ferrix/crates/integration-tests/tests/update_errors.rs

//! End-to-end tests for errors in UPDATE messages (RFC 7606): most of them
//! cost the routes in the UPDATE, not the session.

use std::net::Ipv4Addr;
use std::time::Duration;

use bgp_integration_tests::test_peer::TestPeer;
use bgp_integration_tests::test_server::TestServer;
use bgp_packet::errors::{BgpError, UpdateMsgSubcode};
use bgp_packet::message::{LocalPrefPathAttribute, Message, PathAttribute};
use bgp_primitives::ip_prefix::IpPrefix;
use bgp_rib::SnapshotEntry;
use bgp_server::metrics::UPDATE_ERRORS;
use eyre::{Result, bail};

const LOCAL_ASN: u32 = 65000;
const PEER_ASN: u32 = 65001;
const TIMEOUT: Duration = Duration::from_secs(5);

const WELL_KNOWN: u8 = 0x40;
const OPTIONAL_TRANSITIVE: u8 = 0xc0;

fn attr(flags: u8, type_code: u8, value: &[u8]) -> Vec<u8> {
    let mut wire = vec![flags, type_code, value.len() as u8];
    wire.extend_from_slice(value);
    wire
}

fn origin() -> Vec<u8> {
    attr(WELL_KNOWN, 1, &[0])
}

/// AS_SEQUENCE of the peer's AS, four octets wide.
fn as_path() -> Vec<u8> {
    attr(WELL_KNOWN, 2, &[2, 1, 0, 0, 0xfd, 0xe9])
}

fn next_hop() -> Vec<u8> {
    attr(WELL_KNOWN, 3, &[192, 0, 2, 1])
}

/// A whole UPDATE message announcing `nlri` with `attributes`, as bytes, so
/// that it can be wrong in ways the encoder would not write.
fn update(attributes: &[Vec<u8>], nlri: &[u8]) -> Vec<u8> {
    let attributes = attributes.concat();
    let mut body = vec![0, 0];
    body.extend_from_slice(&(attributes.len() as u16).to_be_bytes());
    body.extend_from_slice(&attributes);
    body.extend_from_slice(nlri);

    let mut wire = vec![0xff; 16];
    wire.extend_from_slice(&((19 + body.len()) as u16).to_be_bytes());
    wire.push(2);
    wire.extend_from_slice(&body);
    wire
}

/// 10.1.0.0/16 and 10.2.0.0/16.
const ROUTE_A: [u8; 3] = [16, 10, 1];
const ROUTE_B: [u8; 3] = [16, 10, 2];

fn prefix(s: &str) -> IpPrefix {
    IpPrefix::try_from(s).unwrap()
}

async fn established(server: &TestServer, asn: u32) -> Result<TestPeer> {
    let mut peer = TestPeer::connect(server.addr).await?;
    peer.handshake(asn, Ipv4Addr::new(10, 0, 0, 1), 90).await?;
    Ok(peer)
}

/// Waits until the server's Adj-RIB-In holds exactly `prefixes`.
async fn wait_for_routes(server: &TestServer, prefixes: &[&str]) -> Result<Vec<SnapshotEntry>> {
    let mut expected: Vec<IpPrefix> = prefixes.iter().map(|p| prefix(p)).collect();
    expected.sort_by_key(|p| p.to_string());
    Ok(tokio::time::timeout(TIMEOUT, async {
        loop {
            let mut routes = server.adj_rib_in.snapshot().await;
            routes.sort_by_key(|r| r.prefix.to_string());
            if routes.iter().map(|r| &r.prefix).eq(expected.iter()) {
                return routes;
            }
            tokio::time::sleep(Duration::from_millis(10)).await;
        }
    })
    .await?)
}

/// The situation RFC 7606 exists for: a route is in the table, and then an
/// UPDATE for it arrives with a broken attribute. The route goes; the
/// session, and the other routes learned over it, stay.
#[tokio::test]
async fn test_a_malformed_attribute_withdraws_the_route_and_keeps_the_session() -> Result<()> {
    let server = TestServer::start(LOCAL_ASN).await?;
    let mut peer = established(&server, PEER_ASN).await?;
    let handled_before = UPDATE_ERRORS
        .with_label_values(&["treat-as-withdraw"])
        .get();

    let good = [origin(), as_path(), next_hop()];
    peer.send_raw(&update(&good, &[ROUTE_A, ROUTE_B].concat()))
        .await?;
    wait_for_routes(&server, &["10.1.0.0/16", "10.2.0.0/16"]).await?;

    // ORIGIN with a value that is not defined.
    let bad = [attr(WELL_KNOWN, 1, &[7]), as_path(), next_hop()];
    peer.send_raw(&update(&bad, &ROUTE_A)).await?;
    wait_for_routes(&server, &["10.2.0.0/16"]).await?;

