perf(wasi): optimize data plane and extend host ABI to v3 (#2455)

Overhaul the WASI guest data plane for throughput and add the host
capabilities it relies on. The externally driven Tokio runtime now
runs its timer pre-turn only when a tracked deadline has expired,
and all WASI-reachable timers (STUN, port mapping, WebClient, UDP
flow cleanup) go through the portable time facade so conditional
timer driving cannot starve them.

Data plane:

- Move read/write deadlines onto TCP and UDP resources with one ABI
  setter per direction, reuse a single expiration timer per
  resource, and drop timeout arguments from the four hot data-plane
  submissions (ABI v3). Checked absolute instants treat
  unrepresentable finite timeouts as unbounded instead of panicking.
- Batch host traffic: vectored TCP frame writes combine queued
  slices into one host operation, and reads request a bounded 64
  KiB while retaining excess bytes in the stream buffer.
- Complete TCP writes inside the guest with cancellation-safe
  writes, reporting the completed prefix before honoring
  cancellation or timeout so hosts never replay bytes.
- Repoll smoltcp egress immediately on zero poll delay, enlarge
  virtual UDP receive queues to 128 KiB payload with 128 metadata
  slots, and bound UDP session receive buffers to 8 KiB plus one
  byte while keeping oversized-datagram detection.

Host integration:

- Add optional algorithm-neutral AEAD seal/open imports with the
  ring backend as fallback, and pin the ring AES-128-GCM wire vector
  so the Go host stays interoperable.
- Forward instance events to hosts through one best-effort,
  synchronous, non-blocking import.
- Add a repository-owned build entry point for the Go host artifact:
  Binaryen 131 at -O4 with cached, SHA-256-verified official
  archives.
This commit is contained in:
KKRainbow
2026-07-28 00:29:08 +08:00
committed by GitHub
parent 7b506e25a7
commit d55e63b88e
40 changed files with 1342 additions and 220 deletions
+1 -1
View File
@@ -447,7 +447,7 @@ pub(super) async fn udp_session_layer_recv_task<S, R>(
{
let control_permits = Arc::new(Semaphore::new(UDP_SESSION_QUEUE_CAPACITY));
loop {
let datagram = match socket.recv_datagram().await {
let datagram = match socket.recv_session_datagram().await {
Ok(datagram) => datagram,
Err(err) => {
tracing::debug!(?err, "udp session recv loop stopped");
+29
View File
@@ -301,6 +301,7 @@ async fn udp_session_listener_reports_bound_local_addr_before_accept() {
struct MockVirtualUdpSocket {
local_addr: SocketAddr,
incoming: Mutex<VecDeque<(Vec<u8>, SocketAddr)>>,
recv_capacities: Mutex<Vec<usize>>,
sent: Mutex<Vec<(Vec<u8>, SocketAddr)>>,
send_attempts: Mutex<Vec<(Vec<u8>, SocketAddr, UdpSocketSendMeta)>>,
reject_preferred_source: AtomicBool,
@@ -311,6 +312,7 @@ impl MockVirtualUdpSocket {
Self {
local_addr,
incoming: Mutex::new(incoming.into()),
recv_capacities: Mutex::new(Vec::new()),
sent: Mutex::new(Vec::new()),
send_attempts: Mutex::new(Vec::new()),
reject_preferred_source: AtomicBool::new(false),
@@ -357,6 +359,7 @@ impl VirtualUdpSocket for MockVirtualUdpSocket {
}
async fn recv_from(&self, buf: &mut [u8]) -> io::Result<(usize, SocketAddr)> {
self.recv_capacities.lock().unwrap().push(buf.len());
let (data, remote_addr) =
self.incoming.lock().unwrap().pop_front().ok_or_else(|| {
io::Error::new(io::ErrorKind::UnexpectedEof, "no incoming datagram")
@@ -367,6 +370,32 @@ impl VirtualUdpSocket for MockVirtualUdpSocket {
}
}
#[tokio::test]
async fn portable_udp_receive_keeps_general_and_session_capacities_separate() {
let local_addr = SocketAddr::from(([127, 0, 0, 1], 12000));
let peer_addr = SocketAddr::from(([127, 0, 0, 1], 12001));
let socket = MockVirtualUdpSocket::new(
local_addr,
vec![
(b"session".to_vec(), peer_addr),
(b"general".to_vec(), peer_addr),
],
);
assert_eq!(
socket.recv_session_datagram().await.unwrap().payload,
b"session".as_slice()
);
assert_eq!(
socket.recv_datagram().await.unwrap().payload,
b"general".as_slice()
);
assert_eq!(
*socket.recv_capacities.lock().unwrap(),
[MAX_UDP_SESSION_DATAGRAM_SIZE + 1, MAX_UDP_DATAGRAM_SIZE]
);
}
fn easytier_stun_request(change_ip: bool, change_port: bool) -> Vec<u8> {
let mut request = Message::<Attribute>::new(MessageClass::Request, BINDING, u32_to_tid(7));
if change_ip || change_port {
+29 -9
View File
@@ -76,16 +76,36 @@ pub trait VirtualUdpSocket: Send + Sync + 'static {
/// override it when their socket API can write directly into owned storage,
/// avoiding a second allocation and copy at the Host boundary.
async fn recv_datagram(&self) -> std::io::Result<UdpSocketDatagram> {
let mut payload = BytesMut::new();
payload.resize(MAX_UDP_DATAGRAM_SIZE, 0);
let (len, remote_addr, meta) = self.recv_from_with_meta(&mut payload).await?;
payload.truncate(len);
Ok(UdpSocketDatagram {
payload,
remote_addr,
meta,
})
recv_portable_datagram(self, MAX_UDP_DATAGRAM_SIZE).await
}
/// Receives one datagram for the UDP session/multiplexer data plane.
///
/// Portable hosts receive one byte past the session limit so a truncated
/// oversized datagram remains distinguishable from a valid maximum-sized
/// datagram. Native hosts may override this when they can detect truncation
/// without the extra byte.
async fn recv_session_datagram(&self) -> std::io::Result<UdpSocketDatagram> {
recv_portable_datagram(self, MAX_UDP_SESSION_DATAGRAM_SIZE + 1).await
}
}
async fn recv_portable_datagram<S>(
socket: &S,
capacity: usize,
) -> std::io::Result<UdpSocketDatagram>
where
S: VirtualUdpSocket + ?Sized,
{
let mut payload = BytesMut::new();
payload.resize(capacity, 0);
let (len, remote_addr, meta) = socket.recv_from_with_meta(&mut payload).await?;
payload.truncate(len);
Ok(UdpSocketDatagram {
payload,
remote_addr,
meta,
})
}
#[async_trait]