v4.9.0

changelog.md

@@ -1,5 +1,309 @@
 # Changelog
 
+## 2026-03-18 - 4.9.0 - feat(protocol)
+add sustained-stream tunnel scheduling to isolate high-throughput traffic
+
+- Introduce a third low-priority sustained queue in TunnelIo with a forced drain budget to prevent long-lived high-bandwidth streams from starving control and normal data frames.
+- Classify upload and download streams as sustained after exceeding the throughput threshold for the minimum duration, and route their DATA and CLOSE frames through the sustained channel.
+- Wire the new sustained channel through edge and hub stream handling so sustained traffic is scheduled consistently on both sides of the tunnel.
+
+## 2026-03-18 - 4.8.19 - fix(remoteingress-protocol)
+reduce per-stream flow control windows and increase control channel buffering
+
+- Lower the initial and maximum per-stream window from 16MB to 4MB and scale adaptive windows against a 200MB total budget with a 1MB minimum.
+- Increase edge and hub control frame channel capacity from 256 to 512 to better handle prioritized control traffic.
+- Update flow-control tests and comments to reflect the new window sizing and budget behavior.
+
+## 2026-03-17 - 4.8.18 - fix(rust-protocol)
+switch tunnel frame buffers from Vec<u8> to Bytes to reduce copying and memory overhead
+
+- Add the bytes crate to core and protocol crates
+- Update frame encoding, reader payloads, channel queues, and stream backchannels to use Bytes
+- Adjust edge and hub data/control paths to send framed payloads as Bytes
+
+## 2026-03-17 - 4.8.17 - fix(protocol)
+increase per-stream flow control windows and remove adaptive read caps
+
+- Raise the initial per-stream window from 4MB to 16MB and expand the adaptive window budget to 800MB with a 4MB floor
+- Stop limiting edge and hub reads by the adaptive per-stream target window, keeping reads capped only by the current window and 32KB chunk size
+- Update protocol tests to match the new adaptive window scaling and budget boundaries
+
+## 2026-03-17 - 4.8.16 - fix(release)
+bump package version to 4.8.15
+
+- Updates the package.json version field from 4.8.13 to 4.8.15.
+
+## 2026-03-17 - 4.8.13 - fix(remoteingress-protocol)
+require a flush after each written frame to bound TLS buffer growth
+
+- Remove the unflushed byte threshold and stop queueing additional writes while a flush is pending
+- Simplify write and flush error logging after dropping unflushed byte tracking
+- Update tunnel I/O comments to reflect the stricter flush behavior that avoids OOM and connection resets
+
+## 2026-03-17 - 4.8.12 - fix(tunnel)
+prevent tunnel backpressure buffering from exhausting memory and cancel stream handlers before TLS shutdown
+
+- stop self-waking and writing new frames while a flush is pending to avoid unbounded TLS session buffer growth under load
+- reorder edge and hub shutdown cleanup so stream cancellation happens before TLS close_notify, preventing handlers from blocking on dead channels
+- add load tests covering sustained large transfers, burst traffic, and rapid stream churn to verify tunnel stability
+
+## 2026-03-17 - 4.8.11 - fix(remoteingress-core)
+stop data frame send loops promptly when stream cancellation is triggered
+
+- Use cancellation-aware tokio::select! around data channel sends in both edge and hub stream forwarding paths
+- Prevent stalled or noisy shutdown behavior when stream or client cancellation happens while awaiting frame delivery
+
+## 2026-03-17 - 4.8.10 - fix(remoteingress-core)
+guard tunnel frame sends with cancellation to prevent async send deadlocks
+
+- Wrap OPEN, CLOSE, CLOSE_BACK, WINDOW_UPDATE, and cleanup channel sends in cancellation-aware tokio::select! blocks.
+- Avoid indefinite blocking when tunnel, stream, or writer tasks are cancelled while awaiting channel capacity.
+- Improve shutdown reliability for edge and hub stream handling under tunnel failure conditions.
+
+## 2026-03-17 - 4.8.9 - fix(repo)
+no changes to commit
+
+
+## 2026-03-17 - 4.8.8 - fix(remoteingress-core)
+cancel stale edge connections when an edge reconnects
+
+- Remove any existing edge entry before registering a reconnected edge
+- Trigger the previous connection's cancellation token so stale sessions shut down immediately instead of waiting for TCP keepalive
+
+## 2026-03-17 - 4.8.7 - fix(remoteingress-core)
+perform graceful TLS shutdown on edge and hub tunnel streams
+
+- Send TLS close_notify before cleanup to avoid peer disconnect warnings on both tunnel endpoints
+- Wrap stream shutdown in a 2 second timeout so connection teardown does not block cleanup
+
+## 2026-03-17 - 4.8.6 - fix(remoteingress-core)
+initialize disconnect reason only when set in hub loop break paths
+
+- Replace the default "unknown" disconnect reason with an explicitly assigned string and document that all hub loop exits set it before use
+- Add an allow attribute for unused assignments to avoid warnings around the deferred initialization pattern
+
+## 2026-03-17 - 4.8.5 - fix(repo)
+no changes to commit
+
+
+## 2026-03-17 - 4.8.4 - fix(remoteingress-core)
+prevent stream stalls by guaranteeing flow-control updates and avoiding bounded per-stream channel overflows
+
+- Replace bounded per-stream data channels with unbounded channels on edge and hub, relying on existing WINDOW_UPDATE flow control to limit bytes in flight
+- Use awaited sends for FRAME_WINDOW_UPDATE and FRAME_WINDOW_UPDATE_BACK so updates are not dropped and streams do not deadlock under backpressure
+- Clean up stream state when channel receivers have already exited instead of closing active streams because a bounded queue filled
+
+## 2026-03-17 - 4.8.3 - fix(protocol,edge)
+optimize tunnel frame handling and zero-copy uploads in edge I/O
+
+- extract hub frame processing into a shared edge handler to remove duplicated tunnel logic
+- add zero-copy frame header encoding and read payloads directly into framed buffers for client-to-hub uploads
+- refactor TunnelIo read/write state to avoid unsafe queue access and reduce buffer churn with incremental parsing
+
+## 2026-03-17 - 4.8.2 - fix(rust-edge)
+refactor tunnel I/O to preserve TLS state and prioritize control frames
+
+- replace split TLS handling with a single-owner TunnelIo to avoid handshake and buffered read corruption
+- prioritize control frames over data frames to prevent WINDOW_UPDATE starvation and flow-control deadlocks
+- improve tunnel reliability with incremental frame parsing, liveness/error events, and corrupt frame header logging
+
+## 2026-03-17 - 4.8.1 - fix(remoteingress-core)
+remove tunnel writer timeouts from edge and hub buffered writes
+
+- Drops the 30 second timeout wrapper around writer.write_all and writer.flush in both edge and hub tunnel writers.
+- Updates error logging to report write failures without referring to stalled writes.
+
+## 2026-03-17 - 4.8.0 - feat(events)
+include disconnect reasons in edge and hub management events
+
+- Add reason fields to tunnelDisconnected and edgeDisconnected events emitted from the Rust core and binary bridge
+- Propagate specific disconnect causes such as EOF, liveness timeout, writer failure, handshake failure, and hub cancellation
+- Update TypeScript edge and hub classes to log and forward disconnect reason data
+- Extend serialization tests to cover the new reason fields
+
+## 2026-03-17 - 4.7.2 - fix(remoteingress-core)
+add tunnel write timeouts and scale initial stream windows by active stream count
+
+- Wrap tunnel frame writes and flushes in a 30-second timeout on both edge and hub to detect stalled writers and trigger faster reconnect or cleanup.
+- Compute each stream's initial send window from the current active stream count instead of using a fixed window to keep total in-flight data within the 32MB budget.
+
+## 2026-03-17 - 4.7.1 - fix(remoteingress-core)
+improve tunnel failure detection and reconnect handling
+
+- Enable TCP keepalive on edge and hub connections to detect silent network failures sooner
+- Trigger immediate reconnect or disconnect when tunnel writer tasks fail instead of waiting for liveness timeouts
+- Prevent active stream counter underflow during concurrent connection cleanup
+
+## 2026-03-16 - 4.7.0 - feat(edge,protocol,test)
+add configurable edge bind address and expand flow-control test coverage
+
+- adds an optional bindAddress configuration for edge TCP listeners, defaulting to 0.0.0.0 when not provided
+- passes bindAddress through the TypeScript edge client and Rust edge runtime so local test setups can bind to localhost
+- adds protocol unit tests for adaptive stream window sizing and window update frame encoding/decoding
+- introduces end-to-end flow-control tests and updates the test script to build before running tests
+
+## 2026-03-16 - 4.6.1 - fix(remoteingress-core)
+avoid spurious tunnel disconnect events and increase control channel capacity
+
+- Emit TunnelDisconnected only after an established connection is actually lost, preventing false disconnect events during failed reconnect attempts.
+- Increase edge and hub control-channel buffer sizes from 64 to 256 to better prioritize control frames under load.
+
+## 2026-03-16 - 4.6.0 - feat(remoteingress-core)
+add adaptive per-stream flow control based on active stream counts
+
+- Track active stream counts on edge and hub connections to size per-stream flow control windows dynamically.
+- Cap WINDOW_UPDATE increments and read sizes to the adaptive window so bandwidth is shared more evenly across concurrent streams.
+- Apply the adaptive logic to both upload and download paths on edge and hub stream handlers.
+
+## 2026-03-16 - 4.5.12 - fix(remoteingress-core)
+improve tunnel liveness handling and enable TCP keepalive for accepted client sockets
+
+- Avoid disconnecting edges when PING or PONG frames cannot be queued because the control channel is temporarily full.
+- Enable TCP_NODELAY and TCP keepalive on accepted client connections to help detect stale or dropped clients.
+
+## 2026-03-16 - 4.5.11 - fix(repo)
+no changes to commit
+
+
+## 2026-03-16 - 4.5.10 - fix(remoteingress-core)
+guard zero-window reads to avoid false EOF handling on stalled streams
+
+- Prevent upload and download loops from calling read on an empty buffer when flow-control window remains at 0 after stall timeout
+- Log a warning and close the affected stream instead of misinterpreting Ok(0) as end-of-file
+
+## 2026-03-16 - 4.5.9 - fix(remoteingress-core)
+delay stream close until downstream response draining finishes to prevent truncated transfers
+
+- Waits for the hub-to-client download task to finish before sending the stream CLOSE frame
+- Prevents upstream reads from being cancelled mid-response during asymmetric transfers such as git fetch
+- Retains the existing timeout so stalled downloads still clean up safely
+
+## 2026-03-16 - 4.5.8 - fix(remoteingress-core)
+ensure upstream writes cancel promptly and reliably deliver CLOSE_BACK frames
+
+- listen for stream cancellation while waiting on upstream write timeouts so FRAME_CLOSE does not block for up to 60 seconds
+- replace try_send with send().await when emitting CLOSE_BACK frames to avoid silently dropping close notifications when the data channel is full
+
+## 2026-03-16 - 4.5.7 - fix(remoteingress-core)
+improve tunnel reconnect and frame write efficiency
+
+- Reuse the TLS connector across edge reconnections to preserve session resumption state and reduce reconnect latency.
+- Buffer hub and edge frame writes to coalesce small control and data frames into fewer TLS records and syscalls while still flushing each frame promptly.
+
+## 2026-03-16 - 4.5.6 - fix(remoteingress-core)
+disable Nagle's algorithm on edge, hub, and upstream TCP sockets to reduce control-frame latency
+
+- Enable TCP_NODELAY on the edge connection to the hub for faster PING/PONG and WINDOW_UPDATE delivery
+- Apply TCP_NODELAY on accepted hub streams before TLS handling
+- Enable TCP_NODELAY on SmartProxy upstream connections before sending the PROXY header
+
+## 2026-03-16 - 4.5.5 - fix(remoteingress-core)
+wait for hub-to-client draining before cleanup and reliably send close frames
+
+- switch CLOSE frame delivery on the data channel from try_send to send().await to avoid dropping it when the channel is full
+- delay stream cleanup until the hub-to-client task finishes or times out so large downstream responses continue after upload EOF
+- add a bounded 5-minute wait for download draining to prevent premature termination of asymmetric transfers such as git fetch
+
+## 2026-03-15 - 4.5.4 - fix(remoteingress-core)
+preserve stream close ordering and add flow-control stall timeouts
+
+- Send CLOSE and CLOSE_BACK frames on the data channel so they arrive after the final stream data frames.
+- Log and abort stalled upload and download paths when flow-control windows stay empty for 120 seconds.
+- Apply a 60-second timeout when writing buffered stream data to the upstream connection to prevent hung streams.
+
+## 2026-03-15 - 4.5.3 - fix(remoteingress-core)
+prioritize control frames over data in edge and hub tunnel writers
+
+- Split tunnel/frame writers into separate control and data channels in edge and hub
+- Use biased select loops so PING, PONG, WINDOW_UPDATE, OPEN, and CLOSE frames are sent before data frames
+- Route stream data through dedicated data channels while keeping OPEN, CLOSE, and flow-control updates on control channels to prevent keepalive starvation under load
+
+## 2026-03-15 - 4.5.2 - fix(remoteingress-core)
+improve stream flow control retries and increase channel buffer capacity
+
+- increase per-stream mpsc channel capacity from 128 to 256 on both edge and hub paths
+- only reset accumulated window update bytes after a successful try_send to avoid dropping flow-control credits when the update channel is busy
+
+## 2026-03-15 - 4.5.1 - fix(protocol)
+increase per-stream flow control window and channel buffers to improve high-RTT throughput
+
+- raise the initial stream window from 256 KB to 4 MB to allow more in-flight data per stream
+- increase edge and hub mpsc channel capacities from 16 to 128 to better absorb throughput under flow control
+
+## 2026-03-15 - 4.5.0 - feat(remoteingress-core)
+add per-stream flow control for edge and hub tunnel data transfer
+
+- introduce WINDOW_UPDATE frame types and protocol helpers for per-stream flow control
+- track per-stream send windows on both edge and hub to limit reads based on available capacity
+- send window updates after downstream writes to reduce channel pressure during large transfers
+
+## 2026-03-15 - 4.4.1 - fix(remoteingress-core)
+prevent stream data loss by applying backpressure and closing saturated channels
+
+- replace non-blocking frame writes with awaited sends in per-stream tasks so large transfers respect backpressure instead of dropping data
+- close and remove streams when back-channel or data channels fill up to avoid TCP stream corruption from silently dropped frames
+
+## 2026-03-03 - 4.4.0 - feat(remoteingress)
+add heartbeat PING/PONG and liveness timeouts; implement fast-reconnect/backoff reset and JS crash-recovery auto-restart
+
+- protocol: add FRAME_PING and FRAME_PONG and unit tests for ping/pong frames
+- edge (Rust): reset backoff after successful connection, respond to PING with PONG, track liveness via deadline and reconnect on timeout, use Duration/Instant helpers
+- hub (Rust): send periodic PING to edges, handle PONGs, enforce liveness timeout and disconnect inactive edges, use tokio interval and time utilities
+- ts: RemoteIngressEdge and RemoteIngressHub: add crash-recovery auto-restart with exponential backoff and max attempts, save/restore config and allowed edges, register/remove exit handlers, ensure stop() marks stopping and cleans up listeners
+- minor API/typing: introduce TAllowedEdge alias and persist allowed edges for restart recovery
+
+## 2026-02-26 - 4.3.0 - feat(hub)
+add optional TLS certificate/key support to hub start config and bridge
+
+- TypeScript: add tls.certPem and tls.keyPem to IHubConfig and include tlsCertPem/tlsKeyPem in startHub bridge command when both are provided
+- TypeScript: extend startHub params with tlsCertPem and tlsKeyPem and conditionally send them
+- Rust: change HubConfig serde attributes for tls_cert_pem and tls_key_pem from skip to default so absent PEM fields deserialize as None
+- Enables optional provisioning of TLS certificate and key to the hub when provided from the JS side
+
+## 2026-02-26 - 4.2.0 - feat(core)
+expose edge peer address in hub events and migrate writers to channel-based, non-blocking framing with stream limits and timeouts
+
+- Add peerAddr to ConnectedEdgeStatus and HubEvent::EdgeConnected and surface it to the TS frontend event (management:edgeConnected).
+- Replace Arc<Mutex<WriteHalf>> writers with dedicated mpsc channel writer tasks in both hub and edge crates to serialize writes off the main tasks.
+- Use non-blocking try_send for data frames to avoid head-of-line blocking and drop frames with warnings when channels are full.
+- Introduce MAX_STREAMS_PER_EDGE semaphore to limit concurrent streams per edge and reject excess opens with a CLOSE_BACK frame.
+- Add a 10s timeout when connecting to SmartProxy to avoid hanging connections.
+- Ensure writer tasks are aborted on shutdown/cleanup and propagate cancellation tokens appropriately.
+
+## 2026-02-26 - 4.1.0 - feat(remoteingress-bin)
+use mimalloc as the global allocator to reduce memory overhead and improve allocation performance
+
+- added mimalloc = "0.1" dependency to rust/crates/remoteingress-bin/Cargo.toml
+- registered mimalloc as the #[global_allocator] in rust/crates/remoteingress-bin/src/main.rs
+- updated Cargo.lock with libmimalloc-sys and mimalloc package entries
+
+## 2026-02-26 - 4.0.1 - fix(hub)
+cancel per-stream tokens on stream close and avoid duplicate StreamClosed events; bump @types/node devDependency to ^25.3.0
+
+- Add CancellationToken to per-stream entries so each stream can be cancelled independently.
+- Ensure StreamClosed event is only emitted when a stream was actually present (guards against duplicate events).
+- Cancel the stream-specific token on FRAME_CLOSE to stop associated tasks and free resources.
+- DevDependency bump: @types/node updated from ^25.2.3 to ^25.3.0.
+
+## 2026-02-19 - 4.0.0 - BREAKING CHANGE(remoteingress-core)
+add cancellation tokens and cooperative shutdown; switch event channels to bounded mpsc and improve cleanup
+
+- Introduce tokio-util::sync::CancellationToken for hub/edge and per-connection/stream cancellation, enabling cooperative shutdown of spawned tasks.
+- Replace unbounded mpsc channels with bounded mpsc::channel(1024) and switch from UnboundedSender/Receiver to Sender/Receiver; use try_send where non-blocking sends are appropriate.
+- Wire cancellation tokens through edge and hub codepaths: child tokens per connection, per-port, per-stream; cancel tokens in stop() and Drop impls to ensure deterministic task termination and cleanup.
+- Reset stream id counters and clear listener state on reconnect; improved error handling around accept/read loops using tokio::select! and cancellation checks.
+- Update Cargo.toml and Cargo.lock to add tokio-util (and related futures entries) as dependencies.
+- BREAKING: public API/types changed — take_event_rx return types and event_tx/event_rx fields now use bounded mpsc::Sender/mpsc::Receiver instead of the unbounded variants; callers must adapt to the new types and bounded behavior.
+
+## 2026-02-18 - 3.3.0 - feat(readme)
+document dynamic port assignment and runtime port updates; clarify TLS multiplexing, frame format, and handshake sequence
+
+- Adds documentation for dynamic port configuration: hub-assigned listen ports, hot-reloadable via FRAME_CONFIG frames
+- Introduces new FRAME type CONFIG (0x06) and describes payload as JSON; notes immediate push of port changes to connected edges
+- Clarifies that the tunnel is a single encrypted TLS multiplexed connection to the hub (preserves PROXY v1 behavior)
+- Specifies frame integer fields are big-endian and that stream IDs are 32-bit unsigned integers
+- Adds new events: portsAssigned and portsUpdated, and updates examples showing updateAllowedEdges usage and live port changes
+
 ## 2026-02-18 - 3.2.1 - fix(tests)
 add comprehensive unit and async tests across Rust crates and TypeScript runtime
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@serve.zone/remoteingress",
-  "version": "3.2.1",
+  "version": "4.9.0",
   "private": false,
   "description": "Edge ingress tunnel for DcRouter - accepts incoming TCP connections at network edge and tunnels them to DcRouter SmartProxy preserving client IP via PROXY protocol v1.",
   "main": "dist_ts/index.js",
@@ -9,7 +9,7 @@
   "author": "Task Venture Capital GmbH",
   "license": "MIT",
   "scripts": {
-    "test": "(tstest test/ --web)",
+    "test": "(pnpm run build && tstest test/ --verbose --logfile --timeout 60)",
     "build": "(tsbuild tsfolders --allowimplicitany && tsrust)",
     "buildDocs": "(tsdoc)"
   },
@@ -20,7 +20,7 @@
     "@git.zone/tsrust": "^1.3.0",
     "@git.zone/tstest": "^3.1.8",
     "@push.rocks/tapbundle": "^6.0.3",
-    "@types/node": "^25.2.3"
+    "@types/node": "^25.3.0"
   },
   "dependencies": {
     "@push.rocks/qenv": "^6.1.3",

pnpm-lock.yaml

@@ -34,8 +34,8 @@ importers:
         specifier: ^6.0.3
         version: 6.0.3(socks@2.8.7)
       '@types/node':
-        specifier: ^25.2.3
+        specifier: ^25.3.0
-        version: 25.2.3
+        version: 25.3.0
 
 packages:
 
