| Internet-Draft | MoFRR based on TILFA | April 2025 |
| Liu, et al. | Expires 3 October 2025 | [Page] |
This document specifies the use of Topology Independent Loop-Free Alternate (TI-LFA) mechanisms with Multicast Only Fast ReRoute (MoFRR) for Protocol Independent Multicast (PIM). The TI-LFA provides fast reroute protection for unicast traffic in IP networks by precomputing backup paths that avoid potential failures. By integrating TI-LFA with MoFRR, this document extends the benefits of fast reroute mechanisms to multicast traffic, enabling enhanced resilience and minimized packet loss in multicast networks. The document outlines an optional approach to implement TI-LFA in conjunction with MoFRR for PIM, ensuring that multicast traffic is rapidly rerouted in the event of a failure.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 3 October 2025.¶
Copyright (c) 2025 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Multicast-only Fast Reroute (MoFRR), as defined in [RFC7431], offers a mechanism to reduce multicast packet loss in the event of node or link failures by introducing simple enhancements to multicast routing protocols, such as Protocol Independent Multicast (PIM) [RFC7761]. However, the [RFC7431] MoFRR mechanism, which selects the secondary multicast next-hop based solely on the loop-free alternate fast reroute defined in [RFC7431], has limitations in certain multicast deployment scenarios (see Section 2).¶
This document introduces a new mechanism for MoFRR using Topology Independent Loop-Free Alternate (TI-LFA) [I-D.ietf-rtgwg-segment-routing-ti-lfa] fast reroute. Unlike conventional methods, TI-LFA is independent of network topology, enabling broader coverage across diverse network environments. The applicability of this mechanism extends to PIM networks, such as scenarios where PIM operates over native IP.¶
The TI-LFA mechanism is designed for standard link-state Interior Gateway Protocol (IGP) shortest path and Segment Routing (SR) scenarios. For each destination advertised by the IGP in a network, TI-LFA pre-installs a backup forwarding entry for the protected destination, ready to be activated upon the detection of a link failure used to reach that destination. This document leverages the backup path computed by TI-LFA through the IGP as a secondary Upstream Multicast Hop (UMH) for PIM. By sending PIM secondary Join messages hop by hop on the TI-LFA backup path, a fast reroute backup path can be created for PIM multicast.¶
The techniques described in this document are limited to protecting links and nodes within a link-state IGP area. Protecting domain exit routers and/or links attached to other routing domains is beyond the scope of this document. All the Segment Identifiers (SIDs) required are contained within the Link State Database (LSDB) of the IGP.¶
The approach does not alter the existing management and operation of LFA, RLFA, and TI-LFA [RFC7916] [RFC8102] [I-D.ietf-rtgwg-segment-routing-ti-lfa]. Also, this approach is not affected by micro-loops [RFC5715] occurring during the network re-converges.¶
Note that this document introduces an optional approach for backup join paths, designed to enhance the protection scope of existing multicast systems. It is fully compatible with current protocol implementations and does not necessitate any changes to the protocols or forwarding functions on intermediate nodes. If there is a choice between vector and non-vector Join messages on the intermediate nodes, the non-vector option should be prioritized, which implies that protection paths will remain inactive. This document does not modify the handling of conflicts in PIM Join messages. For guidance on conflicts in Join attributes, please refer to Section 3.3.3 of [RFC5384].¶
This document utilizes the terminology as defined in [RFC7431] and incorporates the concepts established in [RFC7490]. The definitions of individual terms are not reiterated within this document.¶
Section 3 of [RFC7431] specifies that a secondary UMH in PIM for MoFRR is a Loop-Free Alternate (LFA). However, the conventional LFA mechanism requires that at least one neighbor's next-hop to the destination node is a loop-free next-hop. Due to existing limitations of the LFA mechanism in network deployments, such as topology dependency and incomplete destination coverage, the LFA mechanism can only be deployed in certain network topology environments. In specific network topologies, the secondary UMH cannot be computed in PIM for MoFRR, preventing PIM from establishing a standby multicast tree and thus from implementing MoFRR protection. Consequently, the [RFC7431] MoFRR functionality in PIM is applicable only in network topologies where LFA is feasible.¶
