MGS: Flesh out RFD250-style SP/location discovery

gateway-sp-comms/Cargo.toml

@@ -10,6 +10,7 @@ http = "0.2.6"
 hyper = "0.14.17"
 ringbuffer = "0.8"
 serde = { version = "1.0", features = ["derive"] }
+serde_with = "1.12.0"
 thiserror = "1.0.30"
 tokio-tungstenite = "0.17"
 tokio-stream = "0.1.8"

gateway-sp-comms/src/communicator.rs

@@ -9,41 +9,31 @@ use crate::error::Error;
 use crate::error::SpCommunicationError;
 use crate::error::StartupError;
 use crate::management_switch::ManagementSwitch;
-use crate::management_switch::ManagementSwitchDiscovery;
 use crate::management_switch::SpSocket;
 use crate::management_switch::SwitchPort;
-use crate::recv_handler::RecvHandler;
 use crate::Elapsed;
-use crate::KnownSps;
 use crate::SpIdentifier;
+use crate::SwitchConfig;
 use crate::Timeout;
 use futures::stream::FuturesUnordered;
 use futures::Future;
 use futures::Stream;
-use gateway_messages::version;
 use gateway_messages::BulkIgnitionState;
 use gateway_messages::DiscoverResponse;
 use gateway_messages::IgnitionCommand;
 use gateway_messages::IgnitionState;
-use gateway_messages::Request;
 use gateway_messages::RequestKind;
-use gateway_messages::ResponseError;
 use gateway_messages::ResponseKind;
 use gateway_messages::SerialConsole;
-use gateway_messages::SerializedSize;
 use gateway_messages::SpComponent;
 use gateway_messages::SpState;
 use hyper::header;
 use hyper::upgrade;
 use hyper::Body;
-use omicron_common::backoff;
-use omicron_common::backoff::Backoff;
-use slog::debug;
 use slog::info;
 use slog::o;
 use slog::Logger;
-use std::sync::atomic::AtomicU32;
-use std::sync::atomic::Ordering;
+use tokio::time::Instant;
 use std::sync::Arc;
 use std::time::Duration;
 use tokio_tungstenite::tungstenite::handshake;
@@ -66,28 +56,30 @@ where
 
 #[derive(Debug)]
 pub struct Communicator {
-    log: Logger,
     switch: ManagementSwitch,
-    request_id: AtomicU32,
-    recv_handler: Arc<RecvHandler>,
 }
 
 impl Communicator {
     pub async fn new(
-        known_sps: KnownSps,
+        config: SwitchConfig,
+        discovery_deadline: Instant,
         log: &Logger,
     ) -> Result<Self, StartupError> {
         let log = log.new(o!("component" => "SpCommunicator"));
-        let discovery = ManagementSwitchDiscovery::placeholder_start(
-            known_sps,
-            log.clone(),
-        )
-        .await?;
-
-        let (switch, recv_handler) = RecvHandler::new(discovery, log.clone());
+        let switch =
+            ManagementSwitch::new(config, discovery_deadline, log.clone())
+                .await?;
 
         info!(&log, "started SP communicator");
-        Ok(Self { log, switch, request_id: AtomicU32::new(0), recv_handler })
+        Ok(Self { switch })
+    }
+
+    /// Get the name of our location.
+    ///
+    /// This matches one of the names specified as a possible location in the
+    /// configuration we were given.
+    pub fn location_name(&self) -> &str {
+        &self.switch.location_name()
     }
 
     // convert an identifier to a port number; this is fallible because
@@ -103,13 +95,36 @@ impl Communicator {
         self.switch.switch_port_to_id(port)
     }
 
