedures for the service origination router role, as described in Service origination and termination roles, are modified as follows when the LFA backup is a repair tunnel using the H.Insert.Red encapsulation.

  1. The top SID of the primary path is copied into the LFA SRH.

  2. The top regular SID or the whole LFA SRH micro-segment container of the repair tunnel is copied in the DA field of the outer IPv6 header. The Segments-Left field of the LFA SRH is set to 1.

  3. If the P-Q set consists of a PQ node only and the PQ node is the same as the node that owns the top SID of the primary path, the LFA SRH insertion is skipped. This is not a datapath procedure per se, but IGP compresses the SID list of the backup path to look the same as that of the primary path.

The datapath procedures for the transit router role, as described in Transit router role with or without segment termination, are modified as follows when the LFA backup is a repair tunnel using the H.Insert.Red encapsulation.

  1. The next SID, read from the original SRH after decrementing Segments-Left field, or from the DA field if no SRH exists, is copied into the LFA SRH .

  2. The top regular SID or the whole LFA SRH micro-segment container of the repair tunnel is copied in the DA field of the outer IPv6 header. The Segments-Left field of the LFA SRH is set to 1.

  3. If the P-Q set consists of a PQ node and the PQ node is the same as the node that owns the next SID, the LFA SRH insertion is skipped. This is not a datapath procedure per se, but IGP compresses the SID list of the backup path to look the same as that of the primary path.

SRv6 tunnel metric and MTU settings

IGP sets the metric of the SRv6 remote locator prefix route and tunnel or that of a local adjacency SID route to the metric of the computed path of the corresponding route.

The metric of a local End SID route is set to 0, similar to any local route.

The metric of a BGP IPv4/IPv6 or VPN-IPv4/VPN-IPv6 route resolved to a SRv6 tunnel inherits the value of the locator prefix route metric.

The user must configure the network interfaces at the ingress PE and at transit P routers with an MTU value that accounts for the fixed IPv6 header (40 bytes) and the additional LFA SRHs (24 bytes each).

Special attention is required when a service packet is forwarded over the SRv6 origination FPE interface (interface-b) at ingress PE. The datapath accounts for the fixed 40-byte IPv6 header in the check for fragmenting IPv4 packets (DF=0) or for dropping IPv4 (DF=1) and IPv6 packets that are intended to be forwarded over an SRv6 tunnel. If LFA is enabled in IS-IS, the LFA overhead is not accounted for, and therefore the configured MTU for the SRv6 origination FPE interface must account for it.

In addition, there are network deployments where it is not possible to modify the network interface MTU or to set all network interfaces to the same value. In that case, the user must configure the outgoing network interface and the SRv6 origination FPE interface MTU to reflect this worst case MTU in the network, accounting fixed and variable LFA overhead. The following is the CLI to use for this purpose.

configure 
+--fwd-path-ext 
   +--fpe <fpe-id> [create] 
         +--srv6 <origination | termination>
            +--interface-b
                +--mtu <1280-9786> 

The show router base icmp6 command has a global count for received “Packets Too Big” that captures the dropped packets on network interfaces, including SRv6 origination FPE interface, caused by MTU violation.

MTU configuration examples

The following are examples of MTU settings of the SRv6 origination FPE interface (interface-b):

  • all default values

    By default, LFA is disabled in IS-IS. The network interface default MTU is 9786 bytes based on default Ethernet MTU of 9800 bytes.

    The SRv6 origination FPE interface MTU defaults to the same value of 9786 bytes; there is no need for further adjustments.

  • all default values and RLFA or TI-LFA enabled

    The user must adjust the SRv6 origination FPE interface MTU from the first bullet item case to account for the 24 bytes introduced by LFA.

    SRv6 origination FPE Interface MTU is 9786 - 24 = 9762 bytes.

  • a remote constrained network interface and RLFA or TI-LFA enabled

    Assume a specific remote network interface in the SRv6 network is limited to a maximum value of 1500 bytes. The user must adjust the SRv6 origination FPE interface MTU in all the ingress PEs from the first bullet item case, to account for that constrained value, plus the 24 bytes introduced by LFA repair tunnel (Users can adjust for as many LFA SRHs as they need. The following example assumes 1 LFA SRH).

    SRv6 origination FPE Interface MTU = 1500 - 24 = 1476 bytes.

Service extensions

This section describes the service extensions to originate and terminate SRv6 services.

The extensions also apply to micro-segment SRv6. To distinguish a regular locator from a micro-segment locator, specific CLI contexts matching those for regular SRv6 are available.

SRv6 forwarding path extension

SRv6 processing is performed in a specialized SRv6 FPE when forwarding service packets over a SRv6 policy or when terminating a SRv6 packet with a local node SID in the DA field and the service SID in the SRH. The SRv6 FPE also processes service packets forwarded over a SRv6 shortest path tunnel when the service SID in the DA field matches a non-locator IPv6 route in the FIB. A non-locator IPv6 route is an ISIS route of the locator prefix advertised without a locator TLV, a static route, an OSPF route, or a BGP route of the locator prefix.

The following guidelines apply for the SRv6 FPE:

  • An internal or external Port Cross-Connect (PXC) can be used for the SRv6 FPE.

  • The SRv6 origination or termination FPE cannot be shared with other applications, but the same physical ports can be used when configuring PXC ports for multiple FPEs of different applications.

    As an example of how two FPEs can share the same physical port (therefore the same bandwidth), define two PXC ports, both sharing the same underlaying physical port; for instance:

    FPE 1 is associated with PXC-1 and FPE 2 is associated with PXC-2, where PXC-1 and PXC-2 are both assigned to port 1/1/1

  • Some considerations about the SRv6 termination FPE follow:

    • It is configured per locator.

    • Multiple locators can optionally use the same or different FPE.

    • Received SRv6 traffic for a specific (local) locator is redirected to the SRv6 termination FPE interface-a.

  • Some considerations about the SRv6 origination FPE follow:

    • There is only one SRv6 origination FPE supported per system.

    • The SRv6 origination and termination FPEs are always different.

  • SRv6 FPE redundancy and load-balancing:

    • Each FPE can use a LAG composed of as many PXC ports as needed (there is no specific limitation in the number of PXC members per LAG).

    • LAG members can be PXC ports in the same or a different card.

The following CLI is required to create the FPE of type srv6 origination or termination and apply it to a locator. All locators may be associated with the same or a different FPE.

configure
+--fwd-path-ext
   +--fpe <fpe-id>
      +--application
         +--srv6 <origination|termination> 
            +--interface-a
               +--qos <network-policy-id>
            +--interface-b
               +--mtu <1280-9786>
               +--qos <network-policy-id>
configure
+--router
|    +--segment-routing
|    |    +--segment-routing-v6
|    |    |    +--origination-fpe <fpe>
|    |    |    +--source-address <ipv6-address>
|    |    |    +--locator <locator-name>
|    |    |    |    +--termination-fpe <fpe>

SRv6 VPRN services

VPRN services support SRv6 End.DT4, End.DT6, and End.DT46 behaviors. VPRNs support IPv4 and IPv6 routes that are advertised in VPN-IPv4 and VPN-IPv6 families, as well as in EVPN IP prefix routes in the Interface-less mode.

The following classic CLI commands configure a VPRN for SRv6 with the VPN-IP families.

configure
+--service
|   +---vprn <service-id>
|   |   +---segment-routing-v6 <instance-id>
|   |   |   +---locator <locator-name>
|   |   |   |   +---function
|   |   |   |   |   +---end-dt4 <integer>
|   |   |   |   |   +---end-dt6 <integer>
|   |   |   |   |   +---end-dt46 <integer>
|   |   +---bgp-ipvpn
|   |   |   +---segment-routing-v6 <bgp-instance-id>
|   |   |   |   +---srv6-instance <id> default-locator <name>
|   |   |   |   +---source-address <ipv6-address>
|   |   |   |   +---route-distinguisher <rd>  
|   |   |   |   +---vrf-export  
|   |   |   |   +---vrf-import  
|   |   |   |   +---vrf-target  
|   |   |   |   +---default-route-tag <number>
|   |   |   |   +---shutdown

The following MD-CLI commands configure a VPRN for SRv6 with the VPN-IP families.

configure
+--service
|   +---vprn <service-id>
|   |   +-- segment-routing-v6 <number>
|   |   |   +-- apply-groups <reference>
|   |   |   +-- apply-groups-exclude <reference>
|   |   |   +-- locator <reference>
|   |   |       +-- apply-groups <reference>
|   |   |       +-- apply-groups-exclude <reference>
|   |   |       +-- function
|   |   |           +-- end-dt4
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt46
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt6
|   |   |           |   +-- value <number>
|   |   +---bgp-ipvpn
|   |   |   +-- segment-routing-v6 <number>
|   |   |       +-- admin-state <keyword>
|   |   |       +-- apply-groups <reference>
|   |   |       +-- apply-groups-exclude <reference>
|   |   |       +-- default-route-tag <number>
|   |   |       +-- domain-id <string>
|   |   |       +-- evi <number>
|   |   |       +-- route-distinguisher <string | keyword>
|   |   |       +-- source-address <global-unicast-ipv6-address>
|   |   |       +-- srv6
|   |   |       |   +-- default-locator <reference>
|   |   |       |   +-- instance <reference>
|   |   |       +-- vrf-export
|   |   |       |   +-- apply-groups <reference>
|   |   |       |   +-- apply-groups-exclude <reference>
|   |   |       |   +-- policy <string | reference>
|   |   |       +-- vrf-import
|   |   |       |   +-- apply-groups <reference>
|   |   |       |   +-- apply-groups-exclude <reference>
|   |   |       |   +-- policy <string | reference>
|   |   |       +-- vrf-target
|   |   |           +-- community <string>
|   |   |           +-- export-community <string>
|   |   |           +-- import-community <string>