@@ -1501,8 +1501,8 @@ packages:
   '@types/node@22.19.11':
     resolution: {integrity: sha512-BH7YwL6rA93ReqeQS1c4bsPpcfOmJasG+Fkr6Y59q83f9M1WcBRHR2vM+P9eOisYRcN3ujQoiZY8uk5W+1WL8w==}
 
-  '@types/node@25.2.3':
+  '@types/node@25.3.0':
-    resolution: {integrity: sha512-m0jEgYlYz+mDJZ2+F4v8D1AyQb+QzsNqRuI7xg1VQX/KlKS0qT9r1Mo16yo5F/MtifXFgaofIFsdFMox2SxIbQ==}
+    resolution: {integrity: sha512-4K3bqJpXpqfg2XKGK9bpDTc6xO/xoUP/RBWS7AtRMug6zZFaRekiLzjVtAoZMquxoAbzBvy5nxQ7veS5eYzf8A==}
 
   '@types/parse5@6.0.3':
     resolution: {integrity: sha512-SuT16Q1K51EAVPz1K29DJ/sXjhSQ0zjvsypYJ6tlwVsRV9jwW5Adq2ch8Dq8kDBCkYnELS7N7VNCSB5nC56t/g==}
@@ -3986,8 +3986,8 @@ packages:
   undici-types@6.21.0:
     resolution: {integrity: sha512-iwDZqg0QAGrg9Rav5H4n0M64c3mkR59cJ6wQp+7C4nI0gsmExaedaYLNO44eT4AtBBwjbTiGPMlt2Md0T9H9JQ==}
 
-  undici-types@7.16.0:
+  undici-types@7.18.2:
-    resolution: {integrity: sha512-Zz+aZWSj8LE6zoxD+xrjh4VfkIG8Ya6LvYkZqtUQGJPZjYl53ypCaUwWqo7eI0x66KBGeRo+mlBEkMSeSZ38Nw==}
+    resolution: {integrity: sha512-AsuCzffGHJybSaRrmr5eHr81mwJU3kjw6M+uprWvCXiNeN9SOGwQ3Jn8jb8m3Z6izVgknn1R0FTCEAP2QrLY/w==}
 
   unified@11.0.5:
     resolution: {integrity: sha512-xKvGhPWw3k84Qjh8bI3ZeJjqnyadK+GEFtazSfZv/rKeTkTjOJho6mFqh2SM96iIcZokxiOpg78GazTSg8+KHA==}
@@ -5178,7 +5178,7 @@ snapshots:
       '@jest/schemas': 29.6.3
       '@types/istanbul-lib-coverage': 2.0.6
       '@types/istanbul-reports': 3.0.4
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       '@types/yargs': 17.0.35
       chalk: 4.1.2
 
@@ -6736,14 +6736,14 @@ snapshots:
 
   '@types/accepts@1.3.7':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/babel__code-frame@7.27.0': {}
 
   '@types/body-parser@1.19.6':
     dependencies:
       '@types/connect': 3.4.38
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/buffer-json@2.0.3': {}
 
@@ -6760,17 +6760,17 @@ snapshots:
 
   '@types/clean-css@4.2.11':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       source-map: 0.6.1
 
   '@types/co-body@6.1.3':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       '@types/qs': 6.14.0
 
   '@types/connect@3.4.38':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/content-disposition@0.5.9': {}
 
@@ -6781,11 +6781,11 @@ snapshots:
       '@types/connect': 3.4.38
       '@types/express': 5.0.6
       '@types/keygrip': 1.0.6
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/cors@2.8.19':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/debounce@1.2.4': {}
 
@@ -6797,7 +6797,7 @@ snapshots:
 
   '@types/express-serve-static-core@5.1.1':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       '@types/qs': 6.14.0
       '@types/range-parser': 1.2.7
       '@types/send': 1.2.1
@@ -6811,7 +6811,7 @@ snapshots:
   '@types/fs-extra@11.0.4':
     dependencies:
       '@types/jsonfile': 6.1.4
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/hast@3.0.4':
     dependencies:
@@ -6845,7 +6845,7 @@ snapshots:
 
   '@types/jsonfile@6.1.4':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/keygrip@1.0.6': {}
 
@@ -6862,7 +6862,7 @@ snapshots:
       '@types/http-errors': 2.0.5
       '@types/keygrip': 1.0.6
       '@types/koa-compose': 3.2.9
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/mdast@4.0.4':
     dependencies:
@@ -6876,19 +6876,19 @@ snapshots:
 
   '@types/mute-stream@0.0.4':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/node-forge@1.3.14':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/node@22.19.11':
     dependencies:
       undici-types: 6.21.0
 
-  '@types/node@25.2.3':
+  '@types/node@25.3.0':
     dependencies:
-      undici-types: 7.16.0
+      undici-types: 7.18.2
 
   '@types/parse5@6.0.3': {}
 
@@ -6904,18 +6904,18 @@ snapshots:
 
   '@types/s3rver@3.7.4':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/semver@7.7.1': {}
 
   '@types/send@1.2.1':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/serve-static@2.2.0':
     dependencies:
       '@types/http-errors': 2.0.5
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/sinon-chai@3.2.12':
     dependencies:
@@ -6934,11 +6934,11 @@ snapshots:
 
   '@types/tar-stream@3.1.4':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/through2@2.0.41':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/triple-beam@1.3.5': {}
 
@@ -6966,11 +6966,11 @@ snapshots:
 
   '@types/ws@7.4.7':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/ws@8.18.1':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
 
   '@types/yargs-parser@21.0.3': {}
 
@@ -6980,7 +6980,7 @@ snapshots:
 
   '@types/yauzl@2.10.3':
     dependencies:
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
     optional: true
 
   '@ungap/structured-clone@1.3.0': {}
@@ -7585,7 +7585,7 @@ snapshots:
   engine.io@6.6.4:
     dependencies:
       '@types/cors': 2.8.19
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       accepts: 1.3.8
       base64id: 2.0.0
       cookie: 0.7.2
@@ -8299,7 +8299,7 @@ snapshots:
   jest-util@29.7.0:
     dependencies:
       '@jest/types': 29.6.3
-      '@types/node': 25.2.3
+      '@types/node': 25.3.0
       chalk: 4.1.2
       ci-info: 3.9.0
       graceful-fs: 4.2.11
@@ -9807,7 +9807,7 @@ snapshots:
 
   undici-types@6.21.0: {}
 
-  undici-types@7.16.0: {}
+  undici-types@7.18.2: {}
 
   unified@11.0.5:
     dependencies:

readme.md

@@ -1,6 +1,6 @@
 # @serve.zone/remoteingress
 
-Edge ingress tunnel for DcRouter — accepts incoming TCP connections at the network edge and tunnels them to a DcRouter SmartProxy instance, preserving the original client IP via PROXY protocol v1.
+Edge ingress tunnel for DcRouter — accepts incoming TCP connections at the network edge and tunnels them over a single encrypted TLS connection to a DcRouter SmartProxy instance, preserving the original client IP via PROXY protocol v1.
 
 ## Issue Reporting and Security
 
@@ -22,8 +22,8 @@ pnpm install @serve.zone/remoteingress
 │                     │   (multiplexed frames +        │                     │
 │  RemoteIngressEdge  │    shared-secret auth)         │  RemoteIngressHub   │
 │  Accepts client TCP │                                │  Forwards to        │
-│  connections        │                                │  SmartProxy on      │
+│  connections on     │                                │  SmartProxy on      │
-│                     │                                │  local ports        │
+│  hub-assigned ports │                                │  local ports        │
 └─────────────────────┘                                └─────────────────────┘
         ▲                                                       │
         │  TCP from end users                                   ▼
@@ -32,8 +32,8 @@ pnpm install @serve.zone/remoteingress
 
 | Component | Role |
 |-----------|------|
-| **RemoteIngressEdge** | Deployed at the network edge (e.g. a VPS or cloud instance). Accepts raw TCP connections and multiplexes them over a single TLS tunnel to the hub. |
+| **RemoteIngressEdge** | Deployed at the network edge (e.g. a VPS or cloud instance). Listens on ports assigned by the hub, accepts raw TCP connections, and multiplexes them over a single TLS tunnel to the hub. Ports are hot-reloadable — the hub can change them at runtime. |
-| **RemoteIngressHub** | Deployed alongside DcRouter/SmartProxy in a private cluster. Accepts edge connections, demuxes streams, and forwards each to SmartProxy with a PROXY protocol v1 header so the real client IP is preserved. |
+| **RemoteIngressHub** | Deployed alongside DcRouter/SmartProxy in a private cluster. Accepts edge connections, demuxes streams, and forwards each to SmartProxy with a PROXY protocol v1 header so the real client IP is preserved. Controls which ports each edge listens on. |
 | **Rust Binary** (`remoteingress-bin`) | The performance-critical networking core. Managed via `@push.rocks/smartrust` RustBridge IPC — you never interact with it directly. Cross-compiled for `linux/amd64` and `linux/arm64`. |
 
 ### ✨ Key Features
@@ -45,6 +45,7 @@ pnpm install @serve.zone/remoteingress
 - 🎫 **Connection tokens** — encode all connection details into a single opaque string
 - 📡 **STUN-based public IP discovery** — the edge automatically discovers its public IP via Cloudflare STUN
 - 🔄 **Auto-reconnect** with exponential backoff if the tunnel drops
+- 🎛️ **Dynamic port configuration** — the hub assigns listen ports per edge and can hot-reload them at runtime via `FRAME_CONFIG` frames
 - 📣 **Event-driven** — both Hub and Edge extend `EventEmitter` for real-time monitoring
 - ⚡ **Rust core** — all frame encoding, TLS, and TCP proxying happen in native code for maximum throughput
 
@@ -79,10 +80,28 @@ await hub.start({
   targetHost: '127.0.0.1', // SmartProxy host to forward streams to (default: 127.0.0.1)
 });
 
-// Register which edges are allowed to connect
+// Register which edges are allowed to connect, including their listen ports
 await hub.updateAllowedEdges([
-  { id: 'edge-nyc-01', secret: 'supersecrettoken1' },
+  {
-  { id: 'edge-fra-02', secret: 'supersecrettoken2' },
+    id: 'edge-nyc-01',
+    secret: 'supersecrettoken1',
+    listenPorts: [80, 443],        // ports the edge should listen on
+    stunIntervalSecs: 300,         // STUN discovery interval (default: 300)
+  },
+  {
+    id: 'edge-fra-02',
+    secret: 'supersecrettoken2',
+    listenPorts: [443, 8080],
+  },
+]);
+
+// Dynamically update ports for a connected edge — changes are pushed instantly
+await hub.updateAllowedEdges([
+  {
+    id: 'edge-nyc-01',
+    secret: 'supersecrettoken1',
+    listenPorts: [80, 443, 8443],  // added port 8443 — edge picks it up in real time
+  },
 ]);
 
 // Check status at any time
@@ -116,6 +135,8 @@ const edge = new RemoteIngressEdge();
 edge.on('tunnelConnected', () => console.log('Tunnel established'));
 edge.on('tunnelDisconnected', () => console.log('Tunnel lost — will auto-reconnect'));
 edge.on('publicIpDiscovered', ({ ip }) => console.log(`Public IP: ${ip}`));
+edge.on('portsAssigned', ({ listenPorts }) => console.log(`Listening on ports: ${listenPorts}`));
+edge.on('portsUpdated', ({ listenPorts }) => console.log(`Ports updated: ${listenPorts}`));
 
 // Single token contains hubHost, hubPort, edgeId, and secret
 await edge.start({
@@ -133,6 +154,8 @@ const edge = new RemoteIngressEdge();
 edge.on('tunnelConnected', () => console.log('Tunnel established'));
 edge.on('tunnelDisconnected', () => console.log('Tunnel lost — will auto-reconnect'));
 edge.on('publicIpDiscovered', ({ ip }) => console.log(`Public IP: ${ip}`));
+edge.on('portsAssigned', ({ listenPorts }) => console.log(`Listening on ports: ${listenPorts}`));
+edge.on('portsUpdated', ({ listenPorts }) => console.log(`Ports updated: ${listenPorts}`));
 
 await edge.start({
   hubHost: 'hub.example.com',  // hostname or IP of the hub
@@ -194,7 +217,7 @@ Tokens are base64url-encoded (URL-safe, no padding) — safe to pass as environm
 |-------------------|-------------|
 | `start(config?)` | Spawns the Rust binary and starts the tunnel listener. Config: `{ tunnelPort?: number, targetHost?: string }` |
 | `stop()` | Gracefully shuts down the hub and kills the Rust process. |
-| `updateAllowedEdges(edges)` | Dynamically update which edges are authorized. Each edge: `{ id: string, secret: string }` |
+| `updateAllowedEdges(edges)` | Dynamically update which edges are authorized and what ports they listen on. Each edge: `{ id: string, secret: string, listenPorts?: number[], stunIntervalSecs?: number }`. If ports change for a connected edge, the update is pushed immediately via a `FRAME_CONFIG` frame. |
 | `getStatus()` | Returns current hub status including connected edges and active stream counts. |
 | `running` | `boolean` — whether the Rust binary is alive. |
 
@@ -204,12 +227,12 @@ Tokens are base64url-encoded (URL-safe, no padding) — safe to pass as environm
 
 | Method / Property | Description |
 |-------------------|-------------|
-| `start(config)` | Spawns the Rust binary and connects to the hub. Accepts `{ token: string }` or `IEdgeConfig`. |
+| `start(config)` | Spawns the Rust binary and connects to the hub. Accepts `{ token: string }` or `IEdgeConfig`. Listen ports are received from the hub during handshake. |
 | `stop()` | Gracefully shuts down the edge and kills the Rust process. |
-| `getStatus()` | Returns current edge status including connection state, public IP, and active streams. |
+| `getStatus()` | Returns current edge status including connection state, public IP, listen ports, and active streams. |
 | `running` | `boolean` — whether the Rust binary is alive. |
 
-**Events:** `tunnelConnected`, `tunnelDisconnected`, `publicIpDiscovered`
+**Events:** `tunnelConnected`, `tunnelDisconnected`, `publicIpDiscovered`, `portsAssigned`, `portsUpdated`
 
 ### Token Utilities
 
@@ -246,7 +269,7 @@ interface IConnectionTokenData {
 The tunnel uses a custom binary frame protocol over TLS:
 
 ```
-[stream_id: 4 bytes][type: 1 byte][length: 4 bytes][payload: N bytes]
+[stream_id: 4 bytes BE][type: 1 byte][length: 4 bytes BE][payload: N bytes]
 ```
 
 | Frame Type | Value | Direction | Purpose |
@@ -256,8 +279,18 @@ The tunnel uses a custom binary frame protocol over TLS:
 | `CLOSE` | `0x03` | Edge → Hub | Client closed the connection |
 | `DATA_BACK` | `0x04` | Hub → Edge | Response data flowing downstream |
 | `CLOSE_BACK` | `0x05` | Hub → Edge | Upstream (SmartProxy) closed the connection |
+| `CONFIG` | `0x06` | Hub → Edge | Runtime configuration update (e.g. port changes); payload is JSON |
 
-Max payload size per frame: **16 MB**.
+Max payload size per frame: **16 MB**. Stream IDs are 32-bit unsigned integers.
+
+### Handshake Sequence
+
+1. Edge opens a TLS connection to the hub
+2. Edge sends: `EDGE <edgeId> <secret>\n`
+3. Hub verifies credentials (constant-time comparison) and responds with JSON: `{"listenPorts":[...],"stunIntervalSecs":300}\n`
+4. Edge starts TCP listeners on the assigned ports
+5. Frame protocol begins — `OPEN`/`DATA`/`CLOSE` frames flow in both directions
+6. Hub can push `CONFIG` frames at any time to update the edge's listen ports
 
