feat(api): Phase 6 PR-3 — substrate hardening (SKIP LOCKED + DNS pinning + test debt) (#500)

* feat(api): operations table immutability trigger

BFNAR 2013:2 kap 8 § behandlingshistorik integrity: once an operations
row is in a terminal status (succeeded / failed / cancelled) the audit
record of what happened becomes immutable. Adds the BEFORE UPDATE and
BEFORE DELETE triggers that the webhook_deliveries table already has
(20260515170000 / 20260515190000), mirroring their predicate shape and
error code exactly.

Closes the Phase 4 PR-2 (PR #469) review-round carry-over flagged by
Swedish-compliance: previously a future bug, a privileged operator, or
a compromised service-role caller could rewrite "this year-end close
succeeded" to "failed" by updating an already-terminal row. The
running → succeeded/failed/cancelled transition itself stays legal
because the trigger keys on OLD.status, which is non-terminal at the
moment of the legitimate UPDATE.

pg test covers all transitions (allowed and blocked) plus DELETE on
both terminal and non-terminal rows.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(api): atomic SKIP LOCKED claim for webhook dispatch

Replaces the SELECT-then-UPDATE-intersect pattern in the dispatcher
with a single-roundtrip SQL function using FOR UPDATE SKIP LOCKED.
PostgREST can't express SKIP LOCKED through the JS client, so the
previous shape relied on a CAS guard inside an UPDATE WHERE status IN
('pending','failed') to ensure only one of two overlapping cron ticks
claimed any given row.

The CAS pattern was correct (under load — receivers >60s could push a
batch past the next minute's tick) but burned two round trips and
forced the application to negotiate the locking semantics in JS.
The function form moves the contention to the DB, where SKIP LOCKED
makes a row held by a concurrent tick simply invisible to the second
caller. One round trip, no JS-side intersect.

All filter semantics are preserved verbatim inside the function:
status IN ('pending','failed'), next_attempt_at <= now, webhook_id
IS NOT NULL, ORDER BY next_attempt_at ASC, LIMIT batchSize. p_batch_size
is bounded (0, 1000] to forestall a runaway lock-set in case a caller
misconfigures it.

pg test covers basic claim (pending + failed), future-due skip, dangling-
row (webhook_id IS NULL) skip, terminal-status skip, batch-size limits,
out-of-range argument rejection, and the SKIP LOCKED invariant itself
using two concurrent pool clients in BEGIN — the second caller does not
see the row A locked, no double-delivery.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(api): pinned-IP HTTPS dispatch (close DNS-rebinding window)

The url-guard.ts file header openly flagged the remaining gap:
"a separate DNS-rebinding window (between dispatch-time validation
and the actual fetch) remains; closing that requires a custom HTTPS
agent that pins the resolved IP — tracked for follow-up." This closes it.

The previous shape was:
  1. validateWebhookUrl()  → DNS resolves to [public IP], returns ok
  2. fetch(webhook_url)    → re-resolves DNS; an attacker who flipped
                             the A record in the interval gets a
                             private-IP socket

The new pinnedHttpsFetch helper validates DNS once, then opens a
node:https.request to that pinned IP — but keeps the original hostname
in the TLS SNI extension (so the receiver's cert validates) and in the
HTTP Host header (so vhost routing still works). The request socket
never re-resolves DNS, foreclosing the rebind race entirely.

Built on node:https.request rather than undici's Agent so the project
doesn't take on a new dep — the stdlib API is also more explicit about
the SNI / Host / pinned-IP split. Test seam injects both validateUrl
and httpsRequest so the unit tests verify the pinning shape without
standing up an HTTPS server.

The dispatcher's attemptDelivery is rewritten as a switch over the four
PinnedFetchResult kinds (ok / unsafe_url / redirect_blocked / timeout
/ transport_error). The previous fetch-based code path that distinguished
redirect rejection by string-matching err.message is gone — the new
result type makes the distinction structural.

8 unit tests cover the SNI/Host/pinned-IP shape, port handling,
redirect_blocked, transport_error, timeout, response-body truncation,
first-IP determinism, and the validation short-circuit (never opens a
socket when the URL fails the SSRF guard).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(api): pg tests for webhook substrate triggers (PR-1 test debt)

CLAUDE.md ("Testing" + "Migration Rules") mandates a *.pg.test.ts for
any PR touching a trigger / RPC / RLS / DEFERRABLE constraint. Phase 6
PR-1 (#496) shipped three webhook_deliveries triggers without the
accompanying pg test; this closes that debt.

Triggers covered:
  - enforce_webhook_delivery_immutability  (BEFORE UPDATE)
  - block_webhook_delivery_terminal_delete (BEFORE DELETE)
  - assert_webhook_delivery_company_match  (BEFORE INSERT)

13 cases verify the lifecycle the dispatcher depends on remains mutable
(pending → in_flight, in_flight → failed, failed → in_flight, in_flight
→ delivered) while terminal-status rows (delivered / dead) are write-
locked and the cross-tenant INSERT path is refused with the
ERRCODE=check_violation contract documented in the migration.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(api): integration tests for webhook routes (PR-1 test debt)

CLAUDE.md mandates integration tests under app/api/v1/ for every route.
Phase 6 PR-1 (#496) shipped the eight v1 webhook routes (five under
/companies/{companyId}/webhooks/ + the cross-tenant /webhook-deliveries/
{id}/retry) without them; closes that debt.

19 cases for the /webhooks/ verticals:
  POST     /webhooks                    create + secret-once + payroll-scope gate + SSRF
  GET      /webhooks                    list (no secret) + empty list
  GET      /webhooks/:id                detail (no secret) + 404
  PATCH    /webhooks/:id                update + active=true re-enable + SSRF re-check + empty-body
  DELETE   /webhooks/:id                204 hard delete
  POST     /webhooks/:id/test           enqueue + 404 + disabled-rejection
  GET      /webhooks/:id/deliveries     happy path + ownership 404

7 cases for the retry route:
  POST /webhook-deliveries/:id/retry   dead → fresh pending row, live-status refusal,
                                       cross-tenant 404, disabled-webhook gate,
                                       SSRF re-check, delivery 404, webhook-gone 404

Both files mirror the suppliers/customers integration test pattern:
Proxy-backed Supabase mock with per-table queues, validateApiKey +
validateWebhookUrl stubbed to control auth and DNS deterministically.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(api): address PR-500 review round 1 — pg-real CI fix + 4 review items

1. pg-real CI was red on this PR: the new webhook trigger pg.test.ts and
   claim-due-webhook-deliveries pg.test.ts fixtures tried to INSERT into
   `webhooks.user_id`, which doesn't exist in the migration history. The
   column was never declared in automation_webhooks (20260415000000) nor
   added by webhooks_v2 (20260515170000) — so a fresh schema replay had
   no such column. The webhook create route (`webhooks.create`) was also
   referencing this non-existent column in its INSERT, so the production
   route was latent-broken since PR-1 and never exercised against a fresh
   DB. Drop the `user_id` field from both the route INSERT and the pg
   fixtures. Actor attribution lives on `created_by_api_key_id` (which
   leads back to the owning user via `api_keys.user_id`).