The following classic CLI commands configure a VPRN for SRv6 with EVPN-IFL.

configure
+--service
|   +---vprn <service-id>
|   |   +-- segment-routing-v6 <number>
|   |   |   +-- apply-groups <reference>
|   |   |   +-- apply-groups-exclude <reference>
|   |   |   +-- locator <reference>
|   |   |       +-- apply-groups <reference>
|   |   |       +-- apply-groups-exclude <reference>
|   |   |       +-- function
|   |   |           +-- end-dt4
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt46
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt6
|   |   |           |   +-- value <number>
|   |   +---bgp-evpn
|   |   |   +---segment-routing-v6 <bgp-instance-id>
|   |   |   |   +---srv6-instance <id> default-locator <name>
|   |   |   |   +---source-address <ipv6-address>
|   |   |   |   +---route-distinguisher <rd>  
|   |   |   |   +---vrf-export  
|   |   |   |   +---vrf-import  
|   |   |   |   +---vrf-target  
|   |   |   |   +---default-route-tag <number>
|   |   |   |   +---shutdown

The following MD-CLI commands configure a VPRN for SRv6 with EVPN-IFL.

configure
+--service
|   +---vprn <service-id>
|   |   +-- segment-routing-v6 <number>
|   |   |   +-- apply-groups <reference>
|   |   |   +-- apply-groups-exclude <reference>
|   |   |   +-- locator <reference>
|   |   |       +-- apply-groups <reference>
|   |   |       +-- apply-groups-exclude <reference>
|   |   |       +-- function
|   |   |           +-- end-dt4
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt46
|   |   |           |   +-- value <number>
|   |   |           +-- end-dt6
|   |   |           |   +-- value <number>
|   |   +---bgp-evpn
|   |   |   +-- segment-routing-v6 <number>
|   |   |       +-- admin-state <keyword>
|   |   |       +-- apply-groups <reference>
|   |   |       +-- apply-groups-exclude <reference>
|   |   |       +-- default-route-tag <number>
|   |   |       +-- domain-id <string>
|   |   |       +-- evi <number>
|   |   |       +-- resolution <keyword>
|   |   |       +-- route-distinguisher <string | keyword>
|   |   |       +-- source-address <global-unicast-ipv6-address>
|   |   |       +-- srv6
|   |   |       |   +-- default-locator <reference>
|   |   |       |   +-- instance <reference>
|   |   |       +-- vrf-export
|   |   |       |   +-- apply-groups <reference>
|   |   |       |   +-- apply-groups-exclude <reference>
|   |   |       |   +-- policy <string | reference>
|   |   |       +-- vrf-import
|   |   |       |   +-- apply-groups <reference>
|   |   |       |   +-- apply-groups-exclude <reference>
|   |   |       |   +-- policy <string | reference>
|   |   |       +-- vrf-target
|   |   |           +-- community <string>
|   |   |           +-- export-community <string>
|   |   |           +-- import-community <string>

The associated locator must be configured to enable SRv6 on the VPRN service. In addition, the following rules apply:

  • The function value can either be statically configured or dynamically allocated.

  • Any Layer 3 function behavior can be configured, although VPN-IPv4 and EVPN-IFL IPv4 routes are advertised with end-dt4 or end-dt46 in that preference order (if they exist) and VPN-IPv6 and EVPN-IFL IPv6 routes are advertised with end-dt6 or end-dt46 in that preference order.

  • The VPRN or label mode is not relevant to SRv6 and setting it has no effect on the behavior of the SRv6 feature.

  • The following are supported:

    • BGP-IPVPN and BGP-EVPN (EVPN-IFL) families are simultaneously supported in the same VPRN where SRv6 is enabled.

    • Up to two BGP instances per VPRN are supported.

    • The two BGP instances can be associated with the same family or different families.

    • A family can have two BGP instances of SRv6 encapsulation.

    • In addition, two BGP-IPVPN instances can be configured with SRv6 and MPLS encapsulations respectively. Two BGP-EVPN instances can also be configured with SRv6 and MPLS encapsulations, respectively.

  • The following applies to VPRN feature interaction with SRv6:

    • The following commands under the VPRN context only operate on MPLS encapsulations:

      • class-forwarding

      • entropy-label

      • hash-label

      • label-mode

      • ttl-propagate

    • The following commands under the VPRN context are mutually exclusive with SRv6:

      • carrier-carrier-vpn

      • network-interface

      • export-inactive-bgp

    • The following VPRN commands and features operate on MPLS and SRv6 encapsulations:

      • vprn-type

      • allow-export-bgp-vpn

      • ecmp-unequal-cost

      • bgp-vpn-backup

      • Unicast protocols on PE-CE interfaces

      • Commands such as ipsec, nat, and subscriber-interfaces are supported (no interaction).

      • ECMP and edge PIC are supported for SRv6 and also across routes of the same family with different encapsulations; for example, the same prefix is resolved to SRv6 and MPLS tunnels.

      • Route target-based leaking is supported for SRv6 routes in the VPRN.

SRv6 VPRN and BGP path attribute propagation between route table BGP owners

As is the case with existing MPLS encapsulation, BGP Path Attribute Propagation between BGP owners of the same VPRN route table is supported, including routes with SRv6 encapsulation.

  • BGP Path Attribute propagation for SRv6 routes does not require enabling a CLI knob

  • In cases where multiple BGP owners coexist in the same VPRN route table with a single instance, propagation is supported in the following cases, irrespective of the encapsulation of the route (MPLS or SRv6):

    • VPN-IPv4/6 ← → EVPN-IFL

    • VPN-IPv4/6 ← → VPN-IPv4/6 – when allow-export-bgp-vpn is enabled

    • EVPN-IFL← → EVPN-IFL – when allow-export-bgp-vpn is enabled

    • VPN-IPv4/6 ← → IPv4/v6

    • EVPN-IFL ← → IPv4/v6

    • VPN-IPv4/6 ← → EVPN-IFF (requires iff-attribute-uniform-propagation)

    • EVPN-IFL ← → EVPN-IFF (requires iff-attribute-uniform-propagation)

  • In the case of multi-instance bgp VPRNs, VPN-IPv4/6 and EVPN-IFL routes received in one instance are readvertised into the other instance, including all the BGP Path Attributes of the original route.

  • The following attributes are filtered out before the path attributes are propagated in all the previously described cases:

    • all type 0x06 extended communities

    • the BGP encapsulation extended community

    • the BGP encapsulation attribute

    • the BGP Prefix-SID attribute (which includes SRv6 Service SID TLVs)

    • all Route Target extended communities

Migration from MPLS to SRv6 in VPRN services

To allow a seamless migration from MPLS tunnels to SRv6 in VPRN services, both MPLS auto-bind tunnels and SRv6 are supported in the same VPRN service.

Routes for the same prefix follow regular route table selection in the VPRN. At the end of the selection process, if there are still SRv6 and MPLS routes, the SRv6 route is selected first and the MPLS route is removed from consideration.

SRv6 service SIDs and BGP routes in the base router

In the base router BGP instance, BGP routes may be originated, propagated or received with SRv6 TLVs in the prefix SID attribute. The processing rules are expected to use the following logic:

  • A VPN-IPv4 or VPN-IPv6 route that is imported by the base router BGP instance from a VPRN (through the VRF export process), and that already has a prefix SID attribute with an SRv6 TLV at this stage, is advertised to all VPN-IPv4 and VPN-IPv6 peers of the base router BGP instance, with the next-hop-self as normal. There is no option to strip the SRv6 TLV or prefix SID attribute toward any of these peers, however, there is an option to drop the entire route toward selected peers or groups. This is similar to the effect of a BGP export policy that drops the route, but route policies cannot match routes based on the presence of an SRv6 TLV.

  • An IPv6 route that is imported by the base router BGP instance from another protocol’s route that was added to base router route table (static, OSPF, IS-IS, and so on) does not have a prefix SID attribute carrying the SRv6 TLV added to it by default. Adding a prefix SID can only be achieved by configuring the config>router>bgp>segment-routing-v6>family ipv6 add-srv6-tlvs command; however, this command also has the effect of adding a prefix SID attribute with SRv6 TLV to BGP IPv6 routes received from other peers without the SRv6 TLV and that are propagated to other family IPv6 peers with next-hop-self applied.