 ## 💡 Example Scenarios
 
@@ -292,6 +325,22 @@ const edge = new RemoteIngressEdge();
 await edge.start({ token });
 ```
 
+### 5. Dynamic Port Management
+
+The hub controls which ports each edge listens on. Ports can be changed at runtime without restarting the edge — the hub pushes a `CONFIG` frame and the edge hot-reloads its TCP listeners.
+
+```typescript
+// Initially assign ports 80 and 443
+await hub.updateAllowedEdges([
+  { id: 'edge-nyc-01', secret: 'secret', listenPorts: [80, 443] },
+]);
+
+// Later, add port 8080 — the connected edge picks it up instantly
+await hub.updateAllowedEdges([
+  { id: 'edge-nyc-01', secret: 'secret', listenPorts: [80, 443, 8080] },
+]);
+```
+
 ## License and Legal Information
 
 This repository contains open-source code licensed under the MIT License. A copy of the license can be found in the [LICENSE](./LICENSE) file.

rust/Cargo.lock

@@ -234,6 +234,18 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"
 
+[[package]]
+name = "futures-core"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e3450815272ef58cec6d564423f6e755e25379b217b0bc688e295ba24df6b1d"
+
+[[package]]
+name = "futures-sink"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c39754e157331b013978ec91992bde1ac089843443c49cbc7f46150b0fad0893"
+
 [[package]]
 name = "getrandom"
 version = "0.2.17"
@@ -315,6 +327,16 @@ version = "0.2.182"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6800badb6cb2082ffd7b6a67e6125bb39f18782f793520caee8cb8846be06112"
 
+[[package]]
+name = "libmimalloc-sys"
+version = "0.1.44"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "667f4fec20f29dfc6bc7357c582d91796c169ad7e2fce709468aefeb2c099870"
+dependencies = [
+ "cc",
+ "libc",
+]
+
 [[package]]
 name = "lock_api"
 version = "0.4.14"
@@ -336,6 +358,15 @@ version = "2.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f8ca58f447f06ed17d5fc4043ce1b10dd205e060fb3ce5b979b8ed8e59ff3f79"
 
+[[package]]
+name = "mimalloc"
+version = "0.1.48"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e1ee66a4b64c74f4ef288bcbb9192ad9c3feaad75193129ac8509af543894fd8"
+dependencies = [
+ "libmimalloc-sys",
+]
+
 [[package]]
 name = "mio"
 version = "1.1.1"
@@ -507,6 +538,7 @@ dependencies = [
  "clap",
  "env_logger",
  "log",
+ "mimalloc",
  "remoteingress-core",
  "remoteingress-protocol",
  "rustls",
@@ -519,6 +551,7 @@ dependencies = [
 name = "remoteingress-core"
 version = "2.0.0"
 dependencies = [
+ "bytes",
  "log",
  "rcgen",
  "remoteingress-protocol",
@@ -526,15 +559,20 @@ dependencies = [
  "rustls-pemfile",
  "serde",
  "serde_json",
+ "socket2 0.5.10",
  "tokio",
  "tokio-rustls",
+ "tokio-util",
 ]
 
 [[package]]
 name = "remoteingress-protocol"
 version = "2.0.0"
 dependencies = [
+ "bytes",
+ "log",
  "tokio",
+ "tokio-util",
 ]
 
 [[package]]
@@ -668,6 +706,16 @@ version = "1.15.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "67b1b7a3b5fe4f1376887184045fcf45c69e92af734b7aaddc05fb777b6fbd03"
 
+[[package]]
+name = "socket2"
+version = "0.5.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e22376abed350d73dd1cd119b57ffccad95b4e585a7cda43e286245ce23c0678"
+dependencies = [
+ "libc",
+ "windows-sys 0.52.0",
+]
+
 [[package]]
 name = "socket2"
 version = "0.6.2"
@@ -732,7 +780,7 @@ dependencies = [
  "parking_lot",
  "pin-project-lite",
  "signal-hook-registry",
- "socket2",
+ "socket2 0.6.2",
  "tokio-macros",
  "windows-sys 0.61.2",
 ]
@@ -758,6 +806,19 @@ dependencies = [
  "tokio",
 ]
 
+[[package]]
+name = "tokio-util"
+version = "0.7.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9ae9cec805b01e8fc3fd2fe289f89149a9b66dd16786abd8b19cfa7b48cb0098"
+dependencies = [
+ "bytes",
+ "futures-core",
+ "futures-sink",
+ "pin-project-lite",
+ "tokio",
+]
+
 [[package]]
 name = "unicode-ident"
 version = "1.0.24"

rust/crates/remoteingress-bin/Cargo.toml

@@ -17,3 +17,4 @@ serde_json = "1"
 log = "0.4"
 env_logger = "0.11"
 rustls = { version = "0.23", default-features = false, features = ["ring"] }
+mimalloc = "0.1"

rust/crates/remoteingress-bin/src/main.rs

@@ -1,3 +1,6 @@
+#[global_allocator]
+static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
+
 use clap::Parser;
 use serde::{Deserialize, Serialize};
 use std::sync::Arc;
@@ -164,16 +167,16 @@ async fn handle_request(
                 tokio::spawn(async move {
                     while let Some(event) = event_rx.recv().await {
                         match &event {
-                            HubEvent::EdgeConnected { edge_id } => {
+                            HubEvent::EdgeConnected { edge_id, peer_addr } => {
                                 send_event(
                                     "edgeConnected",
-                                    serde_json::json!({ "edgeId": edge_id }),
+                                    serde_json::json!({ "edgeId": edge_id, "peerAddr": peer_addr }),
                                 );
                             }
-                            HubEvent::EdgeDisconnected { edge_id } => {
+                            HubEvent::EdgeDisconnected { edge_id, reason } => {
                                 send_event(
                                     "edgeDisconnected",
-                                    serde_json::json!({ "edgeId": edge_id }),
+                                    serde_json::json!({ "edgeId": edge_id, "reason": reason }),
                                 );
                             }
                             HubEvent::StreamOpened {
@@ -292,8 +295,8 @@ async fn handle_request(
                             EdgeEvent::TunnelConnected => {
                                 send_event("tunnelConnected", serde_json::json!({}));
                             }
-                            EdgeEvent::TunnelDisconnected => {
+                            EdgeEvent::TunnelDisconnected { reason } => {
-                                send_event("tunnelDisconnected", serde_json::json!({}));
+                                send_event("tunnelDisconnected", serde_json::json!({ "reason": reason }));
                             }
                             EdgeEvent::PublicIpDiscovered { ip } => {
                                 send_event(

rust/crates/remoteingress-core/Cargo.toml

@@ -7,9 +7,12 @@ edition = "2021"
 remoteingress-protocol = { path = "../remoteingress-protocol" }
 tokio = { version = "1", features = ["full"] }
 tokio-rustls = "0.26"
+bytes = "1"
 rustls = { version = "0.23", default-features = false, features = ["ring", "logging", "std", "tls12"] }
 rcgen = "0.13"
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
 log = "0.4"
 rustls-pemfile = "2"
+tokio-util = "0.7"
+socket2 = "0.5"

rust/crates/remoteingress-core/src/hub.rs

@@ -1,22 +1,51 @@
 use std::collections::HashMap;
 use std::sync::Arc;
-use tokio::io::{AsyncBufReadExt, AsyncReadExt, AsyncWriteExt, BufReader};
+use std::sync::atomic::{AtomicU32, Ordering};
+use std::time::Duration;
+use tokio::io::{AsyncReadExt, AsyncWriteExt};
 use tokio::net::{TcpListener, TcpStream};
-use tokio::sync::{mpsc, Mutex, RwLock};
+use tokio::sync::{mpsc, Mutex, Notify, RwLock, Semaphore};
+use tokio::time::{interval, sleep_until, Instant};
 use tokio_rustls::TlsAcceptor;
+use tokio_util::sync::CancellationToken;
 use serde::{Deserialize, Serialize};
 
+use bytes::Bytes;
 use remoteingress_protocol::*;
 
+type HubTlsStream = tokio_rustls::server::TlsStream<TcpStream>;
+
+/// Result of processing a frame.
+#[allow(dead_code)]
+enum FrameAction {
+    Continue,
+    Disconnect(String),
+}
+
+/// Per-stream state tracked in the hub's stream map.
+struct HubStreamState {
+    /// Unbounded channel to deliver FRAME_DATA payloads to the upstream writer task.
+    /// Unbounded because flow control (WINDOW_UPDATE) already limits bytes-in-flight.
+    /// A bounded channel would kill streams instead of applying backpressure.
+    data_tx: mpsc::UnboundedSender<Bytes>,
+    /// Cancellation token for this stream.
+    cancel_token: CancellationToken,
+    /// Send window for FRAME_DATA_BACK (download direction).
+    /// Decremented by the upstream reader, incremented by FRAME_WINDOW_UPDATE from edge.
+    send_window: Arc<AtomicU32>,
+    /// Notifier to wake the upstream reader when the window opens.
+    window_notify: Arc<Notify>,
+}
+
 /// Hub configuration.
 #[derive(Debug, Clone, Deserialize, Serialize)]
 #[serde(rename_all = "camelCase")]
 pub struct HubConfig {
     pub tunnel_port: u16,
     pub target_host: Option<String>,
-    #[serde(skip)]
+    #[serde(default)]
     pub tls_cert_pem: Option<String>,
-    #[serde(skip)]
+    #[serde(default)]
     pub tls_key_pem: Option<String>,
 }
 
@@ -64,6 +93,7 @@ pub struct ConnectedEdgeStatus {
     pub edge_id: String,
     pub connected_at: u64,
     pub active_streams: usize,
+    pub peer_addr: String,
 }
 
 /// Events emitted by the hub.
@@ -72,9 +102,9 @@ pub struct ConnectedEdgeStatus {
 #[serde(tag = "type")]
 pub enum HubEvent {
     #[serde(rename_all = "camelCase")]
-    EdgeConnected { edge_id: String },
+    EdgeConnected { edge_id: String, peer_addr: String },
     #[serde(rename_all = "camelCase")]
-    EdgeDisconnected { edge_id: String },
+    EdgeDisconnected { edge_id: String, reason: String },
     #[serde(rename_all = "camelCase")]
     StreamOpened { edge_id: String, stream_id: u32 },
     #[serde(rename_all = "camelCase")]
@@ -95,21 +125,25 @@ pub struct TunnelHub {
     config: RwLock<HubConfig>,
     allowed_edges: Arc<RwLock<HashMap<String, AllowedEdge>>>,
     connected_edges: Arc<Mutex<HashMap<String, ConnectedEdgeInfo>>>,
-    event_tx: mpsc::UnboundedSender<HubEvent>,
+    event_tx: mpsc::Sender<HubEvent>,
-    event_rx: Mutex<Option<mpsc::UnboundedReceiver<HubEvent>>>,
+    event_rx: Mutex<Option<mpsc::Receiver<HubEvent>>>,
     shutdown_tx: Mutex<Option<mpsc::Sender<()>>>,
     running: RwLock<bool>,
+    cancel_token: CancellationToken,
 }
 
 struct ConnectedEdgeInfo {
     connected_at: u64,
-    active_streams: Arc<Mutex<HashMap<u32, mpsc::Sender<Vec<u8>>>>>,
+    peer_addr: String,
+    edge_stream_count: Arc<AtomicU32>,
     config_tx: mpsc::Sender<EdgeConfigUpdate>,
+    /// Used to cancel the old connection when an edge reconnects.
+    cancel_token: CancellationToken,
 }
 
 impl TunnelHub {
     pub fn new(config: HubConfig) -> Self {
-        let (event_tx, event_rx) = mpsc::unbounded_channel();
+        let (event_tx, event_rx) = mpsc::channel(1024);
         Self {
             config: RwLock::new(config),
             allowed_edges: Arc::new(RwLock::new(HashMap::new())),
@@ -118,11 +152,12 @@ impl TunnelHub {
             event_rx: Mutex::new(Some(event_rx)),
             shutdown_tx: Mutex::new(None),
             running: RwLock::new(false),
+            cancel_token: CancellationToken::new(),
         }
     }
 
     /// Take the event receiver (can only be called once).
-    pub async fn take_event_rx(&self) -> Option<mpsc::UnboundedReceiver<HubEvent>> {
+    pub async fn take_event_rx(&self) -> Option<mpsc::Receiver<HubEvent>> {
         self.event_rx.lock().await.take()
     }
 
@@ -166,11 +201,11 @@ impl TunnelHub {
 
         let mut connected = Vec::new();
         for (id, info) in edges.iter() {
-            let streams = info.active_streams.lock().await;
             connected.push(ConnectedEdgeStatus {
                 edge_id: id.clone(),
                 connected_at: info.connected_at,
-                active_streams: streams.len(),
+                active_streams: info.edge_stream_count.load(Ordering::Relaxed) as usize,
+                peer_addr: info.peer_addr.clone(),
             });
         }
 
@@ -198,6 +233,7 @@ impl TunnelHub {
         let connected = self.connected_edges.clone();
         let event_tx = self.event_tx.clone();
         let target_host = config.target_host.unwrap_or_else(|| "127.0.0.1".to_string());
+        let hub_token = self.cancel_token.clone();
 
         tokio::spawn(async move {
             loop {
@@ -211,9 +247,11 @@ impl TunnelHub {
                                 let connected = connected.clone();
                                 let event_tx = event_tx.clone();
                                 let target = target_host.clone();
+                                let edge_token = hub_token.child_token();
+                                let peer_addr = addr.ip().to_string();
                                 tokio::spawn(async move {
                                     if let Err(e) = handle_edge_connection(
-                                        stream, acceptor, allowed, connected, event_tx, target,
+                                        stream, acceptor, allowed, connected, event_tx, target, edge_token, peer_addr,
                                     ).await {
                                         log::error!("Edge connection error: {}", e);
                                     }
@@ -224,6 +262,10 @@ impl TunnelHub {
                             }
                         }
                     }
+                    _ = hub_token.cancelled() => {
+                        log::info!("Hub shutting down (token cancelled)");
+                        break;
+                    }
                     _ = shutdown_rx.recv() => {
                         log::info!("Hub shutting down");
                         break;
@@ -237,6 +279,7 @@ impl TunnelHub {
 
     /// Stop the hub.
     pub async fn stop(&self) {
+        self.cancel_token.cancel();
         if let Some(tx) = self.shutdown_tx.lock().await.take() {
             let _ = tx.send(()).await;
         }
@@ -246,22 +289,367 @@ impl TunnelHub {
     }
 }
 
+impl Drop for TunnelHub {
+    fn drop(&mut self) {
+        self.cancel_token.cancel();
+    }
+}
+
+/// Maximum concurrent streams per edge connection.
+const MAX_STREAMS_PER_EDGE: usize = 1024;
+
+/// Process a single frame received from the edge side of the tunnel.
+/// Handles FRAME_OPEN, FRAME_DATA, FRAME_WINDOW_UPDATE, FRAME_CLOSE, and FRAME_PONG.
+async fn handle_hub_frame(
+    frame: Frame,
+    tunnel_io: &mut remoteingress_protocol::TunnelIo<HubTlsStream>,
+    streams: &mut HashMap<u32, HubStreamState>,
+    stream_semaphore: &Arc<Semaphore>,
+    edge_stream_count: &Arc<AtomicU32>,
+    edge_id: &str,
+    event_tx: &mpsc::Sender<HubEvent>,
+    ctrl_tx: &mpsc::Sender<Bytes>,
+    data_tx: &mpsc::Sender<Bytes>,
+    sustained_tx: &mpsc::Sender<Bytes>,
+    target_host: &str,
+    edge_token: &CancellationToken,
+    cleanup_tx: &mpsc::Sender<u32>,
+) -> FrameAction {
+    match frame.frame_type {
+        FRAME_OPEN => {
+            // A4: Check stream limit before processing
+            let permit = match stream_semaphore.clone().try_acquire_owned() {
+                Ok(p) => p,
+                Err(_) => {
+                    log::warn!("Edge {} exceeded max streams ({}), rejecting stream {}",
+                        edge_id, MAX_STREAMS_PER_EDGE, frame.stream_id);
+                    let close_frame = encode_frame(frame.stream_id, FRAME_CLOSE_BACK, &[]);
+                    tunnel_io.queue_ctrl(close_frame);
+                    return FrameAction::Continue;
+                }
+            };
+
+            // Payload is PROXY v1 header line
+            let proxy_header = String::from_utf8_lossy(&frame.payload).to_string();
+
+            // Parse destination port from PROXY header
+            let dest_port = parse_dest_port_from_proxy(&proxy_header).unwrap_or(443);
+
+            let stream_id = frame.stream_id;
+            let cleanup = cleanup_tx.clone();
+            let writer_tx = ctrl_tx.clone();   // control: CLOSE_BACK, WINDOW_UPDATE_BACK
+            let data_writer_tx = data_tx.clone(); // data: DATA_BACK
+            let sustained_writer_tx = sustained_tx.clone(); // sustained: DATA_BACK from elephant flows
+            let target = target_host.to_string();
+            let stream_token = edge_token.child_token();
+
+            let _ = event_tx.try_send(HubEvent::StreamOpened {
+                edge_id: edge_id.to_string(),
+                stream_id,
+            });
+
+            // Create channel for data from edge to this stream
+            let (stream_data_tx, mut stream_data_rx) = mpsc::unbounded_channel::<Bytes>();
+            // Adaptive initial window: scale with current stream count
+            // to keep total in-flight data within the 200MB budget.
+            let initial_window = compute_window_for_stream_count(
+                edge_stream_count.load(Ordering::Relaxed),
+            );
+            let send_window = Arc::new(AtomicU32::new(initial_window));
+            let window_notify = Arc::new(Notify::new());
+            streams.insert(stream_id, HubStreamState {
+                data_tx: stream_data_tx,
+                cancel_token: stream_token.clone(),
+                send_window: Arc::clone(&send_window),
+                window_notify: Arc::clone(&window_notify),
+            });
+
+            // Spawn task: connect to SmartProxy, send PROXY header, pipe data
+            let stream_counter = Arc::clone(edge_stream_count);
+            tokio::spawn(async move {
+                let _permit = permit; // hold semaphore permit until stream completes
+                stream_counter.fetch_add(1, Ordering::Relaxed);
+
+                let result = async {
+                    // A2: Connect to SmartProxy with timeout
+                    let mut upstream = tokio::time::timeout(
+                        Duration::from_secs(10),
+                        TcpStream::connect((target.as_str(), dest_port)),
+                    )
+                    .await
+                    .map_err(|_| -> Box<dyn std::error::Error + Send + Sync> {
+                        format!("connect to SmartProxy {}:{} timed out (10s)", target, dest_port).into()
+                    })??;
+
+                    upstream.set_nodelay(true)?;
+                    upstream.write_all(proxy_header.as_bytes()).await?;
+
+                    let (mut up_read, mut up_write) =
+                        upstream.into_split();
+
+                    // Forward data from edge (via channel) to SmartProxy
+                    // After writing to upstream, send WINDOW_UPDATE_BACK to edge
+                    let writer_token = stream_token.clone();
+                    let wub_tx = writer_tx.clone();
+                    let stream_counter_w = Arc::clone(&stream_counter);
+                    let writer_for_edge_data = tokio::spawn(async move {
+                        let mut consumed_since_update: u32 = 0;
+                        loop {
+                            tokio::select! {
+                                data = stream_data_rx.recv() => {
+                                    match data {
+                                        Some(data) => {
+                                            let len = data.len() as u32;
+                                            // Check cancellation alongside the write so we respond
+                                            // promptly to FRAME_CLOSE instead of blocking up to 60s.
+                                            let write_result = tokio::select! {
+                                                r = tokio::time::timeout(
+                                                    Duration::from_secs(60),
+                                                    up_write.write_all(&data),
+                                                ) => r,
+                                                _ = writer_token.cancelled() => break,
+                                            };
+                                            match write_result {
+                                                Ok(Ok(())) => {}
+                                                Ok(Err(_)) => break,
+                                                Err(_) => {
+                                                    log::warn!("Stream {} write to upstream timed out (60s)", stream_id);
