Files
hack-house/cmd_chat/server/views.py
T
leetcrypt 3a54578f0a perf(net): near-real-time sandbox for non-host viewers
Two causes made a shared shell's PTY stream lag for non-host viewers and
then arrive "all at once":

1. Nagle's algorithm — PTY echo/keystrokes/chat are tiny frames; Nagle +
   delayed-ACK (amplified by Tailscale RTT) coalesced them into bursts.
   Set TCP_NODELAY on the client websocket (Plain/--no-tls path) and on
   each server connection's socket (via ws.io_proto.transport).

2. Head-of-line blocking — ConnectionManager.broadcast awaited each
   send() serially under a global lock, so the slowest viewer gated
   delivery to everyone AND stalled the server's read of the broker's
   next frame, backing the stream up. Give each connection its own
   bounded outbound queue + writer task; broadcast now only enqueues and
   never waits on a socket. A peer that overflows its backlog (4096) is
   evicted (close 1013) and resyncs on reconnect rather than buffering
   without bound or being fed a gapped terminal stream. Init snapshot is
   enqueued as the guaranteed-first frame (send_state -> state_frame).

Adds tests/test_manager.py (FIFO, slow-client isolation, wedged
eviction, reconnect replacement). Also fmt-clean sbx.rs/theme.rs.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-06-06 17:11:51 -07:00