  • Any BGP route received with a prefix SID attribute carrying SRv6 TLVs that is not an imported VPN-IPv4, VPN-IPv6, or EVPN route or an IPv6 unlabeled unicast route is treated the same as an existing route, that is, by ignoring the prefix SID attribute contents, resolving the route based only on its BGP next hop, and propagating the prefix SID attribute to other peers unchanged.

  • If an IPv6 unlabeled unicast route is received with a prefix SID attribute carrying an SRv6 TLV and the config>router>bgp>segment-routing-v6>family ipv6 ignore-received-srv6-tlvs command is configured, that route is treated the same as an existing route, that is, by ignoring the prefix SID attribute contents, resolving the route based only on its BGP next hop, and propagating the attribute to other peers unchanged, irrespective of the next-hop-self.

  • If an IPv6 unlabeled unicast route is received with a prefix SID attribute carrying an SRv6 TLV and the config>router>bgp>segment-routing-v6>family ipv6 ignore-received-srv6-tlvs command is set to FALSE, that route is considered resolved only if its BGP next hop is reachable and the locator prefix is reachable. The datapath or FNH programming and IGP cost to reach the next hop (used by the BGP decision process) is based on the route to the locator prefix. The IPv6 route is propagated to other family IPv6 peers with the prefix SID attribute and its SRv6 TLV unchanged, irrespective of the next-hop-self.

  • An IPv6 unlabeled unicast route that is received without a prefix SID attribute containing an SRv6 TLV does not have a prefix-SID attribute carrying the SRv6 TLV added to it by default, even if it is re-advertised with the next-hop-self applied. This can only be achieved by configuring the config>router>bgp>segment-routing-v6>family ipv6 add-srv6-tlvs; however, this command also has the effect of adding a prefix SID attribute with SRv6 TLV to imported or redistributed routes.

configure
+--router
   +--segment-routing
|   |   +--- base-routing-instance
|   |   |   |   locator <locator-name>
|   |   |   |   +---function
|   |   |   |   |   +---end-dt6 <integer>
configure
+--router
   +--bgp
      +--segment-routing-v6
         +--source-address <ipv6-address>
         +--family* 
            +--add-srv6-tlvs
            |  +--locator <locator-name>
            +--ignore-received-srv6-tlvs
Note:
  • When an SRv6 TLV is added to an IPv6 unlabeled unicast route, the signaled behavior is End.DT6 and the SID Structure Sub-Sub-TLV contains a TO and TL of 0.

  • As in the case of VPRNs, a locator is configured for the base router. If the End.DT6 function for that locator is not statically configured, a dynamically-allocated function is reserved for the End.DT6 behavior. This is internally mapped to a label, as in the case of VPRNs.

SRv6 Epipe services

Epipe services support SRv6 End.DX2 behavior. Currently, the SRv6 SID for End.DX2 is signaled by EVPN-VPWS AD per-EVI routes.

Use the following CLI commands to configure an Epipe service for SRv6:

+--epipe
   +--segment-routing-v6 <instance-id> 
   |  +--locator <locator-name>  
   |  |  +--function 
   |  |  |  +--end-dx2 <integer>
   |  +--bgp-evpn
   |  |  +-- segment-routing-v6 <number>
   |  |  |   +-- admin-state <keyword>
   |  |  |   +-- default-route-tag <number>
   |  |  |   +-- ecmp <number>
   |  |  |   +-- force-vc-forwarding <keyword>
   |  |  |   +-- oper-group <reference>
   |  |  |   +-- route-next-hop
   |  |  |   |   +-- ip-address <ip-address>
   |  |  |   |   +-- system-ipv4
   |  |  |   |   +-- system-ipv6
   |  |  |   +-- source-address <global-unicast-ipv6-address>
   |  |  |   +-- srv6
   |  |  |       +-- default-locator <reference>
   |  |  |       +-- instance <reference>

Where the following conditions apply:

  • The SRv6 Epipe is configured with the locator that is used. This determines the SID structure and value that is advertised in the AD per EVI route for the service.

  • A single SRv6 instance with a single locator is supported on Epipes.

  • The SRv6 Epipe uses an End.DX2 function value that, if not configured, is dynamically allocated by the system, out of the dynamic range available for the locator. If the end-dx2 function is configured, then this value is used instead of a dynamic value.

  • The following is supported in SRv6 Epipes:

    ecmp
    used for aliasing on remote SRv6 EVPN ES destinations
    force-vlan-vc-forwarding
    used to preserve one VLAN tag in the payload of the SRv6 tunnel. On termination, the VLAN tag is transparently passed through the termination SRv6 FPE.
    default-route-tag
    used to allow easy matching of the service routes on export policies
    oper-group
    required for fault-propagation and fate-sharing with the monitoring objects
    route-next-hop
    required to control the advertised BGP next hop for the EVPN AD per-EVI route. When EVPN Multihoming is used, this value must match the ES originating-ip and the ES route-next-hop value.

EVPN-VPWS is the control plane technology to signal SRv6 Epipes. The local-attachment-circuit Ethernet-tag value is advertised in an AD per-EVI route that may have a zero or non-zero ESI (if multihoming is used). The AD per-EVI routes are advertised along with the Layer-2 SRv6 Services TLV encoding with an End.DX2 behavior and using the transposition procedures described in Transposition procedures when advertising service routes. Upon reception the ingress PE creates an EVPN destination, as long as the received route includes the remote expected Ethernet-tag and route-target. The following CLI excerpt shows a configuration example and the created EVPN SRv6 destination on the PE:

*A:PE-3# configure service epipe 200 
*A:PE-3>config>service>epipe# info 
----------------------------------------------
            segment-routing-v6 1 create
                locator "LOC-1"
                    function
                        end-dx2 200
                    exit
                exit
            exit
            bgp
            exit
            bgp-evpn
                local-attachment-circuit ac-23 create
                    eth-tag 23
                exit
                remote-attachment-circuit ac-5 create
                    eth-tag 5
                exit
                evi 200
                segment-routing-v6 bgp 1 srv6-instance 1 default-locator "LOC-1" create
                    source-address 2001:db8::3
                    ecmp 2
                    route-next-hop system-ipv6
                    no shutdown
                exit
            exit
            sap lag-1:200 create
                no shutdown
            exit
            no shutdown
----------------------------------------------
*A:PE-3>config>service>epipe# /show service id 200 segment-routing-v6 destinations 

===============================================================================
TEP, SID
===============================================================================
Instance  TEP Address                        Segment Id
-------------------------------------------------------------------------------
1         2001:db8::5                        cafe:1:0:5:c:8000::
-------------------------------------------------------------------------------
Number of TEP, SID: 1
-------------------------------------------------------------------------------
===============================================================================

===============================================================================
Segment Routing v6 Ethernet Segment Dest
===============================================================================
Instance  Eth SegId                       Num. Macs     Last Change
-------------------------------------------------------------------------------
No Matching Entries
===============================================================================

SRv6 Epipes support EVPN Multihoming. Their associated Ethernet Segments (ES) can be shared among MPLS services and other SRv6 services.

EVPN Multihoming for SRv6 Epipes follows these guidelines:

  • The ES used by SRv6 Epipes is only supported in evi-rt mode, therefore separate AD per-ES routes per service are advertised for SRv6 Epipes.

  • If evi-rt-set is configured in the ES, the SRv6 services are skipped when packing route-targets for the AD per-ES route. This is the same behavior as having a Epipe services with a configured vsi-export policy; the system also skips the route-targets for Epipes with vsi-export when packing multiple route-targets in the same AD per-ES route.

  • The advertised AD per-ES route for the SRv6 Epipe includes a Layer-2 SRv6 Services TLV with a SID value of zero. The used behavior code point is End.DT2M. In addition, the AD per-ES route is also advertised with an ESI Label extended community that encodes the implicit-null value in the ESI label field. A debug example of a received AD per-ES route for SRv6 follows:
    22 2021/05/21 18:23:10.247 UTC MINOR: DEBUG #2001 Base Peer 1: 2001:db8::2
    "Peer 1: 2001:db8::2: UPDATE
    Peer 1: 2001:db8::2 - Received BGP UPDATE:
        Withdrawn Length = 0
        Total Path Attr Length = 125
        Flag: 0x90 Type: 14 Len: 48 Multiprotocol Reachable NLRI:
            Address Family EVPN
            NextHop len 16 Global NextHop 2001:db8::2
            Type: EVPN-AD Len: 25 RD: 192.0.2.2:200 ESI: 01:d8:47:ff:00:00:00:00:01:
    00, tag: MAX-ET Label: 0 
        Flag: 0x40 Type: 1 Len: 1 Origin: 0
        Flag: 0x40 Type: 2 Len: 0 AS Path:
        Flag: 0x40 Type: 5 Len: 4 Local Preference: 100
        Flag: 0xc0 Type: 16 Len: 16 Extended Community:
            target:64500:200
            esi-label:3/Single-Active
        Flag: 0xc0 Type: 40 Len: 37 Prefix-SID-attr:
           SRv6 Services TLV (37 bytes):-
             Type: SRV6 L2 Service TLV (6)
             Length: 34 bytes, Reserved: 0x0
             SRv6 Service Information
             Service Information Sub-TLV Type 1
                Type: 1 Len: 30 Rsvd1: 0x0
                SRv6 SID: ::
                SID Flags: 0x0 Endpoint Behavior: 0x18 Rsvd2: 0x0
                SRv6 SID Sub-Sub-TLV
                   Type: 1 Len: 6
                   BL:0 NL:0 FL:0 AL0 TL:0 TO:0
    

