feat(federation): peer_reset_pending guard during limbo window

packages/server/src/routes/dm.peerResetPending.test.ts

@@ -0,0 +1,165 @@
+import { describe, it, expect, beforeEach, vi } from 'vitest';
+import Fastify, { type FastifyInstance } from 'fastify';
+import Database from 'better-sqlite3';
+import { drizzle } from 'drizzle-orm/better-sqlite3';
+import { eq } from 'drizzle-orm';
+import fs from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+import * as schema from '../db/schema.js';
+import { setWorkerId } from '../utils/snowflake.js';
+
+setWorkerId(1);
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+
+type TestDb = ReturnType<typeof drizzle<typeof schema>>;
+let sqlite: Database.Database;
+let testDb: TestDb;
+const currentUserId = 'user-A';
+
+vi.mock('../db/index.js', () => ({
+  getDb: () => testDb,
+  schema,
+}));
+
+vi.mock('../utils/auth.js', () => ({
+  authenticate: async (req: { userId?: string }) => {
+    req.userId = currentUserId;
+  },
+}));
+
+vi.mock('../ws/handler.js', () => ({
+  connectionManager: {
+    sendToUser: vi.fn(),
+    sendToDmMembers: vi.fn(),
+    sendToAdmins: vi.fn(),
+    getAllOnlineUserIds: () => [],
+  },
+}));
+
+vi.mock('../utils/federationOutbox.js', async () => {
+  const actual = await vi.importActual<typeof import('../utils/federationOutbox.js')>('../utils/federationOutbox.js');
+  return {
+    ...actual,
+    isFederationRelayEnabled: () => true,
+    queueDmCloseRelay: vi.fn(),
+    sendTypingRelay: vi.fn(),
+    queueDmRelay: vi.fn(),
+    queueOutboxEvent: vi.fn(),
+    appendMutationLog: vi.fn(),
+  };
+});
+
+function applyMigrations(db: Database.Database): void {
+  const migrationsDir = path.resolve(__dirname, '../../drizzle');
+  const files = fs.readdirSync(migrationsDir).filter(f => f.endsWith('.sql')).sort();
+  for (const f of files) {
+    const sqlText = fs.readFileSync(path.join(migrationsDir, f), 'utf8');
+    const statements = sqlText.split(/-->\s*statement-breakpoint/);
+    for (const stmt of statements) {
+      const clean = stmt.trim();
+      if (clean) db.exec(clean);
+    }
+  }
+}
+
+function seedCaller(): void {
+  testDb.insert(schema.users).values({
+    id: 'user-A',
+    username: 'alice',
+    displayName: 'Alice',
+    passwordHash: 'x',
+    homeUserId: 'user-A',
+    homeInstance: null,
+    createdAt: Date.now(),
+  }).run();
+}
+
+/** The reset peer's persistent row (still present during the limbo window — only
+ *  deleted on admin Re-peer). Its `origin` is the exact string markPeerReset journals. */
+function seedPeer(): void {
+  testDb.insert(schema.federationPeers).values({
+    id: 'peer-remote',
+    origin: 'https://remote.example',
+    hmacSecret: 'secret',
+    status: 'needs_attention',
+    needsAttentionReason: 'peer_reset_detected',
+    peerInstanceId: 'dead-epoch',
+    observedPeerInstanceId: 'new-epoch',
+    createdAt: Date.now(),
+  }).run();
+}
+
+function seedResetEvent(resolvedAt: number | null): void {
+  testDb.insert(schema.federationResetEvents).values({
+    origin: 'https://remote.example',
+    deadEpoch: 'dead-epoch',
+    newEpoch: resolvedAt === null ? null : 'new-epoch',
+    detectedAt: Date.now(),
+    resolvedAt,
+    stubCount: 1,
+    orphanedAccountCount: 0,
+  }).run();
+}
+
+async function buildApp(): Promise<FastifyInstance> {
+  const app = Fastify({ logger: false });
+  const { dmRoutes } = await import('./dm.js');
+  await app.register(dmRoutes);
+  await app.ready();
+  return app;
+}
+
+describe('POST /api/dm — limbo-window peer_reset_pending guard', () => {
+  beforeEach(() => {
+    sqlite = new Database(':memory:');
+    testDb = drizzle(sqlite, { schema });
+    applyMigrations(sqlite);
+    seedCaller();
+    seedPeer();
+  });
+
+  it('returns 409 peer_reset_pending when creating a federated DM to a reset-pending origin', async () => {
+    seedResetEvent(null);
+
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/dm',
+      payload: { homeUserId: 'remote-bob', homeInstance: 'https://remote.example' },
+    });
+
+    expect(res.statusCode).toBe(409);
+    expect(res.json().error).toBe('peer_reset_pending');
+    // No stub created and no DM channel created for the reset-pending peer.
+    expect(testDb.select().from(schema.dmChannels).all()).toHaveLength(0);
+    expect(testDb.select().from(schema.users).where(eq(schema.users.homeUserId, 'remote-bob')).all()).toHaveLength(0);
+  });
+
+  it('proceeds normally when the reset event is RESOLVED', async () => {
+    seedResetEvent(Date.now());
+
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/dm',
+      payload: { homeUserId: 'remote-bob', homeInstance: 'https://remote.example' },
+    });
+
+    expect(res.statusCode).toBe(201);
+    expect(res.json().federatedId).toMatch(/^[a-f0-9]{32}$/);
+  });
+
+  it('proceeds normally when NO reset event exists for the origin', async () => {
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/dm',
+      payload: { homeUserId: 'remote-bob', homeInstance: 'https://remote.example' },
+    });
+
+    expect(res.statusCode).toBe(201);
+    expect(res.json().federatedId).toMatch(/^[a-f0-9]{32}$/);
+  });
+});

