Files
TREK/wiki/Internal-Network-Access.md
T
Maurice 0887f251ce fix(ssrf): connect past a link-local DNS record instead of refusing the host (#2506)
Since 4.3.0 the guard resolves a name to every address it has, and a
single blocked address refused the whole name. LAN DNS servers commonly
publish a host's fe80:: address next to its IPv4, so a self-hosted IdP on
such a name failed every login with "Requests to link-local /
cloud-metadata addresses are not allowed". Immich, Synology and the other
strict-guard integrations on the same kind of name failed the same way.

Both guards now leave a blocked address out of the list the socket is
pinned to and connect over the rest, and a name that resolves to nothing
else is still refused. A private address among the remaining ones still
needs ALLOW_INTERNAL_NETWORK under the strict guard, and
ALLOW_LINK_LOCAL_IPS works as before.

Leaving addresses out is only safe if the blocked set is complete, so the
strict guard now blocks everything the admin lane does. The AWS IMDSv6
(fd00:ec2::/32) and Alibaba (100.100.100.200, 100.100.100.100) metadata
addresses sit in ULA and CGNAT space, which ALLOW_INTERNAL_NETWORK opened
to any URL a user can type. Both guards also read an IPv4-mapped or
IPv4-compatible address as the IPv4 it carries: Node prints an AAAA
record of ::a9fe:a9fe as ::169.254.169.254, which passed both guards, and
::ffff:100.100.100.200 passed the admin lane. Neither guard connects to a
link-local or metadata address now, and the strict guard never connects
to loopback either.
2026-09-26 02:26:39 +00:00

8.9 KiB

Internal Network Access

TREK makes outbound HTTP requests when you configure integrations such as Immich or Synology Photos. By default, it blocks requests to private and local IP ranges to prevent server-side request forgery (SSRF) attacks. You need to allow internal network access when those services are hosted on your LAN.

Default behavior

TREK has two SSRF guards, both in ssrfGuard.ts. Which one applies depends on the call site, not on who configured the URL.

The strict guard (safeFetch / safeFetchFollow, built on checkSsrf) covers most outbound traffic: Immich, Synology Photos, AirTrail, Dawarich, the document stores behind Document-Sync (Paperless-ngx, Papra, Nextcloud, OpenCloud and Synology), notification webhooks, ntfy, Unsplash, and place lookups. It resolves the hostname to an IP address before allowing the connection and blocks loopback, link-local and private ranges. Only the private ranges open up, and only with ALLOW_INTERNAL_NETWORK=true. The two tables below describe this guard.

The relaxed guard (safeFetchAdminConfigured, also exported as safeFetchLlm) covers endpoints that are expected to live on your own network: OIDC (discovery, token, userinfo, JWKS), the LLM providers behind the AI Parsing addon (a local Ollama or any OpenAI-compatible endpoint), and plugin OAuth token exchanges. The self-hosted routing engines an admin sets for the Road-Trip addon (Own routing engine, Own Valhalla instance) go through it too when the server asks them itself, which it does for the MCP road trip tools. The planner asks them from the browser, so they must also be reachable from your users' devices. It deliberately permits loopback and LAN targets, so a model server on localhost or an identity provider on your LAN works without ALLOW_INTERNAL_NETWORK. It still resolves every hostname, re-checks every redirect hop, and always blocks link-local and cloud-metadata addresses (169.254.0.0/16, the full fe80::/10, and the AWS and Alibaba metadata addresses). The only way past is a single address listed in ALLOW_LINK_LOCAL_IPS, see below.

No guard applies to the addresses an admin sets in the environment for place search: TREK_PLACES_URL, NOMINATIM_URL and OVERPASS_URL. They are configuration rather than user input, so a self-run copy of the TREK Places API, a Nominatim or an Overpass instance on your LAN or in the same Docker network is reached without ALLOW_INTERNAL_NETWORK.

Always blocked

Under the strict guard, these ranges are blocked whatever ALLOW_INTERNAL_NETWORK says:

Range Description
127.0.0.0/8, ::1 Loopback
0.0.0.0/8, :: Unspecified
169.254.0.0/16, fe80::/10 Link-local / cloud metadata endpoints
fd00:ec2::/32, 100.100.100.200, 100.100.100.100 AWS (IMDSv6) and Alibaba Cloud metadata. They sit inside the ULA and CGNAT ranges below and stay blocked when those open up
IPv4-mapped, IPv4-compatible and NAT64/6to4/Teredo forms of the above e.g. ::ffff:127.0.0.1, ::169.254.169.254, 64:ff9b::a9fe:a9fe

The IPv6 link-local rule covers the whole fe80::/10 prefix (fe80: to febf:). The link-local and metadata rows hold under the relaxed guard as well.

The one way past this table is ALLOW_LINK_LOCAL_IPS, for a single IPv4 link-local address, see below.

