fix(io): retry interrupted receives

Real signal delivery on Linux (observed via x86_64-musl workspace testing) can interrupt blocking and nonblocking recv/recvmsg calls with EINTR. Several production and test receive paths treated that as a fatal error rather than a transient interruption: - Client::open's blocking receive loop and Client::recv_once/ recv_available's blocking receive returned ClientError::Socket on Interrupted instead of retrying within their existing timeout budget. - AsyncClient::poll_open and poll_recv's nonblocking recvmsg path (Linux ancillary) failed the same way instead of retrying through the existing readiness/work-budget machinery. - The server's ancillary receive loop could terminate the listener on an interrupted nonblocking recvmsg. - The client's MSG_ERRQUEUE TX-timestamp drain turned an interrupted read of optional metadata into a fatal ClientError::Socket. - Fake-server test helpers panicked or misclassified EINTR as a timeout instead of retrying it. Interrupted is now retried transparently everywhere a receive can observe it, while preserving each path's existing timing contract: - Blocking Open reuses its already-absolute per-attempt deadline, so interruptions resume within the same attempt without retransmitting or extending the deadline. - recv_once/recv_available compute one absolute deadline per logical receive (when a timeout is configured) so repeated interruptions cannot extend it past the configured timeout, and restore the socket's configured read timeout afterward so a later call in the same recv_available budget is unaffected. - AsyncClient's poll_open already bounds its retry loop with OPEN_POLL_WORK_BUDGET; poll_recv gains an equivalent RECV_POLL_WORK_BUDGET, since unlike WouldBlock, Interrupted does not clear Tokio's readiness state, so an unbounded retry loop was otherwise possible under sustained signal delivery. - The server's receive loop retries Interrupted the same way it already retries WouldBlock, yielding to the runtime after a bounded number of consecutive interruptions for the same reason. - The error-queue drain treats an interrupted read as no record obtained this drain (deferring to the next natural drain point) rather than a socket error, matching its existing EWOULDBLOCK handling. Deterministic regression tests use a small test-only injection hook (mirroring existing send-injection test hooks) rather than real signal timing, covering: Open completing despite interruptions with no retransmit, Open's deadline not extending under repeated interruptions, recv_once classifying a reply normally despite interruptions, recv_once's timeout not extending under repeated interruptions, and the AsyncClient equivalents for open and recv. No protocol, session, or timeout-classification behavior changes; no new dependencies; send-side EINTR handling is unchanged and out of scope.
2026-08-17 00:23:24 +07:00
1 parent e3815fdeed
commit ca2389174f
@@ -64,6 +64,63 @@ fn open_retries_after_first_timeout() {
    server.join();
 }