packages/server/src/routes/dm.ts

@@ -47,6 +47,7 @@ import {
   normalizeIconForWire,
 } from '../utils/federationOutbox.js';
 import { getOurOrigin, canonicalizeHomeInstance } from '../utils/federationAuth.js';
+import { resolveOriginFromHostname } from '../utils/federationOriginResolve.js';
 import type { FederationRelayEvent } from '@backspace/shared';
 import { resolveLocalUser, resolveOrCreateReplicatedUser } from './federation.js';
 
@@ -928,6 +929,40 @@ export async function dmRoutes(app: FastifyInstance): Promise<void> {
     let targetUser: typeof schema.users.$inferSelect | undefined;
 
     if (homeUserId && homeInstance) {
+      // Limbo-window guard (federation instance-epoch self-healing §5.3).
+      // If the target's home instance was reset-detected but the admin has not yet
+      // re-peered, an UNRESOLVED `federation_reset_events` row exists for its origin.
+      // Creating a DM now would bind to stale, dead-incarnation identity state, so
+      // surface a clear `peer_reset_pending` instead of silently forming a doomed
+      // channel. Checked BEFORE stub creation so no un-flagged stub is left behind.
+      //
+      // The journal is keyed by the peer's `federation_peers.origin` (the exact string
+      // `markPeerReset` stores). `resolveOriginFromHostname` returns that stored origin
+      // verbatim, giving an O(1) point lookup on the origin PRIMARY KEY; the common case
+      // (no reset) is a single indexed miss and the normal path proceeds unchanged.
+      const canon = canonicalizeHomeInstance(homeInstance);
+      let peerOrigin: string | null = null;
+      if (canon) {
+        try {
+          peerOrigin = resolveOriginFromHostname(new URL(canon).host);
+        } catch {
+          peerOrigin = null;
+        }
+      }
+      if (peerOrigin) {
+        const pendingReset = db
+          .select({ origin: schema.federationResetEvents.origin })
+          .from(schema.federationResetEvents)
+          .where(and(
+            eq(schema.federationResetEvents.origin, peerOrigin),
+            isNull(schema.federationResetEvents.resolvedAt),
+          ))
+          .get();
+        if (pendingReset) {
+          return reply.code(409).send({ error: 'peer_reset_pending', statusCode: 409 });
+        }
+      }
+
       // Federated identity: resolve or create a replicated user stub
       targetUser = resolveOrCreateReplicatedUser(homeUserId, homeInstance, db) ?? undefined;
     } else if (userId && typeof userId === 'string') {