The limitations of the [RFC7431] MoFRR applicability can be illustrated using the example network depicted in Figure 1. In this example, the metric of the R1-R4 link is 20, the metric of the R5-R6 link is 100, and the metrics of the other links are 10. All link metrics are bidirectional.¶
For multicast source S1 and receiver R, the primary path of the PIM Join packet is R3->R2->R1, and the secondary path is R3->R4->R1, which corresponds to the LFA path of unicast routing. In this scenario, the [RFC7431] MoFRR operates effectively.¶
For multicast source S2 and receiver R, the primary path of the PIM Join packet is R3->R2. However, an LFA does not exist. If R3 sends the packet to R4, R4 will forward it back to R3 because the IGP shortest path from R4 to R1 is R4->R3->R2. In this case, the [RFC7431] MoFRR cannot calculate a secondary UMH. Similarly, for multicast source S3 and receiver R, the [RFC7431] MoFRR mechanism is ineffective.¶
10 20
[S1]----(R1)-------------(R4)
| |
| |
|10 |10
10 | |
[S2]----(R2)-------------(R3)----[R]
| 10 | 10
| |
|10 |10
10 | |
[S3]----(R5)-----(R6)----(R7)
100 10
The Remote Loop-Free Alternate (RLFA), as defined in [RFC7490], extends the LFA mechanism, to accommodate a broader range of network deployments by utilizing a tunnel as an alternate path. The RLFA requires that there exists at least one node, denoted as node N, in the network where the fault node is neither on the path from the source node to node N nor on the path from node N to the destination node. The source node and the destination node are the tunnel endpoints of the RLFA repair path.¶
[RFC5496] introduces the Reverse Path Forwarding (RPF) Vector attribute, which can be included in PIM Join messages to ensure that the path is selected based on the unicast reachability of the RPF Vector. By combining the RLFA with the RPF Vector, a secondary multicast tree for MoFRR can be established, thereby supporting a wider range of network topologies compared to the [RFC7431] MoFRR implementation with LFA.¶
For instance, in the network illustrated in Figure 1, the secondary path for the PIM Join packet towards multicast source S2 cannot be computed by the [RFC7431] MoFRR, as previously described. Utilizing the RLFA, R3 sends the packet to R4, including an RPF Vector containing the IP address of R1, which serves as the PQ node of R3 with respect to the protected R2-R3 link. Subsequently, R4 forwards the packet to R1 via the R1-R4 link, in accordance with the unicast route associated with the RPF Vector. R1 then continues to forward the packet to R2, thereby establishing the secondary path, R3->R4->R1->R2, using MoFRR with RLFA.¶
While RLFA provides enhanced capabilities over LFA, it remains dependent on the network topology. In the network depicted in Figure 1, the primary path of the PIM Join packet towards multicast source S3 is R3->R2->R5. However, an RLFA path does not exist because the PQ node of R3 with respect to the protected link R2-R3 is absent. If R3 sends the packet to R7 with an RPF Vector containing the IP address of R5, R7 will forward it back to R3, as the IGP shortest path from R7 to R5 is R7->R3->R2->R5. Similarly, if R3 sends the packet to R7 with an RPF Vector containing the IP address of R6, R7 will forward it to R6, but R6 will then forward it back to R7, as the IGP shortest path from R6 to R5 is R6->R7->R3->R2->R5. In this scenario, MoFRR with RLFA is unable to compute a secondary UMH.¶
RLFA offers improvements over LFA but still has inherent limitations. [I-D.ietf-rtgwg-segment-routing-ti-lfa] defines a unicast FRR solution based on the TI-LFA mechanism. The TI-LFA mechanism allows the expression of a backup path using an explicit path and imposes no constraints on the network topology, thus providing a more robust FRR mechanism. Unicast traffic can be forwarded according to an explicit path list as an alternate path to protect unicast traffic, achieving full coverage across various network environments.¶
The alternate path provided by the TI-LFA mechanism is represented as a Segment List, which includes the NodeSID of the P-space node and the Adjacency SIDs of the links between the P-space and Q-space nodes. PIM can look up the corresponding node IP address in the unicast route base according to the NodeSID and the IP addresses of the endpoints of the corresponding link in the unicast route base according to the Adjacency SIDs. However, multicast protocol packets cannot be directly forwarded along the path of the Segment List.¶