2. Greptile P2 #1 — `recoverStuckInFlight` carried a redundant
   `.not('status','in','(delivered,dead)')` filter alongside
   `.eq('status','in_flight')`, with a comment that incorrectly described
   PostgreSQL's UPDATE re-evaluation semantics. Under READ COMMITTED,
   UPDATE re-evaluates WHERE against each row's CURRENT value when it
   acquires the row lock — a row that raced to terminal status will fail
   `status='in_flight'` on re-evaluation and be skipped, no immutability
   trigger fires. Drop the redundant filter and rewrite the comment.

3. Greptile P2 #2 — added explicit pg test verifying `in_flight` rows are
   skipped by `claim_due_webhook_deliveries`. The status filter is what
   prevents double-delivery and is the entire point of the SKIP LOCKED
   substrate; making that invariant load-bearing in the test suite
   forecloses a future filter expansion silently regressing it.

4. Greptile P2 #3 — pinned-fetch registered both `res.on('end', finalize)`
   and `res.on('close', finalize)`. Node fires BOTH on normal completions,
   so finalize ran twice; the outer `settled` guard squashed the
   double-resolve but the header reconstruction still ran twice. Switch
   to `once` + self-removing pair so finalize runs exactly once on
   whichever event fires first (normal: end; truncation: close).

5. Compliance Swarm V8.2.1 — the retry route only checked
   `webhooks:manage` even when retrying `salary_run.* / agi.*` deliveries.
   Mirror the create-route elevated-scope gate so a key with only
   `webhooks:manage` cannot re-emit payroll payloads carrying
   personnummer / lönesummor / skatteavdrag. New integration test verifies
   the gate returns 403 INSUFFICIENT_SCOPE with `required_scope:
   payroll:read`.

35 tests pass locally (+1 vs pre-fix). Type-check clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(api): address PR-500 review round 2 — 2 small precision fixes

1. Compliance Swarm Art.32 / A.8.24 — response_body size cap was enforced
   only at the application layer (pinnedHttpsFetch's maxResponseBytes=4096
   constant). A future refactor that bypassed the truncation, or a non-
   dispatcher write path into webhook_deliveries.response_body, would
   silently land large blobs in a column adjacent to event payloads
   carrying personal data. Add a CHECK constraint at the DB layer with
   a generous ceiling (8 KB — double the application cap so legitimate
   dispatcher writes never hit it; only a regression surfaces as a
   check_violation).

2. Compliance Swarm CC6.6 — pinned-fetch substitutes the validated IP
   for `host` while keeping the original hostname in `servername`. A
   reader could reasonably worry that the IP substitution weakens TLS
   hostname verification. Document explicitly that Node's default
   `checkServerIdentity` matches the cert's SAN/CN against `servername`
   (not `host`), so a forged endpoint at the pinned IP with a valid
   cert for a different hostname would fail the handshake. No code
   change — the default behavior is correct; the comment forecloses
   future "this looks dangerous" review-round noise on the same line.

Items NOT addressed (with rationale documented elsewhere):

- Compliance Swarm V8.2.1 (retry route 404-vs-404 information leak):
  delivery IDs are UUIDs; the "leak" is the ability to probe existence
  of an opaque 128-bit identifier the caller already has, which is not
  meaningfully different from probing for any opaque token. Both
  branches return the same structured 404 envelope.

- Compliance Swarm CC7.2 (restore the .not() defense-in-depth filter):
  direct contradiction of last round's Greptile P2 fix. Greptile's
  PG-semantics analysis is correct — under READ COMMITTED, UPDATE
  re-evaluates WHERE against the row's current value when it acquires
  the lock, so .eq('status','in_flight') already handles the race.
  Adding a redundant .not() restores a misleading comment without
  closing a real gap. This is the documented Compliance Swarm
  oscillation pattern from the project's Phase 4 lessons.

- Compliance Swarm CC6.1 (webhook secret encryption-at-rest):
  architectural choice from PR-1; not in PR-3 (substrate hardening)
  scope. Belongs to a future hardening PR.

- Swedish-compliance review (operations queued/running rows hard-
  deletable): deliberate operability tradeoff — operators need to
  clear stuck/queued entries that crashed mid-flight. Blocking all
  deletes would force a manual DB intervention every time a worker
  crashed before reaching terminal status. The audit trail starts
  at terminal-state mutation, which IS blocked.

