hyperhive/hivectl/src/client.rs

//! Host admin socket client: one request/response round trip over
//! `/run/hyperhive/host.sock`.
//!
//! Connect failures are classified into an actionable message before they
//! reach the operator (see [`connect_hint`]) — the three ways this fails
//! (not in `hive-admin`, no socket, nobody listening) need three different
//! fixes, and the raw `Permission denied (os error 13)` names none of them.

use std::io::ErrorKind;
use std::path::Path;

use anyhow::{Result, bail};
use hive_host_sock::{HostRequest, HostResponse};
use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader};
use tokio::net::UnixStream;

/// Turn a socket-connect `io::ErrorKind` into a message that names the fix.
///
/// The interesting one is `PermissionDenied`: the socket is `0660
/// root:hive-admin` behind a `0751` runtime dir, so a non-member gets EACCES
/// on connect — and so does a *member* whose shell predates the group being
/// granted, because secondary group membership is only applied at login. That
/// second case is the one that reads as "the daemon is down" and isn't.
fn connect_hint(kind: ErrorKind, socket: &Path) -> String {
    let path = socket.display();
    match kind {
        ErrorKind::PermissionDenied => format!(
            "permission denied opening the host admin socket {path} — reaching it requires \
             membership in the `hive-admin` group. Add your user to \
             `services.hyperhive.adminUsers`, then log out and back in: secondary group \
             membership is applied at login, so a shell opened before the change still \
             cannot connect. `id -nG` shows what the running shell actually has; \
             `newgrp hive-admin` picks the group up without a full re-login"
        ),
        ErrorKind::NotFound => format!(
            "no socket at {path} — hive-c0re is not running, or it binds somewhere else \
             (`--socket`). Check `systemctl status hive-c0re.socket hive-c0re.service`"
        ),
        ErrorKind::ConnectionRefused => format!(
            "nothing is listening on {path} — the socket exists but the daemon behind it \
             is down. Check `systemctl status hive-c0re.service` and \
             `journalctl -u hive-c0re`"
        ),
        _ => format!("could not connect to the hive-c0re host admin socket {path}"),
    }
}

pub async fn request(socket: &Path, req: HostRequest) -> Result<HostResponse> {
    let stream = match UnixStream::connect(socket).await {
        Ok(stream) => stream,
        // Keep the io error as the cause and put the actionable line on top,
        // so `Error: <hint>` / `Caused by: <errno>` reads fix-first.
        Err(e) => {
            let hint = connect_hint(e.kind(), socket);
            return Err(anyhow::Error::new(e).context(hint));
        }
    };
    let (read, mut write) = stream.into_split();

    let mut payload = serde_json::to_string(&req)?;
    payload.push('\n');
    write.write_all(payload.as_bytes()).await?;
    write.flush().await?;

    let mut reader = BufReader::new(read);
    let mut line = String::new();
    reader.read_line(&mut line).await?;
    if line.is_empty() {
        bail!("server closed connection without responding");
    }
    let resp: HostResponse = serde_json::from_str(line.trim())?;
    Ok(resp)
}

#[cfg(test)]
mod tests {
    use super::{ErrorKind, Path, connect_hint};

    fn hint(kind: ErrorKind) -> String {
        connect_hint(kind, Path::new("/run/hyperhive/host.sock"))
    }

    /// EACCES is an operator-side fix, and the non-obvious half is the
    /// re-login — so both the group and the login requirement must appear,
    /// and it must not send the operator off checking a daemon that is fine.
    #[test]
    fn permission_denied_names_the_group_and_the_relogin() {
        let h = hint(ErrorKind::PermissionDenied);
        assert!(h.contains("hive-admin"), "{h}");
        assert!(h.contains("adminUsers"), "{h}");
        assert!(h.contains("log out"), "{h}");
        assert!(
            !h.contains("not running"),
            "EACCES must not blame the daemon: {h}"
        );
    }

    /// The inverse: a missing socket is a daemon-side problem, so it must not
    /// send the operator chasing group membership.
    #[test]
    fn missing_socket_blames_the_daemon_not_the_operator() {
        let h = hint(ErrorKind::NotFound);
        assert!(h.contains("not running"), "{h}");
        assert!(!h.contains("hive-admin"), "{h}");
    }

    #[test]
    fn refused_socket_points_at_the_service_not_the_group() {
        let h = hint(ErrorKind::ConnectionRefused);
        assert!(h.contains("nothing is listening"), "{h}");
        assert!(!h.contains("hive-admin"), "{h}");
    }

    #[test]
    fn every_hint_names_the_socket_path() {
        for kind in [
            ErrorKind::PermissionDenied,
            ErrorKind::NotFound,
            ErrorKind::ConnectionRefused,
            ErrorKind::BrokenPipe,
        ] {
            let h = hint(kind);
            assert!(h.contains("/run/hyperhive/host.sock"), "{kind:?}: {h}");
        }
    }
}