To establish a standby multicast tree, PIM Join messages need to be transmitted hop-by-hop. However, not all nodes and links on the unicast alternate path are included in the Segment List. If PIM protocol packets are transmitted solely in unicast mode, they effectively traverse the unicast tunnel like unicast traffic and do not pass through the intermediate nodes of the tunnel. Consequently, the intermediate nodes on the alternate path cannot forward multicast traffic because they lack PIM state entries. PIM must create entries on each device hop-by-hop, generating an incoming interface and an outgoing interface list, to form a complete end-to- end multicast tree for forwarding multicast traffic. Therefore, simply sending PIM Join packets using the Segment List, as done with unicast traffic, is insufficient to establish a standby multicast tree.¶
A secondary UMH serves as a candidate next-hop that can be used to reach the root of a multicast tree. In this document, the secondary UMH is derived from unicast routing, utilizing the Segment List computed by TI-LFA.¶
The path information from the Segment List is incorporated into the PIM packets to guide hop-by-hop RPF selection. The IP address corresponding to the Node SID can be used as the segmented root node, while the IP addresses of the interfaces at both endpoints of the link associated with the Adjacency SID can be used as the local upstream interface and upstream neighbor.¶
[RFC5496] defines the PIM RPF Vector attribute, which can carry the node's IP address corresponding to the Node SID. Additionally, [RFC7891] defines the explicit RPF Vector, which can carry the peer's IP address corresponding to the Adjacency SID.¶
This document leverages the above RPF Vector standards, obviating the need for PIM protocol extensions. This approach allows the establishment of a standby multicast tree based on the Segment List calculated by TI-LFA, thereby providing comprehensive MoFRR protection for multicast services across diverse network environments.¶
Consider a Segment List calculated by TI-LFA as (NodeSID(A), AdjSID(A-B)). Node A belongs to the P-space, and node B belongs to the Q-space. The IP address corresponding to NodeSID(A) can be retrieved from the local link-state database of the IGP and assumed to be IP-a. Similarly, the IP addresses of the two endpoints of the link corresponding to AdjSID(A-B) can also be retrieved from the local link-state database and assumed to be IP-La and IP-Lb.¶
Within the PIM process, IP-a is treated as the standard RPF Vector Attribute and added to the PIM Join packet. IP-La is considered the local address of the incoming interface, and IP-Lb is regarded as the address of the upstream neighbor. Consequently, IP-Lb can be included in the PIM Join packet as the explicit RPF Vector Attribute.¶
The PIM protocol initially selects the RPF incoming interface and upstream neighbor towards IP-a and proceeds hop-by-hop to establish the PIM standby multicast tree until reaching node A. At node A, IP- Lb is treated as the PIM upstream neighbor. Node A identifies the incoming interface in the unicast routing table based on IP-Lb, and IP-Lb is used as the RPF upstream address for the PIM Join packet directed towards node B.¶
Upon receiving the PIM Join packet at node B, the PIM protocol, finding no additional RPF Vector Attributes, selects the RPF incoming interface and upstream neighbor towards the multicast source directly. The protocol, then, continues hop-by-hop to establish the PIM standby multicast tree, extending to the router directly connected to the source.¶
This section provides an illustration of MoFRR based on TI-LFA. The example topology is depicted in Figure 2. The metric for the R3-R4 link is 100, while the metrics for the other links are 10. All link metrics are bidirectional.¶
<-----Primary Path--- (S,G) Join
[S]---(R1)---(R2)******(R6)-------[R]
| |
<--- | | |
| | | |
| | (R5) |
| | | |
| | | |
| | | |
| (R3)------(R4) |
| |
--Secondary Path--
The IP addresses and SIDs involved in the MoFRR calculation are configured as follows:¶
Node IP Address Node SID R4 IP4-R4 Label-R4 Link IP Address Adjacency SID R3->R4 IP4-R3-R4 Label-R3-R4 R4->R3 IP4-R4-R3 Label-R4-R3 IPv6 Data Plane (SRv6 [RFC8986]) Node IP Address Node SID (End) R4 IP6-R4 SID-R4 Link IP Address Adjacency SID (End.X) R3->R4 IP6-R3-R4 SID-R3-R4 R4->R3 IP6-R4-R3 SID-R4-R3¶
The primary path of the PIM Join packet is R6->R2->R1, and the secondary path is R6->R5->R4->R3->R2->R1. However, the traditional LFA does not function properly for the secondary path because the shortest path to R2 from R5 (or from R4) still traverses the R6-R2 link. In this scenario, R6 must calculate the secondary UMH using the proposed MoFRR method based on TI-LFA.¶
According to the TI-LFA algorithm, the P-space and Q-space are illustrated in Figure 3. The TI-LFA repair path consists of the Node SID of R4 and the Adjacency SID of R4->R3. On the MPLS-SR data plane, the repair segment list is (Label-R4, Label-R4-R3). On the SRv6 data plane, the repair segment list is (SID-R4, SID-R4-R3).¶
........