- Swedish-compliance review (salary_run.* / agi.* payload anonymisation
  after 7 years): already on the deferred-list as part of the 90-day
  TTL cleanup cron item from the PR description. Belongs to a
  retention-policy follow-up PR.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Jakob Wennberg
2026-05-15 20:38:19 +02:00
committed by GitHub
co-authored by Claude Opus 4.7
parent 240412fd32
commit afb21ea638
13 changed files with 2601 additions and 204 deletions
@@ -0,0 +1,284 @@
import { randomUUID } from 'node:crypto'
import { describe, expect, it } from 'vitest'
import { getClient, getPool } from '@/tests/pg/setup'
import { seedCompany } from '@/tests/pg/fixtures'
// ──────────────────────────────────────────────────────────────────────
// Fixtures: parent webhook + child delivery
// ──────────────────────────────────────────────────────────────────────
async function insertWebhook(params: {
// userId kept in the signature for parity with seedCompany's return — the
// webhooks table itself has no user_id column (see route comment in
// app/api/v1/companies/[companyId]/webhooks/route.ts).
userId: string
companyId: string
eventType?: string
active?: boolean
}): Promise<string> {
void params.userId
const id = randomUUID()
await getPool().query(
`INSERT INTO public.webhooks
(id, company_id, name, event_type, webhook_url, secret, active)
VALUES ($1, $2, 'pg-test', $3, 'https://example.com/hook', $4, $5)`,
[
id,
params.companyId,
params.eventType ?? 'invoice.paid',
`whsec_${randomUUID().replace(/-/g, '')}`,
params.active ?? true,
],
)
return id
}
async function insertDelivery(params: {
webhookId: string | null
companyId: string
status?: 'pending' | 'in_flight' | 'delivered' | 'failed' | 'dead'
nextAttemptAt?: string
eventType?: string
attempts?: number
}): Promise<string> {
const id = randomUUID()
await getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at, attempts)
VALUES ($1, $2, $3, $4, '{"hello":"world"}'::jsonb, '2026-05-12',
$5, $6, $7)`,
[
id,
params.webhookId,
params.companyId,
params.eventType ?? 'invoice.paid',
params.status ?? 'pending',
params.nextAttemptAt ?? new Date().toISOString(),
params.attempts ?? 0,
],
)
return id
}
async function getDeliveryStatus(id: string): Promise<string | null> {
const r = await getPool().query<{ status: string }>(
`SELECT status FROM public.webhook_deliveries WHERE id = $1`,
[id],
)
return r.rows[0]?.status ?? null
}
// Direct-insert rows for one test isolation. Each it() seeds its own
// company + webhook so the dispatcher sees a clean slate; we just need to
// make sure the function only returns rows we created, which we do by
// asserting on ids.
describe('claim_due_webhook_deliveries.pg — atomic SKIP LOCKED claim', () => {
it('claims a pending row that is due', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'pending' })
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(rows.map((r) => r.id)).toContain(deliveryId)
expect(await getDeliveryStatus(deliveryId)).toBe('in_flight')
})
it('claims a failed row that has reached its retry deadline', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
// next_attempt_at in the past — retry due.
const deliveryId = await insertDelivery({
webhookId,
companyId,
status: 'failed',
nextAttemptAt: new Date(Date.now() - 60_000).toISOString(),
attempts: 2,
})
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(rows.map((r) => r.id)).toContain(deliveryId)
expect(await getDeliveryStatus(deliveryId)).toBe('in_flight')
})
it('skips a row whose next_attempt_at is still in the future', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({
webhookId,
companyId,
status: 'failed',
nextAttemptAt: new Date(Date.now() + 5 * 60_000).toISOString(),
})
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(rows.map((r) => r.id)).not.toContain(deliveryId)
// Status stays pre-claim.
expect(await getDeliveryStatus(deliveryId)).toBe('failed')
})
it('skips dangling rows (webhook_id IS NULL)', async () => {
const { companyId } = await seedCompany()
// webhook deleted between enqueue and dispatch — FK ON DELETE SET NULL.
const deliveryId = await insertDelivery({
webhookId: null,
companyId,
status: 'pending',
})
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(rows.map((r) => r.id)).not.toContain(deliveryId)
})
it('skips terminal-status rows', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveredId = await insertDelivery({
webhookId,
companyId,
status: 'delivered',
})
const deadId = await insertDelivery({
webhookId,
companyId,
status: 'dead',
})
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
const ids = rows.map((r) => r.id)
expect(ids).not.toContain(deliveredId)
expect(ids).not.toContain(deadId)
})
// The status filter `IN ('pending', 'failed')` is what prevents
// double-delivery once a tick has already claimed a row to in_flight.
// recoverStuckInFlight is the ONLY legitimate path back from in_flight
// (sweeps the row to 'failed' after the stuck-threshold), so the claim
// function must NEVER re-pick a row already marked in_flight.
it('skips rows already in in_flight status', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const inFlightId = await insertDelivery({
webhookId,
companyId,
status: 'in_flight',
})
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(rows.map((r) => r.id)).not.toContain(inFlightId)
// Status must NOT have been re-flipped.
expect(await getDeliveryStatus(inFlightId)).toBe('in_flight')
})
it('respects the batch size', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const ids = await Promise.all(
Array.from({ length: 5 }, () =>
insertDelivery({ webhookId, companyId, status: 'pending' }),
),
)
const { rows } = await getPool().query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[2],
)
expect(rows.length).toBe(2)
// The 3 unclaimed rows stay pending.
const unclaimed = ids.filter((id) => !rows.some((r) => r.id === id))
for (const id of unclaimed) {
expect(await getDeliveryStatus(id)).toBe('pending')
}
})
it('rejects out-of-range batch sizes', async () => {
await expect(
getPool().query(`SELECT * FROM public.claim_due_webhook_deliveries($1, now())`, [0]),
).rejects.toThrow(/p_batch_size must be in/i)
await expect(
getPool().query(`SELECT * FROM public.claim_due_webhook_deliveries($1, now())`, [-1]),
).rejects.toThrow(/p_batch_size must be in/i)
await expect(
getPool().query(`SELECT * FROM public.claim_due_webhook_deliveries($1, now())`, [10000]),
).rejects.toThrow(/p_batch_size must be in/i)
})
// SKIP LOCKED is the entire point of this migration. Two transactions
// calling the function at the same moment must not both see the same
// row — the row locked by the first caller is invisible to the second,
// closing the duplicate-delivery window the old SELECT-then-UPDATE-
// intersect pattern documented as load-bearing.
it('SKIP LOCKED: a concurrent caller does not see rows locked by an in-flight transaction', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({
webhookId,
companyId,
status: 'pending',
})
const a = await getClient()
const b = await getClient()
try {
await a.query('BEGIN')
await b.query('BEGIN')
// A claims first. The row is now status='in_flight' AND held under
// a row lock by transaction A (UPDATE sets ROW EXCLUSIVE).
const aClaim = await a.query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(aClaim.rows.map((r) => r.id)).toContain(deliveryId)
// B's call SKIPs the locked row entirely. Without SKIP LOCKED this
// call would BLOCK on the row lock; the test would hang and only
// fail via testTimeout. SKIP LOCKED makes it return promptly with
// the row simply absent from results.
const bClaim = await b.query<{ id: string }>(
`SELECT id FROM public.claim_due_webhook_deliveries($1, now())`,
[10],
)
