Per-route HTTP-level blocking policies for L7 routes. Two rule types: block_user_agent (substring match against User-Agent, returns 403) and require_header (named header must be present, returns 403). Config: L7Policy struct with type/value fields, added as L7Policies slice on Route. Validated in config (type enum, non-empty value, warning if set on L4 routes). DB: Migration 4 creates l7_policies table with route_id FK (cascade delete), type CHECK constraint, UNIQUE(route_id, type, value). New l7policies.go with ListL7Policies, CreateL7Policy, DeleteL7Policy, GetRouteID. Seed updated to persist policies from config. L7 middleware: PolicyMiddleware in internal/l7/policy.go evaluates rules in order, returns 403 on first match, no-op if empty. Composed into the handler chain between context injection and reverse proxy. Server: L7PolicyRule type on RouteInfo with AddL7Policy/RemoveL7Policy mutation methods on ListenerState. handleL7 threads policies into l7.RouteConfig. Startup loads policies per L7 route from DB. Proto: L7Policy message, repeated l7_policies on Route. Three new RPCs: ListL7Policies, AddL7Policy, RemoveL7Policy. All follow the write-through pattern. Client: L7Policy type, ListL7Policies/AddL7Policy/RemoveL7Policy methods. CLI: mcproxyctl policies list/add/remove subcommands. Tests: 6 PolicyMiddleware unit tests (no policies, UA match/no-match, header present/absent, multiple rules). 4 DB tests (CRUD, cascade, duplicate, GetRouteID). 3 gRPC tests (add+list, remove, validation). 2 end-to-end L7 tests (UA block, required header with allow/deny). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
11 KiB
11 KiB