    // The session goes on: the route can be announced again.
    peer.send_raw(&update(&good, &ROUTE_A)).await?;
    wait_for_routes(&server, &["10.1.0.0/16", "10.2.0.0/16"]).await?;
    assert_eq!(server.registry.list().len(), 1);
    assert!(
        UPDATE_ERRORS
            .with_label_values(&["treat-as-withdraw"])
            .get()
            > handled_before
    );
    Ok(())
}

/// An announcement without a mandatory attribute is not a route we can use.
#[tokio::test]
async fn test_a_route_without_as_path_is_not_accepted() -> Result<()> {
    let server = TestServer::start(LOCAL_ASN).await?;
    let mut peer = established(&server, PEER_ASN).await?;

    peer.send_raw(&update(&[origin(), next_hop()], &ROUTE_A))
        .await?;
    // A well-formed one after it shows that the first was dealt with.
    peer.send_raw(&update(&[origin(), as_path(), next_hop()], &ROUTE_B))
        .await?;

    wait_for_routes(&server, &["10.2.0.0/16"]).await?;
    assert_eq!(server.registry.list().len(), 1);
    Ok(())
}

/// A broken attribute that plays no part in choosing a route is dropped,
/// and the route is kept.
#[tokio::test]
async fn test_a_malformed_aggregator_is_dropped_and_the_route_kept() -> Result<()> {
    let server = TestServer::start(LOCAL_ASN).await?;
    let mut peer = established(&server, PEER_ASN).await?;

    let attributes = [
        origin(),
        as_path(),
        next_hop(),
        // AGGREGATOR is eight octets on this session.
        attr(OPTIONAL_TRANSITIVE, 7, &[0, 0, 0xfd, 0xe9, 192]),
    ];
    peer.send_raw(&update(&attributes, &ROUTE_A)).await?;

    let routes = wait_for_routes(&server, &["10.1.0.0/16"]).await?;
    assert_eq!(routes[0].entry.path_attributes.len(), 3);
    assert!(
        !routes[0]
            .entry
            .path_attributes
            .iter()
            .any(|a| matches!(a, PathAttribute::Aggregator(_)))
    );
    Ok(())
}

/// LOCAL_PREF is a matter between routers of one AS. From an external peer
/// it is dropped; from an internal one it is kept.
#[tokio::test]
async fn test_local_pref_is_kept_only_from_an_internal_peer() -> Result<()> {
    let local_pref = attr(WELL_KNOWN, 5, &[0, 0, 0, 200]);
    let has_local_pref = |routes: &[SnapshotEntry]| {
        routes[0]
            .entry
            .path_attributes
            .contains(&PathAttribute::LocalPref(LocalPrefPathAttribute(200)))
    };

    let server = TestServer::start(LOCAL_ASN).await?;
    let mut external = established(&server, PEER_ASN).await?;
    let attributes = [origin(), as_path(), next_hop(), local_pref.clone()];
    external.send_raw(&update(&attributes, &ROUTE_A)).await?;
    let routes = wait_for_routes(&server, &["10.1.0.0/16"]).await?;
    assert!(!has_local_pref(&routes));

    let server = TestServer::start(LOCAL_ASN).await?;
    let mut internal = established(&server, LOCAL_ASN).await?;
    // An internal peer has not added its AS to the path.
    let attributes = [origin(), attr(WELL_KNOWN, 2, &[]), next_hop(), local_pref];
    internal.send_raw(&update(&attributes, &ROUTE_A)).await?;
    let routes = wait_for_routes(&server, &["10.1.0.0/16"]).await?;
    assert!(has_local_pref(&routes));
    Ok(())
}

/// When the prefixes themselves cannot be read there is nothing to
/// withdraw, and the session is reset as before.
#[tokio::test]
async fn test_an_unreadable_prefix_still_resets_the_session() -> Result<()> {
    let server = TestServer::start(LOCAL_ASN).await?;
    let mut peer = established(&server, PEER_ASN).await?;

    // An IPv4 prefix of length 33.
    peer.send_raw(&update(
        &[origin(), as_path(), next_hop()],
        &[33, 10, 1, 2, 3, 4],
    ))
    .await?;

    loop {
        match tokio::time::timeout(TIMEOUT, peer.recv()).await?? {
            Message::KeepAlive => {}
            Message::Notification(n) => {
                assert_eq!(
                    n.error,
                    BgpError::UpdateMsg(UpdateMsgSubcode::InvalidNetworkField)
                );
                break;
            }
            other => bail!("expected NOTIFICATION, got {other:?}"),
        }
    }
    tokio::time::timeout(TIMEOUT, async {
        while !server.registry.list().is_empty() {
            tokio::time::sleep(Duration::from_millis(10)).await;
        }
    })
    .await?;
    Ok(())
}