bgp redistribute

Pre-Alpha. This page describes behavior that may change.

Cross-protocol route redistribution. Imports routes from non-BGP sources (static, connected, kernel, l2tp, ospf) into BGP for advertisement to peers.

Configuration

redistribute {
    destination bgp {
        import static { }
        import kernel { }
        import l2tp {
            family [ ipv4/unicast ipv6/unicast ];
        }
        import connected { }
    }
}

Each destination protocol has its own container. Each import entry names a registered source. The family leaf-list restricts which address families are imported (empty means all).

Available sources

Source	What it redistributes
`static`	Static routes configured under `static { }`.
`kernel`	Externally-installed kernel routes (DHCP, PPP, manual). Filters RTPROT_KERNEL and RTPROT_REDIRECT.
`connected`	Directly connected interface subnets.
`l2tp`	Subscriber routes from L2TP/PPP sessions.

The kernel source uses a shared routewatch infrastructure that subscribes to netlink route notifications. It filters out routes owned by Ze itself (identified by RTPROT_ZE), so only externally-installed routes are redistributed. Routes are tracked and withdrawn automatically when removed from the kernel table.

Architecture

The redistribute system has two plugins:

redistribute-egress (orchestrator): discovers producers at startup, subscribes to route events, dispatches to registered consumers.
redistribute-ingress: ingress filter with loop prevention and family filtering.

Producers (static, kernel, l2tp, connected) register themselves with the orchestrator. Consumers (bgp) subscribe to receive redistributed routes.

The kernel redistribute plugin (internal/plugins/kernel/) registers as a routewatch consumer. The shared routewatch infrastructure (internal/core/routewatch/) provides netlink route subscription with synchronous fanout to all registered consumers, halving syscall and parsing cost compared to per-plugin subscriptions.