bgpd: Fix missing SRv6 advertisement with distribute bgp-fabric-link-state

[cscarpitta]

When distribute bgp-fabric-link-state is enabled via CLI, BGP-LS advertises Node, Link, and Prefix NLRIs, but SRv6 information is missing.

SRv6 capability and SRv6 attributes (locator, End, End.X) should be advertised together with the BGP-fabric topology, but currently are not.

This PR fixes SRv6 advertisement for BGP-LS BGP-fabric export and adds a dedicated topotest that validates SRv6 BGP-LS advertisements.

[greptile-apps]

Greptile Summary

This PR fixes missing SRv6 information in BGP-LS advertisements when distribute bgp-fabric-link-state is enabled. It adds SRv6 capability (locators, End, and End.X SIDs) to BGP-fabric Link State exports and includes a new topotest that validates the full SRv6 BGP-LS NLRI set.

• Adds bgp_ls_originate_srv6_locator_prefix/bgp_ls_withdraw_srv6_locator_prefix to track SRv6 locators and gate the SRv6 Capabilities node attribute on srv6_locator_nlri_count; resets that counter in bgp_ls_withdraw_all, addressing two prior findings.
• Extends the bgp_redistribute_add signature with seg6local_action/seg6local_ctx, propagates seg6local data from zebra redistribute messages, and introduces a static_endx_sids DLIST to associate End.X SIDs with BGP link NLRIs; adds corresponding cleanup in bgp_ls_cleanup.
• Extends srv6_manager_get_locator in lib/zclient and zebra to accept a NULL locator name, requesting all locators at once, used when BGP-LS distribution is enabled.

Confidence Score: 5/5

Safe to merge; the core SRv6 BGP-LS origination logic is sound and two previously reported counter/flag issues are addressed.

The locator-count tracking, DLIST memory management, seg6local data plumbing through zebra to bgp_route to bgp_ls, and the NULL-locator get-all path are all implemented correctly. The two inline comments flag unnecessary node NLRI re-originations when End.X SIDs are manipulated without any active locator, but neither produces incorrect data — only extra BGP updates — and both have straightforward fixes.

bgpd/bgp_ls.c — the upsert/delete End.X SID paths around the bgp_ls_originate_bgp_node call guards.

Important Files Changed

Filename	Overview
bgpd/bgp_ls.c	Core SRv6 BGP-LS logic added: locator prefix origination/withdrawal with srv6_locator_nlri_count tracking, End.X SID DLIST management with proper cleanup, and node capability re-origination. Capability check only considers locator count, causing spurious node re-originations when End.X SIDs exist without locators.
bgpd/bgp_ls.h	Adds PREDECL_DLIST for End.X SID list, new fields to bgp_ls struct, and public prototypes for new SRv6 origination/withdrawal functions.
bgpd/bgp_route.c	Adds seg6local_action/seg6local_ctx parameters to bgp_redistribute_add and hooks BGP-LS route add/delete handlers; signature change is correctly propagated to all callers.
bgpd/bgp_zebra.c	Extracts seg6local data from zebra redistribute messages, hooks locator add/delete events into BGP-LS, and requests all locators when BGP-LS distribution is enabled at instance register time.
bgpd/bgp_vty.c	Subscribes to AFI_IP6/ZEBRA_ROUTE_ALL redistribution and requests all SRv6 locators when distribution is enabled; unsubscribes on disable.
lib/zclient.c	Extends srv6_manager_get_locator to accept NULL name (request all locators), encoding length=0 and skipping the stream_put when no name is given.
zebra/zapi_msg.c	Propagates seg6local action/ctx in zsend_redistribute_route and handles NULL locator name in zread_srv6_manager_get_locator with a length check to guard against buffer overrun.
zebra/zebra_srv6.c	Handles NULL locator_name in srv6_manager_get_srv6_locator_internal by iterating all locators and sending each one to the client.
tests/topotests/bgp_link_state_bgp_fabric_srv6/test_bgp_link_state_bgp_fabric_srv6.py	New topotest validating SRv6 BGP-LS NLRIs on the route reflector against a reference JSON; uses standard FRR convergence + JSON comparison pattern.

Sequence Diagram

sequenceDiagram
    participant Z as Zebra
    participant BZ as bgp_zebra.c
    participant BR as bgp_route.c
    participant BLS as bgp_ls.c

    Note over Z,BLS: distribute bgp-fabric-link-state enabled

    Z->>BZ: ZEBRA_REDISTRIBUTE_ROUTE_ADD (static, seg6local)
    BZ->>BR: bgp_redistribute_add(..., seg6local_action, seg6local_ctx)
    BR->>BLS: bgp_ls_handle_route_add(...)
    BLS->>BLS: "bgp_ls_handle_srv6_localsid_update (is_add=true)"
    alt END action
        BLS->>BLS: bgp_ls_originate_static_srv6_sid_from_seg6local()
        BLS->>BLS: originate SRv6 SID NLRI
    else END_X action
        BLS->>BLS: bgp_ls_upsert_bgp_link_srv6_endx_sid()
        BLS->>BLS: store in static_endx_sids list
        BLS->>BLS: bgp_ls_refresh_bgp_link_endx_attrs(peer)
        BLS->>BLS: originate Link NLRI with End.X SID attr
    end