. .
[S]---(R1)---(R2)******(R6)---[R]
. | . |
. | . +++|++++
. | . + | +
. | . + (R5) +
. | . + | +
. | . + | +
. | . + | +
. (R3)------(R4) +
. . + +
........ ++++++++
Q-Space P-Space
In the PIM process, the IP addresses associated with the repair segment list are retrieved from the IGP link-state database.¶
On the IPv4 data plane, the Node SID Label-R4 corresponds to IP4-R4, which will be carried in the RPF Vector Attribute. The Adjacency SID Label-R4-R3 corresponds to the local address IP4-R4-R3 and the remote peer address IP4-R3-R4, with IP4-R3-R4 carried in the Explicit RPF Vector Attribute.¶
On the IPv6 data plane, the End SID SID-R4 corresponds to IP6-R4, which will be carried in the RPF Vector Attribute. The End.X SID SID-R4-R3 corresponds to the local address IP6-R4-R3 and the remote peer address IP6-R3-R4, with IP6-R3-R4 carried in the Explicit RPF Vector Attribute.¶
+---------+
|Type = 0 |
|IP4-R4 |
+---------+ +---------+
|Type = 4 | |Type = 4 |
|IP4-R3-R4| |IP4-R3-R4|
+---------+ +---------+ No RPF Vector
R6----->R5---->R4------------>R3----->R2---->R1
On the IPv4 data plane, the forwarding of the PIM Join packet along the secondary path is shown in Figure 4.¶
R6 inserts two RPF Vector Attributes in the PIM Join packet: IP4-R4 of Type 0 (RPF Vector Attribute) and IP4-R3-R4 of Type 4 (Explicit RPF Vector Attribute). R6 then forwards the packet along the secondary path.¶
When R5 receives the packet, it performs a unicast route lookup of the first RPF Vector IP4-R4 and sends the packet to R4.¶
R4, being the owner of IP4-R4, removes the first RPF Vector from the packet and forwards it according to the next RPF Vector. R4 sends the packet to R3 based on the next RPF Vector IP4-R3-R4, as its PIM neighbor R3 corresponds to IP4-R3-R4.¶
When R3 receives the packet, as the owner of IP4-R3-R4, it removes the RPF Vector. The packet, now devoid of RPF Vectors, is forwarded to the source through R3->R2->R1 based on unicast routes.¶
After the PIM Join packet reaches R1, a secondary multicast tree, R1->R2->R3->R4->R5->R6, is established hop-by-hop for (S, G) using MoFRR based on TI-LFA.¶
+---------+
|Type = 0 |
|IP6-R4 |
+---------+ +---------+
|Type = 4 | |Type = 4 |
|IP6-R3-R4| |IP6-R3-R4|
+---------+ +---------+ No RPF Vector
R6----->R5---->R4------------>R3----->R2---->R1
On the IPv6 data plane, the forwarding of the PIM Join packet along the secondary path is illustrated in Figure 5. The procedure is analogous to that of the IPv4 data plane.¶
This section illustrates the resolution of PIM Join conflicts. When intermediate nodes must choose between PIM Join messages with an RPF Vector attribute and those without, priority should be given to the Join messages without the RPF Vector attribute. Figure 6 shows the example topology. In this setup, the metric for the R3-R4 link is 100, while the metrics for all other links are 10. Additionally, all link metrics are bidirectional.¶
<---(1)-Primary Path--- (S,G) Join
[S]---(R1)---(R2)******(R6)-------[Receiver1]
| | ^
<--- | | | |
| | | | |
| | (R5) | |
| | | (3)-(S,G) Join |
| | | | |
| | | | |
| (R3)------(R4)--[Receiver2] |
| |
+-(2)-Secondary Path--(S,G)----+
¶
Join Conflict Scenario¶
Figure 6 illustrates that both R6 and R4 have receivers, with R6 utilizing MoFRR based on TI-LFA. At R6 the primary PIM Join packet path is R6->R2->R1, while the secondary path is R6->R5->R4->R3->R2->R1. At R4 the PIM Join packet path is R4->R5->R6->R2->R1.¶
On R5, the PIM forwarding entries are:¶
[R5]:
(S,G) Entry 1 from Receiver1:
Incoming interface: R4-R5
Outgoing interface: R5-R6 (Join with RPF Vector)
(S,G) Entry 2 from Receiver2:
Incoming interface: R6-R5
Outgoing interface: R5-R4 (Join without RPF Vector)
¶
To prevent conflicts, the PIM Join without the RPF Vector should be prioritized. Thus, R5 retains only Entry 2 with R6-R5 as the incoming interface and R5-R4 as the outgoing interface. After selection, the final forwarding entry on R5 is as follows:¶
[R5]:
(S,G) active forwarding Entry:
Incoming interface: R6-R5
Outgoing interface: R5-R4
¶
When using the proposed optional approach for backup join paths in this document, in the event of PIM Join conflicts, Join messages without the RPF Vector attribute are prioritized to ensure that the original PIM forwarding functionality remains unaffected.¶
