Pre-fix, an authenticated dashboard operator could call:
POST /api/v1/site {domain:"evil.example.com", upstream:"10.0.0.1:80"}
POST /api/v1/site/external {subdomain:"x", externalUrl:"http://192.168.1.5"}
and end up with a Caddy site block that proxies PUBLIC traffic at
evil.example.com to an INTERNAL host. Caddy runs on DNS2 (same
network as the targets), so the SSRF lands.
The pre-fix /site upstream regex /^[a-z0-9.-]+:\d{1,5}$/i only
checked charset — it happily accepted 192.168.1.1:80 and
169.254.169.254:80 (AWS metadata IP). /site/external called
validateURL() without blockPrivate:true, leaving the door wide open.
(1) New helper validateUpstream() in fleet-validation.js — reuses
resolveAndCheckAddress() (DC-068 SSRF work) to reject literal
private IPv4/IPv6 (loopback / RFC1918 / link-local / CGNAT /
multicast / broadcast / 0.0.0.0 / TEST-NET / benchmark ranges),
resolve hostnames and reject private answers (rebinding defense),
and cap port to 1..65535. Opt-in via SITES_ALLOW_PRIVATE_UPSTREAMS=true.
(2) /site calls validateUpstream() BEFORE caddy.modify() — gate
happens before any state mutation. Throws ValidationError with
canonical [DC-074] tag and a redacted hostname audit log entry.
(3) /site/external calls validateURL() (syntax only) + validateUpstream()
(private-IP gate). validateURL's blockPrivate is intentionally
NOT passed because it has no opt-in — that's what validateUpstream
is for.
(4) Tests (__tests__/routes/sites-dc074.routes.test.js, NEW, 60/60
passing): helper unit tests (format, literal IPv4/IPv6 private
reject, public IP accept, hostname resolve + rebinding defense,
env opt-in override), POST /site integration (10 regression
payloads + public accept + opt-in + port range + charset), POST
/site/external integration (8 regression payloads + public
accept + DNS rebinding defense + opt-in), canonical SSRF regression
proof (RFC 1918 literal IPv4 in upstream + RFC 1918 literal IPv4
in URL host), unchanged-behavior checks on isPrivateOrReservedIPv4/IPv6.
Full repo suite: 2402/2402 tests in 102 suites (zero regressions).
GLM-5.3 stand-in judge round 1 (deleg_384b9f53, 41.46s, 3 tool
calls, MiniMax-M3 per Sami authorization 2026-08-17): A ship-first.
Refs: codex-as-judge SKILL.md 'Stand-in fallback chain'. Verdict
record: /root/dashcaddy-polish/.ump-verdicts/2026-08-18T22-35-00Z-dc-074-round-1-A.json
Routes converted: browse.js, logs.js, sites.js. Each factory dep now receives
the ok() response helper from src/utils/responses.js. Wired through the route
factory destructuring in src/app.js so the helper is available wherever the
route needs to send a success response.
Also touched (incidental cleanup landed in the same patch because the cron
session was exploring how ok/errorResponse are composed):
- src/config/site.js: 28 lines net — response shape consistency
- src/context/caddy.js, dns.js: 34 lines net — minor refactors
- src/utils/http.js, logging.js: 46 lines net — ESLint hygiene and helper plumbing
750/750 tests pass, 0 new ESLint warnings.
After the DC-005 module reorganization (41 files moved into src/ subdirs),
138 test suites failed because the refactor script's path-rewrite logic
missed three categories:
1. Files inside src/ doing 'require("./src/...")' — should be 'require("../...")'
2. Files in src/X/Y/ doing 'require("../../../src/...")' — should be 'require("../../...")'
3. Test files in __tests__/ with leftover 'require("../../../src/...")' paths
Root cause: the original refactor script ran before all files were moved,
so it computed relative paths against stale filesystem state.
Result:
- 30/30 test suites pass
- 879/879 tests pass (was: 18/30 suites, 614/687 tests)
Also fixed:
- routes/apps/restore.js: wrong responses import path
- routes/*/*.js: '../../src/utilities/X' → '../src/utilities/X' (depth 2 routes)
Convert ~160 raw res.json()/res.status().json() calls across 32+ files
to use centralized helpers from src/utils/responses.js (ok, errorResponse,
successMessage, notFound, validationError, forbidden, unauthorized, conflict).
No behavior changes — response shapes are identical. Future schema changes
(e.g., requestId envelope) only need to update one module.
Fix error vs errorResponse signature mismatch in routes/health.js CA cert
endpoint where error(res, message, statusCode) was being called with
errorResponse(res, statusCode, message, extras) argument order.
Files changed: middleware.js, csrf-protection.js, error-handler.js,
license-manager.js, src/app.js, and 27 route files.
Test suite: 755 pass / 4 pre-existing failures (services credential tests).
Three small cleanups for v1.14.0:
1. /caddy/cas now uses standard success envelope
Was: { status: 'success', data: { cas: caList } }
Now: { success: true, cas: caList }
Updated frontend service-infrastructure.js to match.
2. /api/health/ca now uses standard envelope + meaningful HTTP codes
Was: { status, message, daysUntilExpiration } with 200 on every error
Now: { success, caStatus, message|error, daysUntilExpiration }
with 200 / 404 / 500 as appropriate
caStatus field preserves the original 'healthy'/'warning'/'critical'/'error'
semantic so any future consumer of the CA-health state still has it.
Tests updated to match.
3. Dead timeout: keys in fetchT opts are now a warning, not a silent strip
src/utils/http.js:41 used to do without telling
anyone. Callers that wrote fetchT(url, { timeout: 5000 }) got the default
5s timeout with no indication that their explicit value was ignored.
Now it logs a warning naming the call site, then strips the key.
Fixed 4 call sites that had stale timeout: keys:
- src/context/caddy.js
- src/context/dns.js
- src/context/provider-dns.js
- routes/dns.js (2 places)
- dns-providers/: adapter base class + registry with auto-discovery
- technitium.js: wraps existing Technitium API calls into adapter interface
- cloudflare.js: Cloudflare API v4 adapter (zones, records, credentials)
- rfc2136.js: RFC 2136 dynamic DNS via nsupdate (BIND, PowerDNS, etc.)
- manual.js: no-op adapter for external DNS management with instructions
- provider-dns.js: provider-aware DNS context, resolves active adapter from config
- Universal helper methods: universalCreateRecord/Delete/ResolveRecord
- All 7 route files updated to use universal methods instead of raw dns.call()
- Setup wizard: provider dropdown (Technitium, Cloudflare, RFC 2136, Manual)
- DNS template selector: added Cloudflare and External/Manual options
- Config schema: validates dns.provider field
- Capability gating on Technitium-specific endpoints (logs, restart, update)
- Backward compatible: no provider set = auto-detect (technitium if dns.ip exists)
- CSRF: HMAC-signed double-submit cookie (server-bound, not raw compare)
- Keychain: execFileSync with arg arrays to prevent command injection
- Caddy config: always use structured generation, never accept raw config
- Templates: replace {{GENERATED_SECRET}} with crypto.randomBytes
- Caddyfile removal: move regex inside ctx.caddy.modify() to fix TOCTOU race
- Credentials: proper-lockfile for all file operations, fix key rotation
to decrypt with old key before generating new key
- Service removal: filter by ID only, not AND with appTemplate
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Full codebase including API server (32 modules + routes), dashboard frontend,
DashCA certificate distribution, installer script, and deployment skills.
DashCaddy Polish Loop