+                                                    break;
+                                                }
+                                            }
+                                            // Track consumption for adaptive flow control.
+                                            // Increment capped to adaptive window to limit per-stream in-flight data.
+                                            consumed_since_update += len;
+                                            let adaptive_window = remoteingress_protocol::compute_window_for_stream_count(
+                                                stream_counter_w.load(Ordering::Relaxed),
+                                            );
+                                            let threshold = adaptive_window / 2;
+                                            if consumed_since_update >= threshold {
+                                                let increment = consumed_since_update.min(adaptive_window);
+                                                let frame = encode_window_update(stream_id, FRAME_WINDOW_UPDATE_BACK, increment);
+                                                // Use send().await for guaranteed delivery — dropping WINDOW_UPDATEs
+                                                // causes permanent flow stalls. Safe: runs in per-stream task, not main loop.
+                                                tokio::select! {
+                                                    result = wub_tx.send(frame) => {
+                                                        if result.is_ok() {
+                                                            consumed_since_update -= increment;
+                                                        }
+                                                    }
+                                                    _ = writer_token.cancelled() => break,
+                                                }
+                                            }
+                                        }
+                                        None => break,
+                                    }
+                                }
+                                _ = writer_token.cancelled() => break,
+                            }
+                        }
+                        // Send final window update for remaining consumed bytes
+                        if consumed_since_update > 0 {
+                            let frame = encode_window_update(stream_id, FRAME_WINDOW_UPDATE_BACK, consumed_since_update);
+                            tokio::select! {
+                                _ = wub_tx.send(frame) => {}
+                                _ = writer_token.cancelled() => {}
+                            }
+                        }
+                        let _ = up_write.shutdown().await;
+                    });
+
+                    // Forward data from SmartProxy back to edge via writer channel
+                    // with per-stream flow control (check send_window before reading).
+                    // Zero-copy: read payload directly after the header, then prepend header.
+                    let mut buf = vec![0u8; FRAME_HEADER_SIZE + 32768];
+                    let mut dl_bytes_sent: u64 = 0;
+                    let dl_start = tokio::time::Instant::now();
+                    let mut is_sustained = false;
+                    loop {
+                        // Wait for send window to have capacity (with stall timeout).
+                        // Safe pattern: register notified BEFORE checking the condition
+                        // to avoid missing a notify_one that fires between load and select.
+                        loop {
+                            let notified = window_notify.notified();
+                            tokio::pin!(notified);
+                            notified.as_mut().enable();
+                            let w = send_window.load(Ordering::Acquire);
+                            if w > 0 { break; }
+                            tokio::select! {
+                                _ = notified => continue,
+                                _ = stream_token.cancelled() => break,
+                                _ = tokio::time::sleep(Duration::from_secs(120)) => {
+                                    log::warn!("Stream {} download stalled (window empty for 120s)", stream_id);
+                                    break;
+                                }
+                            }
+                        }
+                        if stream_token.is_cancelled() { break; }
+
+                        // Limit read size to available window.
+                        // IMPORTANT: if window is 0 (stall timeout fired), we must NOT
+                        // read into an empty buffer — read(&mut buf[..0]) returns Ok(0)
+                        // which would be falsely interpreted as EOF.
+                        let w = send_window.load(Ordering::Acquire) as usize;
+                        if w == 0 {
+                            log::warn!("Stream {} download: window still 0 after stall timeout, closing", stream_id);
+                            break;
+                        }
+                        let max_read = w.min(32768);
+
+                        tokio::select! {
+                            read_result = up_read.read(&mut buf[FRAME_HEADER_SIZE..FRAME_HEADER_SIZE + max_read]) => {
+                                match read_result {
+                                    Ok(0) => break,
+                                    Ok(n) => {
+                                        send_window.fetch_sub(n as u32, Ordering::Release);
+                                        encode_frame_header(&mut buf, stream_id, FRAME_DATA_BACK, n);
+                                        let frame = Bytes::copy_from_slice(&buf[..FRAME_HEADER_SIZE + n]);
+                                        // Sustained classification: >2.5 MB/s for >10 seconds
+                                        dl_bytes_sent += n as u64;
+                                        if !is_sustained {
+                                            let elapsed = dl_start.elapsed().as_secs();
+                                            if elapsed >= remoteingress_protocol::SUSTAINED_MIN_DURATION_SECS
+                                                && dl_bytes_sent / elapsed >= remoteingress_protocol::SUSTAINED_THRESHOLD_BPS
+                                            {
+                                                is_sustained = true;
+                                                log::debug!("Stream {} classified as sustained (download, {} bytes in {}s)",
+                                                    stream_id, dl_bytes_sent, elapsed);
+                                            }
+                                        }
+                                        let tx = if is_sustained { &sustained_writer_tx } else { &data_writer_tx };
+                                        let sent = tokio::select! {
+                                            result = tx.send(frame) => result.is_ok(),
+                                            _ = stream_token.cancelled() => false,
+                                        };
+                                        if !sent { break; }
+                                    }
+                                    Err(_) => break,
+                                }
+                            }
+                            _ = stream_token.cancelled() => break,
+                        }
+                    }
+
+                    // Send CLOSE_BACK via same channel as DATA_BACK (must arrive AFTER last DATA_BACK).
+                    // select! with cancellation guard prevents indefinite blocking if tunnel dies.
+                    if !stream_token.is_cancelled() {
+                        let close_frame = encode_frame(stream_id, FRAME_CLOSE_BACK, &[]);
+                        let tx = if is_sustained { &sustained_writer_tx } else { &data_writer_tx };
+                        tokio::select! {
+                            _ = tx.send(close_frame) => {}
+                            _ = stream_token.cancelled() => {}
+                        }
+                    }
+
+                    writer_for_edge_data.abort();
+                    Ok::<(), Box<dyn std::error::Error + Send + Sync>>(())
+                }
+                .await;
+
+                if let Err(e) = result {
+                    log::error!("Stream {} error: {}", stream_id, e);
+                    // Send CLOSE_BACK on error (must arrive after any DATA_BACK).
+                    // Error path: is_sustained not available here, use data channel (safe —
+                    // if error occurs before classification, no sustained frames were sent).
+                    if !stream_token.is_cancelled() {
+                        let close_frame = encode_frame(stream_id, FRAME_CLOSE_BACK, &[]);
+                        tokio::select! {
+                            _ = data_writer_tx.send(close_frame) => {}
+                            _ = stream_token.cancelled() => {}
+                        }
+                    }
+                }
+
+                // Signal main loop to remove stream from the map.
+                // Cancellation guard prevents indefinite blocking if cleanup channel is full.
+                tokio::select! {
+                    _ = cleanup.send(stream_id) => {}
+                    _ = stream_token.cancelled() => {}
+                }
+                stream_counter.fetch_sub(1, Ordering::Relaxed);
+            });
+        }
+        FRAME_DATA => {
+            // Dispatch to per-stream unbounded channel. Flow control (WINDOW_UPDATE)
+            // limits bytes-in-flight, so the channel won't grow unbounded. send() only
+            // fails if the receiver is dropped (stream handler already exited).
+            if let Some(state) = streams.get(&frame.stream_id) {
+                if state.data_tx.send(frame.payload).is_err() {
+                    // Receiver dropped — stream handler already exited, clean up
+                    streams.remove(&frame.stream_id);
+                }
+            }
+        }
+        FRAME_WINDOW_UPDATE => {
+            // Edge consumed data — increase our send window for this stream
+            if let Some(increment) = decode_window_update(&frame.payload) {
+                if increment > 0 {
+                    if let Some(state) = streams.get(&frame.stream_id) {
+                        let prev = state.send_window.fetch_add(increment, Ordering::Release);
+                        if prev + increment > MAX_WINDOW_SIZE {
+                            state.send_window.store(MAX_WINDOW_SIZE, Ordering::Release);
+                        }
+                        state.window_notify.notify_one();
+                    }
+                }
+            }
+        }
+        FRAME_CLOSE => {
+            if let Some(state) = streams.remove(&frame.stream_id) {
+                state.cancel_token.cancel();
+                let _ = event_tx.try_send(HubEvent::StreamClosed {
+                    edge_id: edge_id.to_string(),
+                    stream_id: frame.stream_id,
+                });
+            }
+        }
+        FRAME_PONG => {
+            log::debug!("Received PONG from edge {}", edge_id);
+        }
+        _ => {
+            log::warn!("Unexpected frame type {} from edge", frame.frame_type);
+        }
+    }
+    FrameAction::Continue
+}
+
 /// Handle a single edge connection: authenticate, then enter frame loop.
 async fn handle_edge_connection(
     stream: TcpStream,
     acceptor: TlsAcceptor,
     allowed: Arc<RwLock<HashMap<String, AllowedEdge>>>,
     connected: Arc<Mutex<HashMap<String, ConnectedEdgeInfo>>>,
-    event_tx: mpsc::UnboundedSender<HubEvent>,
+    event_tx: mpsc::Sender<HubEvent>,
     target_host: String,
+    edge_token: CancellationToken,
+    peer_addr: String,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    let tls_stream = acceptor.accept(stream).await?;
+    // Disable Nagle's algorithm for low-latency control frames (PING/PONG, WINDOW_UPDATE)
-    let (read_half, mut write_half) = tokio::io::split(tls_stream);
+    stream.set_nodelay(true)?;
-    let mut buf_reader = BufReader::new(read_half);
+    // TCP keepalive detects silent network failures (NAT timeout, path change)
+    // faster than the 45s application-level liveness timeout.
+    let ka = socket2::TcpKeepalive::new()
+        .with_time(Duration::from_secs(30));
+    #[cfg(target_os = "linux")]
+    let ka = ka.with_interval(Duration::from_secs(10));
+    let _ = socket2::SockRef::from(&stream).set_tcp_keepalive(&ka);
+    let mut tls_stream = acceptor.accept(stream).await?;
 
-    // Read auth line: "EDGE <edgeId> <secret>\n"
+    // Byte-by-byte auth line reading (no BufReader).
-    let mut auth_line = String::new();
+    // Auth line: "EDGE <edgeId> <secret>\n"
-    buf_reader.read_line(&mut auth_line).await?;
+    let mut auth_buf = Vec::with_capacity(512);
+    loop {
+        let mut byte = [0u8; 1];
+        tls_stream.read_exact(&mut byte).await?;
+        if byte[0] == b'\n' {
+            break;
+        }
+        auth_buf.push(byte[0]);
+        if auth_buf.len() > 4096 {
+            return Err("auth line too long".into());
+        }
+    }
+    let auth_line = String::from_utf8(auth_buf)
+        .map_err(|_| "auth line not valid UTF-8")?;
     let auth_line = auth_line.trim();
 
     let parts: Vec<&str> = auth_line.splitn(3, ' ').collect();
@@ -288,9 +676,10 @@ async fn handle_edge_connection(
         }
     };
 
-    log::info!("Edge {} authenticated", edge_id);
+    log::info!("Edge {} authenticated from {}", edge_id, peer_addr);
-    let _ = event_tx.send(HubEvent::EdgeConnected {
+    let _ = event_tx.try_send(HubEvent::EdgeConnected {
         edge_id: edge_id.clone(),
+        peer_addr: peer_addr.clone(),
     });
 
     // Send handshake response with initial config before frame protocol begins
@@ -300,11 +689,15 @@ async fn handle_edge_connection(
     };
     let mut handshake_json = serde_json::to_string(&handshake)?;
     handshake_json.push('\n');
-    write_half.write_all(handshake_json.as_bytes()).await?;
+    tls_stream.write_all(handshake_json.as_bytes()).await?;
+    tls_stream.flush().await?;
 
     // Track this edge
-    let streams: Arc<Mutex<HashMap<u32, mpsc::Sender<Vec<u8>>>>> =
+    let mut streams: HashMap<u32, HubStreamState> = HashMap::new();
-        Arc::new(Mutex::new(HashMap::new()));
+    // Per-edge active stream counter for adaptive flow control
+    let edge_stream_count = Arc::new(AtomicU32::new(0));
+    // Cleanup channel: spawned stream tasks send stream_id here when done
+    let (cleanup_tx, mut cleanup_rx) = mpsc::channel::<u32>(256);
     let now = std::time::SystemTime::now()
         .duration_since(std::time::UNIX_EPOCH)
         .unwrap_or_default()
@@ -315,169 +708,182 @@ async fn handle_edge_connection(
 
     {
         let mut edges = connected.lock().await;
+        // If this edge already has an active connection (reconnect scenario),
+        // cancel the old connection so it shuts down immediately instead of
+        // lingering until TCP keepalive detects the dead socket.
+        if let Some(old) = edges.remove(&edge_id) {
+            log::info!("Edge {} reconnected, cancelling old connection", edge_id);
+            old.cancel_token.cancel();
+        }
         edges.insert(
             edge_id.clone(),
             ConnectedEdgeInfo {
                 connected_at: now,
-                active_streams: streams.clone(),
+                peer_addr,
+                edge_stream_count: edge_stream_count.clone(),
                 config_tx,
+                cancel_token: edge_token.clone(),
             },
         );
     }
 
-    // Shared writer for sending frames back to edge
+    // QoS dual-channel: ctrl frames have priority over data frames.
-    let write_half = Arc::new(Mutex::new(write_half));
+    // Stream handlers send through these channels -> TunnelIo drains them.
+    let (ctrl_tx, mut ctrl_rx) = mpsc::channel::<Bytes>(512);
+    let (data_tx, mut data_rx) = mpsc::channel::<Bytes>(4096);
+    let (sustained_tx, mut sustained_rx) = mpsc::channel::<Bytes>(4096);
 
     // Spawn task to forward config updates as FRAME_CONFIG frames
-    let config_writer = write_half.clone();
+    let config_writer_tx = ctrl_tx.clone();
     let config_edge_id = edge_id.clone();
+    let config_token = edge_token.clone();
     let config_handle = tokio::spawn(async move {
-        while let Some(update) = config_rx.recv().await {
+        loop {
-            if let Ok(payload) = serde_json::to_vec(&update) {
+            tokio::select! {
-                let frame = encode_frame(0, FRAME_CONFIG, &payload);
+                update = config_rx.recv() => {
-                let mut w = config_writer.lock().await;
+                    match update {
-                if w.write_all(&frame).await.is_err() {
+                        Some(update) => {
-                    log::error!("Failed to send config update to edge {}", config_edge_id);
+                            if let Ok(payload) = serde_json::to_vec(&update) {
-                    break;
+                                let frame = encode_frame(0, FRAME_CONFIG, &payload);
+                                if config_writer_tx.send(frame).await.is_err() {
+                                    log::error!("Failed to send config update to edge {}", config_edge_id);
+                                    break;
+                                }
+                                log::info!("Sent config update to edge {}: ports {:?}", config_edge_id, update.listen_ports);
+                            }
+                        }
+                        None => break,
+                    }
                 }
-                log::info!("Sent config update to edge {}: ports {:?}", config_edge_id, update.listen_ports);
+                _ = config_token.cancelled() => break,
             }
         }
     });
 
-    // Frame reading loop
+    // A4: Semaphore to limit concurrent streams per edge
-    let mut frame_reader = FrameReader::new(buf_reader);
+    let stream_semaphore = Arc::new(Semaphore::new(MAX_STREAMS_PER_EDGE));
 
-    loop {
+    // Heartbeat: periodic PING and liveness timeout
-        match frame_reader.next_frame().await {
+    let ping_interval_dur = Duration::from_secs(15);
-            Ok(Some(frame)) => {
+    let liveness_timeout_dur = Duration::from_secs(45);
-                match frame.frame_type {
+    let mut ping_ticker = interval(ping_interval_dur);
-                    FRAME_OPEN => {
+    ping_ticker.tick().await; // consume the immediate first tick
-                        // Payload is PROXY v1 header line
+    let mut last_activity = Instant::now();
-                        let proxy_header = String::from_utf8_lossy(&frame.payload).to_string();
+    let mut liveness_deadline = Box::pin(sleep_until(last_activity + liveness_timeout_dur));
 
-                        // Parse destination port from PROXY header
+    // Single-owner I/O engine — no tokio::io::split, no mutex
-                        let dest_port = parse_dest_port_from_proxy(&proxy_header).unwrap_or(443);
+    let mut tunnel_io = remoteingress_protocol::TunnelIo::new(tls_stream, Vec::new());
 
-                        let stream_id = frame.stream_id;
-                        let edge_id_clone = edge_id.clone();
-                        let event_tx_clone = event_tx.clone();
-                        let streams_clone = streams.clone();
-                        let writer_clone = write_half.clone();
-                        let target = target_host.clone();
 
-                        let _ = event_tx.send(HubEvent::StreamOpened {