expect(bClaim.rows.map((r) => r.id)).not.toContain(deliveryId)
// Commit A and roll back B (no-op since B claimed nothing).
await a.query('COMMIT')
await b.query('ROLLBACK')
} finally {
a.release()
b.release()
}
// Final state: row is in_flight, claimed exactly once.
expect(await getDeliveryStatus(deliveryId)).toBe('in_flight')
})
})
+271
View File
@@ -0,0 +1,271 @@
import { describe, expect, it, vi } from 'vitest'
import { EventEmitter } from 'node:events'
import type { ClientRequest, IncomingMessage } from 'node:http'
import type { RequestOptions } from 'node:https'
import { pinnedHttpsFetch } from '@/lib/webhooks/pinned-fetch'
// Stand up a minimal stub for node:https.request that captures the args
// we want to assert on (pinned IP, SNI, Host header) and lets us synthesise
// a controlled response back to the caller.
function makeStubRequest(args: {
status: number
headers?: Record<string, string>
body?: string
emitError?: Error
emitTimeout?: boolean
}) {
const captured: {
options: RequestOptions | null
bodyWritten: string
} = { options: null, bodyWritten: '' }
const fakeRequest = (
options: RequestOptions,
callback: (res: IncomingMessage) => void,
): ClientRequest => {
captured.options = options
const req = new EventEmitter() as ClientRequest & EventEmitter
// ClientRequest API surface we touch in pinned-fetch:
req.write = ((chunk: string) => {
captured.bodyWritten += chunk
return true
}) as ClientRequest['write']
req.end = (() => {
// Dispatch the response (or error) asynchronously to mimic real
// network timing — pinned-fetch attaches handlers BEFORE end().
queueMicrotask(() => {
if (args.emitError) {
req.emit('error', args.emitError)
return
}
if (args.emitTimeout) {
req.emit('timeout')
return
}
const res = new EventEmitter() as IncomingMessage & EventEmitter
;(res as unknown as { statusCode: number }).statusCode = args.status
;(res as unknown as { headers: Record<string, string> }).headers =
args.headers ?? { 'content-type': 'application/json' }
// IncomingMessage stubs need stream-shaped methods that pinned-fetch
// calls (resume on redirect-drain, destroy on size truncation).
;(res as unknown as { resume: () => unknown }).resume = () => {
/* no-op — body is already buffered in args.body */
}
;(res as unknown as { destroy: () => unknown }).destroy = () => {
// Truncation path — emit `close` so finalize() runs.
queueMicrotask(() => res.emit('close'))
}
callback(res)
// Emit body bytes then `end`.
queueMicrotask(() => {
if (args.body) res.emit('data', Buffer.from(args.body, 'utf8'))
res.emit('end')
})
})
return req
}) as ClientRequest['end']
req.destroy = (() => {
// no-op: tests don't read the socket after destroy.
return req
}) as ClientRequest['destroy']
req.setTimeout = (() => req) as ClientRequest['setTimeout']
return req
}
return { captured, fakeRequest }
}
function makeStubValidator(addresses: string[]) {
return vi.fn(async () => ({
ok: true as const,
hostname: 'example.com',
resolvedAddresses: addresses,
}))
}
describe('pinnedHttpsFetch', () => {
it('pins the socket to the validated IP while keeping SNI + Host on the hostname', async () => {
const { captured, fakeRequest } = makeStubRequest({
status: 200,
body: 'ok',
})
const result = await pinnedHttpsFetch(
'https://example.com/hooks',
{
method: 'POST',
headers: { 'X-Gnubok-Event': 'invoice.paid' },
body: '{"hello":"world"}',
timeoutMs: 1000,
maxResponseBytes: 1024,
},
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(result.kind).toBe('ok')
if (result.kind !== 'ok') throw new Error('unreachable')
expect(result.status).toBe(200)
expect(result.body).toBe('ok')
expect(result.pinnedAddress).toBe('203.0.113.42')
// Socket goes to the IP — DNS does not re-resolve.
expect(captured.options?.host).toBe('203.0.113.42')
// SNI carries the hostname so the receiver's TLS cert validates.
expect(captured.options?.servername).toBe('example.com')
// HTTP Host header carries the hostname for vhost routing.
const headers = captured.options?.headers as Record<string, string>
expect(headers.host).toBe('example.com')
// Custom dispatcher header survives.
expect(headers['X-Gnubok-Event']).toBe('invoice.paid')
expect(captured.bodyWritten).toBe('{"hello":"world"}')
})
it('includes the port in the Host header when non-default', async () => {
const { captured, fakeRequest } = makeStubRequest({ status: 204 })
await pinnedHttpsFetch(
'https://example.com:8443/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(captured.options?.port).toBe(8443)
const headers = captured.options?.headers as Record<string, string>
expect(headers.host).toBe('example.com:8443')
})
it('returns unsafe_url when validation rejects the hostname', async () => {
const { fakeRequest } = makeStubRequest({ status: 200 })
const spyRequest = vi.fn(fakeRequest)
const validator = vi.fn(async () => ({
ok: false as const,
reason: 'private_address' as const,
detail: '10.0.0.1 is private',
}))
const result = await pinnedHttpsFetch(
'https://internal.example/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{ validateUrl: validator, httpsRequest: spyRequest },
)
expect(result.kind).toBe('unsafe_url')
if (result.kind === 'unsafe_url') {
expect(result.reason).toBe('private_address')
expect(result.pinnedAddress).toBeNull()
}
// Critically — we never opened a socket.
expect(spyRequest).not.toHaveBeenCalled()
})
it('treats 3xx responses as redirect_blocked', async () => {
const { fakeRequest } = makeStubRequest({
status: 302,
headers: { location: 'https://elsewhere.example/' },
})
const result = await pinnedHttpsFetch(
'https://example.com/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(result.kind).toBe('redirect_blocked')
if (result.kind === 'redirect_blocked') {
expect(result.status).toBe(302)
expect(result.pinnedAddress).toBe('203.0.113.42')
}
})
it('maps transport errors to transport_error', async () => {
const { fakeRequest } = makeStubRequest({
status: 0,
emitError: new Error('ECONNREFUSED 203.0.113.42:443'),
})
const result = await pinnedHttpsFetch(
'https://example.com/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(result.kind).toBe('transport_error')
if (result.kind === 'transport_error') {
expect(result.detail).toContain('ECONNREFUSED')
expect(result.pinnedAddress).toBe('203.0.113.42')
}
})
it('maps timeout events to timeout', async () => {
const { fakeRequest } = makeStubRequest({
status: 0,
emitTimeout: true,
})
const result = await pinnedHttpsFetch(
'https://example.com/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(result.kind).toBe('timeout')
})
it('truncates response body at maxResponseBytes', async () => {
const big = 'x'.repeat(10_000)
const { fakeRequest } = makeStubRequest({
status: 200,
headers: { 'content-type': 'text/plain' },
body: big,
})
const result = await pinnedHttpsFetch(
'https://example.com/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 100 },
{
validateUrl: makeStubValidator(['203.0.113.42']),
httpsRequest: fakeRequest,
},
)
expect(result.kind).toBe('ok')
if (result.kind === 'ok') {
expect(result.body.length).toBe(100)
expect(result.bodyTruncated).toBe(true)
}
})
it('picks the first resolved IP deterministically', async () => {
const { captured, fakeRequest } = makeStubRequest({ status: 200 })
await pinnedHttpsFetch(
'https://example.com/hooks',
{ method: 'POST', headers: {}, body: '', timeoutMs: 1000, maxResponseBytes: 1024 },
{
validateUrl: makeStubValidator(['203.0.113.42', '198.51.100.55']),
httpsRequest: fakeRequest,
},
)
// First entry, not the second, not random.
expect(captured.options?.host).toBe('203.0.113.42')
})
})
@@ -0,0 +1,261 @@
import { randomUUID } from 'node:crypto'
import { describe, expect, it } from 'vitest'
import { getPool } from '@/tests/pg/setup'
import { seedCompany } from '@/tests/pg/fixtures'