+    /// Returns true if we've discovered the IP address of our local ignition
+    /// controller.
+    ///
+    /// This method exists to be polled during test setup (to wait for discovery
+    /// to happen); it should not be called outside tests.
+    pub fn local_ignition_controller_address_known(&self) -> bool {
+        self.switch.ignition_controller().is_some()
+    }
+
+    /// Returns true if we've discovered the IP address of the specified SP.
+    ///
+    /// This method exists to be polled during test setup (to wait for discovery
+    /// to happen); it should not be called outside tests. In particular, it
+    /// panics instead of running an error if `sp` describes an SP that isn't
+    /// known to this communicator.
+    pub fn address_known(&self, sp: SpIdentifier) -> bool {
+        let port = self.switch.switch_port(sp).unwrap();
+        self.switch.sp_socket(port).is_some()
+    }
+
     /// Ask the local ignition controller for the ignition state of a given SP.
     pub async fn get_ignition_state(
         &self,
         sp: SpIdentifier,
         timeout: Timeout,
     ) -> Result<IgnitionState, Error> {
-        let controller = self.switch.ignition_controller();
+        let controller = self
+            .switch
+            .ignition_controller()
+            .ok_or(Error::LocalIgnitionControllerAddressUnknown)?;
         let port = self.id_to_port(sp)?;
         let request =
             RequestKind::IgnitionState { target: port.as_ignition_target() };
@@ -129,7 +144,10 @@ impl Communicator {
         &self,
         timeout: Timeout,
     ) -> Result<Vec<(SpIdentifier, IgnitionState)>, Error> {
-        let controller = self.switch.ignition_controller();
+        let controller = self
+            .switch
+            .ignition_controller()
+            .ok_or(Error::LocalIgnitionControllerAddressUnknown)?;
         let request = RequestKind::BulkIgnitionState;
 
         let bulk_state = self
@@ -170,7 +188,10 @@ impl Communicator {
         command: IgnitionCommand,
         timeout: Timeout,
     ) -> Result<(), Error> {
-        let controller = self.switch.ignition_controller();
+        let controller = self
+            .switch
+            .ignition_controller()
+            .ok_or(Error::LocalIgnitionControllerAddressUnknown)?;
         let target = self.id_to_port(target_sp)?.as_ignition_target();
         let request = RequestKind::IgnitionCommand { target, command };
 
@@ -196,6 +217,11 @@ impl Communicator {
     // via UDP. SPs will continuously broadcast any serial console data, even if
     // there is no attached client. Maybe this is fine, since the serial console
     // shouldn't be noisy without a corresponding client driving it?
+    //
+    // TODO Because this method doesn't contact the target SP, it succeeds even
+    // if we don't know the IP address of that SP (yet, or possibly ever)! The
+    // connection will start working if we later discover the address, but this
+    // is probably not the behavior we want.
     pub async fn serial_console_attach(
         self: &Arc<Self>,
         request: &mut http::Request<Body>,
@@ -250,7 +276,7 @@ impl Communicator {
             .map(|key| handshake::derive_accept_key(key))
             .ok_or(Error::BadWebsocketConnection("missing websocket key"))?;
 
-        self.recv_handler.serial_console_attach(
+        self.switch.serial_console_attach(
             Arc::clone(self),
             port,
             component,
@@ -281,7 +307,7 @@ impl Communicator {
         component: &SpComponent,
     ) -> Result<(), Error> {
         let port = self.id_to_port(sp)?;
-        self.recv_handler.serial_console_detach(port, component)
+        self.switch.serial_console_detach(port, component)
     }
 
     /// Send `packet` to the given SP component's serial console.
@@ -398,102 +424,17 @@ impl Communicator {
             .collect::<FuturesUnordered<_>>()
     }
 