240 lines
7.9 KiB
Python

import hashlib
import hmac
import json
import base64
import socket
from dataclasses import asdict
from sanic import Sanic, Request, response, Websocket
from sanic.response import HTTPResponse, json as json_response
from .models import Message, UserSession
from .helpers import get_client_ip, state_frame, utcnow
def _disable_nagle(ws: Websocket) -> None:
"""Set TCP_NODELAY on a websocket's underlying socket. Small interactive
frames (PTY echo, keystrokes, chat) must not wait on Nagle coalescing +
delayed-ACK before reaching a viewer. Best-effort: any transport without a
raw socket (or where the option can't be set) is simply left as-is."""
try:
sock = ws.io_proto.transport.get_extra_info("socket")
if sock is not None:
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
except Exception:
pass
# Hard cap on a single relayed WS frame. The largest legitimate frame is one
# Fernet-encrypted 64 KB file chunk (~120 KB after base64 + token overhead), so
# 256 KB leaves headroom while bounding per-message memory and the 1000-message
# store. Oversized frames are dropped, not stored or broadcast.
MAX_FRAME_SIZE = 256 * 1024
def generate_ws_token(user_id: str, secret: bytes) -> str:
return hmac.new(secret, user_id.encode(), hashlib.sha256).hexdigest()
def _roster_frame(app: Sanic) -> str:
"""Authoritative presence snapshot — all clergy members converge on this."""
users = app.ctx.session_store.get_all()
return json.dumps(
{
"type": "roster",
"users": [{"user_id": u.user_id, "username": u.username} for u in users],
"capacity": app.ctx.max_users,
}
)
async def srp_init(request: Request, app: Sanic) -> HTTPResponse:
try:
client_ip = get_client_ip(request)
if not app.ctx.rate_limiter.is_allowed(client_ip):
return response.json({"error": "Rate limited"}, status=429)
data = request.json or {}
username = data.get("username", "unknown")
client_public_b64 = data.get("A")
if not client_public_b64:
return response.json({"error": "Missing A"}, status=400)
client_public = base64.b64decode(client_public_b64)
if app.ctx.session_store.username_exists(username):
return response.json({"error": "Username taken"}, status=409)
if app.ctx.session_store.count() >= app.ctx.max_users:
return response.json({"error": "Clergy full"}, status=409)
user_id, B, salt = app.ctx.srp_manager.init_auth(username, client_public)
return response.json(
{
"user_id": user_id,
"B": base64.b64encode(B).decode(),
"salt": base64.b64encode(salt).decode(),
"room_salt": base64.b64encode(app.ctx.room_salt).decode(),
}
)
except Exception:
return response.json({"error": "SRP init failed"}, status=500)
async def srp_verify(request: Request, app: Sanic) -> HTTPResponse:
try:
client_ip = get_client_ip(request)
if not app.ctx.rate_limiter.is_allowed(client_ip):
return response.json({"error": "Rate limited"}, status=429)
data = request.json or {}
user_id = data.get("user_id")
client_proof_b64 = data.get("M")
username = data.get("username", "unknown")
if not user_id or not client_proof_b64:
return response.json({"error": "Missing user_id or M"}, status=400)
client_proof = base64.b64decode(client_proof_b64)
# Authoritative capacity gate — the slot is only consumed once a session
# is actually added here (init is best-effort / racy).
if app.ctx.session_store.count() >= app.ctx.max_users:
return response.json({"error": "Clergy full"}, status=409)
H_AMK, session_key = app.ctx.srp_manager.verify_auth(user_id, client_proof)
fernet_key = base64.urlsafe_b64encode(session_key[:32])
session = UserSession(
user_id=user_id,
ip=get_client_ip(request),
username=username,
fernet_key=fernet_key,
)
app.ctx.session_store.add(session)
ws_token = generate_ws_token(user_id, app.ctx.ws_secret)
return response.json(
{
"H_AMK": base64.b64encode(H_AMK).decode(),
"ws_token": ws_token,
}
)
except ValueError as e:
return response.json({"error": str(e)}, status=401)
except Exception:
return response.json({"error": "SRP verify failed"}, status=500)
async def chat_ws(request: Request, ws: Websocket, app: Sanic) -> None:
user_id = request.args.get("user_id")
ws_token = request.args.get("ws_token")
if not user_id or not ws_token:
await ws.close(code=4002, reason="user_id and ws_token required")
return
expected_token = generate_ws_token(user_id, app.ctx.ws_secret)
if not hmac.compare_digest(ws_token, expected_token):
await ws.close(code=4003, reason="Invalid token")
return
session = app.ctx.session_store.get(user_id)
if not session:
await ws.close(code=4002, reason="Invalid session")
return
manager = app.ctx.connection_manager
_disable_nagle(ws)
# Enqueue this client's init snapshot as its first outbound frame, then
# register it so broadcasts can target it — guaranteeing init arrives first.
await manager.connect(user_id, ws, initial=state_frame(app))
try:
# Announce arrival to everyone already present, then a fresh roster.
await manager.broadcast(
json.dumps(
{
"type": "user_joined",
"user_id": user_id,
"username": session.username,
}
),
exclude_user=user_id,
)
await manager.broadcast(_roster_frame(app))
async for data in ws:
if data is None:
break
text = str(data)
# Drop oversized frames before they reach the store/broadcast: this
# bounds memory and stops a single client from flooding the room.
if len(text) > MAX_FRAME_SIZE:
continue
app.ctx.session_store.update_activity(user_id)
message = Message(
text=text,
username=session.username,
)
app.ctx.message_store.add(message)
await manager.broadcast(
json.dumps(
{
"type": "message",
"data": asdict(message),
}
)
)
except Exception:
pass
finally:
await manager.disconnect(user_id)
# Free the slot + username so the clergy can be rejoined (was previously
# held until the 1h stale sweep, which also blocked the name).
app.ctx.session_store.remove(user_id)
await manager.broadcast(
json.dumps(
{
"type": "user_left",
"user_id": user_id,
}
)
)
await manager.broadcast(_roster_frame(app))
async def health(request: Request, app: Sanic) -> HTTPResponse:
return json_response(
{
"status": "ok",
"messages": app.ctx.message_store.count(),
"users": app.ctx.session_store.count(),
"timestamp": utcnow().isoformat(),
}
)
async def clear_messages(request: Request, app: Sanic) -> HTTPResponse:
auth_header = request.headers.get("authorization", "")
if not auth_header.startswith("Bearer "):
return response.json({"error": "Unauthorized"}, status=401)
token = auth_header[7:]
if not hmac.compare_digest(token, app.ctx.admin_token):
return response.json({"error": "Unauthorized"}, status=401)
app.ctx.message_store.clear()
return json_response({"status": "cleared"})