The creation of ES destinations follows the same rules as in EVPN-VPWS services with MPLS transport, except that the ES destination is resolved to remote SRv6 SIDs, as shown in the following example:

*A:PE-5# show service id 200 segment-routing-v6 destinations 

===============================================================================
TEP, SID
===============================================================================
Instance  TEP Address                        Segment Id
-------------------------------------------------------------------------------
No Matching Entries
===============================================================================

===============================================================================
Segment Routing v6 Ethernet Segment Dest
===============================================================================
Instance  Eth SegId                       Num. Macs     Last Change
-------------------------------------------------------------------------------
1         01:d8:47:ff:00:00:00:00:01:00   0             06/18/2021 17:04:15
-------------------------------------------------------------------------------
Number of entries: 1
-------------------------------------------------------------------------------
===============================================================================
*A:PE-5# show service id 200 segment-routing-v6 esi 01:d8:47:ff:00:00:00:00:01:00 

===============================================================================
Segment Routing v6 Ethernet Segment Dest
===============================================================================
Instance  Eth SegId                       Num. Macs     Last Change
-------------------------------------------------------------------------------
1         01:d8:47:ff:00:00:00:00:01:00   0             06/18/2021 17:04:15
-------------------------------------------------------------------------------
Number of entries: 1
-------------------------------------------------------------------------------
===============================================================================

===============================================================================
Segment Routing v6 Dest TEP Info
===============================================================================
Instance  TEP Address                   Segment Id          Last Change
-------------------------------------------------------------------------------
1         2001:db8::2                   cafe:1:0:2:c:8000:: 06/18/2021 17:04:15
-------------------------------------------------------------------------------
Number of entries : 1
-------------------------------------------------------------------------------
===============================================================================

SRv6 VPLS services

VPLS services support SRv6 End.DT2M behavior for BUM traffic and End.DT2U behavior for known unicast traffic. The SRv6 SID for End.DT2M is signaled by EVPN Inclusive Multicast Ethernet Tag (IMET) routes, whereas the SID for End.DT2U is signaled along with EVPN MAC/IP routes.

Use the following contexts to configure a VPLS service for SRv6:

  • MD-CLI
    configure service vpls "1" segment-routing <instance-id>   
    configure service vpls "1" bgp-evpn segment-routing-v6 <number>  
  • classic CLI
    configure service vpls "1" segment-routing-v6 <instance-id>
    configure service vpls "1" bgp-evpn segment-routing-v6 srv6-instance 1 default-locator 

The following conditions apply for a VPLS service for SRv6:

  • The SRv6 VPLS is configured using the locator. This determines the SID structure and value that is advertised along with the EVPN IMET and MAC/IP routes for the service.

  • A single SRv6 instance with a single locator is supported on VPLS services.

  • SRv6 VPLS services use End.DT2M and End.DT2U SIDs. Their corresponding function values, if not configured, are allocated automatically by the router. Dynamic or static, both function values (End.DT2M and End.DT2U) are needed so the EVPN routes for the services are advertised. While both values are different locally on the service, the router accepts the same value for End.DT2M and End.DT2U functions from a remote PE for the same service.

  • If the following command is configured on the PEs attached to the same service, the service MTU value is advertised in the EVPN Layer 2 Attributes extended community along with the IMET routes.
    • MD-CLI
      configure service vpls bgp-evpn routes incl-mcast advertise-l2-attributes
    • classic CLI
      configure service vpls bgp-evpn incl-mcast-l2-attributes-advertisement
    Upon receiving the signaled MTU from an egress PE, the ingress PE compares the received and local MTU. In case of a mismatch, the EVPN SRv6 destination goes operationally down, and the operational flag MTU-Mismatch shows the reason. Use the following command to configure the router to ignore the MTU signaled by the remote PE and bring up the SRv6 destination if there are no other reasons to keep it down.
    configure service vpls bgp-evpn ignore-mtu-mismatch
  • The following commands are supported in SRv6 VPLS services:

    • force-vlan-vc-forwarding

      This command preserves one VLAN tag in the payload of the SRv6 tunnel.

    • default-route-tag

      This command matches BGP EVPN routes for the service on export policies, typically to add or modify BGP attributes.

    • oper-group

      This command is required for fault-propagation purposes.

    • protected-src-mac-violation-action discard

      This command is required for loop avoidance, along with mac-duplication blackhole.

    • split-horizon-group

      This command is used for seamless integration with spoke SDPs and migration from EVPN-MPLS services (in multi-instance services).

    • route-next-hop

      This command is used to control the BGP next-hop used for service routes. If not configured, it is system-ipv4 by default.

    • source-address

      If not configured, it is inherited from the locator’s source address. The source address does not need to be reachable or even exist on a local interface. This is possible because the source address is not looked up in the datapath at the remote PE.

    • resolution {route-table | tunnel-table | fallback-tunnel-to-route-table}

      This command is used to set the resolution of SRv6 routes in the route table or tunnel table (needed for SRv6 policy), and even a fallback from tunnel to route-table resolution.

Configuration example and the allocated functions on the PE

A:PE-2# configure service vpls 1900 
A:PE-2>config>service>vpls# info 
            segment-routing-v6 1 create
                locator “LOC-1"
                    function
                        end-dt2u
                        end-dt2m
                    exit
                exit
            exit
            bgp
            exit
            bgp-evpn
                evi 1900
                segment-routing-v6 bgp 1 srv6-instance 1 default-locator "LOC-1" create
                    source-address 2001:db8::2
                    route-next-hop 2001:db8::2
                    no shutdown
                exit
            exit
            stp
                shutdown
            exit
            no shutdown            

show service id 1900 segment-routing-v6 instance 1
===============================================================================
Segment Routing v6 Instance 1 Service 1900
===============================================================================
Locator                                                           
  Type         Function  SID                                     Status
-------------------------------------------------------------------------------
LOC-1
  End.DT2U       *504273 cafe:1:0:2:7b1d:1000::                  ok
  End.DT2M       *504272 cafe:1:0:2:7b1d::                       ok
===============================================================================
show router segment-routing-v6 local-sid end-dt2u end-dt2m 
===============================================================================
Segment Routing v6 Local SIDs
===============================================================================
SID                                               Type           Function
  Locator                                                        
  Context                                                        
-------------------------------------------------------------------------------
cafe:1:0:2:7b1d::                                 End.DT2M       504272
  LOC-1
  SvcId: 1900 Name: bd-1900-srv6
cafe:1:0:2:7b1d:1000::                            End.DT2U       504273
  LOC-1
  SvcId: 1900 Name: bd-1900-srv6
-------------------------------------------------------------------------------
SIDs : 2
-------------------------------------------------------------------------------
===============================================================================

Remote PEs

If the remote PEs attached to the same VPLS services are configured in a similar way, the EVPN destinations for BUM and unicast traffic are created and can be displayed as follows:

show service id 1900 segment-routing-v6 destinations
===============================================================================
TEP, SID (Instance 1)
===============================================================================
TEP Address                   Segment Id                       Oper  Mcast Num
                                                               State       MACs
-------------------------------------------------------------------------------
2001:db8::3                   cafe:1:0:3:7b1d:7000::           Up    BUM   1
2001:db8::4                   cafe:1:0:4:7:b1db::              Up    BUM   0
2001:db8::5                   cafe:1:0:5:7b1d:d000::           Up    BUM   0 
2001:db8::5                   cafe:1:0:5:7b1d:e000::           Up    None  1
-------------------------------------------------------------------------------
Number of TEP, SID: 4
-------------------------------------------------------------------------------
===============================================================================

===============================================================================
Segment Routing v6 Ethernet Segment Dest
===============================================================================
Instance  Eth SegId                       Num. Macs     Last Change
-------------------------------------------------------------------------------
No Matching Entries
===============================================================================

IMET and MAC/IP routes for SRv6 transport, received on a router

The EVPN IMET and MAC/IP routes for SRv6 are advertised and expected to be received with the SRv6 Services TLV, which includes all the information related to the SRv6 SID and behavior that is used by the receiving router. For other services, transposition procedures are followed as described in Transposition procedures when advertising service routes.

show router bgp routes evpn incl-mcast community target:64500:1900 next-hop 2001:db8::5 hunt 
===============================================================================
 BGP Router ID:192.0.2.2        AS:64500       Local AS:64500      
===============================================================================
 Legend -
 Status codes  : u - used, s - suppressed, h - history, d - decayed, * - valid
                 l - leaked, x - stale, > - best, b - backup, p - purge
 Origin codes  : i - IGP, e - EGP, ? - incomplete