+    // Assigned in every break path of the hub_loop before use at the end.
-                            edge_id: edge_id.clone(),
+    #[allow(unused_assignments)]
-                            stream_id,
+    let mut disconnect_reason = String::new();
-                        });
 
-                        // Create channel for data from edge to this stream
+    'hub_loop: loop {
-                        let (data_tx, mut data_rx) = mpsc::channel::<Vec<u8>>(256);
+        // Drain completed stream cleanups from spawned tasks
-                        {
+        while let Ok(stream_id) = cleanup_rx.try_recv() {
-                            let mut s = streams.lock().await;
+            if streams.remove(&stream_id).is_some() {
-                            s.insert(stream_id, data_tx);
+                let _ = event_tx.try_send(HubEvent::StreamClosed {
-                        }
+                    edge_id: edge_id.clone(),
+                    stream_id,
+                });
+            }
+        }
 
-                        // Spawn task: connect to SmartProxy, send PROXY header, pipe data
+        // Drain any buffered frames
-                        tokio::spawn(async move {
+        loop {
-                            let result = async {
+            let frame = match tunnel_io.try_parse_frame() {
-                                let mut upstream =
+                Some(Ok(f)) => f,
-                                    TcpStream::connect((target.as_str(), dest_port)).await?;
+                Some(Err(e)) => {
-                                upstream.write_all(proxy_header.as_bytes()).await?;
+                    log::error!("Edge {} frame error: {}", edge_id, e);
+                    disconnect_reason = format!("edge_frame_error: {}", e);
+                    break 'hub_loop;
+                }
+                None => break,
+            };
+            last_activity = Instant::now();
+            liveness_deadline.as_mut().reset(last_activity + liveness_timeout_dur);
+            if let FrameAction::Disconnect(reason) = handle_hub_frame(
+                frame, &mut tunnel_io, &mut streams, &stream_semaphore, &edge_stream_count,
+                &edge_id, &event_tx, &ctrl_tx, &data_tx, &sustained_tx, &target_host, &edge_token,
+                &cleanup_tx,
+            ).await {
+                disconnect_reason = reason;
+                break 'hub_loop;
+            }
+        }
 
-                                let (mut up_read, mut up_write) =
+        // Poll I/O: write(ctrl->data), flush, read, channels, timers
-                                    upstream.into_split();
+        let event = std::future::poll_fn(|cx| {
+            // Queue PING if ticker fires
+            if ping_ticker.poll_tick(cx).is_ready() {
+                tunnel_io.queue_ctrl(encode_frame(0, FRAME_PING, &[]));
+            }
+            tunnel_io.poll_step(cx, &mut ctrl_rx, &mut data_rx, &mut sustained_rx, &mut liveness_deadline, &edge_token)
+        }).await;
 
-                                // Forward data from edge (via channel) to SmartProxy
+        match event {
-                                let writer_for_edge_data = tokio::spawn(async move {
+            remoteingress_protocol::TunnelEvent::Frame(frame) => {
-                                    while let Some(data) = data_rx.recv().await {
+                last_activity = Instant::now();
-                                        if up_write.write_all(&data).await.is_err() {
+                liveness_deadline.as_mut().reset(last_activity + liveness_timeout_dur);
-                                            break;
+                if let FrameAction::Disconnect(reason) = handle_hub_frame(
-                                        }
+                    frame, &mut tunnel_io, &mut streams, &stream_semaphore, &edge_stream_count,
-                                    }
+                    &edge_id, &event_tx, &ctrl_tx, &data_tx, &sustained_tx, &target_host, &edge_token,
-                                    let _ = up_write.shutdown().await;
+                    &cleanup_tx,
-                                });
+                ).await {
-
+                    disconnect_reason = reason;
-                                // Forward data from SmartProxy back to edge
+                    break;
-                                let mut buf = vec![0u8; 32768];
-                                loop {
-                                    match up_read.read(&mut buf).await {
-                                        Ok(0) => break,
-                                        Ok(n) => {
-                                            let frame =
-                                                encode_frame(stream_id, FRAME_DATA_BACK, &buf[..n]);
-                                            let mut w = writer_clone.lock().await;
-                                            if w.write_all(&frame).await.is_err() {
-                                                break;
-                                            }
-                                        }
-                                        Err(_) => break,
-                                    }
-                                }
-
-                                // Send CLOSE_BACK to edge
-                                let close_frame = encode_frame(stream_id, FRAME_CLOSE_BACK, &[]);
-                                let mut w = writer_clone.lock().await;
-                                let _ = w.write_all(&close_frame).await;
-
-                                writer_for_edge_data.abort();
-                                Ok::<(), Box<dyn std::error::Error + Send + Sync>>(())
-                            }
-                            .await;
-
-                            if let Err(e) = result {
-                                log::error!("Stream {} error: {}", stream_id, e);
-                                // Send CLOSE_BACK on error
-                                let close_frame = encode_frame(stream_id, FRAME_CLOSE_BACK, &[]);
-                                let mut w = writer_clone.lock().await;
-                                let _ = w.write_all(&close_frame).await;
-                            }
-
-                            // Clean up stream
-                            {
-                                let mut s = streams_clone.lock().await;
-                                s.remove(&stream_id);
-                            }
-                            let _ = event_tx_clone.send(HubEvent::StreamClosed {
-                                edge_id: edge_id_clone,
-                                stream_id,
-                            });
-                        });
-                    }
-                    FRAME_DATA => {
-                        let s = streams.lock().await;
-                        if let Some(tx) = s.get(&frame.stream_id) {
-                            let _ = tx.send(frame.payload).await;
-                        }
-                    }
-                    FRAME_CLOSE => {
-                        let mut s = streams.lock().await;
-                        s.remove(&frame.stream_id);
-                    }
-                    _ => {
-                        log::warn!("Unexpected frame type {} from edge", frame.frame_type);
-                    }
                 }
             }
-            Ok(None) => {
+            remoteingress_protocol::TunnelEvent::Eof => {
                 log::info!("Edge {} disconnected (EOF)", edge_id);
+                disconnect_reason = "edge_eof".to_string();
                 break;
             }
-            Err(e) => {
+            remoteingress_protocol::TunnelEvent::ReadError(e) => {
                 log::error!("Edge {} frame error: {}", edge_id, e);
+                disconnect_reason = format!("edge_frame_error: {}", e);
+                break;
+            }
+            remoteingress_protocol::TunnelEvent::WriteError(e) => {
+                log::error!("Tunnel write error to edge {}: {}", edge_id, e);
+                disconnect_reason = format!("tunnel_write_error: {}", e);
+                break;
+            }
+            remoteingress_protocol::TunnelEvent::LivenessTimeout => {
+                log::warn!("Edge {} liveness timeout (no frames for {}s), disconnecting",
+                    edge_id, liveness_timeout_dur.as_secs());
+                disconnect_reason = "liveness_timeout".to_string();
+                break;
+            }
+            remoteingress_protocol::TunnelEvent::Cancelled => {
+                log::info!("Edge {} cancelled by hub", edge_id);
+                disconnect_reason = "cancelled_by_hub".to_string();
                 break;
             }
         }
     }
 
-    // Cleanup
+    // Cancel stream tokens FIRST so stream handlers exit immediately.
+    // If we TLS-shutdown first, stream handlers are stuck sending to dead channels
+    // for up to 2 seconds while the shutdown times out on a dead connection.
+    edge_token.cancel();
     config_handle.abort();
+
+    // Graceful TLS shutdown: send close_notify so the edge sees a clean disconnect.
+    // Stream handlers are already cancelled, so no new data is being produced.
+    let mut tls_stream = tunnel_io.into_inner();
+    let _ = tokio::time::timeout(
+        Duration::from_secs(2),
+        tls_stream.shutdown(),
+    ).await;
     {
         let mut edges = connected.lock().await;
         edges.remove(&edge_id);
     }
-    let _ = event_tx.send(HubEvent::EdgeDisconnected {
+    let _ = event_tx.try_send(HubEvent::EdgeDisconnected {
         edge_id: edge_id.clone(),
+        reason: disconnect_reason,
     });
 
     Ok(())
@@ -688,20 +1094,24 @@ mod tests {
     fn test_hub_event_edge_connected_serialize() {
         let event = HubEvent::EdgeConnected {
             edge_id: "edge-1".to_string(),
+            peer_addr: "203.0.113.5".to_string(),
         };
         let json = serde_json::to_value(&event).unwrap();
         assert_eq!(json["type"], "edgeConnected");
         assert_eq!(json["edgeId"], "edge-1");
+        assert_eq!(json["peerAddr"], "203.0.113.5");
     }
 
     #[test]
     fn test_hub_event_edge_disconnected_serialize() {
         let event = HubEvent::EdgeDisconnected {
             edge_id: "edge-2".to_string(),
+            reason: "liveness_timeout".to_string(),
         };
         let json = serde_json::to_value(&event).unwrap();
         assert_eq!(json["type"], "edgeDisconnected");
         assert_eq!(json["edgeId"], "edge-2");
+        assert_eq!(json["reason"], "liveness_timeout");
     }
 
     #[test]

rust/crates/remoteingress-protocol/Cargo.toml

@@ -4,4 +4,10 @@ version = "2.0.0"
 edition = "2021"
 
 [dependencies]
-tokio = { version = "1", features = ["io-util"] }
+tokio = { version = "1", features = ["io-util", "sync", "time"] }
+tokio-util = "0.7"
+bytes = "1"
+log = "0.4"
+
+[dev-dependencies]
+tokio = { version = "1", features = ["io-util", "macros", "rt"] }

rust/crates/remoteingress-protocol/src/lib.rs

@@ -1,4 +1,11 @@
-use tokio::io::{AsyncRead, AsyncReadExt};
+use std::collections::VecDeque;
+use std::future::Future;
+use std::pin::Pin;
+use std::task::{Context, Poll};
+use std::time::Duration;
+use bytes::{Bytes, BytesMut, BufMut};
+use tokio::io::{AsyncRead, AsyncReadExt, AsyncWrite, ReadBuf};
+use tokio::time::Instant;
 
 // Frame type constants
 pub const FRAME_OPEN: u8 = 0x01;
@@ -7,6 +14,10 @@ pub const FRAME_CLOSE: u8 = 0x03;
 pub const FRAME_DATA_BACK: u8 = 0x04;
 pub const FRAME_CLOSE_BACK: u8 = 0x05;
 pub const FRAME_CONFIG: u8 = 0x06;     // Hub -> Edge: configuration update
+pub const FRAME_PING: u8 = 0x07;       // Hub -> Edge: heartbeat probe
+pub const FRAME_PONG: u8 = 0x08;       // Edge -> Hub: heartbeat response
+pub const FRAME_WINDOW_UPDATE: u8 = 0x09;      // Edge -> Hub: per-stream flow control
+pub const FRAME_WINDOW_UPDATE_BACK: u8 = 0x0A;  // Hub -> Edge: per-stream flow control
 
 // Frame header size: 4 (stream_id) + 1 (type) + 4 (length) = 9 bytes
 pub const FRAME_HEADER_SIZE: usize = 9;
@@ -14,23 +25,72 @@ pub const FRAME_HEADER_SIZE: usize = 9;
 // Maximum payload size (16 MB)
 pub const MAX_PAYLOAD_SIZE: u32 = 16 * 1024 * 1024;
 
+// Per-stream flow control constants
+/// Initial (and maximum) per-stream window size (4 MB).
+pub const INITIAL_STREAM_WINDOW: u32 = 4 * 1024 * 1024;
+/// Send WINDOW_UPDATE after consuming this many bytes (half the initial window).
+pub const WINDOW_UPDATE_THRESHOLD: u32 = INITIAL_STREAM_WINDOW / 2;
+/// Maximum window size to prevent overflow.
+pub const MAX_WINDOW_SIZE: u32 = 4 * 1024 * 1024;
+
+// Sustained stream classification constants
+/// Throughput threshold for sustained classification (2.5 MB/s = 20 Mbit/s).
+pub const SUSTAINED_THRESHOLD_BPS: u64 = 2_500_000;
+/// Minimum duration before a stream can be classified as sustained.
+pub const SUSTAINED_MIN_DURATION_SECS: u64 = 10;
+/// Fixed window for sustained streams (1 MB — the floor).
+pub const SUSTAINED_WINDOW: u32 = 1 * 1024 * 1024;
+/// Maximum bytes written from sustained queue per forced drain (1 MB/s guarantee).
+pub const SUSTAINED_FORCED_DRAIN_CAP: usize = 1_048_576;
+
+/// Encode a WINDOW_UPDATE frame for a specific stream.
+pub fn encode_window_update(stream_id: u32, frame_type: u8, increment: u32) -> Bytes {
+    encode_frame(stream_id, frame_type, &increment.to_be_bytes())
+}
+
+/// Compute the target per-stream window size based on the number of active streams.
+/// Total memory budget is ~200MB shared across all streams. Up to 50 streams get the
+/// full 4MB window; above that the window scales down to a 1MB floor at 200+ streams.
+pub fn compute_window_for_stream_count(active: u32) -> u32 {
+    let per_stream = (200 * 1024 * 1024u64) / (active.max(1) as u64);
+    per_stream.clamp(1 * 1024 * 1024, INITIAL_STREAM_WINDOW as u64) as u32
+}
+
+/// Decode a WINDOW_UPDATE payload into a byte increment. Returns None if payload is malformed.
+pub fn decode_window_update(payload: &[u8]) -> Option<u32> {
+    if payload.len() != 4 {
+        return None;
+    }
+    Some(u32::from_be_bytes([payload[0], payload[1], payload[2], payload[3]]))
+}
+
 /// A single multiplexed frame.
 #[derive(Debug, Clone)]
 pub struct Frame {
     pub stream_id: u32,
     pub frame_type: u8,
-    pub payload: Vec<u8>,
+    pub payload: Bytes,
 }
 
 /// Encode a frame into bytes: [stream_id:4][type:1][length:4][payload]
-pub fn encode_frame(stream_id: u32, frame_type: u8, payload: &[u8]) -> Vec<u8> {
+pub fn encode_frame(stream_id: u32, frame_type: u8, payload: &[u8]) -> Bytes {
     let len = payload.len() as u32;
-    let mut buf = Vec::with_capacity(FRAME_HEADER_SIZE + payload.len());
+    let mut buf = BytesMut::with_capacity(FRAME_HEADER_SIZE + payload.len());
-    buf.extend_from_slice(&stream_id.to_be_bytes());
+    buf.put_slice(&stream_id.to_be_bytes());
-    buf.push(frame_type);
+    buf.put_u8(frame_type);
-    buf.extend_from_slice(&len.to_be_bytes());
+    buf.put_slice(&len.to_be_bytes());
-    buf.extend_from_slice(payload);
+    buf.put_slice(payload);
-    buf
+    buf.freeze()
+}
+
+/// Write a frame header into `buf[0..FRAME_HEADER_SIZE]`.
+/// The caller must ensure payload is already at `buf[FRAME_HEADER_SIZE..FRAME_HEADER_SIZE + payload_len]`.
+/// This enables zero-copy encoding: read directly into `buf[FRAME_HEADER_SIZE..]`, then
+/// prepend the header without copying the payload.
+pub fn encode_frame_header(buf: &mut [u8], stream_id: u32, frame_type: u8, payload_len: usize) {
+    buf[0..4].copy_from_slice(&stream_id.to_be_bytes());
+    buf[4] = frame_type;
+    buf[5..9].copy_from_slice(&(payload_len as u32).to_be_bytes());
 }
 
 /// Build a PROXY protocol v1 header line.
@@ -85,13 +145,17 @@ impl<R: AsyncRead + Unpin> FrameReader<R> {
         ]);
 
         if length > MAX_PAYLOAD_SIZE {
+            log::error!(
+                "CORRUPT FRAME HEADER: raw={:02x?} stream_id={} type=0x{:02x} length={}",
+                self.header_buf, stream_id, frame_type, length
+            );
             return Err(std::io::Error::new(
                 std::io::ErrorKind::InvalidData,
-                format!("frame payload too large: {} bytes", length),
+                format!("frame payload too large: {} bytes (header={:02x?})", length, self.header_buf),
             ));
         }
 
-        let mut payload = vec![0u8; length as usize];
+        let mut payload = BytesMut::zeroed(length as usize);
         if length > 0 {
             self.reader.read_exact(&mut payload).await?;
         }
@@ -99,7 +163,7 @@ impl<R: AsyncRead + Unpin> FrameReader<R> {
         Ok(Some(Frame {
             stream_id,
             frame_type,
-            payload,
+            payload: payload.freeze(),
         }))
     }
 
@@ -109,10 +173,409 @@ impl<R: AsyncRead + Unpin> FrameReader<R> {
     }
 }
 
+// ---------------------------------------------------------------------------
+// TunnelIo: single-owner I/O multiplexer for the TLS tunnel connection
+// ---------------------------------------------------------------------------
+
+/// Events produced by the TunnelIo event loop.
+#[derive(Debug)]
+pub enum TunnelEvent {
+    /// A complete frame was read from the remote side.
+    Frame(Frame),
+    /// The remote side closed the connection (EOF).
+    Eof,
+    /// A read error occurred.
+    ReadError(std::io::Error),
+    /// A write error occurred.
+    WriteError(std::io::Error),
+    /// No frames received for the liveness timeout duration.
+    LivenessTimeout,
+    /// The cancellation token was triggered.
+    Cancelled,
+}
+
+/// Write state extracted into a sub-struct so the borrow checker can see
+/// disjoint field access between `self.write` and `self.stream`.
+struct WriteState {
+    ctrl_queue: VecDeque<Bytes>,      // PONG, WINDOW_UPDATE, CLOSE, OPEN — always first
+    data_queue: VecDeque<Bytes>,      // DATA, DATA_BACK — only when ctrl is empty
+    sustained_queue: VecDeque<Bytes>, // DATA, DATA_BACK from sustained streams — lowest priority
+    offset: usize,                    // progress within current frame being written
+    flush_needed: bool,
+    // Sustained starvation prevention: guaranteed 1 MB/s drain
+    sustained_last_drain: Instant,
+    sustained_bytes_this_period: usize,
+}
+
+impl WriteState {
+    fn has_work(&self) -> bool {
+        !self.ctrl_queue.is_empty() || !self.data_queue.is_empty() || !self.sustained_queue.is_empty()
+    }
+}
+
+/// Single-owner I/O engine for the tunnel TLS connection.
+///
+/// Owns the TLS stream directly — no `tokio::io::split()`, no mutex.
+/// Uses three priority write queues:
+///   1. ctrl  (PONG, WINDOW_UPDATE, CLOSE, OPEN) — always first
+///   2. data  (DATA, DATA_BACK from normal streams) — when ctrl empty
+///   3. sustained (DATA, DATA_BACK from sustained streams) — lowest priority,
+///      drained freely when ctrl+data empty, or forced 1MB/s when they're not
+pub struct TunnelIo<S> {
+    stream: S,
+    // Read state: accumulate bytes, parse frames incrementally
+    read_buf: Vec<u8>,
+    read_pos: usize,
+    parse_pos: usize,
+    // Write state: extracted sub-struct for safe disjoint borrows
+    write: WriteState,
+}
+
+impl<S: AsyncRead + AsyncWrite + Unpin> TunnelIo<S> {
+    pub fn new(stream: S, initial_data: Vec<u8>) -> Self {
+        let read_pos = initial_data.len();
+        let mut read_buf = initial_data;
+        if read_buf.capacity() < 65536 {
+            read_buf.reserve(65536 - read_buf.len());
+        }
+        Self {
+            stream,
+            read_buf,
+            read_pos,
+            parse_pos: 0,
+            write: WriteState {
+                ctrl_queue: VecDeque::new(),
+                data_queue: VecDeque::new(),
+                sustained_queue: VecDeque::new(),
+                offset: 0,
+                flush_needed: false,
+                sustained_last_drain: Instant::now(),
+                sustained_bytes_this_period: 0,
