fix(sandbox): relay WebSocket frames after HTTP 101 Switching Protocols

The L7 REST proxy treats 101 Switching Protocols as a generic 1xx
informational response via is_bodiless_response(), forwarding the
headers and returning to the HTTP parsing loop. After a 101, the
connection has been upgraded (e.g. to WebSocket) and subsequent bytes
are protocol frames, not HTTP requests. The relay loop either blocks
or silently drops them.

This patch:
- Adds RelayOutcome::Upgraded variant to signal protocol upgrades
- Detects 101 responses before the generic 1xx handler in
  relay_response(), capturing any overflow bytes read past the headers
- Switches relay_rest() and relay_passthrough_with_credentials() to
  raw bidirectional TCP copy (tokio::io::copy_bidirectional) after
  receiving an Upgraded outcome
- Adds a test verifying 101 response handling and overflow capture

This enables WebSocket connections (OpenClaw node meshes, Discord/Slack
bots) to work from inside fully sandboxed environments.

Fixes: #652
Related: NVIDIA/NemoClaw#409

Signed-off-by: David Peden <davidpeden3@gmail.com>

Author: davidpeden3

crates/openshell-sandbox/src/l7/provider.rs

@@ -14,6 +14,21 @@ use std::collections::HashMap;
 use std::future::Future;
 use tokio::io::{AsyncRead, AsyncWrite};
 
+/// Outcome of relaying a single HTTP request/response pair.
+#[derive(Debug)]
+pub enum RelayOutcome {
+    /// Connection is reusable for further HTTP requests (keep-alive).
+    Reusable,
+    /// Connection was consumed (e.g. read-until-EOF or `Connection: close`).
+    Consumed,
+    /// Server responded with 101 Switching Protocols.
+    /// The connection has been upgraded (e.g. to WebSocket) and must be
+    /// relayed as raw bidirectional TCP from this point forward.
+    /// Contains any overflow bytes read from upstream after the 101 headers
+    /// that must be forwarded to the client before switching to copy mode.
+    Upgraded { overflow: Vec<u8> },
+}
+
 /// Body framing for HTTP requests/responses.
 #[derive(Debug, Clone, Copy)]
 pub enum BodyLength {
@@ -57,14 +72,15 @@ pub trait L7Provider: Send + Sync {
 
     /// Forward an allowed request to upstream and relay the response back.
     ///
-    /// Returns `true` if the upstream connection is reusable (keep-alive),
-    /// `false` if it was consumed (e.g. read-until-EOF or `Connection: close`).
+    /// Returns a [`RelayOutcome`] indicating whether the connection is
+    /// reusable (keep-alive), consumed, or has been upgraded (101 Switching
+    /// Protocols) and must be relayed as raw bidirectional TCP.
     fn relay<C, U>(
         &self,
         req: &L7Request,
         client: &mut C,
         upstream: &mut U,
-    ) -> impl Future<Output = Result<bool>> + Send
+    ) -> impl Future<Output = Result<RelayOutcome>> + Send
     where
         C: AsyncRead + AsyncWrite + Unpin + Send,
         U: AsyncRead + AsyncWrite + Unpin + Send;

crates/openshell-sandbox/src/l7/relay.rs

@@ -7,12 +7,12 @@
 //! Parses each request within the tunnel, evaluates it against OPA policy,
 //! and either forwards or denies the request.
 
-use crate::l7::provider::L7Provider;
+use crate::l7::provider::{L7Provider, RelayOutcome};
 use crate::l7::{EnforcementMode, L7EndpointConfig, L7Protocol, L7RequestInfo};
 use crate::secrets::SecretResolver;
 use miette::{IntoDiagnostic, Result, miette};
 use std::sync::{Arc, Mutex};
-use tokio::io::{AsyncRead, AsyncWrite};
+use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 use tracing::{debug, info, warn};
 
 /// Context for L7 request policy evaluation.
@@ -136,20 +136,42 @@ where
 
         if allowed || config.enforcement == EnforcementMode::Audit {
             // Forward request to upstream and relay response
-            let reusable = crate::l7::rest::relay_http_request_with_resolver(
+            let outcome = crate::l7::rest::relay_http_request_with_resolver(
                 &req,
                 client,
                 upstream,
                 ctx.secret_resolver.as_deref(),
             )
             .await?;
-            if !reusable {
-                debug!(
-                    host = %ctx.host,
-                    port = ctx.port,
-                    "Upstream connection not reusable, closing L7 relay"
-                );
-                return Ok(());
+            match outcome {
+                RelayOutcome::Reusable => {} // continue loop
+                RelayOutcome::Consumed => {
+                    debug!(
+                        host = %ctx.host,
+                        port = ctx.port,
+                        "Upstream connection not reusable, closing L7 relay"
+                    );
+                    return Ok(());
+                }
+                RelayOutcome::Upgraded { overflow } => {
+                    info!(
+                        host = %ctx.host,
+                        port = ctx.port,
+                        overflow_bytes = overflow.len(),
+                        "101 Switching Protocols — switching to raw bidirectional relay"
+                    );
+                    // Forward any overflow bytes from the upgrade response
+                    if !overflow.is_empty() {
+                        client.write_all(&overflow).await.into_diagnostic()?;
+                        client.flush().await.into_diagnostic()?;
+                    }
+                    // Switch to raw bidirectional TCP copy for the upgraded
+                    // protocol (WebSocket, HTTP/2, etc.)
+                    tokio::io::copy_bidirectional(client, upstream)
+                        .await
+                        .into_diagnostic()?;
+                    return Ok(());
+                }
             }
         } else {
             // Enforce mode: deny with 403 and close connection
@@ -281,12 +303,29 @@ where
         // Forward request with credential rewriting and relay the response.
         // relay_http_request_with_resolver handles both directions: it sends
         // the request upstream and reads the response back to the client.
-        let reusable =
+        let outcome =
             crate::l7::rest::relay_http_request_with_resolver(&req, client, upstream, resolver)
                 .await?;
 
-        if !reusable {
-            break;
+        match outcome {
+            RelayOutcome::Reusable => {} // continue loop
+            RelayOutcome::Consumed => break,
+            RelayOutcome::Upgraded { overflow } => {
+                info!(
+                    host = %ctx.host,
+                    port = ctx.port,
+                    overflow_bytes = overflow.len(),
+                    "101 Switching Protocols — switching to raw bidirectional relay"
+                );
+                if !overflow.is_empty() {
+                    client.write_all(&overflow).await.into_diagnostic()?;
+                    client.flush().await.into_diagnostic()?;
+                }
+                tokio::io::copy_bidirectional(client, upstream)
+                    .await
+                    .into_diagnostic()?;
+                return Ok(());
+            }
         }
     }