ospfd,tests: fix OSPF connected overlapping prefix bug

[rzalamena]

When two distinct interfaces have overlapping prefixes (example: 10.0.0.1/24 and 10.0.0.129/30) the more specific prefix gets omitted due to the wrong comparison in ospf_distribute_check_connected: if (prefix_match(oi->address, (struct prefix *)&ei->p)).

Fix the ospf_distribute_check_connected to use prefix_same() in order to rightfully generate the external route for this case.

[greptile-apps]

Greptile Summary

Fixes the OSPF connected-route suppression logic in ospf_distribute_check_connected so that a more-specific prefix (e.g. 10.0.0.128/30) on a non-OSPF interface is no longer wrongly suppressed when a broader OSPF-enabled interface (10.0.0.1/24) is present. The old prefix_match performed a containment check (is /30 inside the /24?), which returned true and suppressed redistribution; the new code masks the interface address with apply_mask before calling prefix_same, which requires an exact network and prefix-length match. A new topotest reproduces the bug with r1 advertising both a /24 OSPF intra-area route and a /30 redistributed external route, and verifies the correct behavior on r2.

Confidence Score: 5/5

Safe to merge; the fix is correct and the test suite covers both the regression and the suppression guard

The apply_mask + prefix_same pattern correctly replaces prefix_match by zeroing host bits before the exact-match comparison, fixing the root cause without introducing new regressions. The topotest validates the positive case (external /30) and the suppression case (intra-area /24). No P0 or P1 issues remain.

No files require special attention

Important Files Changed

Filename	Overview
ospfd/ospf_zebra.c	Replaces prefix_match with apply_mask + prefix_same to fix incorrect suppression of more-specific overlapping connected prefixes; the approach is correct and safe
tests/topotests/ospf_connected_overlapping_prefix/test_ospf_connected_overlapping_prefix.py	New topotest covering the bug scenario; tests neighbor convergence, external advertisement of the /30, and intra-area presence of the /24; topology dict "s2": "r1" is handled correctly by Topogen's is_string guard
tests/topotests/ospf_connected_overlapping_prefix/r1/frr.conf	r1 config: eth0 in OSPF area 0 (/24), eth1 not in OSPF (/30 overlapping), redistribute connected — correctly models the bug scenario
tests/topotests/ospf_connected_overlapping_prefix/r2/frr.conf	r2 config: eth0 in OSPF area 0 (/24), acts as the observer verifying that the /30 external LSA and /24 intra-area route are received correctly

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["ospf_distribute_check_connected(ospf, ei)"] --> B["for each oi in ospf->oiflist"]
    B --> C["prefix_copy(&address, oi->address)\napply_mask(&address)\n(zeros host bits: 10.0.0.1/24 → 10.0.0.0/24)"]
    C --> D{"prefix_same(address, ei->p)?"}
    D -- "Yes\n(exact network match)\ne.g. 10.0.0.0/24 == 10.0.0.0/24" --> E["return 0\n(suppress redistribution)"]
    D -- "No\n(different network or prefix len)\ne.g. 10.0.0.0/24 ≠ 10.0.0.128/30" --> F["next oi"]
    F --> B
    B -- "exhausted" --> G["return 1\n(allow redistribution)"]

    style E fill:#f99,color:#000
    style G fill:#9f9,color:#000

Loading

_{Reviews (3): Last reviewed commit: "ospfd: fix redistribution for overlappin..." | Re-trigger Greptile}

[rzalamena]

@greptileai

[rzalamena]

@greptileai