+            },
+        }
+    }
+
+    /// Queue a high-priority control frame (PONG, WINDOW_UPDATE, CLOSE, OPEN).
+    pub fn queue_ctrl(&mut self, frame: Bytes) {
+        self.write.ctrl_queue.push_back(frame);
+    }
+
+    /// Queue a lower-priority data frame (DATA, DATA_BACK).
+    pub fn queue_data(&mut self, frame: Bytes) {
+        self.write.data_queue.push_back(frame);
+    }
+
+    /// Queue a lowest-priority sustained data frame.
+    pub fn queue_sustained(&mut self, frame: Bytes) {
+        self.write.sustained_queue.push_back(frame);
+    }
+
+    /// Try to parse a complete frame from the read buffer.
+    /// Uses a parse_pos cursor to avoid drain() on every frame.
+    pub fn try_parse_frame(&mut self) -> Option<Result<Frame, std::io::Error>> {
+        let available = self.read_pos - self.parse_pos;
+        if available < FRAME_HEADER_SIZE {
+            return None;
+        }
+
+        let base = self.parse_pos;
+        let stream_id = u32::from_be_bytes([
+            self.read_buf[base], self.read_buf[base + 1],
+            self.read_buf[base + 2], self.read_buf[base + 3],
+        ]);
+        let frame_type = self.read_buf[base + 4];
+        let length = u32::from_be_bytes([
+            self.read_buf[base + 5], self.read_buf[base + 6],
+            self.read_buf[base + 7], self.read_buf[base + 8],
+        ]);
+
+        if length > MAX_PAYLOAD_SIZE {
+            let header = [
+                self.read_buf[base], self.read_buf[base + 1],
+                self.read_buf[base + 2], self.read_buf[base + 3],
+                self.read_buf[base + 4], self.read_buf[base + 5],
+                self.read_buf[base + 6], self.read_buf[base + 7],
+                self.read_buf[base + 8],
+            ];
+            log::error!(
+                "CORRUPT FRAME HEADER: raw={:02x?} stream_id={} type=0x{:02x} length={}",
+                header, stream_id, frame_type, length
+            );
+            return Some(Err(std::io::Error::new(
+                std::io::ErrorKind::InvalidData,
+                format!("frame payload too large: {} bytes (header={:02x?})", length, header),
+            )));
+        }
+
+        let total_frame_size = FRAME_HEADER_SIZE + length as usize;
+        if available < total_frame_size {
+            return None;
+        }
+
+        let payload = Bytes::copy_from_slice(
+            &self.read_buf[base + FRAME_HEADER_SIZE..base + total_frame_size],
+        );
+        self.parse_pos += total_frame_size;
+
+        // Compact when parse_pos > half the data to reclaim memory
+        if self.parse_pos > self.read_pos / 2 && self.parse_pos > 0 {
+            self.read_buf.drain(..self.parse_pos);
+            self.read_pos -= self.parse_pos;
+            self.parse_pos = 0;
+        }
+
+        Some(Ok(Frame { stream_id, frame_type, payload }))
+    }
+
+    /// Poll-based I/O step. Returns Ready on events, Pending when idle.
+    ///
+    /// Order: write(ctrl->data->sustained) -> flush -> read -> channels -> timers
+    pub fn poll_step(
+        &mut self,
+        cx: &mut Context<'_>,
+        ctrl_rx: &mut tokio::sync::mpsc::Receiver<Bytes>,
+        data_rx: &mut tokio::sync::mpsc::Receiver<Bytes>,
+        sustained_rx: &mut tokio::sync::mpsc::Receiver<Bytes>,
+        liveness_deadline: &mut Pin<Box<tokio::time::Sleep>>,
+        cancel_token: &tokio_util::sync::CancellationToken,
+    ) -> Poll<TunnelEvent> {
+        // 1. WRITE: 3-tier priority — ctrl first, then data, then sustained.
+        //    Sustained drains freely when ctrl+data are empty.
+        //    Write one frame, set flush_needed, then flush must complete before
+        //    writing more. This prevents unbounded TLS session buffer growth.
+        //    Safe: `self.write` and `self.stream` are disjoint fields.
+        let mut writes = 0;
+        while self.write.has_work() && writes < 16 && !self.write.flush_needed {
+            // Pick queue: ctrl > data > sustained
+            let queue_id = if !self.write.ctrl_queue.is_empty() {
+                0 // ctrl
+            } else if !self.write.data_queue.is_empty() {
+                1 // data
+            } else {
+                2 // sustained
+            };
+            let frame = match queue_id {
+                0 => self.write.ctrl_queue.front().unwrap(),
+                1 => self.write.data_queue.front().unwrap(),
+                _ => self.write.sustained_queue.front().unwrap(),
+            };
+            let remaining = &frame[self.write.offset..];
+
+            match Pin::new(&mut self.stream).poll_write(cx, remaining) {
+                Poll::Ready(Ok(0)) => {
+                    log::error!("TunnelIo: poll_write returned 0 (write zero), ctrl_q={} data_q={} sustained_q={}",
+                        self.write.ctrl_queue.len(), self.write.data_queue.len(), self.write.sustained_queue.len());
+                    return Poll::Ready(TunnelEvent::WriteError(
+                        std::io::Error::new(std::io::ErrorKind::WriteZero, "write zero"),
+                    ));
+                }
+                Poll::Ready(Ok(n)) => {
+                    self.write.offset += n;
+                    self.write.flush_needed = true;
+                    if self.write.offset >= frame.len() {
+                        match queue_id {
+                            0 => { self.write.ctrl_queue.pop_front(); }
+                            1 => { self.write.data_queue.pop_front(); }
+                            _ => {
+                                self.write.sustained_queue.pop_front();
+                                self.write.sustained_last_drain = Instant::now();
+                                self.write.sustained_bytes_this_period = 0;
+                            }
+                        }
+                        self.write.offset = 0;
+                        writes += 1;
+                    }
+                }
+                Poll::Ready(Err(e)) => {
+                    log::error!("TunnelIo: poll_write error: {} (ctrl_q={} data_q={} sustained_q={})",
+                        e, self.write.ctrl_queue.len(), self.write.data_queue.len(), self.write.sustained_queue.len());
+                    return Poll::Ready(TunnelEvent::WriteError(e));
+                }
+                Poll::Pending => break,
+            }
+        }
+
+        // 1b. FORCED SUSTAINED DRAIN: when ctrl/data have work but sustained is waiting,
+        //     guarantee at least 1 MB/s by draining up to SUSTAINED_FORCED_DRAIN_CAP
+        //     once per second.
+        if !self.write.sustained_queue.is_empty()
+            && (!self.write.ctrl_queue.is_empty() || !self.write.data_queue.is_empty())
+            && !self.write.flush_needed
+        {
+            let now = Instant::now();
+            if now.duration_since(self.write.sustained_last_drain) >= Duration::from_secs(1) {
+                self.write.sustained_bytes_this_period = 0;
+                self.write.sustained_last_drain = now;
+
+                while !self.write.sustained_queue.is_empty()
+                    && self.write.sustained_bytes_this_period < SUSTAINED_FORCED_DRAIN_CAP
+                    && !self.write.flush_needed
+                {
+                    let frame = self.write.sustained_queue.front().unwrap();
+                    let remaining = &frame[self.write.offset..];
+                    match Pin::new(&mut self.stream).poll_write(cx, remaining) {
+                        Poll::Ready(Ok(0)) => {
+                            return Poll::Ready(TunnelEvent::WriteError(
+                                std::io::Error::new(std::io::ErrorKind::WriteZero, "write zero"),
+                            ));
+                        }
+                        Poll::Ready(Ok(n)) => {
+                            self.write.offset += n;
+                            self.write.flush_needed = true;
+                            self.write.sustained_bytes_this_period += n;
+                            if self.write.offset >= frame.len() {
+                                self.write.sustained_queue.pop_front();
+                                self.write.offset = 0;
+                            }
+                        }
+                        Poll::Ready(Err(e)) => {
+                            return Poll::Ready(TunnelEvent::WriteError(e));
+                        }
+                        Poll::Pending => break,
+                    }
+                }
+            }
+        }
+
+        // 2. FLUSH: push encrypted data from TLS session to TCP.
+        if self.write.flush_needed {
+            match Pin::new(&mut self.stream).poll_flush(cx) {
+                Poll::Ready(Ok(())) => {
+                    self.write.flush_needed = false;
+                }
+                Poll::Ready(Err(e)) => {
+                    log::error!("TunnelIo: poll_flush error: {}", e);
+                    return Poll::Ready(TunnelEvent::WriteError(e));
+                }
+                Poll::Pending => {} // TCP waker will notify us
+            }
+        }
+
+        // 3. READ: drain stream until Pending to ensure the TCP waker is always registered.
+        //    Without this loop, a Ready return with partial frame data would consume
+        //    the waker without re-registering it, causing the task to sleep until a
+        //    timer or channel wakes it (potentially 15+ seconds of lost reads).
+        loop {
+            // Compact if needed to make room for reads
+            if self.parse_pos > 0 && self.read_buf.len() - self.read_pos < 32768 {
+                self.read_buf.drain(..self.parse_pos);
+                self.read_pos -= self.parse_pos;
+                self.parse_pos = 0;
+            }
+            if self.read_buf.len() < self.read_pos + 32768 {
+                self.read_buf.resize(self.read_pos + 32768, 0);
+            }
+            let mut rbuf = ReadBuf::new(&mut self.read_buf[self.read_pos..]);
+            match Pin::new(&mut self.stream).poll_read(cx, &mut rbuf) {
+                Poll::Ready(Ok(())) => {
+                    let n = rbuf.filled().len();
+                    if n == 0 {
+                        return Poll::Ready(TunnelEvent::Eof);
+                    }
+                    self.read_pos += n;
+                    if let Some(result) = self.try_parse_frame() {
+                        return match result {
+                            Ok(frame) => Poll::Ready(TunnelEvent::Frame(frame)),
+                            Err(e) => Poll::Ready(TunnelEvent::ReadError(e)),
+                        };
+                    }
+                    // Partial data — loop to call poll_read again so the TCP
+                    // waker is re-registered when it finally returns Pending.
+                }
+                Poll::Ready(Err(e)) => {
+                    log::error!("TunnelIo: poll_read error: {}", e);
+                    return Poll::Ready(TunnelEvent::ReadError(e));
+                }
+                Poll::Pending => break,
+            }
+        }
+
+        // 4. CHANNELS: drain ctrl (always — priority), data (only if queue is small).
+        //    Ctrl frames must never be delayed — always drain fully.
+        //    Data frames are gated: keep data in the bounded channel for proper
+        //    backpressure when TLS writes are slow. Without this gate, the internal
+        //    data_queue (unbounded VecDeque) grows to hundreds of MB under throttle -> OOM.
+        let mut got_new = false;
+        loop {
+            match ctrl_rx.poll_recv(cx) {
+                Poll::Ready(Some(frame)) => { self.write.ctrl_queue.push_back(frame); got_new = true; }
+                Poll::Ready(None) => {
+                    return Poll::Ready(TunnelEvent::WriteError(
+                        std::io::Error::new(std::io::ErrorKind::BrokenPipe, "ctrl channel closed"),
+                    ));
+                }
+                Poll::Pending => break,
+            }
+        }
+        if self.write.data_queue.len() < 64 {
+            loop {
+                match data_rx.poll_recv(cx) {
+                    Poll::Ready(Some(frame)) => { self.write.data_queue.push_back(frame); got_new = true; }
+                    Poll::Ready(None) => {
+                        return Poll::Ready(TunnelEvent::WriteError(
+                            std::io::Error::new(std::io::ErrorKind::BrokenPipe, "data channel closed"),
+                        ));
+                    }
+                    Poll::Pending => break,
+                }
+            }
+        }
+        // Sustained channel: drain when sustained_queue is small (same backpressure pattern).
+        // Channel close is non-fatal — not all connections have sustained streams.
+        if self.write.sustained_queue.len() < 64 {
+            loop {
+                match sustained_rx.poll_recv(cx) {
+                    Poll::Ready(Some(frame)) => { self.write.sustained_queue.push_back(frame); got_new = true; }
+                    Poll::Ready(None) | Poll::Pending => break,
+                }
+            }
+        }
+
+        // 5. TIMERS
+        if liveness_deadline.as_mut().poll(cx).is_ready() {
+            return Poll::Ready(TunnelEvent::LivenessTimeout);
+        }
+        if cancel_token.is_cancelled() {
+            return Poll::Ready(TunnelEvent::Cancelled);
+        }
+
+        // 6. SELF-WAKE: only when flush is complete AND we have work.
+        //    When flush is Pending, the TCP write-readiness waker will notify us.
+        //    CRITICAL: do NOT self-wake when flush_needed — poll_write always returns
+        //    Ready (TLS buffers in-memory), so self-waking causes a tight spin loop
+        //    that fills the TLS session buffer unboundedly -> OOM -> ECONNRESET.
+        if !self.write.flush_needed && (got_new || self.write.has_work()) {
+            cx.waker().wake_by_ref();
+        }
+
+        Poll::Pending
+    }
+
+    pub fn into_inner(self) -> S {
+        self.stream
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
 
+    #[test]
+    fn test_encode_frame_header() {
+        let payload = b"hello";
+        let mut buf = vec![0u8; FRAME_HEADER_SIZE + payload.len()];
+        buf[FRAME_HEADER_SIZE..].copy_from_slice(payload);
+        encode_frame_header(&mut buf, 42, FRAME_DATA, payload.len());
+        assert_eq!(buf, &encode_frame(42, FRAME_DATA, payload)[..]);
+    }
+
+    #[test]
+    fn test_encode_frame_header_empty_payload() {
+        let mut buf = vec![0u8; FRAME_HEADER_SIZE];
+        encode_frame_header(&mut buf, 99, FRAME_CLOSE, 0);
+        assert_eq!(buf, &encode_frame(99, FRAME_CLOSE, &[])[..]);
+    }
+
     #[test]
     fn test_encode_frame() {
         let data = b"hello";
@@ -261,6 +724,8 @@ mod tests {
             FRAME_DATA_BACK,
             FRAME_CLOSE_BACK,
             FRAME_CONFIG,
+            FRAME_PING,
+            FRAME_PONG,
         ];
 
         let mut data = Vec::new();
@@ -276,7 +741,7 @@ mod tests {
             let frame = reader.next_frame().await.unwrap().unwrap();
             assert_eq!(frame.stream_id, i as u32);
             assert_eq!(frame.frame_type, ft);
-            assert_eq!(frame.payload, format!("payload_{}", i).as_bytes());
+            assert_eq!(&frame.payload[..], format!("payload_{}", i).as_bytes());
         }
 
         assert!(reader.next_frame().await.unwrap().is_none());
@@ -285,7 +750,7 @@ mod tests {
     #[tokio::test]
     async fn test_frame_reader_zero_length_payload() {
         let data = encode_frame(42, FRAME_CLOSE, &[]);
-        let cursor = std::io::Cursor::new(data);
+        let cursor = std::io::Cursor::new(data.to_vec());
         let mut reader = FrameReader::new(cursor);
 
         let frame = reader.next_frame().await.unwrap().unwrap();
@@ -293,4 +758,116 @@ mod tests {
         assert_eq!(frame.frame_type, FRAME_CLOSE);
         assert!(frame.payload.is_empty());
     }
+
+    #[test]
+    fn test_encode_frame_ping_pong() {
+        // PING: stream_id=0, empty payload (control frame)
+        let ping = encode_frame(0, FRAME_PING, &[]);
+        assert_eq!(ping[4], FRAME_PING);
+        assert_eq!(&ping[0..4], &0u32.to_be_bytes());
+        assert_eq!(ping.len(), FRAME_HEADER_SIZE);
+
+        // PONG: stream_id=0, empty payload (control frame)
+        let pong = encode_frame(0, FRAME_PONG, &[]);
+        assert_eq!(pong[4], FRAME_PONG);
+        assert_eq!(&pong[0..4], &0u32.to_be_bytes());
+        assert_eq!(pong.len(), FRAME_HEADER_SIZE);
+    }
+
+    // --- compute_window_for_stream_count tests ---
+
+    #[test]
+    fn test_adaptive_window_zero_streams() {
+        // 0 streams treated as 1: 200MB/1 -> clamped to 4MB max
+        assert_eq!(compute_window_for_stream_count(0), INITIAL_STREAM_WINDOW);
+    }
+
+    #[test]
+    fn test_adaptive_window_one_stream() {
+        assert_eq!(compute_window_for_stream_count(1), INITIAL_STREAM_WINDOW);
+    }
+
+    #[test]
+    fn test_adaptive_window_50_streams_full() {
+        // 200MB/50 = 4MB = exactly INITIAL_STREAM_WINDOW
+        assert_eq!(compute_window_for_stream_count(50), INITIAL_STREAM_WINDOW);
+    }
+
+    #[test]
+    fn test_adaptive_window_51_streams_starts_scaling() {
+        // 200MB/51 < 4MB — first value below max
+        let w = compute_window_for_stream_count(51);
+        assert!(w < INITIAL_STREAM_WINDOW);
+        assert_eq!(w, (200 * 1024 * 1024u64 / 51) as u32);
+    }
+
+    #[test]
+    fn test_adaptive_window_100_streams() {
+        // 200MB/100 = 2MB
+        assert_eq!(compute_window_for_stream_count(100), 2 * 1024 * 1024);
+    }
+
+    #[test]
+    fn test_adaptive_window_200_streams_at_floor() {
+        // 200MB/200 = 1MB = exactly the floor
+        assert_eq!(compute_window_for_stream_count(200), 1 * 1024 * 1024);
+    }
+
+    #[test]
+    fn test_adaptive_window_500_streams_clamped() {
+        // 200MB/500 = 0.4MB -> clamped up to 1MB floor
+        assert_eq!(compute_window_for_stream_count(500), 1 * 1024 * 1024);
+    }
+
+    #[test]
+    fn test_adaptive_window_max_u32() {
+        // Extreme: u32::MAX streams -> tiny value -> clamped to 1MB
+        assert_eq!(compute_window_for_stream_count(u32::MAX), 1 * 1024 * 1024);
+    }
+
+    #[test]
+    fn test_adaptive_window_monotonically_decreasing() {
+        let mut prev = compute_window_for_stream_count(1);
+        for n in [2, 10, 50, 51, 100, 200, 500, 1000] {
+            let w = compute_window_for_stream_count(n);
+            assert!(w <= prev, "window increased from {} to {} at n={}", prev, w, n);
+            prev = w;
+        }
+    }
+
+    #[test]
+    fn test_adaptive_window_total_budget_bounded() {
+        // active x per_stream_window should never exceed 200MB (+ clamp overhead for high N)
+        for n in [1, 10, 50, 100, 200] {
+            let w = compute_window_for_stream_count(n);
+            let total = w as u64 * n as u64;
+            assert!(total <= 200 * 1024 * 1024, "total {}MB exceeds budget at n={}", total / (1024*1024), n);
+        }
+    }
+
+    // --- encode/decode window_update roundtrip ---
+
+    #[test]
+    fn test_window_update_roundtrip() {
+        for &increment in &[0u32, 1, 64 * 1024, INITIAL_STREAM_WINDOW, MAX_WINDOW_SIZE, u32::MAX] {
+            let frame = encode_window_update(42, FRAME_WINDOW_UPDATE, increment);
+            assert_eq!(frame[4], FRAME_WINDOW_UPDATE);
+            let decoded = decode_window_update(&frame[FRAME_HEADER_SIZE..]);
+            assert_eq!(decoded, Some(increment));
+        }
+    }
+
+    #[test]
+    fn test_window_update_back_roundtrip() {
+        let frame = encode_window_update(7, FRAME_WINDOW_UPDATE_BACK, 1234567);
+        assert_eq!(frame[4], FRAME_WINDOW_UPDATE_BACK);
+        assert_eq!(decode_window_update(&frame[FRAME_HEADER_SIZE..]), Some(1234567));
+    }
+
+    #[test]
+    fn test_decode_window_update_malformed() {