===============================================================================
BGP EVPN Inclusive-Mcast Routes
===============================================================================
-------------------------------------------------------------------------------
RIB In Entries
-------------------------------------------------------------------------------
Network        : n/a
Nexthop        : 2001:db8::5
Path Id        : None                   
From           : 2001:db8::5
Res. Nexthop   : fe80::b449:1ff:fe01:1f
Local Pref.    : 100                    Interface Name : int-PE-2-PE-5
Aggregator AS  : None                   Aggregator     : None
Atomic Aggr.   : Not Atomic             MED            : None
AIGP Metric    : None                   IGP Cost       : 10
Connector      : None
Community      : target:64500:1900
Cluster        : No Cluster Members
Originator Id  : None                   Peer Router Id : 192.0.2.5
Flags          : Used Valid Best IGP 
Route Source   : Internal
AS-Path        : No As-Path
EVPN type      : INCL-MCAST             
Tag            : 0                      
Originator IP  : 2001:db8::5
Route Dist.    : 192.0.2.5:1900
Route Tag      : 0                      
Neighbor-AS    : n/a
Orig Validation: N/A                    
Source Class   : 0                      Dest Class     : 0
Add Paths Send : Default                
Last Modified  : 00h14m18s              
SRv6 TLV Type  : SRv6 L2 Service TLV (6)
SRv6 SubTLV    : SRv6 SID Information (1)
Sid            : cafe:1:0:5::
Full Sid       : cafe:1:0:5:7b1d:d000::
Behavior       : End.DT2M (24)
SRv6 SubSubTLV : SRv6 SID Structure (1)
Loc-Block-Len  : 32                     Loc-Node-Len   : 32
Func-Len       : 20                     Arg-Len        : 0
Tpose-Len      : 20                     Tpose-offset   : 64
-------------------------------------------------------------------------------
PMSI Tunnel Attributes : 
Tunnel-type    : Ingress Replication    
Flags          : Type: RNVE(0) BM: 0 U: 0 Leaf: not required
MPLS Label     : 8068560                
Tunnel-Endpoint: 2001:db8::5
-------------------------------------------------------------------------------
 
-------------------------------------------------------------------------------
RIB Out Entries
-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
Routes : 1
===============================================================================      

show router bgp routes evpn mac community target:64500:1900 next-hop 2001:db8::5 hunt
===============================================================================
 BGP Router ID:192.0.2.2        AS:64500       Local AS:64500      
===============================================================================
 Legend -
 Status codes  : u - used, s - suppressed, h - history, d - decayed, * - valid
                 l - leaked, x - stale, > - best, b - backup, p - purge
 Origin codes  : i - IGP, e - EGP, ? - incomplete

===============================================================================
BGP EVPN MAC Routes
===============================================================================
-------------------------------------------------------------------------------
RIB In Entries
-------------------------------------------------------------------------------
Network        : n/a
Nexthop        : 2001:db8::5
Path Id        : None                   
From           : 2001:db8::5
Res. Nexthop   : fe80::b449:1ff:fe01:1f
Local Pref.    : 100                    Interface Name : int-PE-2-PE-5
Aggregator AS  : None                   Aggregator     : None
Atomic Aggr.   : Not Atomic             MED            : None
AIGP Metric    : None                   IGP Cost       : 10
Connector      : None
Community      : target:64500:1900
                 mac-mobility:Seq:0/Static
Cluster        : No Cluster Members
Originator Id  : None                   Peer Router Id : 192.0.2.5
Flags          : Used Valid Best IGP 
Route Source   : Internal
AS-Path        : No As-Path
EVPN type      : MAC                    
ESI            : ESI-0
Tag            : 0                      
IP Address     : n/a
Route Dist.    : 192.0.2.5:1900         
Mac Address    : 00:ca:ca:de:ba:ca      
MPLS Label1    : LABEL 504286           MPLS Label2    : n/a
Route Tag      : 0                      
Neighbor-AS    : n/a
Orig Validation: N/A                    
Source Class   : 0                      Dest Class     : 0
Add Paths Send : Default                
Last Modified  : 00h14m26s              
SRv6 TLV Type  : SRv6 L2 Service TLV (6)
SRv6 SubTLV    : SRv6 SID Information (1)
Sid            : cafe:1:0:5::
Full Sid       : cafe:1:0:5:7b1d:e000::
Behavior       : End.DT2U (23)
SRv6 SubSubTLV : SRv6 SID Structure (1)
Loc-Block-Len  : 32                     Loc-Node-Len   : 32
Func-Len       : 20                     Arg-Len        : 0
Tpose-Len      : 20                     Tpose-offset   : 64
 
-------------------------------------------------------------------------------
RIB Out Entries
-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
Routes : 1
=============================================================================== 

VPLS and EVPN SRv6 seamless integration

EVPN and VPLS seamless integration (RFC 8560) is supported for SRv6 VPLS services, as in the case of VPLS services with EVPN-MPLS.

This seamless integration comprises the following aspects:

  • SDP binds signaled by TLDP (manual, BGP-AD) or BGP can coexist in VPLS services where EVPN SRv6 is enabled.

    If an EVPN destination and an SDP binding to the same far end exist, the router brings down the SDP binding. The EVPN SRv6 destination far end is specified by the route’s next hop (shown as TEP Address in the show service id id segment-routing-v6 destinations command), and not the locator. The router allows the creation of the EVPN destination and the SDP binding to the same far end but the SDP binding is kept operationally down (with a flag indicating there is an EVPN route conflict):

    show service id 10 sdp 32765:4294967291 detail | match Flag 
    Flags              : EvpnRouteConflict
    

    In case of an SDP binding and EVPN endpoint to different far-end IPs on the same remote PE, both links are operationally up. This can happen if the SDP binding is terminated in an IPv6 address or IPv4 address different from the system address where the EVPN endpoint is terminated.

  • The user can configure spoke SDPs and all the EVPN endpoints in the same split horizon group (SHG).

    The split-horizon-group command is added under the bgp-evpn>segment-routing-v6 context so that the EVPN SRv6 endpoints are added to a split-horizon-group.

    Note: the BGP EVPN SRv6 split horizon group must reference user-configured SHGs and not auto-created SHGs.

    If the split-horizon-group command for bgp-evpn>segment-routing-v6> context is not used, the default split horizon group (in which all the EVPN endpoints are) is still used, but it is not possible to reference it on SAP or spoke SDPs. User-configured split horizon groups can be configured within the service context. The same group-name can be associated to SAPs, spoke SDPs, pw-templates, pw-template-bindings and EVPN-SRv6 destinations. The bgp-evpn segment-routing-v6 split-horizon-group configuration is only allowed if the bgp-evpn> segment-routing-v6 is disabled. No changes are allowed when bgp-evpn> segment-routing-v6 is enabled.

  • The advertisement of MACs learned on spoke SDPs or SAPs that are part of an EVPN split horizon group are disabled.

    When the SAPs, spoke SDPs, or both (manual or BGP-AD/VPLS discovered) are configured within the same split-horizon-group as the EVPN destinations, MAC addresses are still learned on them, but they are not advertised in EVPN.

    The previously mentioned is also true if proxy-ARP or ND is enabled and an IP-MAC pair is learned on a SAP or SDP binding that belongs to the EVPN split-horizon-group.

EVPN SRv6 multihoming

SR OS supports EVPN multihoming in VPLS services where SRv6 is enabled. In EVPN-MPLS VPLS services, all-active multihoming (and single-active multihoming by default) makes use of the ESI label to identity the source Ethernet Segment from which the packet was sent. At the egress PE, a lookup on the ESI label determines if the router can send the packet to a particular local Ethernet Segment. In EVPN-SRv6 VPLS services, the identification of the source Ethernet Segment is based on the addition of an argument (of type arg.FE2 as per RFC9252) that is appended immediately after the function when forwarding end.dt2m or u.dt2m packets.

Before configuring Ethernet Segments that require arguments for split-horizon filtering, use the following command to enable the use of 16-bit arguments in base SRv6 locators:

  • MD-CLI
    configure router segment-routing segment-routing-v6 locator argument-length 16   
  • classic CLI
    configure router segment-routing segment-routing-v6 locator argument-length 16

For micro segments the use of arguments is enabled as follows:

  • MD-CLI
    configure router segment-routing segment-routing-v6 micro-segment argument-length 16   
  • classic CLI
    configure router segment-routing segment-routing-v6 micro-segment argument-length 16
The following considerations apply to EVPN multihoming in VPLS SRv6 services:
  • You can associate Ethernet Segments used in SRv6 VPLS services with ports, lags or SDPs, and can be configured as type virtual or regular.
  • The VPLS SRv6 instances using Ethernet Segments are configured with the mh-mode network command.
  • When the Ethernet Segment is configured as:
    • MD-CLI
      configure service system bgp evpn ethernet-segment multi-homing-mode single-active-no-esi-label    
    • classic CLI
      configure service system bgp-evpn ethernet-segment multi-homing single-active no-esi-label 

    No argument is allocated by the router. In any other mode, an argument in the range 1 to 4095 is dynamically allocated by the router.

  • When the argument-length command is set to 16 (from the default 0), the SRv6 prefix plus function lengths must be a multiple of 8, as enforced by the CLI. The length of each individual field, locator, and function, may be different from a multiple of 8, if the sum is a multiple of 8, so that the argument is pushed or read at the octet boundary. For micro-segments, this octet boundary is guaranteed.

