About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
If a BGP session is Established but you receive no routes, the usual cause on an eBGP session is that no inbound policy is applied. RFC 8212 changed the default: without an explicit import policy, routes from an eBGP peer are not eligible, and FRR shows (Policy) in the prefix column of show bgp summary. Two other causes look the same from a distance: the address family is not activated for that neighbour, or an inbound filter rejects everything.
Read the summary line first. The text in the prefix column tells you which of the three you have before you change any configuration.
What the error means
A BGP session that reaches Established has exchanged OPEN and KEEPALIVE messages. Established says nothing about whether any routes are accepted. Route acceptance is decided afterwards, per address family, by policy.
RFC 8212 (July 2017) updates the BGP specification with two rules. "Routes contained in an Adj-RIB-In associated with an EBGP peer SHALL NOT be considered eligible in the Decision Process if no explicit Import Policy has been applied." And "Routes SHALL NOT be added to an Adj-RIB-Out associated with an EBGP peer if no explicit Export Policy has been applied." The RFC also allows an implementation to offer a configuration option to deviate from this behaviour, so whether you see it depends on the router software and its settings.
FRR implements it as bgp ebgp-requires-policy. The FRR BGP documentation says: "This command requires incoming and outgoing filters to be applied for eBGP sessions as part of RFC-8212 compliance. Without the incoming filter, no routes will be accepted. Without the outgoing filter, no routes will be announced." It is enabled by default for the traditional configuration and turned off by default for the datacenter configuration.
FRR also documents that "If PfxRcd and/or PfxSnt is shown as (Policy), that means that the EBGP default policy is turned on, but you don't have any filters applied for incoming/outgoing directions."
Address families are separate. FRR states that by default only the IPv4 unicast address family is enabled for a neighbour, and that no bgp default ipv4-unicast means every address family must be enabled explicitly. RFC 4760 requires each side to advertise the capability for an address family before routes of that family are exchanged in both directions. FRR brings peering up with the minimal common capability, so a mismatch can leave the session Established while one address family carries nothing.
The output and commands below are FRR's. Other vendors may show the same state with different words.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| eBGP default reject (RFC 8212): no inbound policy | (Policy) in State/PfxRcd on your side |
show bgp summary; show bgp neighbor says "Inbound updates discarded due to missing policy" |
| Peer has no outbound policy | You show 0 received; the peer shows (Policy) under PfxSnt |
Run show bgp summary on the peer; check its advertised routes |
| Address family not activated on your side or the peer's | Session Established, 0 prefixes for the family you expect | show bgp ipv6 unicast summary (or the family in question); check for activate under the address family |
| Inbound filter denies everything | A number (not (Policy)) that is zero, while the peer says it sent routes |
received-routes lists routes, routes is empty |
| Peer sends nothing for that family | Zero received, no filters on your side | Peer's advertised-routes is empty |
| Maximum-prefix limit exceeded | Not Established: the session is torn down | Different fault; see the note below |
Prove the cause before changing anything
These commands are taken from the FRR documentation and have not been run or lab-tested. They only read state. In FRR they run inside vtysh, for example vtysh -c 'show bgp summary'.
1. Read the summary.
show bgp summary
The FRR documentation shows this layout:
Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd PfxSnt Desc
192.168.0.2 4 65002 8 10 0 0 0 00:03:09 5 (Policy) N/A
fe80:1::2222 4 65002 9 11 0 0 0 00:03:09 (Policy) (Policy) N/A
On the first row the neighbour has sent 5 prefixes and the local router is sending none, shown as (Policy) under PfxSnt. On the second row both columns read (Policy). A bad result is a row that has an Up/Down timer (so it is Established) and (Policy) or 0 under State/PfxRcd. A good result is a number greater than zero for the family you expect.
To restrict the output to one family, the documented form is show bgp [afi] [safi] [all] summary, for example show bgp ipv4 unicast summary or show bgp ipv6 unicast summary.
2. Read the neighbour block for that family.
show bgp neighbor
In the "For address family:" block FRR documents these lines when a filter is missing:
For address family: IPv4 Unicast
Inbound updates discarded due to missing policy
Outbound updates discarded due to missing policy
0 accepted prefixes
If you see those lines, you have the RFC 8212 case. If the lines are absent and 0 accepted prefixes still appears, go on to the filter and address-family checks. FRR's page does not list the capability lines of this command, so use it to read the per-family block rather than to prove negotiation.
3. Separate "not sent" from "sent and rejected".
show bgp ipv4 unicast neighbors 192.168.0.2 routes
show bgp ipv4 unicast neighbors 192.168.0.2 received-routes
FRR documents routes as the routes received by this peer and accepted by inbound policy, and received-routes as all routes received prior to inbound policy. If received-routes is full and routes is empty, an inbound filter is rejecting the routes. FRR documents neighbor PEER soft-reconfiguration inbound as the setting that stores all received updates, including those rejected by inbound policy, so a peer without it may not be able to show you what was filtered.
4. Check the other side. Run show bgp summary on the peer. If its PfxSnt shows (Policy) for your neighbour, the peer is the one with no outbound policy. If PfxSnt is a number, the peer is sending, and the fault is on your side.
5. Read the configuration for the neighbour.
show run
Look for no bgp ebgp-requires-policy, no bgp default ipv4-unicast, neighbor ... activate under the address family, and any neighbor ... route-map, prefix-list or filter-list ending in in.
Fix
Cause 1: no inbound or outbound policy (RFC 8212)
The correct fix is to apply a policy that states what you want to accept. FRR documents these commands: neighbor PEER route-map NAME [in|out], neighbor PEER prefix-list NAME [in|out] and neighbor PEER filter-list NAME [in|out]. A route-map that matches a prefix list uses the syntax FRR gives in its default-originate example (route-map NAME permit 10 with match ip address prefix-list NAME).