The management of the proposed approach is consistent with [RFC7916]. But, in the operation of this approach, the node on the backup Join paths must have independent configuration strategy for installing RPF Vector Attributes in the PIM Join packet and controlling the sending of this PIM Join messages.¶
All nodes on the backup Join paths must be able to parse the PIM Join message with RPF Vector attribute. If the nodes do not understand the RPF Vector attribute in the PIM Join packet, then it must discard the RPF Vector attribute because failing to remove the RPF Vectors could cause upstream nodes to send the Join back toward these nodes causing loops.¶
If an administrator is manually specifying the path that the Join messages need to be sent on, it is recommended that the administrator computes the path to include nodes that support the Explicit RPF Vector and check that the state is created correctly on each node along the path. Tools like mtrace [RFC8487] can be used for debugging and to ensure that the Join state is setup correctly.¶
This document has no IANA actions.¶
This document does not introduce additional security concerns. It does not change the security properties of PIM. For general PIM-SM protocol security considerations, see [RFC7761]. The security considerations of LFA, R-LFA, and MoFRR described in [RFC5286], [RFC7490], and [RFC7431] should apply to this document.¶
When deploying TI-LFA, packets may be sent over nodes and links they were not transported through before, potentially raising the following security issues:¶
Spoofing and False Route Advertisements¶
Dependencies of LFA/R-LFA/TI-LFA on Routing Information¶
LFAs depend on accurate routing information to determine alternate paths. If an attacker can inject false routing information (e.g., by spoofing link-state advertisements), it could cause the network to select suboptimal or malicious paths for LFAs.¶
R-LFA and TI-LFA also depend on accurate routing information, particularly for determining the tunneling paths or explicit paths. False route advertisements could mislead the network into using insecure or compromised paths.¶
On-path Attacks¶
Use of Alternate Paths¶
By rerouting traffic through alternate paths, especially those that traverse multiple hops (as in R-LFA and TI-LFA), the risk of On-path attacks increases if any of the intermediate routers on the alternate path is compromised.¶
TI-LFA, which uses explicit paths, might expose traffic to routers that were not originally part of the primary path, potentially allowing for interception or alteration of the traffic.¶
Confidentiality and Integrity¶
Traffic Encapsulation¶
R-LFA and TI-LFA involve encapsulating traffic, which may expose it to vulnerabilities if the encapsulation mechanisms are not secure. For instance, if IPsec or another secure encapsulation method is not used, an attacker might be able to intercept or alter the traffic in transit.¶
Protection of Explicit Paths¶
TI-LFA relies on explicit paths that are typically defined using segment routing. If these paths are not properly protected, an attacker could manipulate the segment list to reroute traffic through malicious nodes.¶
Increased Attack Surface¶
To address security issues #1 and #2 mentioned above, control plane protocols need to provide security protection. To mitigate the risks associated with false route advertisements and On-path attacks, it is recommended to use secure routing protocols (e.g., OSPFv3 with IPsec, ISIS HMAC-SHA256, or PIM with IPsec) that provide authentication and integrity protection for routing updates.¶
To address security issues #3 and #4 mentioned above, these mechanisms need to run within a trusted network. The security of LFA, R-LFA, and TI-LFA mechanisms heavily relies on the trustworthiness of the underlying routing infrastructure. As the solution described in the document is based on Segment Routing technology, readers should be aware of the security considerations related to this technology ([RFC8402]) and its data plane instantiations ([RFC8660], [RFC8754], and [RFC8986]).¶
Mengxiao Chen New H3C Technologies China Email: chen.mengxiao@h3c.com¶