Segment routing policies with an IPv6 data plane

Segment routing policies with an IPv6 data plane (SRv6 policies) can be used as a part of traffic engineering in SRv6. SRv6 policies build on the concepts introduced with SR policies with an MPLS data plane:
  • The SRv6 policies use SRv6 SIDs instead of MPLS labels as SIDs in the segment list and as binding SID. The SIDs can be encapsulated in a segment routing header (SRH).
  • SRv6 polices can be statically configured on the router or programmed through BGP. In both cases, the user should use the following command to configure a source IPv6 address for use by SRv6 Polices:
    configure router segment-routing segment-routing-v6 source-address

See Segment routing policies for more information about SR policies.

The router supports uncompressed 128-bit SRv6 SIDs as the binding SID and as SIDs for the segments in the segment lists. The binding SID and all segment list SIDs must be of the same type, that is, MPLS labels or SRv6 SIDs. The SRv6 SID can correspond to a 128-bit SID of any function type (for example, node SID, adjacency SID, binding SID).
Note: An SRv6 SID can also represent a SID in an IGP flexible algorithm.
At the head end of an SRv6 policy, service packets are encapsulated in the SRv6 packet using a head end behavior type. The following table lists the head end behavior types specified in RFC 8986.
H.Encaps SR head end with encapsulation in an SR policy
H.Encaps.Red H.Encaps with reduced encapsulation
H.Encaps.L2 H.Encaps applied to received L2 frames
H.Encaps.L2.Red H.Encaps.Red applied to received L2 frames
SR OS supports the following head end behavior types:
  • H.Encaps.Red
  • H.Encaps.L2.Red
The supported types are an optimization of the corresponding H.Encaps behavior types, that is, they exclude the first SID in the SRH of the pushed IPv6 header to reduce the length of the SRH. The Destination Address field of the pushed IPv6 header contains the first SID, while the SRH contains the other SIDs and the service SID.
Figure 18. Datapath example for a VPRNv4 service over SRv6

Static SRv6 policies

Use the following commands to configure a static SRv6 policy:
  • configure router segment-routing sr-policies static-policy type
    Set the type to srv6 and only commands relevant to SRv6 policies can be configured. The default value is sr-mpls. If the type is set to sr-mpls, only MPLS labels can be configured for the binding SID and the segments in the segment lists.
  • configure router segment-routing sr-policies static-policy segment-list segment srv6-sid
    Configure SRv6 SIDS for each segment in the segment lists of the SRv6 policy. SRv6 SIDs are 128-bit IPv6 addresses.
    Note: The static SRv6 policy can only be administratively enabled if its type is SRv6 and all its segments have SRv6 SIDs.
  • configure router segment-routing sr-policies static-policy segment-routing-v6 binding-sid
    Configure the binding SID. The binding SID is mandatory. The system supports only one binding SID per SRv6 policy. The locator or the ip-address command in the binding-sid context configure the value of the binding SID. The locator and the ip-address commands are mutually exclusive. The command to use depends on the configuration of the configure router segment-routing sr-policies static-policy head-end command:
    • If the head end is a non-local IP address, use the ip-address command to configure an SRv6 binding SID for a remote SRv6 policy. The address is a 128-bit IPv6 address. The router does not perform any local checks on the configured address.
      Note: The non-local head end can only be an IPv4-formatted address to match the head end BGP router ID, which can only be IPv4.
    • If the head end is configured as local, configure the locator name, function, and function value for the SRv6 binding SID in the following context:
      configure router segment-routing sr-policies static-policy segment-routing-v6 binding-sid locator

      The router checks that a corresponding locator exists in the configure router segment-routing segment-routing-v6 context.

      The function is a binding SID function for the static SRv6 policy and can only be end-b6-encaps-red. The datapath implements the End.B6.Encaps.Red as the End.B6 function for the binding SID.

      The function value is optional. If configured, the router checks the availability of the value for the configured locator. If a function value is not configured (default), the router dynamically allocates one for the function.

BGP SRv6 policies

The router supports BGP extensions for SRv6 policies as defined in draft-ietf-idr-segment-routing-te-policy-11 Advertising Segment Routing Policies in BGP. BGP uses the sr-policy-ipv6 address family for SR policies with SAFI SR Policy (73).

SRv6 segment lists are signaled using a series of SRv6 SID sub-TLVs, as follows:

  • Segment sub-TLV Type B – SID only, in the form of an IPv6 address
  • SRv6 binding SID sub-TLV

The SRv6 endpoint behavior sub-TLV is not included.

SRv6 binding SID procedures

The router supports one binding SID per SRv6 policy. The application of binding SIDs for SRv6 policies is similar to those for SR policies with an MPLS data plane. The binding SID is programmed using the function type End.B6.Encaps.Red (as defined in RFC 8986).

Binding SIDs for local static SRv6 policies

If the head end is local, the router attempts to program the SR policy. The binding SID is mandatory and the configuration consists of:
  • a locator name
  • a function type
  • a function value (optional)
Only one binding SID can be configured per SRv6 policy. The router performs the following checks and actions:
  • The locator name must be configured.
  • The function type must be configured. Only end-b6-encaps-red is supported.
    Note: The procedures defined in RFC 8986 for the End.B6.Encaps.Red function mean that a maximum of two SRv6 policies can be concatenated to provide an end-to-end path using binding SIDs with this function.
  • A function value can optionally be configured. If a value is configured, it is considered static. If no value is configured, the router tries to allocate a dynamic value for the function within the named locator.
  • The binding SID function value uses the rules for the named locator:
    • The named locator has a static and a dynamic range.

      If a static value is configured, the router checks and allocates within the static range for the locator. If no static value is configured, the router allocates a value within the dynamic range.

    • The named locator uses a reserved global function block from which statically and dynamically allocated function values of all locators can be drawn.

      The router checks and allocates the value within the reserved global function block.

    • All above checks fail.

      The SRv6 policy is not programmed. The router retries the checks at a later time. The allocation scheme for the binding SID function values is first-come-first-served.

Binding SIDs for non-local static SRv6 policies

If the head end is not local, the router advertises the SRv6 policy in BGP toward the head end router.

The binding SID is a 128-bit IPv6 address. The router performs no local checks on the binding SID value. The router advertises the binding SID in BGP using the SRv6 binding SID sub-TLV, see BGP SRv6 policies. The SRv6 endpoint behavior sub-TLV is not included.

Receiving and programing BGP SRv6 policy routes

In general, the behavior for received SRv6 policy routes is similar to IPv4 SR-MPLS policies. For an SRv6 policy imported through BGP, the binding SID type is the indicator of the type of the SR policy. If the binding SID is an SRv6 binding SID, the router performs a longest prefix match against the route.

Note: The longest prefix match fails for SIDs with non-zero argument bits. SR OS requires that the argument bits are set to zero in the incoming SRv6 policy binding SID.

If there is no match, the SRv6 policy (candidate path) is invalid.

If there is a match, the router performs the following checks on the locator and the function value:
  • If the function value collides with an already allocated function value within the static range for the locator matching the binding SID (or within the static part of a reserved global function block for the locator, if this is configured), the programming fails. If there is no collision with an already allocated static function value (regardless of the function type), the router allocates the function value to the End.B6.Encaps.Red function for the binding SID. The allocation is on a first-come-first-served basis.
  • If a received SRv6 policy includes a SID structure TLV associated with the binding SID sub-TLV or the SRv6 segment sub-TLV, the router ignores it.
  • If a received SRv6 policy route contains more than one SRv6 binding SID TLV, the router treats the path as invalid.

If the binding SID is valid, the router programs the SRv6 policy. The binding SID is programmed as a full /128 address.

Tunnel table support for SRv6 policies

Each tunnel in the tunnel table has a protocol owner. The router programs SRv6 polices in TTMv6 with srv6-pol as protocol owner.

If two SR policies have the same color and endpoint but a different technology (MPLS and SRv6), the router programs a separate tunnel for each.

The router treats the TTM preference for SRv6 policies the same as for MPLS-based SR policies. The router populates the TTMv6 tunnel table with the SRv6 tunnel for the SRv6 policy. IGP provides the most accurate MTU value and programs the SR tunnel MTU value in the TTM. The MTU value includes the lowest MTU among the ECMP next hops or the primary/LFA next hops for each remote locator tunnel and local adjacency SID tunnel.