// Verifies the three webhook-substrate DB guards shipped in Phase 6 PR-1:
// - enforce_webhook_delivery_immutability (BEFORE UPDATE)
// - block_webhook_delivery_terminal_delete (BEFORE DELETE)
// - assert_webhook_delivery_company_match (BEFORE INSERT)
//
// CLAUDE.md ("Migration Rules" + Testing section) mandates a *.pg.test.ts
// for any PR that touches a trigger / RPC / RLS / DEFERRABLE constraint.
// PR-1 (#496) shipped the triggers without the accompanying pg test; this
// closes that test debt.
async function insertWebhook(params: {
// userId kept in the signature for parity with seedCompany's return — the
// webhooks table itself has no user_id column (see route comment in
// app/api/v1/companies/[companyId]/webhooks/route.ts).
userId: string
companyId: string
eventType?: string
}): Promise<string> {
void params.userId
const id = randomUUID()
await getPool().query(
`INSERT INTO public.webhooks
(id, company_id, name, event_type, webhook_url, secret, active)
VALUES ($1, $2, 'pg-test', $3, 'https://example.com/hook', $4, true)`,
[
id,
params.companyId,
params.eventType ?? 'invoice.paid',
`whsec_${randomUUID().replace(/-/g, '')}`,
],
)
return id
}
async function insertDelivery(params: {
webhookId: string | null
companyId: string
status?: 'pending' | 'in_flight' | 'delivered' | 'failed' | 'dead'
}): Promise<string> {
const id = randomUUID()
await getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at)
VALUES ($1, $2, $3, 'invoice.paid', '{"hello":"world"}'::jsonb,
'2026-05-12', $4, now())`,
[id, params.webhookId, params.companyId, params.status ?? 'pending'],
)
return id
}
describe('webhook_deliveries triggers — immutability + DELETE block', () => {
// The lifecycle that dispatcher.ts depends on must remain mutable:
// pending → in_flight (claim), in_flight → failed (retry-pending),
// failed → in_flight (re-claim). Only `delivered` and `dead` are
// terminal and locked.
it('allows pending → in_flight (claim) transition', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'pending' })
const result = await getPool().query(
`UPDATE public.webhook_deliveries SET status = 'in_flight' WHERE id = $1`,
[deliveryId],
)
expect(result.rowCount).toBe(1)
})
it('allows in_flight → failed (retry-pending) transition', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'in_flight' })
const result = await getPool().query(
`UPDATE public.webhook_deliveries SET status = 'failed', attempts = 1 WHERE id = $1`,
[deliveryId],
)
expect(result.rowCount).toBe(1)
})
it('allows failed → in_flight (re-claim) transition', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'failed' })
const result = await getPool().query(
`UPDATE public.webhook_deliveries SET status = 'in_flight' WHERE id = $1`,
[deliveryId],
)
expect(result.rowCount).toBe(1)
})
it('allows in_flight → delivered (success terminal) transition', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'in_flight' })
const result = await getPool().query(
`UPDATE public.webhook_deliveries
SET status = 'delivered', delivered_at = now(), response_status = 200
WHERE id = $1`,
[deliveryId],
)
expect(result.rowCount).toBe(1)
})
it('rejects UPDATE on a delivered row (status flip-back blocked)', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'delivered' })
await expect(
getPool().query(
`UPDATE public.webhook_deliveries SET status = 'pending' WHERE id = $1`,
[deliveryId],
),
).rejects.toThrow(/terminal status \(delivered\) is immutable/i)
})
it('rejects UPDATE on a dead row (response rewrite blocked)', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'dead' })
await expect(
getPool().query(
`UPDATE public.webhook_deliveries
SET response_body = 'tampered' WHERE id = $1`,
[deliveryId],
),
).rejects.toThrow(/terminal status \(dead\) is immutable/i)
})
it('rejects DELETE on a delivered row', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'delivered' })
await expect(
getPool().query(
`DELETE FROM public.webhook_deliveries WHERE id = $1`,
[deliveryId],
),
).rejects.toThrow(/terminal status \(delivered\) cannot be deleted/i)
})
it('rejects DELETE on a dead row', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'dead' })
await expect(
getPool().query(
`DELETE FROM public.webhook_deliveries WHERE id = $1`,
[deliveryId],
),
).rejects.toThrow(/terminal status \(dead\) cannot be deleted/i)
})
// Non-terminal rows are still deletable — the queue-cleanup path
// (operator clears a stuck pending row, dev environment wipes,
// companies CASCADE delete) keeps working.
it('allows DELETE on a pending row', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const deliveryId = await insertDelivery({ webhookId, companyId, status: 'pending' })
const result = await getPool().query(
`DELETE FROM public.webhook_deliveries WHERE id = $1`,
[deliveryId],
)
expect(result.rowCount).toBe(1)
})
})
describe('webhook_deliveries triggers — cross-tenant INSERT guard', () => {
it('rejects INSERT when delivery.company_id != webhooks.company_id', async () => {
// Tenant A owns the webhook; tenant B owns the company on the
// delivery row. A compromised service-role caller (or future bug)
// attempting to enqueue a delivery against another tenant's webhook
// is refused at write time — closes cases (a) (cross-tenant
// visibility) and (c) (existence leak) flagged in migration
// 20260515190000's comment.
const a = await seedCompany()
const b = await seedCompany()
const webhookId = await insertWebhook({ userId: a.userId, companyId: a.companyId })
await expect(
getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at)
VALUES (gen_random_uuid(), $1, $2, 'invoice.paid',
'{"hello":"world"}'::jsonb, '2026-05-12', 'pending', now())`,
[webhookId, b.companyId],
),
).rejects.toThrow(/company_id .+ does not match parent webhooks\.company_id/i)
})
it('accepts INSERT when delivery.company_id == webhooks.company_id', async () => {
const { userId, companyId } = await seedCompany()
const webhookId = await insertWebhook({ userId, companyId })
const result = await getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at)
VALUES (gen_random_uuid(), $1, $2, 'invoice.paid',
'{"hello":"world"}'::jsonb, '2026-05-12', 'pending', now())
RETURNING id`,
[webhookId, companyId],
)
expect(result.rowCount).toBe(1)
})
it('allows INSERT with webhook_id IS NULL (dangling row after webhook delete)', async () => {
// ON DELETE SET NULL on the FK leaves these rows after a webhook is
// deleted. New inserts with webhook_id IS NULL aren't a normal write
// path (the handler never inserts a null webhook_id) but the trigger
// explicitly bypasses the check rather than blocking — leaving room
// for an admin-side audit-replay tool that recreates an archived
// delivery for forensic export.
const { companyId } = await seedCompany()
const result = await getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at)
VALUES (gen_random_uuid(), NULL, $1, 'invoice.paid',
'{"hello":"world"}'::jsonb, '2026-05-12', 'pending', now())
RETURNING id`,
[companyId],
)
expect(result.rowCount).toBe(1)
})
it('rejects INSERT pointing at a non-existent webhook_id', async () => {
// Bypass-check path: the trigger early-returns when the parent
// lookup yields NULL so the FK constraint surfaces the bad
// reference rather than the more-confusing company_match error.
// This test pins the FK-error pathway.
const { companyId } = await seedCompany()
const ghostWebhookId = randomUUID()
await expect(
getPool().query(
`INSERT INTO public.webhook_deliveries
(id, webhook_id, company_id, event_type, payload, api_version,
status, next_attempt_at)
VALUES (gen_random_uuid(), $1, $2, 'invoice.paid',
'{"hello":"world"}'::jsonb, '2026-05-12', 'pending', now())`,
[ghostWebhookId, companyId],
),
).rejects.toThrow(/webhook_deliveries_webhook_id_fkey|foreign key/i)
