bgpd: Fix SRv6 SID export route-map update not taking effect

[cscarpitta]

When sid export ... route-map <name> is reconfigured with a different route-map on an already-configured SID export, the new route-map is silently ignored. The code skips the early-return (rmap_str != rmap_name), but then calls bgp_srv6_unicast_announce() without updating rmap_name, so the old route-map remains in effect.

Fix this by replacing rmap_name before triggering the re-announce: decrement the reference count on the old route-map, free its name string, assign the new name, and increment the reference count on the new route-map.

[greptile-apps]

Greptile Summary

This PR fixes a bug where reconfiguring sid export ... route-map <name> with a different route-map on an already-configured SID export silently kept the old route-map in effect. The fix correctly updates rmap_name — decrementing the old route-map's reference count, freeing its name string, duplicating the new name, and incrementing the new route-map's reference count — before calling bgp_srv6_unicast_announce(). A new topotest validates the fix end-to-end.

Changes:

• bgpd/bgp_vty.c: In the sid_export DEFPY handler, adds the missing rmap_name update steps (ref decrement, free, strdup, ref increment) in the "already configured, rmap changed" branch before triggering re-announcement.
• tests/topotests/bgp_srv6_unicast/test_bgp_srv6_unicast.py: Adds test_bgp_srv6_sid_rmap_update() which swaps the route-map on a live SID export and confirms the policy change propagates to peers R2 and R3, then restores the original and confirms recovery.

Confidence Score: 5/5

• This PR is safe to merge — it is a minimal, well-targeted bug fix with correct memory and reference-count management, mirroring the established pattern used elsewhere in the same function.
• The fix is logically sound: at the point the new code executes, rmap_str is guaranteed non-NULL (the early-return gate handles the NULL case), and the if (bgp->srv6_unicast[afi].rmap_name) guard safely covers the case where no previous route-map was set. The new code exactly mirrors the decrement/free/strdup/increment pattern used in the no branch and the initial-configuration path in the same function. The topotest exercises the precise failure scenario from the bug report and also verifies rollback. No unrelated code is touched.
• No files require special attention.

Important Files Changed

Filename	Overview
bgpd/bgp_vty.c	Inserts proper rmap_name update (decrement old ref, free, strdup new, increment new ref) before re-announcing, fixing the bug where a reconfigured route-map was silently ignored.
tests/topotests/bgp_srv6_unicast/test_bgp_srv6_unicast.py	Adds test_bgp_srv6_sid_rmap_update() to verify that replacing a route-map on an existing SID export actually takes effect, directly exercising the fixed code path.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[sid export route-map VTY command] --> B{SID already configured?}
    B -- No --> C[Initial setup: XSTRDUP rmap_name, increment ref, ensure SID]
    B -- Yes --> D{rmap_str same as current rmap_name?}
    D -- Same or NULL --> E[Return CMD_SUCCESS no-op]
    D -- Different --> F[Decrement old rmap ref count]
    F --> G[XFREE old rmap_name]
    G --> H[XSTRDUP new rmap_name]
    H --> I[Increment new rmap ref count]
    I --> J[bgp_srv6_unicast_announce with new route-map]
    J --> K[Return CMD_SUCCESS - new route-map in effect]

Loading

_{Reviews (1): Last reviewed commit: "tests: Add test for sid export route-map..." | Re-trigger Greptile}