Next the MTU of the SRv6 policy is derived as follows:

SRv6Policy_ MTU= SR_Tunnel_MTU - numSIDs * 16

SRv6 policy support for Layer 2 and Layer 3 services

The following services can resolve their next hop over an SRv6 policy in TTMv6:

  • VPRNv4
  • VPRNv6
  • EVPN VPLS and VPWS
  • BGP Shortcuts
  • EVPN IFL

As with MPLS SR policies, the next hop resolution is based on the color (if present) and on the comparison of the next hop with the SR policy endpoint. The color, the endpoint, and the head end define a matching policy, but a matching policy can be an SR MPLS policy or an SRv6 policy. Therefore, the service also uses the data plane technology of the policy to select the tunnel type to resolve over. The following restrictions apply:

  • MPLS services cannot resolve over SRv6 tunnels
  • SRv6 services can only resolve over SRv6 tunnels

An SRv6 service cannot fall back to an MPLS tunnel type. Suppose the configuration of the resolution allows an SRv6 policy, but an SRv6 policy with a matching color and endpoint does not exist in TTMv6. In that case, the SRv6 service can fall back to any SRv6 shortest path tunnel in TTMv6 or RTM, if the locator of the tunnel matches the locator of the SRv6 service.

Use the resolution command to control the automatic binding of SRv6 services to SRv6 tunnels. The resolution options are:
  • tunnel-table

    This command option resolves the route directly to a tunnel in TTMv6. The system tries to find an SRv6 policy with a matching color and endpoint for BGP routes received with an SRv6 TLV and containing an SRv6 service SID in the IPv6 tunnel table. If no such SRv6 policy is found, the resolution fails.

  • route-table

    This command option resolves the route to a shortest-path SRv6 tunnel. If none is found, the resolution fails. This is the default behavior.

  • fallback-tunnel-to-route-table

    This command options first tries resolving the route directly to a tunnel in the IPv6 tunnel table. If none is found, fall back to the shortest-path SRv6 resolution. If no such SRv6 policy is found, the resolution fails.

Use the resolution command in the following contexts to configure the resolution options:

  • For BGP shortcuts
    configure router bgp segment-routing-v6 family
  • For VPRN services
    configure service vprn bgp-ipvpn segment-routing-v6
  • For EVPN services
    configure service epipe bgp-evpn segment-routing-v6
    configure service vpls bgp-evpn segment-routing-v6
  • For EVPN-IFL services
    configure service vprn bgp-evpn segment-routing-v6

Seamless BFD and end-to-end protection for SRv6 policies

S-BFD and end-to-end protection, using Linear or ECMP protection modes, are supported for static and BGP SRv6 policies. To configure S-BFD and, optionally, a protection mode, apply a maintenance policy to a static or BGP SRv6 policy. The router attempts to establish an S-BFD session on each segment list of the SRv6 policy.

See Seamless BFD and end-to-end protection for SR policies for information about how to configure S-BFD and protection. S-BFD with a routed return path and S-BFD with a controlled return path are supported. In routed return path, the S-BFD reply packet is sent using a routed path from the reflector. In controlled return path, the S-BFD packet is returned to the initiator via a specified traffic engineered path such as an SRv6 Policy.

In routed return path, S-BFD packets are encapsulated in the SRv6 policy packet using a segment routing header which includes a SID containing the endpoint address of the SRv6 policy as the last SID of the header. The following figure shows S-BFD encapsulation over an SRv6 policy with routed return path.

Figure 19. S-BFD encapsulation over an SRv6 policy with routed return path

In controlled return path, a SID specified at the initiator router is also inserted in the SRH for S-BFD packets. This SID refers to a binding SID on an SRv6 policy configured at the far-end of the S-BFD session (usually the endpoint of the SRv6 policy). Echo mode is also used for the S-BFD session. The following figure shows S-BFD encapsulation over an SRv6 policy with controlled return path.

Figure 20. S-BFD encapsulation over an SRv6 policy with controlled return path

Traffic statistics

SRv6 policies support egress statistics similar to SR MPLS policies. See Traffic statistics, Segment routing with MPLS data plane (SR-MPLS), and Segment routing policies.

SRv6 policies do not support ingress statistics.

Traffic statistics related commands that are defined in the following context affect both SRv6 and segment routing MPLS policies. It is not possible to enable or disable statistics for only one type of policy.

configure router segment-routing sr-policies

Micro-segment support for SRv6 policies

SRv6 policies support micro-segment SRv6. With the seamless introduction of micro-segment SRv6 , the term SRv6 policy now refers to policies using regular (128-bit) SIDs or uSIDs (16-bit), or a combination of both.

To support uSIDs, SRv6 policies are enhanced to include the following additional functionalities:
  • compression
  • configuring a micro-binding SID

All services supported on SRv6 policies before the introduction of micro-segment support continue to be supported. The same applies to applications interacting with SRv6 policies: maintenance policies, sBFD, OAM, PBR, and traffic statistics.

Compression

Compression is supported for both static and BGP-signaled policies. The following compression algorithm is critical in understanding the system behavior, particularly for configuring or signaling SRv6 policies.

The compression algorithm treats each segment list the same, whether it was signaled from a peer (other node or controller) or locally configured. The algorithm parses the ordered list of segments and only compresses SIDs if they meet all the conditions of any of the three following sets of conditions:

  • Set 1: B ⊂ [43, 50] AND (BL, NL, FL, AL) == ({8, 16, 24, 32, 40, 48, 56, 64}, 16, 0, 128-BL-16) AND ipv6[0, BL-1] ≠ 0 AND ipv6[BL, BL+15] ≠ 0 AND ipv6[BL+16, 127] == 0

  • Set 2: (B ⊂ [52, 59] OR B ⊂ [85, 92] OR B ⊂ [93, 94]) AND (BL, NL, FL, AL) == ({8, 16, 24, 32, 40, 48, 56, 64}, 16, 16, 128-BL-32) AND ipv6[0, BL-1] ≠ 0 AND ipv6[BL, BL+15] ≠ 0 AND ipv6[BL+16, BL+31] ≠ 0 AND ipv6[BL+32, 127] == 0

  • Set 3: (B ⊂ [52, 59] OR B ⊂ [85, 92] OR B ⊂ [93, 94]) AND (BL, NL, FL, AL) == ({8, 16, 24, 32, 40, 48, 56, 64}, 0, 16, 128-BL-16) AND ipv6[0, BL-1] ≠ 0 AND ipv6[BL, BL+15] ≠ 0 AND ipv6[BL+16, BL+127] == 0

where

  • B = Behaviour

  • BL = Block Length

  • NL = Node Length

  • FL = Function Length

  • AL = Argument Length

  • ipv6[X, Y] = the bits number X up to number Y of the IPv6 address (the SID), with bit number 0 being the MSB

B, BL, NL, FL, and AL are either obtained from the static policy configuration or the "Endpoint Behavior and Structure" sub-TLV signaled with the SID in BGP. The availability of these values to the compression algorithm is a necessary (but not sufficient) condition for compression to be applied to the associated SID.

The range of values between square brackets corresponds to the IANA-registered values of END with NEXT-CSID, END.X with NEXT-CSID, END.T with NEXT-CSID, and END.B6.Encaps with NEXT-CSID (and their respective flavors). The compression algorithm only considers these behaviors for possible compression.

A detailed list of behaviors supported by the compression algorithm is provided in the following table.

Table 11. Behaviors supported by the compression algorithm
Value Description

43

End with NEXT-CSID

44

End with NEXT-CSID & PSP

45

End with NEXT-CSID & USP

46

End with NEXT-CSID, PSP & USP

47

End with NEXT-CSID & USD

48

End with NEXT-CSID, PSP & USD

49

End with NEXT-CSID, USP & USD

50

End with NEXT-CSID, PSP, USP & USD

52

End.X with NEXT-CSID

53

End.X with NEXT-CSID & PSP

54

End.X with NEXT-CSID & USP

55

End.X with NEXT-CSID, PSP & USP

56

End.X with NEXT-CSID & USD

57

End.X with NEXT-CSID, PSP & USD

58

End.X with NEXT-CSID, USP & USD

59

End.X with NEXT-CSID, PSP, USP & USD

85

End.T with NEXT-CSID

86

End.T with NEXT-CSID & PSP

87

End.T with NEXT-CSID & USP

88

End.T with NEXT-CSID, PSP & USP

89

End.T with NEXT-CSID & USD

90

End.T with NEXT-CSID, PSP & USD

91

End.T with NEXT-CSID, USP & USD

92

End.T with NEXT-CSID, PSP, USP & USD

93

End.B6.Encaps with NEXT-CSID

94

End.B6.Encaps.Red with NEXT-CSID

The values between curly brackets correspond to a specific condition set on SR OS that the Block Length corresponds to one of the block length-configurable values.

To ensure backward compatibility with the regular SIDs based implementation, SIDs are not compressed in the following cases:

  • any SID, including uSIDs, that do not satisfy all of the conditions of any set

  • SIDs without an associated behavior and structure

  • SIDs with a behavior corresponding to a regular SID

The compression algorithm does not conduct additional verifications on the received SIDs. However, if a uN segment followed by a uA segment (encoded without a uN) is received, the algorithm ensures that they are part of the same container, instead of the uN being the last SID of a container and the uA the first SID of the next container.

The system treats the segment list as programmable if, after compression (if any compression was possible), it fits in the datapath capabilities in terms of SRH size. However, other, non-micro-segment specific conditions, exist for effectively programming a segment list.

Policy segment configuration

Knowing how the compression algorithm works is critical for the configuration phase. The user should configure a policy with the understanding of whether the segment list will fit in the SRH. This applies to policies configured on SR OS, and those originated by peers or controllers and destined for an SR OS device.