})
})
+125 -203
View File
@@ -23,7 +23,7 @@
import type { SupabaseClient } from '@supabase/supabase-js'
import { signPayload } from './signing'
import { validateWebhookUrl } from './url-guard'
import { pinnedHttpsFetch, type PinnedFetchResult } from './pinned-fetch'
import { createLogger } from '@/lib/logger'
const log = createLogger('webhooks/dispatcher')
@@ -80,12 +80,12 @@ export async function dispatchDueDeliveries(args: {
batchSize?: number
/** Override for tests. */
now?: Date
/** Override for tests; injected fetch implementation. */
fetchImpl?: typeof fetch
/** Override for tests; injected pinned-fetch implementation. */
pinnedFetchImpl?: typeof pinnedHttpsFetch
}): Promise<DispatchSummary> {
const batchSize = args.batchSize ?? 50
const now = args.now ?? new Date()
const fetchImpl = args.fetchImpl ?? fetch
const pinnedFetchImpl = args.pinnedFetchImpl ?? pinnedHttpsFetch
const summary: DispatchSummary = { picked: 0, delivered: 0, failed: 0, dead: 0 }
@@ -138,7 +138,7 @@ export async function dispatchDueDeliveries(args: {
const outcome = await attemptDelivery({
delivery,
webhook,
fetchImpl,
pinnedFetchImpl,
now,
})
@@ -198,16 +198,12 @@ export async function dispatchDueDeliveries(args: {
*/
async function recoverStuckInFlight(supabase: SupabaseClient, now: Date): Promise<void> {
const stuckBefore = new Date(now.getTime() - 2 * REQUEST_TIMEOUT_MS)
// The status='in_flight' filter alone is not sufficient — a row could
// race between this SELECT and the UPDATE and reach 'delivered' or
// 'dead' in the interim. Postgres applies the status filter to the
// CURRENT (post-race) state, so the row would slip through and the
// immutability trigger would raise check_violation, aborting the
// entire bulk UPDATE and leaving legitimately stuck rows unrecovered.
//
// Defense-in-depth: explicitly exclude terminal status values. The
// partial guard makes a successful sweep on a mixed batch safe even
// when one row terminalized mid-flight.
// Under READ COMMITTED (Postgres default), UPDATE re-evaluates the WHERE
// clause against each row's current value when it acquires the row lock.
// A row that raced from 'in_flight' to 'delivered'/'dead' between scan
// and lock will fail status='in_flight' on re-evaluation and be skipped
// entirely — the immutability trigger never fires, so a mid-flight
// terminal flip cannot abort the bulk update.
const { data, error } = await supabase
.from('webhook_deliveries')
.update({
@@ -216,7 +212,6 @@ async function recoverStuckInFlight(supabase: SupabaseClient, now: Date): Promis
error: 'recovered_from_in_flight_timeout',
})
.eq('status', 'in_flight')
.not('status', 'in', '(delivered,dead)')
.lt('updated_at', stuckBefore.toISOString())
.select('id')
@@ -234,61 +229,26 @@ async function claimDueDeliveries(
batchSize: number,
now: Date,
): Promise<DueDelivery[]> {
// PostgREST cannot express FOR UPDATE SKIP LOCKED through the JS client.
// The cleaner long-term shape is a SQL claim function — tracked for a
// follow-up commit. Until then we SELECT candidate rows, then UPDATE
// with a CAS guard and `.select('id')` to learn which rows the UPDATE
// actually claimed. The dispatch loop runs ONLY against the intersection
// of (selected, claimed) — so an overlapping cron tick that picked up
// the same SELECT can never double-deliver: at most one tick wins the
// CAS update for any given row.
// Atomic FOR UPDATE SKIP LOCKED claim via the SQL function shipped in
// migration 20260515220000. PostgREST can't express SKIP LOCKED through
// the JS client, so the function form is the documented entry point —
// see the migration comment for the full rationale (one round trip,
// no CAS contention, rows locked by a concurrent tick are simply
// invisible to the second caller).
//
// Per-minute Vercel cron has best-effort single-instance semantics, but
// the documented contract is "at-least-once" not "at-most-once" — under
// load (e.g. a 50-row batch with mostly slow receivers > 60s) the next
// tick can fire while this one is still running, so the CAS-then-
// intersect pattern is load-bearing, not defensive.
const { data, error } = await supabase
.from('webhook_deliveries')
.select('id, webhook_id, company_id, event_type, payload, previous_attributes, api_version, attempts')
.in('status', ['pending', 'failed'])
.lte('next_attempt_at', now.toISOString())
// Skip dangling rows (webhook deleted between enqueue and dispatch).
// The webhook_deliveries.webhook_id FK is ON DELETE SET NULL
// (migration 20260515170000) so terminal rows survive webhook deletion
// for BFNAR 2013:2 kap 8 § audit retention; non-terminal rows for a
// deleted webhook have no receiver to deliver to and stay dormant in
// the audit trail.
.not('webhook_id', 'is', null)
.order('next_attempt_at', { ascending: true })
.limit(batchSize)
// All filter semantics from the previous JS path are preserved inside
// the function: status IN ('pending','failed'), next_attempt_at <= now,
// webhook_id IS NOT NULL, ORDER BY next_attempt_at ASC, LIMIT batchSize.
const { data, error } = await supabase.rpc('claim_due_webhook_deliveries', {
p_batch_size: batchSize,
p_now: now.toISOString(),
})
if (error || !data) {
log.error('claim due deliveries failed', error as Error)
if (error) {
log.error('claim_due_webhook_deliveries rpc failed', error as Error)
return []
}
if (data.length === 0) return []
const candidates = data as DueDelivery[]
const candidateIds = candidates.map((d) => d.id)
const { data: claimed, error: updateErr } = await supabase
.from('webhook_deliveries')
.update({ status: 'in_flight' })
.in('id', candidateIds)
.in('status', ['pending', 'failed']) // CAS guard
.select('id')
if (updateErr) {
log.error('claim deliveries update failed', updateErr as Error)
return []
}
// Trust the UPDATE's returned set as authoritative — anything not in
// `claimed` was lost to a competing tick (or had its status flipped
// out from under us between SELECT and UPDATE).
const claimedIds = new Set(((claimed ?? []) as { id: string }[]).map((r) => r.id))
return candidates.filter((d) => claimedIds.has(d.id))
return (data ?? []) as DueDelivery[]
}
async function loadWebhooksByIds(
@@ -439,10 +399,10 @@ type AttemptOutcome = DeliveredOutcome | FailedOutcome | DeadOutcome
async function attemptDelivery(args: {
delivery: DueDelivery
webhook: WebhookForDelivery
fetchImpl: typeof fetch
pinnedFetchImpl: typeof pinnedHttpsFetch
now: Date
}): Promise<AttemptOutcome> {
const { delivery, webhook, fetchImpl, now } = args
const { delivery, webhook, pinnedFetchImpl, now } = args
const attempts = delivery.attempts + 1
const requestId = `whdel_${delivery.id}`
@@ -455,135 +415,109 @@ async function attemptDelivery(args: {
previous_attributes: delivery.previous_attributes,
})
// Re-validate the URL at dispatch time as defense in depth — DNS records
// can change between webhook creation and dispatch (DNS rebinding,
// hijack, A-record swap to internal IP), so the create-time check alone
// is insufficient. A failure here marks the delivery dead with a
// distinct reason so the operator can investigate without thinking it's
// a transient receiver issue.
const urlCheck = await validateWebhookUrl(webhook.webhook_url)
if (!urlCheck.ok) {
return {
kind: 'dead',
reason: `url_unsafe:${urlCheck.reason}`,
disableWebhook: true,
attempts,
responseStatus: null,
responseBody: null,
responseHeaders: null,
error: urlCheck.detail,
}
}
const { header } = signPayload({
body,
secret: webhook.secret,
timestamp: Math.floor(now.getTime() / 1000),
})
const { header } = signPayload({ body, secret: webhook.secret, timestamp: Math.floor(now.getTime() / 1000) })
// pinnedHttpsFetch performs DNS validation AND opens the socket against