Example (illustrative): router fw-a (AS 65001) peers with 192.168.0.2 (AS 65002) and should accept only 198.51.100.0/24.
configure terminal
ip prefix-list PEER-IN seq 5 permit 198.51.100.0/24
route-map PEER-IN permit 10
match ip address prefix-list PEER-IN
exit
router bgp 65001
address-family ipv4 unicast
neighbor 192.168.0.2 route-map PEER-IN in
exit-address-family
Write the matching outbound policy in the same way if you also need to announce routes. Then ask the router to apply the new inbound policy. FRR documents clear bgp ipv4|ipv6 PEER soft|in|out, and notes that with route-refresh support on both peers a soft reconfiguration works without extra settings.
Risk: a prefix list with the wrong prefix accepts nothing, or too much. Accepting too much from a transit provider is how route leaks happen, which is the harm RFC 8212 was written to prevent. Check the list against what the peer should send before you apply it. Back out with no neighbor 192.168.0.2 route-map PEER-IN in, which returns the session to the default-reject state.
Alternative: switch the default off. FRR documents no bgp ebgp-requires-policy. Disruptive step: FRR states that when you enable or disable this option you MUST clear the session, so the session resets and traffic on it stops until it comes back. It also removes the protection RFC 8212 provides, so every eBGP peer then accepts and announces everything unless you filter it. Back out with bgp ebgp-requires-policy and another session clear.
Cause 2: address family not activated
Under the address family, activate the neighbour. FRR documents neighbor A.B.C.D activate, with the note that by default only IPv4 unicast is enabled, and shows an IPv6 neighbour activated under address-family ipv6 unicast. Do the same on the peer. Ask for the matching activate on the other router if you do not run it.
Risk: activating a family adds that family's routes to the session. With ebgp-requires-policy still on, a newly activated family also needs its own inbound and outbound policy before anything is accepted. Back out with no neighbor ... activate under that family.
Cause 3: an inbound filter rejects everything
Compare received-routes with the prefix list, AS-path list or route-map it hits. Correct the entry, then run clear bgp ipv4 PEER soft in (or the IPv6 form). A soft reset asks for or reprocesses the received routes without dropping the session.
Cause 4: the peer is sending nothing
Ask the peer's operator to check its outbound policy and its advertised-routes for your address. Nothing on your router can create routes the peer does not send.
Not this: maximum-prefix
A prefix limit does not leave the session Established with zero routes. FRR documents that when neighbor PEER maximum-prefix NUMBER is exceeded, "the BGP session will be destroyed". On Cisco IOS XE the maximum-prefix page shows the log %BGP-4-MAXPFX: Number of prefixes received from 10.0.0.1 (afi 0) reaches 9, max 10 and a session state of Idle (PfxCt). If you see that state, you have a different fault. With the warning-only option the session stays established.
Confirm the fix
show bgp summaryshows a number under State/PfxRcd for the family, and(Policy)is gone from both columns.show bgp neighborno longer lists "Inbound updates discarded due to missing policy".show bgp ipv4 unicast neighbors 192.168.0.2 routeslists the prefixes you expect and only those.- The routes appear in the routing table, and a ping or traceroute to a prefix learned from the peer takes the intended path.
What to watch afterwards
- Keep the policy in version control. A router rebuilt from an old configuration can return to default reject if the policy lines are missing.
- Watch PfxRcd against the number you expect from each peer, and alert when it falls to zero while the session is still Established.
- Add a prefix limit deliberately if you want protection from a leak. It will tear the session down, so choose the number with that in mind. FRR itself says a prefix list is preferable to a limit because "tearing down the BGP session when a limit is exceeded is far more destructive than merely rejecting undesired prefixes".
Read next
Related articles you may want next.
- VMwareNSX Tier-0 BGP Neighbour Down: Read the Session State Before Resetting AnythingNSX 4.2 Tier-0 BGP neighbour down? Record the session state, BFD state and Edge log lines per neighbour, then match them to Broadcom's documented causes.Read article
- Backup & RecoveryBGP policy recovery: restore matches, actions and dependenciesHow RFC 4271 policy and FRRouting prefix lists and route maps decide which routes survive a restore, and which mistakes withdraw the only path.Read article
- CloudAWS Site-to-Site VPN: BGP against static, tunnel failover and asymmetric returnHow AWS picks a tunnel, what BGP gives that static routing does not, what happens during failover, and how to handle asymmetric return traffic.Read article
Risks and limits
- The output samples and the
(Policy)behaviour come from the FRR documentation. Other routers (Cisco, Juniper, Arista and others) may use other column text for the same state, or do not apply default reject unless configured. RFC 8212 allows a speaker to offer a way to deviate from it. - The FRR documentation page shows no date, so confirm behaviour against the FRR version you run. The default depends on the configuration mode (traditional or datacenter).
- Soft reconfiguration stores a second copy of received updates and uses extra memory, as the FRR documentation notes.
- Clearing a session ends traffic carried over it. Do it in a maintenance window for a production peer.
Sources
- BGP — FRR documentation (latest) — FRRouting project, date not shown (copyright line 2017), accessed 17 November 2026
- RFC 8212: Default External BGP (EBGP) Route Propagation Behavior without Policies — IETF, July 2017, accessed 17 November 2026
- RFC 4760: Multiprotocol Extensions for BGP-4 — IETF, January 2007, accessed 17 November 2026
- How to Configure BGP Maximum-Prefix (document 25160) — Cisco, updated 20 August 2026, accessed 17 November 2026
This article was originally published on .