+        assert_eq!(decode_window_update(&[]), None);
+        assert_eq!(decode_window_update(&[0, 0, 0]), None);
+        assert_eq!(decode_window_update(&[0, 0, 0, 0, 0]), None);
+    }
 }

test/test.flowcontrol.node.ts

@@ -0,0 +1,475 @@
+import { expect, tap } from '@push.rocks/tapbundle';
+import * as net from 'net';
+import * as crypto from 'crypto';
+import { RemoteIngressHub, RemoteIngressEdge } from '../ts/index.js';
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+/** Find N free ports by binding to port 0 and collecting OS-assigned ports. */
+async function findFreePorts(count: number): Promise<number[]> {
+  const servers: net.Server[] = [];
+  const ports: number[] = [];
+  for (let i = 0; i < count; i++) {
+    const server = net.createServer();
+    await new Promise<void>((resolve) => server.listen(0, '127.0.0.1', resolve));
+    ports.push((server.address() as net.AddressInfo).port);
+    servers.push(server);
+  }
+  await Promise.all(servers.map((s) => new Promise<void>((resolve) => s.close(() => resolve()))));
+  return ports;
+}
+
+type TrackingServer = net.Server & { destroyAll: () => void };
+
+/** Start a TCP echo server that tracks connections for force-close. */
+function startEchoServer(port: number, host: string): Promise<TrackingServer> {
+  return new Promise((resolve, reject) => {
+    const connections = new Set<net.Socket>();
+    const server = net.createServer((socket) => {
+      connections.add(socket);
+      socket.on('close', () => connections.delete(socket));
+
+      // Skip PROXY protocol v1 header line before echoing
+      let proxyHeaderParsed = false;
+      let pendingBuf = Buffer.alloc(0);
+      socket.on('data', (data: Buffer) => {
+        if (!proxyHeaderParsed) {
+          pendingBuf = Buffer.concat([pendingBuf, data]);
+          const idx = pendingBuf.indexOf('\r\n');
+          if (idx !== -1) {
+            proxyHeaderParsed = true;
+            const remainder = pendingBuf.subarray(idx + 2);
+            if (remainder.length > 0) {
+              socket.write(remainder);
+            }
+          }
+          return;
+        }
+        socket.write(data);
+      });
+      socket.on('error', () => {});
+    }) as TrackingServer;
+
+    server.destroyAll = () => {
+      for (const conn of connections) conn.destroy();
+      connections.clear();
+    };
+
+    server.on('error', reject);
+    server.listen(port, host, () => resolve(server));
+  });
+}
+
+/**
+ * Start a server that sends a large response immediately on first data received.
+ * Does NOT wait for end (the tunnel protocol has no half-close).
+ * On receiving first data chunk after PROXY header, sends responseSize bytes then closes.
+ */
+function startLargeResponseServer(port: number, host: string, responseSize: number): Promise<TrackingServer> {
+  return new Promise((resolve, reject) => {
+    const connections = new Set<net.Socket>();
+    const server = net.createServer((socket) => {
+      connections.add(socket);
+      socket.on('close', () => connections.delete(socket));
+
+      let proxyHeaderParsed = false;
+      let pendingBuf = Buffer.alloc(0);
+      let responseSent = false;
+
+      socket.on('data', (data: Buffer) => {
+        if (!proxyHeaderParsed) {
+          pendingBuf = Buffer.concat([pendingBuf, data]);
+          const idx = pendingBuf.indexOf('\r\n');
+          if (idx !== -1) {
+            proxyHeaderParsed = true;
+            const remainder = pendingBuf.subarray(idx + 2);
+            if (remainder.length > 0 && !responseSent) {
+              responseSent = true;
+              sendLargeResponse(socket, responseSize);
+            }
+          }
+          return;
+        }
+        if (!responseSent) {
+          responseSent = true;
+          sendLargeResponse(socket, responseSize);
+        }
+      });
+      socket.on('error', () => {});
+    }) as TrackingServer;
+
+    server.destroyAll = () => {
+      for (const conn of connections) conn.destroy();
+      connections.clear();
+    };
+
+    server.on('error', reject);
+    server.listen(port, host, () => resolve(server));
+  });
+}
+
+function sendLargeResponse(socket: net.Socket, totalBytes: number) {
+  const chunkSize = 32 * 1024;
+  let sent = 0;
+  const writeChunk = () => {
+    while (sent < totalBytes) {
+      const toWrite = Math.min(chunkSize, totalBytes - sent);
+      // Use a deterministic pattern for verification
+      const chunk = Buffer.alloc(toWrite, (sent % 256) & 0xff);
+      const canContinue = socket.write(chunk);
+      sent += toWrite;
+      if (!canContinue) {
+        socket.once('drain', writeChunk);
+        return;
+      }
+    }
+    socket.end();
+  };
+  writeChunk();
+}
+
+/** Force-close a server: destroy all connections, then close. */
+async function forceCloseServer(server: TrackingServer): Promise<void> {
+  server.destroyAll();
+  await new Promise<void>((resolve) => server.close(() => resolve()));
+}
+
+interface TestTunnel {
+  hub: RemoteIngressHub;
+  edge: RemoteIngressEdge;
+  edgePort: number;
+  cleanup: () => Promise<void>;
+}
+
+/**
+ * Start a full hub + edge tunnel.
+ * Edge binds to 127.0.0.1, upstream server binds to 127.0.0.2.
+ * Hub targetHost = 127.0.0.2 so hub -> upstream doesn't loop back to edge.
+ */
+async function startTunnel(edgePort: number, hubPort: number): Promise<TestTunnel> {
+  const hub = new RemoteIngressHub();
+  const edge = new RemoteIngressEdge();
+
+  await hub.start({
+    tunnelPort: hubPort,
+    targetHost: '127.0.0.2',
+  });
+
+  await hub.updateAllowedEdges([
+    { id: 'test-edge', secret: 'test-secret', listenPorts: [edgePort] },
+  ]);
+
+  const connectedPromise = new Promise<void>((resolve, reject) => {
+    const timeout = setTimeout(() => reject(new Error('Edge did not connect within 10s')), 10000);
+    edge.once('tunnelConnected', () => {
+      clearTimeout(timeout);
+      resolve();
+    });
+  });
+
+  await edge.start({
+    hubHost: '127.0.0.1',
+    hubPort,
+    edgeId: 'test-edge',
+    secret: 'test-secret',
+    bindAddress: '127.0.0.1',
+  });
+
+  await connectedPromise;
+  await new Promise((resolve) => setTimeout(resolve, 500));
+
+  return {
+    hub,
+    edge,
+    edgePort,
+    cleanup: async () => {
+      await edge.stop();
+      await hub.stop();
+    },
+  };
+}
+
+/**
+ * Send data through the tunnel and collect the echoed response.
+ */
+function sendAndReceive(port: number, data: Buffer, timeoutMs = 30000): Promise<Buffer> {
+  return new Promise((resolve, reject) => {
+    const chunks: Buffer[] = [];
+    let totalReceived = 0;
+    const expectedLength = data.length;
+    let settled = false;
+
+    const client = net.createConnection({ host: '127.0.0.1', port }, () => {
+      client.write(data);
+      client.end();
+    });
+
+    const timer = setTimeout(() => {
+      if (!settled) {
+        settled = true;
+        client.destroy();
+        reject(new Error(`Timeout after ${timeoutMs}ms — received ${totalReceived}/${expectedLength} bytes`));
+      }
+    }, timeoutMs);
+
+    client.on('data', (chunk: Buffer) => {
+      chunks.push(chunk);
+      totalReceived += chunk.length;
+      if (totalReceived >= expectedLength && !settled) {
+        settled = true;
+        clearTimeout(timer);
+        client.destroy();
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('end', () => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('error', (err) => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        reject(err);
+      }
+    });
+  });
+}
+
+/**
+ * Connect to the tunnel, send a small request, and collect a large response.
+ * Does NOT call end() — the tunnel has no half-close.
+ * Instead, collects until expectedResponseSize bytes arrive.
+ */
+function sendAndReceiveLarge(
+  port: number,
+  data: Buffer,
+  expectedResponseSize: number,
+  timeoutMs = 60000,
+): Promise<Buffer> {
+  return new Promise((resolve, reject) => {
+    const chunks: Buffer[] = [];
+    let totalReceived = 0;
+    let settled = false;
+
+    const client = net.createConnection({ host: '127.0.0.1', port }, () => {
+      client.write(data);
+      // Do NOT call client.end() — the server will respond immediately
+      // and the tunnel CLOSE will happen when the download finishes
+    });
+
+    const timer = setTimeout(() => {
+      if (!settled) {
+        settled = true;
+        client.destroy();
+        reject(new Error(`Timeout after ${timeoutMs}ms — received ${totalReceived}/${expectedResponseSize} bytes`));
+      }
+    }, timeoutMs);
+
+    client.on('data', (chunk: Buffer) => {
+      chunks.push(chunk);
+      totalReceived += chunk.length;
+      if (totalReceived >= expectedResponseSize && !settled) {
+        settled = true;
+        clearTimeout(timer);
+        client.destroy();
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('end', () => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('error', (err) => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        reject(err);
+      }
+    });
+  });
+}
+
+function sha256(buf: Buffer): string {
+  return crypto.createHash('sha256').update(buf).digest('hex');
+}
+
+// ---------------------------------------------------------------------------
+// Tests
+// ---------------------------------------------------------------------------
+
+let tunnel: TestTunnel;
+let echoServer: TrackingServer;
+let hubPort: number;
+let edgePort: number;
+
+tap.test('setup: start echo server and tunnel', async () => {
+  [hubPort, edgePort] = await findFreePorts(2);
+
+  echoServer = await startEchoServer(edgePort, '127.0.0.2');
+  tunnel = await startTunnel(edgePort, hubPort);
+
+  expect(tunnel.hub.running).toBeTrue();
+});
+
+tap.test('single stream: 32MB transfer exceeding initial 4MB window (multiple refills)', async () => {
+  const size = 32 * 1024 * 1024;
+  const data = crypto.randomBytes(size);
+  const expectedHash = sha256(data);
+
+  const received = await sendAndReceive(edgePort, data, 60000);
+
+  expect(received.length).toEqual(size);
+  expect(sha256(received)).toEqual(expectedHash);
+});
+
+tap.test('200 concurrent streams with 64KB each', async () => {
+  const streamCount = 200;
+  const payloadSize = 64 * 1024;
+
+  const promises = Array.from({ length: streamCount }, () => {
+    const data = crypto.randomBytes(payloadSize);
+    const hash = sha256(data);
+    return sendAndReceive(edgePort, data, 30000).then((received) => ({
+      sent: hash,
+      received: sha256(received),
+      sizeOk: received.length === payloadSize,
+    }));
+  });
+
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.sent !== r.received);
+
+  expect(failures.length).toEqual(0);
+});
+
+tap.test('512 concurrent streams at minimum window boundary (16KB each)', async () => {
+  const streamCount = 512;
+  const payloadSize = 16 * 1024;
+
+  const promises = Array.from({ length: streamCount }, () => {
+    const data = crypto.randomBytes(payloadSize);
+    const hash = sha256(data);
+    return sendAndReceive(edgePort, data, 60000).then((received) => ({
+      sent: hash,
+      received: sha256(received),
+      sizeOk: received.length === payloadSize,
+    }));
+  });
+
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.sent !== r.received);
+
+  expect(failures.length).toEqual(0);
+});
+
+tap.test('asymmetric transfer: 4KB request -> 4MB response', async () => {
+  // Swap to large-response server
+  await forceCloseServer(echoServer);
+  const responseSize = 4 * 1024 * 1024; // 4 MB
+  const largeServer = await startLargeResponseServer(edgePort, '127.0.0.2', responseSize);
+
+  try {
+    const requestData = crypto.randomBytes(4 * 1024); // 4 KB
+    const received = await sendAndReceiveLarge(edgePort, requestData, responseSize, 60000);
+    expect(received.length).toEqual(responseSize);
+  } finally {
+    // Always restore echo server even on failure
+    await forceCloseServer(largeServer);
+    echoServer = await startEchoServer(edgePort, '127.0.0.2');
+  }
+});
+
+tap.test('100 streams x 1MB each (100MB total exceeding 200MB budget)', async () => {
+  const streamCount = 100;
+  const payloadSize = 1 * 1024 * 1024;
+
+  const promises = Array.from({ length: streamCount }, () => {
+    const data = crypto.randomBytes(payloadSize);
+    const hash = sha256(data);
+    return sendAndReceive(edgePort, data, 120000).then((received) => ({
+      sent: hash,
+      received: sha256(received),
+      sizeOk: received.length === payloadSize,
+    }));
+  });
+
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.sent !== r.received);
+
+  expect(failures.length).toEqual(0);
+});
+
+tap.test('active stream counter tracks concurrent connections', async () => {
+  const N = 50;
+
+  // Open N connections and keep them alive (send data but don't close)
+  const sockets: net.Socket[] = [];
+  const connectPromises = Array.from({ length: N }, () => {
+    return new Promise<net.Socket>((resolve, reject) => {
+      const sock = net.createConnection({ host: '127.0.0.1', port: edgePort }, () => {
+        resolve(sock);
+      });
+      sock.on('error', () => {});
+      setTimeout(() => reject(new Error('connect timeout')), 5000);
+    });
+  });
+
+  const connected = await Promise.all(connectPromises);
+  sockets.push(...connected);
+
+  // Brief delay for stream registration to propagate
+  await new Promise((resolve) => setTimeout(resolve, 500));
+
+  // Verify the edge reports >= N active streams.
+  // This counter is the input to compute_window_for_stream_count(),
+  // so its accuracy determines whether adaptive window sizing is correct.
+  const status = await tunnel.edge.getStatus();
+  expect(status.activeStreams).toBeGreaterThanOrEqual(N);
+
+  // Clean up: destroy all sockets (the tunnel's 300s stream timeout will handle cleanup)
+  for (const sock of sockets) {
+    sock.destroy();
+  }
+});
+
+tap.test('50 streams x 2MB each (forces multiple window refills per stream)', async () => {
+  // At 50 concurrent streams: adaptive window = 200MB/50 = 4MB per stream
+  // Each stream sends 2MB → needs ~3 WINDOW_UPDATE refill cycles per stream
+  const streamCount = 50;
+  const payloadSize = 2 * 1024 * 1024;
+
+  const promises = Array.from({ length: streamCount }, () => {
+    const data = crypto.randomBytes(payloadSize);
+    const hash = sha256(data);
+    return sendAndReceive(edgePort, data, 120000).then((received) => ({
+      sent: hash,
+      received: sha256(received),
+      sizeOk: received.length === payloadSize,
+    }));
+  });
+
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.sent !== r.received);
+
+  expect(failures.length).toEqual(0);
+});
+
+tap.test('teardown: stop tunnel and echo server', async () => {
+  await tunnel.cleanup();
+  await forceCloseServer(echoServer);
+});
+
+export default tap.start();

test/test.loadtest.node.ts

@@ -0,0 +1,402 @@
+import { expect, tap } from '@push.rocks/tapbundle';
+import * as net from 'net';
+import * as stream from 'stream';
+import * as crypto from 'crypto';
+import { RemoteIngressHub, RemoteIngressEdge } from '../ts/index.js';
+
+// ---------------------------------------------------------------------------
+// Helpers (self-contained — same patterns as test.flowcontrol.node.ts)
+// ---------------------------------------------------------------------------
+
+async function findFreePorts(count: number): Promise<number[]> {
+  const servers: net.Server[] = [];
+  const ports: number[] = [];
+  for (let i = 0; i < count; i++) {
+    const server = net.createServer();
+    await new Promise<void>((resolve) => server.listen(0, '127.0.0.1', resolve));
+    ports.push((server.address() as net.AddressInfo).port);
+    servers.push(server);
+  }
+  await Promise.all(servers.map((s) => new Promise<void>((resolve) => s.close(() => resolve()))));
+  return ports;
+}
+
+type TrackingServer = net.Server & { destroyAll: () => void };
+
+function startEchoServer(port: number, host: string): Promise<TrackingServer> {
+  return new Promise((resolve, reject) => {
+    const connections = new Set<net.Socket>();
+    const server = net.createServer((socket) => {
+      connections.add(socket);
+      socket.on('close', () => connections.delete(socket));
+      let proxyHeaderParsed = false;
+      let pendingBuf = Buffer.alloc(0);
+      socket.on('data', (data: Buffer) => {
+        if (!proxyHeaderParsed) {
+          pendingBuf = Buffer.concat([pendingBuf, data]);
+          const idx = pendingBuf.indexOf('\r\n');
+          if (idx !== -1) {
+            proxyHeaderParsed = true;
+            const remainder = pendingBuf.subarray(idx + 2);
+            if (remainder.length > 0) socket.write(remainder);
+          }
+          return;
+        }
+        socket.write(data);
+      });
+      socket.on('error', () => {});
+    }) as TrackingServer;
+    server.destroyAll = () => {
+      for (const conn of connections) conn.destroy();
+      connections.clear();
+    };
+    server.on('error', reject);
+    server.listen(port, host, () => resolve(server));
+  });
+}
+
+function sendAndReceive(port: number, data: Buffer, timeoutMs = 30000): Promise<Buffer> {
+  return new Promise((resolve, reject) => {
+    const chunks: Buffer[] = [];
+    let totalReceived = 0;
+    const expectedLength = data.length;
+    let settled = false;
+
+    const client = net.createConnection({ host: '127.0.0.1', port }, () => {
+      client.write(data);
+      client.end();
+    });
+
+    const timer = setTimeout(() => {
+      if (!settled) {
+        settled = true;
+        client.destroy();
+        reject(new Error(`Timeout after ${timeoutMs}ms — received ${totalReceived}/${expectedLength} bytes`));
+      }
+    }, timeoutMs);
+
+    client.on('data', (chunk: Buffer) => {
+      chunks.push(chunk);
+      totalReceived += chunk.length;
+      if (totalReceived >= expectedLength && !settled) {
+        settled = true;
+        clearTimeout(timer);
+        client.destroy();
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('end', () => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        resolve(Buffer.concat(chunks));
+      }
+    });
+
+    client.on('error', (err) => {
+      if (!settled) {
+        settled = true;
+        clearTimeout(timer);
+        reject(err);
+      }
+    });
+  });
+}
+
+function sha256(buf: Buffer): string {
+  return crypto.createHash('sha256').update(buf).digest('hex');
+}
+
+// ---------------------------------------------------------------------------
+// Throttle Proxy: rate-limits TCP traffic between edge and hub
+// ---------------------------------------------------------------------------
+
+class ThrottleTransform extends stream.Transform {
+  private bytesPerSec: number;
+  private bucket: number;
+  private lastRefill: number;
+  private destroyed_: boolean = false;
+
+  constructor(bytesPerSecond: number) {
+    super();
+    this.bytesPerSec = bytesPerSecond;
+    this.bucket = bytesPerSecond;
+    this.lastRefill = Date.now();
+  }
+
+  _transform(chunk: Buffer, _encoding: BufferEncoding, callback: stream.TransformCallback) {
+    if (this.destroyed_) return;
+
+    const now = Date.now();
+    const elapsed = (now - this.lastRefill) / 1000;
+    this.bucket = Math.min(this.bytesPerSec, this.bucket + elapsed * this.bytesPerSec);
+    this.lastRefill = now;
+
+    if (chunk.length <= this.bucket) {
+      this.bucket -= chunk.length;
+      callback(null, chunk);
+    } else {
+      // Not enough budget — delay the entire chunk (don't split)
+      const deficit = chunk.length - this.bucket;
+      this.bucket = 0;
+      const delayMs = Math.min((deficit / this.bytesPerSec) * 1000, 1000);
+      setTimeout(() => {
+        if (this.destroyed_) { callback(); return; }
+        this.lastRefill = Date.now();
+        this.bucket = 0;
+        callback(null, chunk);
+      }, delayMs);
+    }
+  }
+
+  _destroy(err: Error | null, callback: (error: Error | null) => void) {