// the validated IP in a single call. The previous shape (separate
// validateWebhookUrl + fetch calls) left a DNS-rebinding window between
// the two — closed here. SNI + Host header continue to carry the
// original hostname so receiver-side TLS + vhost routing still work.
const result = await pinnedFetchImpl(webhook.webhook_url, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'X-Gnubok-Signature': header,
'X-Gnubok-Event': delivery.event_type,
'X-Gnubok-Delivery': delivery.id,
'X-Gnubok-Api-Version': delivery.api_version,
'X-Request-Id': requestId,
'User-Agent': 'gnubok-webhook/1',
},
body,
timeoutMs: REQUEST_TIMEOUT_MS,
maxResponseBytes: MAX_RESPONSE_BODY_BYTES,
})
const controller = new AbortController()
const timeout = setTimeout(() => controller.abort(), REQUEST_TIMEOUT_MS)
let response: Response
try {
response = await fetchImpl(webhook.webhook_url, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'X-Gnubok-Signature': header,
'X-Gnubok-Event': delivery.event_type,
'X-Gnubok-Delivery': delivery.id,
'X-Gnubok-Api-Version': delivery.api_version,
'X-Request-Id': requestId,
'User-Agent': 'gnubok-webhook/1',
},
body,
signal: controller.signal,
// Reject 3xx responses entirely. Following a redirect would let a
// receiver bounce the dispatcher to a private/internal address
// AFTER the SSRF guard (which validated the original webhook_url's
// hostname) has cleared. Receivers that legitimately move endpoints
// should ask integrators to update the webhook URL.
redirect: 'error',
})
} catch (err) {
clearTimeout(timeout)
const message = err instanceof Error ? err.message : String(err)
// Distinguish redirect-rejection errors from generic transport
// failures. With redirect: 'error' the runtime fetch throws when the
// receiver returns 3xx — that's an SSRF-bypass attempt (or a
// misconfigured receiver), not a transient failure. Treating it as
// 'failed' would burn 8 retry attempts over ~72h before going dead.
// Mirror the HTTP 410 treatment: terminal + auto-disable so the
// operator surfaces the misbehaving receiver immediately.
//
// Node's undici (the runtime fetch) raises 'unexpected redirect'
// / 'redirect mode is set to error' messages; check both shapes
// since the exact wording has changed across Node versions.
const isRedirectError = /redirect/i.test(message)
if (isRedirectError) {
switch (result.kind) {
case 'unsafe_url':
return {
kind: 'dead',
reason: 'redirect_blocked',
reason: `url_unsafe:${result.reason}`,
disableWebhook: true,
attempts,
responseStatus: null,
responseBody: null,
responseHeaders: null,
error: message.length > 500 ? `${message.slice(0, 497)}...` : message,
error: result.detail,
}
case 'redirect_blocked':
return {
kind: 'dead',
reason: 'redirect_blocked',
disableWebhook: true,
attempts,
responseStatus: result.status,
responseBody: null,
responseHeaders: null,
error: truncateError(result.detail),
}
case 'timeout':
case 'transport_error':
return {
kind: 'failed',
attempts,
responseStatus: null,
responseBody: null,
responseHeaders: null,
error: truncateError(result.detail),
}
case 'ok': {
const responseHeaders = filterResponseHeaders(result.headers)
const responseBody = isSafeContentType(result.headers['content-type'] ?? '')
? result.body
: null
// HTTP 410 — receiver explicitly asks us to stop. Auto-disable.
if (result.status === 410) {
return {
kind: 'dead',
reason: 'http_410_gone',
disableWebhook: true,
attempts,
responseStatus: 410,
responseBody,
responseHeaders,
}
}
if (result.status >= 200 && result.status < 300) {
return {
kind: 'delivered',
attempts,
responseStatus: result.status,
responseBody,
responseHeaders,
}
}
return {
kind: 'failed',
attempts,
responseStatus: result.status,
responseBody,
responseHeaders,
error: `HTTP ${result.status}`,
}
}
return {
kind: 'failed',
attempts,
responseStatus: null,
responseBody: null,
responseHeaders: null,
error: message.length > 500 ? `${message.slice(0, 497)}...` : message,
}
}
}
// Keep the abort timeout armed across the body read — a slow body
// stream can stall the entire dispatch batch otherwise. Clear only
// after readBoundedText returns (or aborts).
let responseBody: string | null
try {
responseBody = await readBoundedText(response)
} finally {
clearTimeout(timeout)
}
const responseHeaders = headersToObject(response.headers)
// HTTP 410 — receiver explicitly asks us to stop. Auto-disable the
// webhook + mark this delivery dead.
if (response.status === 410) {
return {
kind: 'dead',
reason: 'http_410_gone',
disableWebhook: true,
attempts,
responseStatus: 410,
responseBody,
responseHeaders,
}
}
if (response.status >= 200 && response.status < 300) {
return {
kind: 'delivered',
attempts,
responseStatus: response.status,
responseBody,
responseHeaders,
}
}
return {
kind: 'failed',
attempts,
responseStatus: response.status,
responseBody,
responseHeaders,
error: `HTTP ${response.status}`,
}
function truncateError(message: string): string {
return message.length > 500 ? `${message.slice(0, 497)}...` : message
}
// Content-Type prefixes for which we persist response_body verbatim. Other
@@ -593,21 +527,9 @@ async function attemptDelivery(args: {
// when the operator can see the response_status and response_headers.
const SAFE_BODY_CONTENT_TYPE_PREFIXES = ['text/plain', 'application/json']
async function readBoundedText(response: Response): Promise<string | null> {
const contentType = response.headers.get('content-type')?.toLowerCase() ?? ''
const isSafe = SAFE_BODY_CONTENT_TYPE_PREFIXES.some((p) => contentType.startsWith(p))
if (!isSafe) {
// Drain the body so the connection can be reused, but discard the bytes.
try { await response.text() } catch { /* ignore */ }
return null
}
try {
const text = await response.text()
if (text.length <= MAX_RESPONSE_BODY_BYTES) return text
return text.slice(0, MAX_RESPONSE_BODY_BYTES)
} catch {
return null
}
function isSafeContentType(contentType: string): boolean {
const lower = contentType.toLowerCase()
return SAFE_BODY_CONTENT_TYPE_PREFIXES.some((p) => lower.startsWith(p))
}
// Allowlist for response_headers persistence. Receiver-side headers like
@@ -627,13 +549,13 @@ const SAFE_RESPONSE_HEADERS = new Set([
'cf-ray',
])
function headersToObject(headers: Headers): Record<string, string> {
function filterResponseHeaders(headers: Record<string, string>): Record<string, string> {
const obj: Record<string, string> = {}
headers.forEach((v, k) => {
for (const [k, v] of Object.entries(headers)) {
if (SAFE_RESPONSE_HEADERS.has(k.toLowerCase())) {
obj[k] = v
}
})
}
return obj
}
+275
View File
@@ -0,0 +1,275 @@
/**
* Pinned-IP HTTPS POST for webhook dispatch.
*
* Closes the DNS-rebinding window between url-guard validation and the
* actual HTTPS request. The previous shape was:
*
* 1. validateWebhookUrl() → DNS resolves to [public IP], returns ok
* 2. fetch(webhook_url) → re-resolves DNS; an attacker who flipped
* the A record in the interval gets a
* private-IP socket
*
* The new shape pins the request to the IP validated in step 1, with the
* original hostname carried in:
* - the TLS SNI extension (so the receiver's cert continues to match)
* - the HTTP Host header (so vhost routing on the receiver continues to
* work)
*
* The request socket therefore never re-resolves DNS, foreclosing the
* rebind race. Documented openly per the url-guard.ts file header
* ("closing that requires a custom HTTPS agent that pins the resolved IP").
*
* Built on `node:https.request` rather than undici's Agent because (a) the
* project doesn't take a dependency on undici, (b) the stdlib API is more
* explicit about the SNI / Host / IP split, (c) https.request is enough
* for HTTP/1.1 + TLS, which every webhook receiver supports.
*
* Inversion seam: `httpsRequest` injectable for tests so we don't need to