packages/server/src/routes/social.federated.test.ts

@@ -457,3 +457,100 @@ describe('POST /api/social/requests — federated branch (authority + self-frien
     expect(body.requestId).toBe('incoming-req');
   });
 });
+
+describe('POST /api/social/requests — federated branch (limbo-window peer_reset_pending)', () => {
+  beforeEach(() => {
+    seedSelf();
+    resolveOriginFromHostnameMock.mockReturnValue('https://orbit.test');
+  });
+
+  function seedResetEvent(resolvedAt: number | null): void {
+    testDb.insert(schema.federationResetEvents).values({
+      origin: 'https://orbit.test',
+      deadEpoch: 'dead-epoch',
+      newEpoch: resolvedAt === null ? null : 'new-epoch',
+      detectedAt: Date.now(),
+      resolvedAt,
+      stubCount: 1,
+      orphanedAccountCount: 0,
+    }).run();
+  }
+
+  it('returns 409 peer_reset_pending when an UNRESOLVED reset event exists for the target origin', async () => {
+    seedResetEvent(null);
+    // Even a stale friendship must NOT surface as `already_friends` during the limbo window.
+    testDb.insert(schema.users).values({
+      id: 'stub-alice',
+      username: 'remote-alice@orbit.test',
+      displayName: 'Alice',
+      passwordHash: '!federation-replicated',
+      status: 'offline',
+      isAdmin: 0,
+      homeInstance: 'orbit.test',
+      homeUserId: 'remote-alice-old',
+      createdAt: Date.now(),
+    }).run();
+    testDb.insert(schema.friends).values({
+      userId: CALLER_ID,
+      friendId: 'stub-alice',
+      createdAt: Date.now(),
+    }).run();
+
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/social/requests',
+      payload: { username: 'alice@orbit.test' },
+    });
+
+    expect(res.statusCode).toBe(409);
+    expect(JSON.parse(res.body).error).toBe('peer_reset_pending');
+    // Short-circuits before peering/lookup — neither is consulted.
+    expect(ensurePeeredMock).not.toHaveBeenCalled();
+    expect(lookupRemoteUserMock).not.toHaveBeenCalled();
+    // No new request row created.
+    expect(testDb.select().from(schema.friendRequests).all()).toHaveLength(0);
+  });
+
+  it('proceeds normally when the reset event is RESOLVED (resolved_at set)', async () => {
+    seedResetEvent(Date.now());
+    ensurePeeredMock.mockResolvedValue({ status: 'active', peerId: 'peer-1' });
+    lookupRemoteUserMock.mockResolvedValue({
+      ok: true,
+      homeUserId: 'remote-alice',
+      username: 'alice',
+      profile: { displayName: 'Alice', avatar: null, avatarColor: 'mint', banner: null, bio: null },
+    });
+
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/social/requests',
+      payload: { username: 'alice@orbit.test' },
+    });
+
+    expect(res.statusCode).toBe(201);
+    expect(ensurePeeredMock).toHaveBeenCalled();
+    expect(JSON.parse(res.body).success).toBe(true);
+  });
+
+  it('proceeds normally when NO reset event exists for the origin', async () => {
+    ensurePeeredMock.mockResolvedValue({ status: 'active', peerId: 'peer-1' });
+    lookupRemoteUserMock.mockResolvedValue({
+      ok: true,
+      homeUserId: 'remote-alice',
+      username: 'alice',
+      profile: { displayName: 'Alice', avatar: null, avatarColor: 'mint', banner: null, bio: null },
+    });
+
+    const app = await buildApp();
+    const res = await app.inject({
+      method: 'POST',
+      url: '/api/social/requests',
+      payload: { username: 'alice@orbit.test' },
+    });
+
+    expect(res.statusCode).toBe(201);
+    expect(ensurePeeredMock).toHaveBeenCalled();
+  });
+});

packages/server/src/routes/social.ts

@@ -1,5 +1,5 @@
 import type { FastifyInstance, FastifyRequest, FastifyReply } from 'fastify';
-import { eq, and, or, ne, like, sql, inArray } from 'drizzle-orm';
+import { eq, and, or, ne, like, sql, inArray, isNull } from 'drizzle-orm';
 import { getDb, getRawDb, schema } from '../db/index.js';
 import { authenticate } from '../utils/auth.js';
 import { generateSnowflake } from '../utils/snowflake.js';
@@ -183,6 +183,32 @@ async function handleFederatedFriendRequest(
     return reply.code(400).send({ error: 'invalid_target_domain', statusCode: 400, domain: targetDomain });
   }
 
+  // 1a. Limbo-window guard (federation instance-epoch self-healing §5.3).
+  //     If this peer's home instance was reset-detected but the admin has not yet
+  //     re-peered, an UNRESOLVED `federation_reset_events` row exists for its origin
+  //     (the peer sits in `needs_attention`, its local friendship/stub graph still
+  //     bound to the dead incarnation). Without this guard the re-add would surface
+  //     a confusing `already_friends` (stale friendship) or `peer_rejected` (the
+  //     needs_attention peer) — neither of which tells the user what to do. Return a
+  //     clear `peer_reset_pending` instead.
+  //
+  //     `resolveOriginFromHostname` returns the peer's stored `federation_peers.origin`
+  //     verbatim, which is exactly the string `markPeerReset` journals as
+  //     `federation_reset_events.origin` (its PRIMARY KEY), so this is an O(1) indexed
+  //     point lookup. The common case — no reset in progress — is a single indexed miss
+  //     and the normal path proceeds unchanged.
+  const pendingReset = db
+    .select({ origin: schema.federationResetEvents.origin })
+    .from(schema.federationResetEvents)
+    .where(and(
+      eq(schema.federationResetEvents.origin, peerOrigin),
+      isNull(schema.federationResetEvents.resolvedAt),
+    ))
+    .get();
+  if (pendingReset) {
+    return reply.code(409).send({ error: 'peer_reset_pending', statusCode: 409 });
+  }
+
   // 2. ensurePeered — block until 'active', or surface peer status as error
   const peering = await ensurePeered(peerOrigin, {
     kind: 'user_action',