tools/windows: scoped DNS resolver for the FIFA 17 roster hostname

FIFA 17's ProtoSSL verifies the roster certificate by dNSName only, so the
client must reach the roster as winter15.gosredirector.ea.com. Retested on
Windows 2026-08-23: an IP-addressed roster host is refused even though the
certificate carries IP Address:10.10.0.120 as a SAN.

Public DNS points that name at EA's dead 159.153.51.20, so the client has to
resolve it to us. scoped-dns.py pins exactly that one name and forwards every
other query upstream verbatim, so a client pointed at it cannot lose general
resolution -- verified against www.microsoft.com, github.com and
www.msftconnecttest.com.

Paired with a Windows NRPT rule rather than a hosts entry: per-name, auditable
via Get-DnsClientNrptRule, and revertible in one command. A hosts edit on this
machine had previously taken its whole internet down.
This commit is contained in:
funman300
2026-08-23 01:34:13 +00:00
parent 8c353c6c66
commit 286a44461d
2 changed files with 180 additions and 0 deletions
+35
View File
@@ -85,6 +85,41 @@ Copy-Item 'C:\FIFA 17\version.dll.stale-849k.bak' 'C:\FIFA 17\version.dll' -Forc
Always keep a `*.bak` of the live hook before redeploying (the preflight checks
that a rollback backup exists and differs from the live DLL).
## Roster / "FUT Squad Update" — the client MUST reach the roster by HOSTNAME
If FUT fails with **"An error occurred downloading the FUT Squad Update"**, the
client could not fetch `https://<roster>/fifa17/fut/rosterupdate.xml`.
FIFA 17's ProtoSSL verifies that certificate by **dNSName only**. Pointing the
roster at an IP does **not** work even though our certificate carries
`IP Address:10.10.0.120` in its SANs — this was retested on Windows on
2026-08-23 and rejected. Do not retry an IP roster host, and do not reissue the
certificate for an IP SAN. The roster must be reached as
`winter15.gosredirector.ea.com`, which our certificate does carry as a dNSName.
Public DNS resolves that name to EA's dead `159.153.51.20`, so the client has to
be told to resolve it to us. **Use an NRPT rule, not the hosts file.** A hosts
edit on this machine previously took its entire internet down; NRPT is per-name,
auditable with `Get-DnsClientNrptRule`, and reverts in one command.
On the server, run the scoped resolver (needs root for UDP 53):
```bash
sudo python3 tools/windows/scoped-dns.py
```
It answers **only** `winter15.gosredirector.ea.com` and forwards every other
query upstream verbatim, so it cannot strand a client that is pointed at it.
Then start Blaze with `OPENFUT_ROSTER_HOST=winter15.gosredirector.ea.com:8081`.
On the client, as Administrator:
```powershell
Add-DnsClientNrptRule -Namespace "winter15.gosredirector.ea.com" -NameServers "10.10.0.120"
# revert:
Get-DnsClientNrptRule | Where-Object Namespace -eq "winter15.gosredirector.ea.com" | Remove-DnsClientNrptRule -Force
```
## Preflight
`openfut-client-preflight.ps1` is **read-only**: it never launches the game,
+145
View File
@@ -0,0 +1,145 @@
#!/usr/bin/env python3
"""A deliberately tiny DNS responder that answers ONE name and forwards the rest.
WHY THIS EXISTS. FIFA 17's ProtoSSL verifies the roster certificate by dNSName
only, so the client must reach the roster as `winter15.gosredirector.ea.com`
rather than by IP -- an IP-addressed roster is rejected and the player gets
"An error occurred downloading the FUT Squad Update". The usual fix is a hosts
entry on the client, which on this operator's machine previously took the whole
machine's internet down.
So instead of editing the client's hosts file, the client points ONE name at
this resolver through a Windows NRPT rule. Everything else on the client keeps