+#[test]
+fn open_survives_interrupted_receive_within_same_attempt() {
+    let server = start_fake_server(|socket, tx| {
+        let (first, peer) = recv_request(&socket, &tx);
+        let params = Params::decode(&first[4..]).unwrap();
+        let reply = open_reply(FLAG_OPEN | FLAG_REPLY, TOKEN, &params, None);
+        socket.send_to(&reply, peer).unwrap();
+    });
+    let config = ClientConfig {
+        open_timeouts: vec![Duration::from_millis(500)],
+        ..default_test_config(server.addr)
+    };
+    let mut client = Client::connect(config).unwrap();
+    // Several interrupted receive attempts before the real reply arrives
+    // must not be treated as a timeout, must not retransmit the open
+    // request, and must not advance to a next configured attempt.
+    client.test_hooks.inject_recv_interrupted.set(5);
+
+    let outcome = client.open().unwrap();
+    assert_open_started(outcome);
+    // Exactly one request reached the server: interruptions were absorbed
+    // inside the single configured attempt.
+    assert_eq!(server.rx.iter().take(1).count(), 1);
+    server.join();
+}
+
+#[test]
+fn open_interrupted_receive_does_not_extend_the_attempt_deadline() {
+    // A fake server that never replies: with a short attempt timeout, the
+    // real question is whether interruptions can push the effective wait
+    // past the configured deadline. They must not: the deadline stays
+    // absolute across any number of interruptions.
+    let server = start_fake_server(|socket, tx| {
+        let _ = recv_request(&socket, &tx);
+    });
+    let config = ClientConfig {
+        open_timeouts: vec![Duration::from_millis(200)],
+        ..default_test_config(server.addr)
+    };
+    let mut client = Client::connect(config).unwrap();
+    // Each interruption carries no timing information of its own, so the
+    // loop must keep re-checking the real deadline rather than treating an
+    // interruption as "more time available" or a reason to fail outright.
+    client.test_hooks.inject_recv_interrupted.set(20);
+
+    let started = Instant::now();
+    let result = client.open();
+    let elapsed = started.elapsed();
+
+    assert!(matches!(result, Err(ClientError::OpenTimeout)));
+    assert!(
+        elapsed < Duration::from_millis(800),
+        "interrupted receives must not extend the attempt deadline, took {elapsed:?}"
+    );
+    server.join();
+}
+
 #[test]
 fn open_timeout_after_all_timeouts() {
    let server = timeout_server(Duration::from_millis(700));
@@ -0,0 +1,80 @@
+use super::*;
+
+/// An interrupted receive (`io::ErrorKind::Interrupted`, e.g. `EINTR`) must be
+/// invisible to the caller: retried transparently, with the eventual reply
+/// classified exactly as if no interruption had occurred.
+#[test]
+fn interrupted_receive_is_retried_and_the_reply_is_still_classified() {
+    let params = default_params();
+    let server = start_fake_server(move |socket, tx| {
+        let (_, peer) = recv_request(&socket, &tx);
+        let reply = open_reply(FLAG_OPEN | FLAG_REPLY, TOKEN, &params, None);
+        socket.send_to(&reply, peer).unwrap();
+
+        let mut buf = [0_u8; 2048];
+        let (size, _) = socket.recv_from(&mut buf).unwrap();
+        let seq = u32::from_le_bytes(buf[12..16].try_into().unwrap());
+        let _ = size;
+        let ts = TimestampFields::default();
+        let reply_packet = echo_reply_packet(TOKEN, seq, &params, &ts, None);
+        socket.send_to(&reply_packet, peer).unwrap();
+    });
+    let config = ClientConfig {
+        socket_config: crate::SocketConfig {
+            recv_timeout: Some(Duration::from_secs(2)),
+            ..Default::default()
+        },
+        ..default_test_config(server.addr)
+    };
+    let mut client = Client::connect(config).unwrap();
+    assert_open_started(client.open().unwrap());
+
+    client.send_probe().unwrap();
+    thread::sleep(Duration::from_millis(50));
+
+    client.test_hooks.inject_recv_interrupted.set(10);
+    let events = client.recv_once().unwrap();
+
+    assert_eq!(events.len(), 1);
+    assert!(
+        matches!(&events[0], ClientEvent::EchoReply { seq: 0, .. }),
+        "expected a normal EchoReply despite interrupted receive attempts, got {:?}",
+        events[0]
+    );
+
+    client.close().unwrap();
+    server.join();
+}
+
+/// One logical receive keeps one timeout budget: repeated interruptions must
+/// not extend a `recv_once` call past its configured receive timeout.
+#[test]
+fn interrupted_receive_does_not_extend_the_configured_timeout() {
+    let params = default_params();
+    let server = silent_open_server(params);
+    let config = ClientConfig {
+        socket_config: crate::SocketConfig {
+            recv_timeout: Some(Duration::from_millis(100)),
+            ..Default::default()
+        },
+        ..default_test_config(server.addr)
+    };
+    let mut client = Client::connect(config).unwrap();
+    assert_open_started(client.open().unwrap());
+
+    // No reply is ever coming; only interrupted attempts occur before the
+    // real socket takes over and legitimately times out.
+    client.test_hooks.inject_recv_interrupted.set(20);
+    let started = Instant::now();
+    let events = client.recv_once().unwrap();
+    let elapsed = started.elapsed();
+
+    assert!(events.is_empty());
+    assert!(
+        elapsed < Duration::from_millis(500),
+        "interrupted receives must not extend the configured receive timeout, took {elapsed:?}"
+    );
+
+    client.close().unwrap();
+    server.join();
+}
@@ -20,6 +20,7 @@ fn kernel_rx_timestamp_unavailable_skip(test_name: &str) {
 }

 mod burst_recv;
+mod interrupted_receive;
 mod packet_length;
 mod probe_schedule;
 mod reply_classification;