    Z->>BZ: ZEBRA_SRV6_LOCATOR_ADD
    BZ->>BLS: bgp_ls_originate_srv6_locator_prefix(default_bgp, loc)
    BLS->>BLS: originate IPv6 Prefix NLRI with SRv6 Locator attr
    BLS->>BLS: srv6_locator_nlri_count++
    BLS->>BLS: bgp_ls_originate_bgp_node() with SRv6 Capabilities attr

    Z->>BZ: ZEBRA_SRV6_LOCATOR_DEL
    BZ->>BLS: bgp_ls_withdraw_srv6_locator_prefix(default_bgp, loc)
    BLS->>BLS: bgp_ls_withdraw() prefix NLRI
    BLS->>BLS: srv6_locator_nlri_count--
    alt count reaches 0
        BLS->>BLS: bgp_ls_originate_bgp_node() without SRv6 Capabilities
    end

Loading

Prompt To Fix All With AI

Fix the following 2 code review issues. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 2
bgpd/bgp_ls.c:1685-1718
`bgp_ls_has_srv6_capability` only counts SRv6 locator NLRIs, so `had_srv6_cap` is always `false` when there are no locators — regardless of how many End.X SIDs are already in `static_endx_sids`. As a result, every call to `bgp_ls_upsert_bgp_link_srv6_endx_sid` when no locator is present unconditionally re-originates the node NLRI without any SRv6 capability attribute set (a no-op advertisement). The guard should fire only when SRv6 capability is actually gained — i.e., only when the End.X SID list was previously empty as well.

```suggestion
	had_srv6_cap = bgp_ls_has_srv6_capability(bgp);
	bool endx_was_empty = bgp_ls_endx_sid_list_empty(&bgp->ls_info->static_endx_sids);

	entry = bgp_ls_static_endx_sid_lookup(bgp, sid, prefixlen);
	if (!entry) {
		entry = XCALLOC(MTYPE_BGP_LS, sizeof(*entry));
		entry->sid = *sid;
		entry->prefixlen = prefixlen;
		bgp_ls_endx_sid_list_add_tail(&bgp->ls_info->static_endx_sids, entry);
	} else {
		old_peer = entry->peer;
	}

	entry->behavior = behavior;
	entry->lb_len = ctx->block_len;
	entry->ln_len = ctx->node_len;
	entry->fn_len = ctx->function_len;
	entry->arg_len = ctx->argument_len;
	entry->nh6 = ctx->nh6;
	entry->ifindex = ctx->ifindex;
	entry->peer = NULL;

	peer = bgp_ls_find_peer_by_nh6(bgp, &ctx->nh6, ctx->ifindex);
	if (!peer) {
		if (BGP_DEBUG(linkstate, LINKSTATE))
			zlog_debug("BGP-LS: %pI6/%u stored pending End.X SID for nh6=%pI6 ifindex=%u until link exists",
				   sid, prefixlen, &ctx->nh6, ctx->ifindex);
	} else {
		entry->peer = peer;
		if (BGP_DEBUG(linkstate, LINKSTATE))
			zlog_debug("BGP-LS: %pI6/%u matched peer %s", sid, prefixlen, peer->host);
	}

	if (!had_srv6_cap && endx_was_empty)
		bgp_ls_originate_bgp_node(bgp);
```

### Issue 2 of 2
bgpd/bgp_ls.c:1742-1747
The symmetrical problem exists on deletion: `bgp_ls_has_srv6_capability` only tests the locator count, so this guard fires whenever there are no locators — even when the `static_endx_sids` list still has other entries. That generates a spurious node re-origination (without SRv6 capability) for every End.X SID removal when no locator is configured. The re-origination should only happen when the last End.X SID is removed and there are no locators.

```suggestion
	peer = entry->peer;
	bgp_ls_endx_sid_list_del(&bgp->ls_info->static_endx_sids, entry);
	XFREE(MTYPE_BGP_LS, entry);

	if (!bgp_ls_has_srv6_capability(bgp) &&
	    bgp_ls_endx_sid_list_empty(&bgp->ls_info->static_endx_sids))
		bgp_ls_originate_bgp_node(bgp);
```

_{Reviews (3): Last reviewed commit: "tests: Add topotest for BGP-LS SRv6 exte..." | Re-trigger Greptile}

[cscarpitta]

@greptile review

[cscarpitta]

@greptile review

[Jafaral]

@Mergifyio backport stable/10.6 stable/10.5

[mergify]

backport stable/10.6 stable/10.5

🟠 Waiting for conditions to match

Details

• merged [📌 backport requirement]