-    pub(crate) async fn request_response<F, T>(
+    async fn request_response<F, T>(
         &self,
         sp: &SpSocket<'_>,
-        mut kind: RequestKind,
-        mut map_response_kind: F,
+        kind: RequestKind,
+        map_response_kind: F,
         timeout: Option<Timeout>,
     ) -> Result<T, Error>
     where
         F: FnMut(ResponseKind) -> Result<T, BadResponseType>,
     {
-        // helper to wrap a future in a timeout if we have one
-        async fn maybe_with_timeout<F, U>(
-            timeout: Option<Timeout>,
-            fut: F,
-        ) -> Result<U, Elapsed>
-        where
-            F: Future<Output = U>,
-        {
-            match timeout {
-                Some(t) => t.timeout_at(fut).await,
-                None => Ok(fut.await),
-            }
-        }
-
-        // We'll use exponential backoff if and only if the SP responds with
-        // "busy"; any other error will cause the loop below to terminate.
-        let mut backoff = backoff::internal_service_policy();
-
-        loop {
-            // It would be nicer to use `backoff::retry()` instead of manually
-            // stepping the backoff policy, but the dance we do with `kind` to
-            // avoid cloning it is hard to map into `retry()` in a way that
-            // satisfies the borrow checker. ("The dance we do with `kind` to
-            // avoid cloning it" being that we move it into `request` below, and
-            // on a busy response from the SP we move it back out into the
-            // `kind` local var.)
-            let duration = backoff
-                .next_backoff()
-                .expect("internal backoff policy gave up");
-            maybe_with_timeout(timeout, tokio::time::sleep(duration))
-                .await
-                .map_err(|err| Error::Timeout {
-                    timeout: err.duration(),
-                    sp: self.port_to_id(sp.port()),
-                })?;
-
-            // request IDs will eventually roll over; since we enforce timeouts
-            // this should be a non-issue in practice. does this need testing?
-            let request_id = self.request_id.fetch_add(1, Ordering::Relaxed);
-
-            // update our recv_handler to expect a response for this request ID
-            let response_fut =
-                self.recv_handler.register_request_id(sp.port(), request_id);
-
-            // Serialize and send our request. We know `buf` is large enough for
-            // any `Request`, so unwrapping here is fine.
-            let request = Request { version: version::V1, request_id, kind };
-            let mut buf = [0; Request::MAX_SIZE];
-            let n = gateway_messages::serialize(&mut buf, &request).unwrap();
-            let serialized_request = &buf[..n];
-
-            // Actual communication, guarded by `timeout` if it's not `None`.
-            let result = maybe_with_timeout(timeout, async {
-                debug!(&self.log, "sending {:?} to SP {:?}", request, sp);
-                sp.send(serialized_request).await.map_err(|err| {
-                    SpCommunicationError::UdpSend { addr: sp.addr(), err }
-                })?;
-
-                Ok::<ResponseKind, SpCommunicationError>(response_fut.await?)
-            })
-            .await
-            .map_err(|err| Error::Timeout {
-                timeout: err.duration(),
-                sp: self.port_to_id(sp.port()),
-            })?;
-
-            match result {
-                Ok(response_kind) => {
-                    return map_response_kind(response_kind)
-                        .map_err(SpCommunicationError::from)
-                        .map_err(Error::from)
-                }
-                Err(SpCommunicationError::SpError(ResponseError::Busy)) => {
-                    debug!(
-                        &self.log,
-                        "SP busy; sleeping before retrying send";
-                        "sp" => ?sp,
-                    );
-
-                    // move `kind` back into local var; required to satisfy
-                    // borrow check of this loop
-                    kind = request.kind;
-                }
-                Err(err) => return Err(err.into()),
-            }
-        }
+        self.switch.request_response(sp, kind, map_response_kind, timeout).await
     }
 }
 

gateway-sp-comms/src/error.rs

@@ -35,8 +35,10 @@ pub enum Error {
         .0.slot,
     )]
     SpAddressUnknown(SpIdentifier),
-    #[error("timeout ({timeout:?}) elapsed communicating with {sp:?}")]
-    Timeout { timeout: Duration, sp: SpIdentifier },
+    #[error(
+        "timeout ({timeout:?}) elapsed communicating with {sp:?} on port {port}"
+    )]
+    Timeout { timeout: Duration, port: usize, sp: Option<SpIdentifier> },
     #[error("error communicating with SP: {0}")]
     SpCommunicationFailed(#[from] SpCommunicationError),
     #[error("serial console is already attached")]

gateway-sp-comms/src/lib.rs

@@ -24,13 +24,11 @@ pub mod error;
 
 pub use communicator::Communicator;
 pub use communicator::FuturesUnorderedImpl;
+pub use management_switch::LocationConfig;
+pub use management_switch::LocationDeterminationConfig;
 pub use management_switch::SpIdentifier;
 pub use management_switch::SpType;
+pub use management_switch::SwitchConfig;
+pub use management_switch::SwitchPortConfig;
 pub use timeout::Elapsed;
 pub use timeout::Timeout;
-
-// TODO these will remain public for a while, but eventually will be removed
-// altogther; currently these provide a way to hard-code the rack topology,
-// which is not what we want.
-pub use management_switch::KnownSp;
-pub use management_switch::KnownSps;