* stand up an HTTPS server to verify the pinning / SNI / Host shape. The
* dispatcher's tests pass a stub through `pinnedFetchImpl`.
*/
import {
request as httpsRequestDefault,
type RequestOptions as HttpsRequestOptions,
} from 'node:https'
import type { ClientRequest, IncomingMessage } from 'node:http'
import { validateWebhookUrl as validateWebhookUrlDefault } from './url-guard'
export type PinnedFetchResult =
| {
kind: 'ok'
status: number
headers: Record<string, string>
body: string
bodyTruncated: boolean
pinnedAddress: string
}
| { kind: 'unsafe_url'; reason: string; detail: string; pinnedAddress: null }
| { kind: 'redirect_blocked'; status: number; detail: string; pinnedAddress: string }
| { kind: 'timeout'; detail: string; pinnedAddress: string }
| { kind: 'transport_error'; detail: string; pinnedAddress: string | null }
export interface PinnedFetchInit {
method: string
headers: Record<string, string>
body: string
timeoutMs: number
/** Max bytes captured from response body — receivers returning long error pages get truncated. */
maxResponseBytes: number
}
export interface PinnedFetchDeps {
/** DNS validation seam. Defaults to url-guard's validateWebhookUrl. */
validateUrl?: typeof validateWebhookUrlDefault
/** Raw HTTPS request seam. Defaults to node:https.request. */
httpsRequest?: (
options: HttpsRequestOptions,
callback: (res: IncomingMessage) => void,
) => ClientRequest
}
export async function pinnedHttpsFetch(
rawUrl: string,
init: PinnedFetchInit,
deps: PinnedFetchDeps = {},
): Promise<PinnedFetchResult> {
const validateUrl = deps.validateUrl ?? validateWebhookUrlDefault
const httpsRequest = deps.httpsRequest ?? httpsRequestDefault
let parsed: URL
try {
parsed = new URL(rawUrl)
} catch {
return {
kind: 'unsafe_url',
reason: 'invalid_url',
detail: 'URL did not parse.',
pinnedAddress: null,
}
}
const validation = await validateUrl(rawUrl)
if (!validation.ok) {
return {
kind: 'unsafe_url',
reason: validation.reason,
detail: validation.detail,
pinnedAddress: null,
}
}
// Pick the first vetted address. validateWebhookUrl rejects the whole
// set when ANY entry is unsafe, so the first is safe by construction.
// Deterministic choice keeps log output stable across retries.
const pinnedAddress = validation.resolvedAddresses[0]
if (!pinnedAddress) {
// Defensive — validateWebhookUrl returns ok only when there's at least
// one address, but a future refactor could regress this and we want
// the failure to be loud, not a silent DNS-lookup-by-empty-host.
return {
kind: 'transport_error',
detail: 'No resolved address from validateWebhookUrl',
pinnedAddress: null,
}
}
const port = parsed.port ? Number(parsed.port) : 443
return new Promise<PinnedFetchResult>((resolve) => {
let settled = false
const settle = (r: PinnedFetchResult) => {
if (settled) return
settled = true
resolve(r)
}
// The HTTP Host header must carry the original hostname (vhost routing
// on the receiver). Include the port only when non-default — RFC 7230
// §5.4 says the port is omitted when it matches the scheme default.
const hostHeader = port === 443 ? parsed.hostname : `${parsed.hostname}:${port}`
const requestOptions: HttpsRequestOptions = {
protocol: 'https:',
// Pin the socket to the validated IP. node:https accepts the
// address directly — no further DNS lookup happens.
host: pinnedAddress,
port,
path: parsed.pathname + parsed.search,
method: init.method,
// SNI carries the original hostname so the receiver's TLS cert
// (which is issued for the hostname, not the IP) validates.
//
// Cert-vs-hostname verification: Node's default checkServerIdentity
// matches the cert's SAN/CN against `servername` (or `host` when
// servername is unset). Because `servername` is set to the original
// hostname, the IP substitution above does NOT weaken the hostname-
// verification step — a forged endpoint at the pinned IP presenting
// a valid cert for a DIFFERENT hostname would fail the handshake.
// No explicit checkServerIdentity override is needed; relying on
// the default is the documented contract.
servername: parsed.hostname,
headers: {
...init.headers,
// Lowercase 'host' — Node's https.request would synthesise one
// from `host` (the pinned IP) if we didn't set it explicitly,
// which would break vhost routing on the receiver.
host: hostHeader,
},
// Fresh socket per call — webhook delivery doesn't benefit from
// Keep-Alive (the dispatcher serializes and the IP changes per
// dispatch from re-validation). agent:false also forecloses any
// accidental pool-level reuse across pinned IPs.
agent: false,
}
let absoluteTimer: NodeJS.Timeout | null = null
const req = httpsRequest(requestOptions, (res) => {
// Receivers MUST return a non-redirect. Following a 3xx would let
// them bounce the dispatcher to a private address AFTER the SSRF
// guard cleared. We don't follow redirects; treat as terminal here
// and let the dispatcher mark the row dead with reason='redirect_
// blocked' for consistency with the old fetch path's behavior.
const status = res.statusCode ?? 0
if (status >= 300 && status < 400) {
// Drain body so the socket cleans up; ignore errors.
res.resume()
req.destroy()
if (absoluteTimer) clearTimeout(absoluteTimer)
return settle({
kind: 'redirect_blocked',
status,
detail: `Receiver returned ${status}; redirects are refused.`,
pinnedAddress,
})
}
const chunks: Buffer[] = []
let total = 0
let truncated = false
res.on('data', (chunk: Buffer) => {
if (truncated) return
if (total + chunk.length > init.maxResponseBytes) {
const remaining = init.maxResponseBytes - total
if (remaining > 0) chunks.push(chunk.subarray(0, remaining))
total = init.maxResponseBytes
truncated = true
// Destroy the stream — no point pulling the rest over the wire.
res.destroy()
} else {
chunks.push(chunk)
total += chunk.length
}
})
const finalize = () => {
if (absoluteTimer) clearTimeout(absoluteTimer)
const headers: Record<string, string> = {}
for (const [k, v] of Object.entries(res.headers)) {
if (typeof v === 'string') headers[k] = v
else if (Array.isArray(v)) headers[k] = v.join(', ')
}
settle({
kind: 'ok',
status,
headers,
body: Buffer.concat(chunks).toString('utf8'),
bodyTruncated: truncated,
pinnedAddress,
})
}
// Two completion paths to handle: 'end' (normal completion) and
// 'close' (when we destroyed the stream for size truncation, where
// 'end' does not fire). Node emits BOTH 'end' and 'close' on normal
// completions, so `once()` + a self-removing pair keeps finalize
// single-shot without relying on the outer `settled` guard to
// squash duplicate header reconstruction.
const finalizeOnce = () => {
res.removeListener('end', finalizeOnce)
res.removeListener('close', finalizeOnce)
finalize()
}
res.once('end', finalizeOnce)
res.once('close', finalizeOnce)
res.on('error', (err) => {
if (absoluteTimer) clearTimeout(absoluteTimer)
settle({ kind: 'transport_error', detail: err.message, pinnedAddress })
})
})
// Two-layer timeout: socket-idle timeout via Node's built-in, plus a
// wall-clock absolute timeout. node:https `timeout` is idle-only and
// wouldn't fire if a slow receiver dribbles bytes; the absolute timer
// is the hard cap.
req.setTimeout(init.timeoutMs)
req.on('timeout', () => {
req.destroy()
if (absoluteTimer) clearTimeout(absoluteTimer)
settle({
kind: 'timeout',
detail: `Socket idle for ${init.timeoutMs} ms`,
pinnedAddress,
})
})
absoluteTimer = setTimeout(() => {
req.destroy()
settle({
kind: 'timeout',
detail: `Request exceeded ${init.timeoutMs} ms wall-clock`,
pinnedAddress,
})
}, init.timeoutMs)
req.on('error', (err) => {
if (absoluteTimer) clearTimeout(absoluteTimer)
settle({ kind: 'transport_error', detail: err.message, pinnedAddress })
})
if (init.body) req.write(init.body)
req.end()
})
}