using its normal DNS and is never routed here at all.
SAFETY PROPERTY THAT MATTERS: even for the queries that do arrive, an unknown
name is FORWARDED upstream verbatim rather than refused or NXDOMAINed. So the
worst case if the NRPT scope is ever widened by accident is added latency, not a
broken resolver -- this cannot repeat the earlier outage.
Answers A records only; every other qtype for the pinned name is forwarded too,
because inventing an answer for a type we do not model is how a resolver starts
lying.
"""
import os
import socket
import socketserver
import struct
import sys
PINNED_NAME = os.environ.get("SCOPED_DNS_NAME", "winter15.gosredirector.ea.com").rstrip(".").lower()
PINNED_ADDR = os.environ.get("SCOPED_DNS_ADDR", "10.10.0.120")
BIND_ADDR = os.environ.get("SCOPED_DNS_BIND", "0.0.0.0")
BIND_PORT = int(os.environ.get("SCOPED_DNS_PORT", "53"))
TTL = 60
QTYPE_A = 1
QCLASS_IN = 1
def upstream_servers() -> list[str]:
"""Real resolvers, read from resolv.conf, minus ourselves."""
out = []
try:
with open("/etc/resolv.conf") as fh:
for line in fh:
parts = line.split()
if len(parts) >= 2 and parts[0] == "nameserver" and parts[1] != PINNED_ADDR:
out.append(parts[1])
except OSError:
pass
return out or ["1.1.1.1", "8.8.8.8"]
UPSTREAM = upstream_servers()
def parse_question(data: bytes):
"""Return (qname, qtype, end_offset) or None if the packet is not parseable."""
if len(data) < 12:
return None
qdcount = struct.unpack("!H", data[4:6])[0]
if qdcount < 1:
return None
labels, off = [], 12
while off < len(data):
ln = data[off]
if ln == 0:
off += 1
break
# A pointer in the QUESTION section is malformed; forward and let the
# upstream decide rather than guessing.
if ln & 0xC0:
return None
off += 1
labels.append(data[off:off + ln].decode("ascii", "replace"))
off += ln
if off + 4 > len(data):
return None
qtype, _qclass = struct.unpack("!HH", data[off:off + 4])
return ".".join(labels).lower(), qtype, off + 4
def build_answer(query: bytes, qend: int) -> bytes:
"""Echo the question and append one A record for the pinned address."""
txid = query[:2]
# QR=1, RD copied from the query, RA=1.
rd = query[2] & 0x01
flags = struct.pack("!H", 0x8000 | (rd << 8) | 0x0080)
counts = struct.pack("!HHHH", 1, 1, 0, 0)
question = query[12:qend]
rr = (
b"\xc0\x0c" # name -> pointer to the question
+ struct.pack("!HHIH", QTYPE_A, QCLASS_IN, TTL, 4)
+ socket.inet_aton(PINNED_ADDR)
)
return txid + flags + counts + question + rr
def forward(query: bytes) -> bytes | None:
for server in UPSTREAM:
try:
with socket.socket(socket.AF_INET, socket.SOCK_DGRAM) as s:
s.settimeout(3.0)
s.sendto(query, (server, 53))
return s.recv(4096)
except OSError:
continue
return None
class Handler(socketserver.BaseRequestHandler):
def handle(self):
data, sock = self.request
parsed = parse_question(data)
if parsed:
qname, qtype, qend = parsed
if qname == PINNED_NAME and qtype == QTYPE_A:
sock.sendto(build_answer(data, qend), self.client_address)
print(f"ANSWER {qname} A -> {PINNED_ADDR} (from {self.client_address[0]})", flush=True)
return
reply = forward(data)
if reply:
sock.sendto(reply, self.client_address)
if parsed:
print(f"forward {parsed[0]} qtype={parsed[1]} (from {self.client_address[0]})", flush=True)
class Server(socketserver.ThreadingUDPServer):
allow_reuse_address = True
daemon_threads = True
if __name__ == "__main__":
print(
f"scoped-dns: pinning {PINNED_NAME} -> {PINNED_ADDR}; "
f"forwarding everything else to {', '.join(UPSTREAM)}",
flush=True,
)
try:
with Server((BIND_ADDR, BIND_PORT), Handler) as srv:
print(f"scoped-dns: LISTENING on {BIND_ADDR}:{BIND_PORT}", flush=True)
srv.serve_forever()
except PermissionError:
sys.exit(f"scoped-dns: cannot bind {BIND_ADDR}:{BIND_PORT} (needs root)")