On SR OS, to express the intent for compression, use the commands in the following context.

configure router segment-routing sr-policies static-policy segment-list segment behavior-and-structure

These commands configure behavior and structure parameters of the SID. Either all command options must be configured or none may be configured; there is no default value for any command option. The values for the behavior are limited to the base (flavorless) behaviors, as flavors play no role in compression; only micro-segment behaviors are supported. However, as indicated above, the compression algorithm can process any flavor value. The three length values correspond to the structure of the SID. The Argument Length is not configurable and always set to 128-BL-NL-FL.

These command options are associated with the IPv6 address whether the policy is local or BGP-signaled. In the latter case, the system adds these values in the Endpoint Behavior and Structure sub-TLV.

The user must configure the IPv6 address and its behavior and structure consistently.

The introduction of micro-segment support comes with the ability to specify up to 24 segments. More precisely, the SR OS allows a user to configure 24 segments if these are micro-segments (for example, an IPv6 address with a behavior and structure) but limits the number of segments to 7 in case of regular segments (for example, an IPv6 address without a behavior and structure). No compressibility verification is made at configuration or validation. The user must ensure that whatever is configured fits into the datapath capabilities.

Note:

The SR OS BGP implementation does not issue a BGP Update Message if it exceeds 4096 bytes. This limitation sets a restriction on the number of segment lists and the number of segments (especially segments with behavior and structure information) per-segment list that the user can configure. Considering segment lists only made of uSIDs (for example, an IPv6 address with behavior and structure information), the message size limit is exceeded by 13 segment lists of 8 uSIDs.

Note:

SR OS systems do not support End.T (any flavor) as a locally-instantiated SID. The CLI for configuring SRv6 policy segments supports that behavior only to allow interoperation with systems which support locally-instantiated End.T SIDs.

Note:

SR OS allows a user to configure segments of segment-lists which are in fact precompressed uSIDs. These segments must be configured without a behavior and structure. Considering a system with the ability to create an SRH with 7 segments and a typical block length of 32 bits and a micro SID length of 16, SR OS allows paths containing as many as 42 uSIDs to form.

Microbinding SID

Micro-segment support includes the ability to configure a microbinding SID, which can either be remote (for a BGP-signaled policy) or local (for a local policy). For a remote policy, the IPv6 address of the SID can be configured, but behavior and structure are not configurable. The SR OS head end ignores this information if received with a binding SID. Use the CLI commands in the following context to configure a local microbinding SID in the context of a micro-segment locator.

configure router segment-routing sr-policies static-policy segment-routing-v6 binding-sid micro-segment-locator
Use the CLI commands in the following context to configure remote binding SIDs.
configure router segment-routing sr-policies static-policy segment-routing-v6 binding-sid ip-address

The function value, if configured, must be within the static range of local SID values. If it is not configured, the system selects a function value within the dynamic range of local SID values. Only a single behavior is supported: End.B6.Encaps.Red with NEXT-CSID. A single binding SID is supported per-SRv6 policy.

If a policy is received from BGP, the SR OS performs specific micro-segment checks. Having determined that a binding SID IPv6 address is covered by a locally configured micro-segment SID block, the system verifies that:

  • the IPv6 address has the following format <block><uN><uB6> (the <block><uB6> format is not supported)

  • the uN value corresponds to a locally configured uN

  • the uB6 value is available under the block

Ultimately, the system programs two FIB entries:

  • <block><uN><uB6>::/(block-length+32)

  • <block><uB6>::/(block-length+16)

Assignment of loopback interface addresses from an SRv6 locator subnet

SR OS supports assigning IPv6 addresses to a system or a loopback interface drawn from a classic or micro-segment SRv6 locator prefix subnet.

To enable this feature, the user must configure an IPv6 address for the system or loopback interface from a classic or a micro-segment locator by setting bits in the least significant octet of the locator, as described in Assigning an IPv6 address from a classic SRv6 locator and Assigning an IPv6 address from a micro-segment SRv6 locator. Only a /128 subnet mask value is allowed for the IPv6 address. The locator must be assigned to algorithm 0. Addresses from a flex-algo locator are not allowed.

A number of checks are performed in CPM to enforce the correct setting of the prefix bits and mask, and to verify the overlap with a locator subnet. See CPM support with classic SRv6 locator and CPM support with micro-segment SRv6 locator for more information.

A system or a loopback interface that is configured to use an IPv6 address from an algorithm 0 locator address space behaves like any other system or loopback interface. It can be injected into IGP or BGP routing protocols and into a VRF. Its address can be used as the source or destination address of control plane protocol and OAM packets.

Note:

This type of system or loopback interface cannot be added into an IES service.

Assigning an IPv6 address from a classic SRv6 locator

When assigning an IPv6 address from a classic SRv6 locator, the feature reserves the least significant octet of the locator subnet for IPv6 address of system and loopback interfaces. This is the main subnet of the locator with bits to the right of the node ID set to zero, except for the interface bits.

The following is an example of the address assignment.

  • locator subnet: fc01:1:0:105::/64
  • locator subnet identifier: fc01:1:0:105:0:0:0:0
  • system/loopback interface addresses allowed are: fc01:1:0:105:0:0:0:00[01-FF]/128

This feature supports a mask value of /128 only. A maximum of 255 addresses can be allocated from the last octet of the locator. A value of zero for that octet represents the identifier of the entire subnet and is not allocatable.

CPM support with classic SRv6 locator

The following checks are performed in the CPM when assigning an IPv6 address from an SRv6 locator subnet. If any of the checks fails, the configuration fails.

  • An address of the system interface or a loopback interface can only be drawn from the subnet of a classic SRv6 locator prefix assigned to algorithm 0.

  • An address of the system interface or a loopback interface can only be drawn from a classic SRv6 locator prefix subnet that satisfies the following condition:

    {Prefix-length + function-length} < 120 bits
  • An address of the system interface or a loopback interface must have all bits to the right of the classic SRv6 locator’s node ID field set to zero, except for the bits of the least significant octet.

  • Creating a new locator or changing the prefix, function-length, or prefix-length of a configured locator are not allowed if an address from the locator prefix subnet is already assigned to the system or a loopback interface. Users must first delete the corresponding addresses from the system interface or loopback interfaces.

7750 SR and 7950 XRS datapath support

The following procedures are supported in datapath:

  • When the user configures an IPv6 address of the system interface or of a loopback interface and that address is drawn from the subnet of a locally configured locator, the CPM adds the /128 prefix into the route table and into the FIB.
    • A packet matching the more specific route of the interface in the FIB is extracted to the CPM for local interface processing.
    • A packet matching the less specific locator route of the locator in the FIB undergoes SRv6 tunnel termination processing, as usual.
  • Ingress PE and transit P routers forward packets destined for this system or loopback interface using either the more specific interface route, if advertised in IGP or BGP by the owner router, or the less specific locator route the interface address is drawn from.

  • A packet destined for the system or a loopback interface which IPv6 address is drawn from a local locator prefix subnet can be received with an SRH in its encapsulation. An example use case os a packet received from a transit router that forwarded the packet over the remote LFA or TI-LFA repair tunnel of the corresponding less specific locator route when the SIDs of the repair tunnel are of USP SRH-mode.

Assigning an IPv6 address from a micro-segment SRv6 locator

When assigning an IPv6 address from a micro-segment SRv6 locator, the feature reserves the least significant octet of the micro-segment locator subnet for IPv6 address of system and loopback interfaces. This is the main subnet of the locator with bits to the right of the uN set to zero, except the interface bits.

The following is an example of the address assignment.

  • locator subnet: fc00:0:105::/48
  • locator subnet identifier: fc00:0:105:0:0:0:0:0
  • system/loopback interface addresses allowed are: fc00:0:105:0:0:0:0:00[01-FF]/128

This feature supports a mask value of /128 only. A maximum of 255 addresses can be allocated from the last octet of the locator. A value of zero for that octet represents the identifier of the entire subnet and is not allocatable.

CPM support with micro-segment SRv6 locator

The following checks are performed in CPM when assigning an IPv6 address from a micro-segment SRv6 locator subnet. If any of the checks fail, the configuration fails.

  • An address of the system interface or a loopback interface can only be drawn from the subnet of a micro-segment SRv6 locator prefix assigned to algorithm 0.
  • An address of the system interface or a loopback interface can only be drawn from a micro-segment SRv6 locator prefix subnet that satisfies the following condition:
    {Block-length + sid-length(uN) + sid-length(uA or uDT)} < 120 bits
  • An address of the system interface or a loopback interface must have all bits to the right of the micro-segment SRv6 locator’s uN field set to zero except bits of the least significant octet.
  • Creating a new locator, changing the uN value of a configured locator, or changing the block length of a configured locator are not allowed if an address from the locator prefix subnet is already assigned to the system or a loopback interface. The user must first delete the corresponding addresses from the system interface or loopback interfaces.

7750 SR and 7950 XRS datapath support

The datapath behavior for a micro-segment SRv6 locator is the same as that of a classic SRv6 locator. See 7750 SR and 7950 XRS datapath support for more information.