Blocked unless ALLOW_INTERNAL_NETWORK=true

Range / Hostname Description
10.0.0.0/8 RFC-1918 private
172.16.0.0/12 RFC-1918 private
192.168.0.0/16 RFC-1918 private
100.64.0.0/10 CGNAT / Tailscale shared address space
fc00::/7 IPv6 ULA
IPv4-mapped RFC-1918 variants e.g. ::ffff:10.x, ::ffff:192.168.x
*.local, *.internal, localhost hostnames mDNS / internal DNS suffixes (e.g. Docker service names, LAN hosts) and the literal localhost

The hostname localhost is matched at the hostname stage too, but it normally resolves to a loopback address (127.0.0.1 or ::1), which the always-blocked loopback rule catches first, so under the strict guard it is blocked no matter how ALLOW_INTERNAL_NETWORK is set. On a host that maps localhost somewhere else, the hostname rule still applies and it stays blocked unless ALLOW_INTERNAL_NETWORK=true. The relaxed guard allows localhost outright, which is what makes a local Ollama the supported default for AI Parsing.

*.local and *.internal hostnames are permitted when ALLOW_INTERNAL_NETWORK=true: the guard still resolves them to an IP and enforces all IP-level rules, so TREK never connects to a loopback or link-local address such a hostname resolves to, and a hostname with no other address remains blocked regardless.

When to enable

Set ALLOW_INTERNAL_NETWORK=true when a service reached through the strict guard (Immich, Synology Photos, AirTrail, Dawarich, a document store bound through Document-Sync, or a notification webhook) is hosted on your local network and you need TREK to reach it. You do not need it for a local or LAN Ollama, an OpenAI-compatible endpoint, an OIDC provider on your LAN, your own OSRM or Valhalla, or plugin OAuth; those go through the relaxed guard and already work. Nor do you need it for your own TREK Places API, Nominatim or Overpass, which are not guarded at all. Leave the flag off if only those need internal access, since turning it on widens the surface for every strict-guard integration at once.

A service in another container on the same Docker network counts as internal too, because its container name resolves to a private address; so do a Tailscale address (100.64.0.0/10) and a .local name. A service on loopback (localhost, 127.0.0.1) is never reached through the strict guard, whatever the flag says: use the host's LAN address or the container name instead.

See Environment-Variables for how to set environment variables.

Admin: Set ALLOW_INTERNAL_NETWORK=true in Environment-Variables before configuring Immich, Synology Photos, AirTrail, Dawarich or a document store on a LAN.

A rootless Podman container reaches its host through 169.254.1.2: AddHost=keycloak.example.com:host-gateway writes that address into the container's /etc/hosts. With the identity provider, or the reverse proxy in front of it, on the host, the OIDC login then fails with [OIDC] Login error: Requests to link-local / cloud-metadata addresses are not allowed, because both guards block 169.254.0.0/16.

Name that one address in ALLOW_LINK_LOCAL_IPS:

ALLOW_LINK_LOCAL_IPS=169.254.1.2
  • The relaxed guard (OIDC, AI Parsing, plugin OAuth, your own routing engines) then reaches it with nothing else set.
  • The strict guard treats it as internal, like a 192.168.x address, so Immich, a document store or any other integration behind it also needs ALLOW_INTERNAL_NETWORK=true.
  • Every other link-local address stays blocked. 169.254.169.x and 169.254.170.x, where AWS, GCP, Azure and the container services built on them hand out instance metadata and credentials, cannot be listed at all. TREK refuses to start with one of them in the list, and with any entry that is not a single IPv4 address such as 169.254.1.2, so IPv6 link-local (fe80::/10) cannot be listed either.

Several addresses are separated by commas. The list is read at startup, so restart TREK after changing it.

DNS rebinding protection

Even with ALLOW_INTERNAL_NETWORK=true, TREK pins the DNS resolution to prevent rebinding attacks. When the guard checks a URL, it resolves the hostname once and records the addresses it may use. The outbound connection is then made directly to those addresses using a pinned dispatcher (via undici), so the hostname cannot re-resolve to a different address between the check and the actual request.

A hostname can resolve to several addresses, and a LAN DNS server often hands out a host's IPv6 link-local address (fe80::) next to its IPv4 one. TREK never connects to an address the guard blocks: it leaves that address out and connects over the ones that remain. Only a hostname whose addresses are all blocked is refused. Under the strict guard, a private address among the rest still needs ALLOW_INTERNAL_NETWORK=true.

Audit log

When a user saves an Immich URL that resolves to a private IP, TREK records an immich.private_ip_configured entry in the Audit-Log including the URL and the resolved IP address. AirTrail does the same with airtrail.private_ip_configured, and Dawarich with dawarich.private_ip_configured. Synology Photos does not emit an equivalent event, and neither does document sync.

See also