+    this.destroyed_ = true;
+    callback(err);
+  }
+}
+
+interface ThrottleProxy {
+  server: net.Server;
+  close: () => Promise<void>;
+}
+
+async function startThrottleProxy(
+  listenPort: number,
+  targetHost: string,
+  targetPort: number,
+  bytesPerSecond: number,
+): Promise<ThrottleProxy> {
+  const connections = new Set<net.Socket>();
+  const server = net.createServer((clientSock) => {
+    connections.add(clientSock);
+    const upstream = net.createConnection({ host: targetHost, port: targetPort });
+    connections.add(upstream);
+
+    const throttleUp = new ThrottleTransform(bytesPerSecond);
+    const throttleDown = new ThrottleTransform(bytesPerSecond);
+
+    clientSock.pipe(throttleUp).pipe(upstream);
+    upstream.pipe(throttleDown).pipe(clientSock);
+
+    let cleaned = false;
+    const cleanup = (source: string, err?: Error) => {
+      if (cleaned) return;
+      cleaned = true;
+      if (err) {
+        console.error(`[ThrottleProxy] cleanup triggered by ${source}: ${err.message}`);
+      } else {
+        console.error(`[ThrottleProxy] cleanup triggered by ${source} (no error)`);
+      }
+      console.error(`[ThrottleProxy] stack:`, new Error().stack);
+      throttleUp.destroy();
+      throttleDown.destroy();
+      clientSock.destroy();
+      upstream.destroy();
+      connections.delete(clientSock);
+      connections.delete(upstream);
+    };
+    clientSock.on('error', (e) => cleanup('clientSock.error', e));
+    upstream.on('error', (e) => cleanup('upstream.error', e));
+    throttleUp.on('error', (e) => cleanup('throttleUp.error', e));
+    throttleDown.on('error', (e) => cleanup('throttleDown.error', e));
+    clientSock.on('close', () => cleanup('clientSock.close'));
+    upstream.on('close', () => cleanup('upstream.close'));
+  });
+
+  await new Promise<void>((resolve) => server.listen(listenPort, '127.0.0.1', resolve));
+  return {
+    server,
+    close: async () => {
+      for (const c of connections) c.destroy();
+      connections.clear();
+      await new Promise<void>((resolve) => server.close(() => resolve()));
+    },
+  };
+}
+
+// ---------------------------------------------------------------------------
+// Test state
+// ---------------------------------------------------------------------------
+
+let hub: RemoteIngressHub;
+let edge: RemoteIngressEdge;
+let echoServer: TrackingServer;
+let throttle: ThrottleProxy;
+let hubPort: number;
+let proxyPort: number;
+let edgePort: number;
+
+// ---------------------------------------------------------------------------
+// Tests
+// ---------------------------------------------------------------------------
+
+tap.test('setup: start throttled tunnel (100 Mbit/s)', async () => {
+  [hubPort, proxyPort, edgePort] = await findFreePorts(3);
+
+  echoServer = await startEchoServer(edgePort, '127.0.0.2');
+
+  // Throttle proxy: edge → proxy → hub at 100 Mbit/s (12.5 MB/s)
+  throttle = await startThrottleProxy(proxyPort, '127.0.0.1', hubPort, 12.5 * 1024 * 1024);
+
+  hub = new RemoteIngressHub();
+  edge = new RemoteIngressEdge();
+
+  await hub.start({ tunnelPort: hubPort, targetHost: '127.0.0.2' });
+  await hub.updateAllowedEdges([
+    { id: 'test-edge', secret: 'test-secret', listenPorts: [edgePort] },
+  ]);
+
+  const connectedPromise = new Promise<void>((resolve, reject) => {
+    const timeout = setTimeout(() => reject(new Error('Edge did not connect within 10s')), 10000);
+    edge.once('tunnelConnected', () => {
+      clearTimeout(timeout);
+      resolve();
+    });
+  });
+
+  // Edge connects through throttle proxy
+  await edge.start({
+    hubHost: '127.0.0.1',
+    hubPort: proxyPort,
+    edgeId: 'test-edge',
+    secret: 'test-secret',
+    bindAddress: '127.0.0.1',
+  });
+
+  await connectedPromise;
+  await new Promise((resolve) => setTimeout(resolve, 500));
+
+  const status = await edge.getStatus();
+  expect(status.connected).toBeTrue();
+});
+
+tap.test('throttled: 5 streams x 20MB each through 100Mbit tunnel', async () => {
+  const streamCount = 5;
+  const payloadSize = 20 * 1024 * 1024; // 20MB per stream = 100MB total round-trip
+
+  const payloads = Array.from({ length: streamCount }, () => crypto.randomBytes(payloadSize));
+  const promises = payloads.map((data) => {
+    const hash = sha256(data);
+    return sendAndReceive(edgePort, data, 300000).then((received) => ({
+      sent: hash,
+      received: sha256(received),
+      sizeOk: received.length === payloadSize,
+    }));
+  });
+
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.sent !== r.received);
+  expect(failures.length).toEqual(0);
+
+  const status = await edge.getStatus();
+  expect(status.connected).toBeTrue();
+});
+
+tap.test('throttled: slow consumer with 20MB does not kill other streams', async () => {
+  // Open a connection that creates download-direction backpressure:
+  // send 20MB but DON'T read the response — client TCP receive buffer fills
+  const slowSock = net.createConnection({ host: '127.0.0.1', port: edgePort });
+  await new Promise<void>((resolve) => slowSock.on('connect', resolve));
+  const slowData = crypto.randomBytes(20 * 1024 * 1024);
+  slowSock.write(slowData);
+  slowSock.end();
+  // Don't read — backpressure builds on the download path
+
+  // Wait for backpressure to develop
+  await new Promise((r) => setTimeout(r, 2000));
+
+  // Meanwhile, 5 normal echo streams with 20MB each must complete
+  const payload = crypto.randomBytes(20 * 1024 * 1024);
+  const hash = sha256(payload);
+  const promises = Array.from({ length: 5 }, () =>
+    sendAndReceive(edgePort, payload, 300000).then((r) => ({
+      hash: sha256(r),
+      sizeOk: r.length === payload.length,
+    }))
+  );
+  const results = await Promise.all(promises);
+  const failures = results.filter((r) => !r.sizeOk || r.hash !== hash);
+  expect(failures.length).toEqual(0);
+
+  // Tunnel still alive
+  const status = await edge.getStatus();
+  expect(status.connected).toBeTrue();
+
+  slowSock.destroy();
+});
+
+tap.test('throttled: rapid churn — 3 x 20MB long + 50 x 1MB short streams', async () => {
+  // 3 long streams (20MB each) running alongside 50 short streams (1MB each)
+  const longPayload = crypto.randomBytes(20 * 1024 * 1024);
+  const longHash = sha256(longPayload);
+  const longPromises = Array.from({ length: 3 }, () =>
+    sendAndReceive(edgePort, longPayload, 300000).then((r) => ({
+      hash: sha256(r),
+      sizeOk: r.length === longPayload.length,
+    }))
+  );
+
+  const shortPayload = crypto.randomBytes(1024 * 1024);
+  const shortHash = sha256(shortPayload);
+  const shortPromises = Array.from({ length: 50 }, () =>
+    sendAndReceive(edgePort, shortPayload, 300000).then((r) => ({
+      hash: sha256(r),
+      sizeOk: r.length === shortPayload.length,
+    }))
+  );
+
+  const [longResults, shortResults] = await Promise.all([
+    Promise.all(longPromises),
+    Promise.all(shortPromises),
+  ]);
+
+  const longFails = longResults.filter((r) => !r.sizeOk || r.hash !== longHash);
+  const shortFails = shortResults.filter((r) => !r.sizeOk || r.hash !== shortHash);
+  expect(longFails.length).toEqual(0);
+  expect(shortFails.length).toEqual(0);
+
+  const status = await edge.getStatus();
+  expect(status.connected).toBeTrue();
+});
+
+tap.test('throttled: 3 burst waves of 5 streams x 20MB each', async () => {
+  for (let wave = 0; wave < 3; wave++) {
+    const streamCount = 5;
+    const payloadSize = 20 * 1024 * 1024; // 20MB per stream = 100MB per wave
+
+    const promises = Array.from({ length: streamCount }, () => {
+      const data = crypto.randomBytes(payloadSize);
+      return sendAndReceive(edgePort, data, 300000).then((r) => r.length === payloadSize);
+    });
+
+    const results = await Promise.all(promises);
+    const ok = results.filter(Boolean).length;
+    expect(ok).toEqual(streamCount);
+
+    // Brief pause between waves
+    await new Promise((r) => setTimeout(r, 500));
+
+    const status = await edge.getStatus();
+    expect(status.connected).toBeTrue();
+  }
+});
+
+tap.test('throttled: tunnel still works after all load tests', async () => {
+  const data = crypto.randomBytes(1024);
+  const hash = sha256(data);
+  const received = await sendAndReceive(edgePort, data, 30000);
+  expect(sha256(received)).toEqual(hash);
+
+  const status = await edge.getStatus();
+  expect(status.connected).toBeTrue();
+});
+
+tap.test('teardown: stop tunnel', async () => {
+  await edge.stop();
+  await hub.stop();
+  if (throttle) await throttle.close();
+  await new Promise<void>((resolve) => echoServer.close(() => resolve()));
+});
+
+export default tap.start();

ts/00_commitinfo_data.ts

@@ -3,6 +3,6 @@
  */
 export const commitinfo = {
   name: '@serve.zone/remoteingress',
-  version: '3.2.1',
+  version: '4.9.0',
   description: 'Edge ingress tunnel for DcRouter - accepts incoming TCP connections at network edge and tunnels them to DcRouter SmartProxy preserving client IP via PROXY protocol v1.'
 }

ts/classes.remoteingressedge.ts

@@ -14,6 +14,7 @@ type TEdgeCommands = {
       hubPort: number;
       edgeId: string;
       secret: string;
+      bindAddress?: string;
     };
     result: { started: boolean };
   };
@@ -38,11 +39,19 @@ export interface IEdgeConfig {
   hubPort?: number;
   edgeId: string;
   secret: string;
+  bindAddress?: string;
 }
 
+const MAX_RESTART_ATTEMPTS = 10;
+const MAX_RESTART_BACKOFF_MS = 30_000;
+
 export class RemoteIngressEdge extends EventEmitter {
   private bridge: InstanceType<typeof plugins.smartrust.RustBridge<TEdgeCommands>>;
   private started = false;
+  private stopping = false;
+  private savedConfig: IEdgeConfig | null = null;
+  private restartBackoffMs = 1000;
+  private restartAttempts = 0;
   private statusInterval: ReturnType<typeof setInterval> | undefined;
 
   constructor() {
@@ -74,8 +83,10 @@ export class RemoteIngressEdge extends EventEmitter {
     this.bridge.on('management:tunnelConnected', () => {
       this.emit('tunnelConnected');
     });
-    this.bridge.on('management:tunnelDisconnected', () => {
+    this.bridge.on('management:tunnelDisconnected', (data: { reason?: string }) => {
-      this.emit('tunnelDisconnected');
+      const reason = data?.reason ?? 'unknown';
+      console.log(`[RemoteIngressEdge] Tunnel disconnected: ${reason}`);
+      this.emit('tunnelDisconnected', data);
     });
     this.bridge.on('management:publicIpDiscovered', (data: { ip: string }) => {
       this.emit('publicIpDiscovered', data);
@@ -109,19 +120,28 @@ export class RemoteIngressEdge extends EventEmitter {
       edgeConfig = config;
     }
 
+    this.savedConfig = edgeConfig;
+    this.stopping = false;
+
     const spawned = await this.bridge.spawn();
     if (!spawned) {
       throw new Error('Failed to spawn remoteingress-bin');
     }
 
+    // Register crash recovery handler
+    this.bridge.on('exit', this.handleCrashRecovery);
+
     await this.bridge.sendCommand('startEdge', {
       hubHost: edgeConfig.hubHost,
       hubPort: edgeConfig.hubPort ?? 8443,
       edgeId: edgeConfig.edgeId,
       secret: edgeConfig.secret,
+      ...(edgeConfig.bindAddress ? { bindAddress: edgeConfig.bindAddress } : {}),
     });
 
     this.started = true;
+    this.restartAttempts = 0;
+    this.restartBackoffMs = 1000;
 
     // Start periodic status logging
     this.statusInterval = setInterval(async () => {
@@ -142,6 +162,7 @@ export class RemoteIngressEdge extends EventEmitter {
    * Stop the edge and kill the Rust process.
    */
   public async stop(): Promise<void> {
+    this.stopping = true;
     if (this.statusInterval) {
       clearInterval(this.statusInterval);
       this.statusInterval = undefined;
@@ -152,6 +173,7 @@ export class RemoteIngressEdge extends EventEmitter {
       } catch {
         // Process may already be dead
       }
+      this.bridge.removeListener('exit', this.handleCrashRecovery);
       this.bridge.kill();
       this.started = false;
     }
@@ -170,4 +192,56 @@ export class RemoteIngressEdge extends EventEmitter {
   public get running(): boolean {
     return this.bridge.running;
   }
+
+  /**
+   * Handle unexpected Rust binary crash — auto-restart with backoff.
+   */
+  private handleCrashRecovery = async (code: number | null, signal: string | null) => {
+    if (this.stopping || !this.started || !this.savedConfig) {
+      return;
+    }
+
+    console.error(
+      `[RemoteIngressEdge] Rust binary crashed (code=${code}, signal=${signal}), ` +
+      `attempt ${this.restartAttempts + 1}/${MAX_RESTART_ATTEMPTS}`
+    );
+
+    this.started = false;
+
+    if (this.restartAttempts >= MAX_RESTART_ATTEMPTS) {
+      console.error('[RemoteIngressEdge] Max restart attempts reached, giving up');
+      this.emit('crashRecoveryFailed');
+      return;
+    }
+
+    await new Promise(resolve => setTimeout(resolve, this.restartBackoffMs));
+    this.restartBackoffMs = Math.min(this.restartBackoffMs * 2, MAX_RESTART_BACKOFF_MS);
+    this.restartAttempts++;
+
+    try {
+      const spawned = await this.bridge.spawn();
+      if (!spawned) {
+        console.error('[RemoteIngressEdge] Failed to respawn binary');
+        return;
+      }
+
+      this.bridge.on('exit', this.handleCrashRecovery);
+
+      await this.bridge.sendCommand('startEdge', {
+        hubHost: this.savedConfig.hubHost,
+        hubPort: this.savedConfig.hubPort ?? 8443,
+        edgeId: this.savedConfig.edgeId,
+        secret: this.savedConfig.secret,
+        ...(this.savedConfig.bindAddress ? { bindAddress: this.savedConfig.bindAddress } : {}),
+      });
+
+      this.started = true;
+      this.restartAttempts = 0;
+      this.restartBackoffMs = 1000;
+      console.log('[RemoteIngressEdge] Successfully recovered from crash');
+      this.emit('crashRecovered');
+    } catch (err) {
+      console.error(`[RemoteIngressEdge] Crash recovery failed: ${err}`);
+    }
+  };
 }

ts/classes.remoteingresshub.ts

@@ -11,6 +11,8 @@ type THubCommands = {
     params: {
       tunnelPort: number;
       targetHost?: string;
+      tlsCertPem?: string;
+      tlsKeyPem?: string;
     };
     result: { started: boolean };
   };
@@ -33,6 +35,7 @@ type THubCommands = {
         edgeId: string;
         connectedAt: number;
         activeStreams: number;
+        peerAddr: string;
       }>;
     };
   };
@@ -41,11 +44,25 @@ type THubCommands = {
 export interface IHubConfig {
   tunnelPort?: number;
   targetHost?: string;
+  tls?: {
+    certPem?: string;
+    keyPem?: string;
+  };
 }
 
+type TAllowedEdge = { id: string; secret: string; listenPorts?: number[]; stunIntervalSecs?: number };
+
+const MAX_RESTART_ATTEMPTS = 10;
+const MAX_RESTART_BACKOFF_MS = 30_000;
+
 export class RemoteIngressHub extends EventEmitter {
   private bridge: InstanceType<typeof plugins.smartrust.RustBridge<THubCommands>>;
   private started = false;
+  private stopping = false;
+  private savedConfig: IHubConfig | null = null;
+  private savedEdges: TAllowedEdge[] = [];
+  private restartBackoffMs = 1000;
+  private restartAttempts = 0;
 
   constructor() {
     super();
@@ -73,10 +90,12 @@ export class RemoteIngressHub extends EventEmitter {
     });
 
     // Forward events from Rust binary
-    this.bridge.on('management:edgeConnected', (data: { edgeId: string }) => {
+    this.bridge.on('management:edgeConnected', (data: { edgeId: string; peerAddr: string }) => {
       this.emit('edgeConnected', data);
     });
-    this.bridge.on('management:edgeDisconnected', (data: { edgeId: string }) => {
+    this.bridge.on('management:edgeDisconnected', (data: { edgeId: string; reason?: string }) => {
+      const reason = data?.reason ?? 'unknown';
+      console.log(`[RemoteIngressHub] Edge ${data.edgeId} disconnected: ${reason}`);
       this.emit('edgeDisconnected', data);
     });
     this.bridge.on('management:streamOpened', (data: { edgeId: string; streamId: number }) => {
@@ -91,29 +110,42 @@ export class RemoteIngressHub extends EventEmitter {
    * Start the hub — spawns the Rust binary and starts the tunnel server.
    */
   public async start(config: IHubConfig = {}): Promise<void> {
+    this.savedConfig = config;
+    this.stopping = false;
+
     const spawned = await this.bridge.spawn();
     if (!spawned) {
       throw new Error('Failed to spawn remoteingress-bin');
     }
 
+    // Register crash recovery handler
+    this.bridge.on('exit', this.handleCrashRecovery);
+
     await this.bridge.sendCommand('startHub', {
       tunnelPort: config.tunnelPort ?? 8443,
       targetHost: config.targetHost ?? '127.0.0.1',
+      ...(config.tls?.certPem && config.tls?.keyPem
+        ? { tlsCertPem: config.tls.certPem, tlsKeyPem: config.tls.keyPem }
+        : {}),
     });
 
     this.started = true;
+    this.restartAttempts = 0;
+    this.restartBackoffMs = 1000;
   }
 
   /**
    * Stop the hub and kill the Rust process.
    */
   public async stop(): Promise<void> {
+    this.stopping = true;
     if (this.started) {
       try {
         await this.bridge.sendCommand('stopHub', {} as Record<string, never>);
       } catch {
         // Process may already be dead
       }
+      this.bridge.removeListener('exit', this.handleCrashRecovery);
       this.bridge.kill();
       this.started = false;
     }
@@ -122,7 +154,8 @@ export class RemoteIngressHub extends EventEmitter {
   /**
    * Update the list of allowed edges that can connect to this hub.
    */
-  public async updateAllowedEdges(edges: Array<{ id: string; secret: string; listenPorts?: number[]; stunIntervalSecs?: number }>): Promise<void> {
+  public async updateAllowedEdges(edges: TAllowedEdge[]): Promise<void> {
+    this.savedEdges = edges;
     await this.bridge.sendCommand('updateAllowedEdges', { edges });
   }
 
@@ -139,4 +172,62 @@ export class RemoteIngressHub extends EventEmitter {
   public get running(): boolean {
     return this.bridge.running;
   }
+
+  /**
+   * Handle unexpected Rust binary crash — auto-restart with backoff.
+   */
+  private handleCrashRecovery = async (code: number | null, signal: string | null) => {
+    if (this.stopping || !this.started || !this.savedConfig) {
+      return;
+    }
+
+    console.error(
+      `[RemoteIngressHub] Rust binary crashed (code=${code}, signal=${signal}), ` +
+      `attempt ${this.restartAttempts + 1}/${MAX_RESTART_ATTEMPTS}`
+    );
+
+    this.started = false;
+
+    if (this.restartAttempts >= MAX_RESTART_ATTEMPTS) {
+      console.error('[RemoteIngressHub] Max restart attempts reached, giving up');
+      this.emit('crashRecoveryFailed');
+      return;
+    }
+
+    await new Promise(resolve => setTimeout(resolve, this.restartBackoffMs));
+    this.restartBackoffMs = Math.min(this.restartBackoffMs * 2, MAX_RESTART_BACKOFF_MS);
+    this.restartAttempts++;
+
+    try {
+      const spawned = await this.bridge.spawn();
+      if (!spawned) {
+        console.error('[RemoteIngressHub] Failed to respawn binary');
+        return;
+      }
+
+      this.bridge.on('exit', this.handleCrashRecovery);
+
+      const config = this.savedConfig;
+      await this.bridge.sendCommand('startHub', {
+        tunnelPort: config.tunnelPort ?? 8443,
+        targetHost: config.targetHost ?? '127.0.0.1',
+        ...(config.tls?.certPem && config.tls?.keyPem
+          ? { tlsCertPem: config.tls.certPem, tlsKeyPem: config.tls.keyPem }
+          : {}),
+      });
+
+      // Restore allowed edges
+      if (this.savedEdges.length > 0) {
+        await this.bridge.sendCommand('updateAllowedEdges', { edges: this.savedEdges });
+      }
+
+      this.started = true;
+      this.restartAttempts = 0;
+      this.restartBackoffMs = 1000;
+      console.log('[RemoteIngressHub] Successfully recovered from crash');
+      this.emit('crashRecovered');
+    } catch (err) {
+      console.error(`[RemoteIngressHub] Crash recovery failed: ${err}`);
+    }
+  };
 }