feat(minvmd): implement DM1 gvproxy tap relay for OwnIp PTasks - #540

crates/minvmd/src/net/relay.rs (1)

176-181: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick win

Relay task failures are silently swallowed and the two directions aren't cross-cancelled.

Both JoinHandle<io::Result<()>> results are only ever abort()ed on Drop; they are never awaited or logged. If a relay loop returns an Err (or relay_switch_to_tap returns Ok(()) on EOF), the sibling task keeps running and the PTask loses connectivity with no diagnostic. Consider logging on task exit and tearing down the partner direction when either side ends so a half-broken relay becomes observable.

♻️ One option: log terminal state on each direction

-    let (sock_rx, sock_tx) = sock.into_split();
-    let tap_to_switch = tokio::spawn(relay_tap_to_switch(Arc::clone(&tap), sock_tx));
-    let switch_to_tap = tokio::spawn(relay_switch_to_tap(sock_rx, tap));
+    let (sock_rx, sock_tx) = sock.into_split();
+    let tap_to_switch = tokio::spawn(async move {
+        let r = relay_tap_to_switch(Arc::clone(&tap), sock_tx).await;
+        if let Err(ref e) = r {
+            tracing::warn!(error = %e, "tap→switch relay terminated with error");
+        }
+        r
+    });
+    let switch_to_tap = tokio::spawn(async move {
+        let r = relay_switch_to_tap(sock_rx, tap).await;
+        if let Err(ref e) = r {
+            tracing::warn!(error = %e, "switch→tap relay terminated with error");
+        }
+        r
+    });

Note tap is moved into the first closure above; keep the Arc::clone/move ordering consistent with the surrounding borrow of tap. Cross-cancelling the sibling (e.g. via a shared CancellationToken or aborting the other handle on exit) would additionally avoid a lingering task after one side closes.

Want me to draft the cross-cancellation variant?

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@crates/minvmd/src/net/relay.rs` around lines 176 - 181, The relay task
spawning in SwitchRelay does not implement proper monitoring or
cross-cancellation between the two relay directions. Currently, the JoinHandles
for relay_tap_to_switch and relay_switch_to_tap are never awaited or logged,
meaning failures or EOF on either side silently swallow errors with no
diagnostics, while the sibling task continues running orphaned. Implement
cross-cancellation by either sharing a CancellationToken between both relay
closures or by having each spawned task monitor its sibling handle and abort it
when it exits. Additionally, add logging at task completion for both
relay_tap_to_switch and relay_switch_to_tap to capture success or error states.
Ensure the Arc::clone and move ordering for tap remains consistent with the
current pattern when passing it to the relay closures.

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes