path: root/crates/daemon/src/main.rs

#[allow(unused_imports)]
use clap::Parser;
use eyre::{Result, WrapErr, bail, eyre};
use fs4::fs_std::FileExt;
use std::fs::{self, File, OpenOptions};
use std::io::{ErrorKind, Write};
#[cfg(unix)]
use std::os::unix::net::UnixStream as StdUnixStream;
use std::path::{Path, PathBuf};
use std::process::{Command, Stdio};
use std::time::{Duration, Instant};
use tokio::time::sleep;
use turtle_daemon::{
    DaemonEvent,
    aclient::{
        database::ClientSqlite, history::History, record::sqlite_store::SqliteStore,
        settings::Settings,
    },
    client::{ControlClient, DaemonClientErrorKind, HistoryClient, classify_error},
};

#[derive(Parser, Debug)]
#[command(infer_subcommands = true)]
pub(crate) enum Cmd {
    /// Start the daemon server
    Start {
        #[arg(long, hide = true)]
        daemonize: bool,

        /// Also write daemon logs to the console (useful for debugging)
        #[arg(long)]
        show_logs: bool,
    },

    /// Show the daemon's current status
    Status,

    /// Stop the daemon gracefully
    Stop,
}

impl Cmd {
    pub(crate) async fn run(
        self,
        settings: Settings,
        store: SqliteStore,
        history_db: ClientSqlite,
    ) -> Result<()> {
        match self {
            Cmd::Start { .. } => run(settings, store, history_db).await,
            Cmd::Status => status_cmd(&settings).await,
            Cmd::Stop => stop_cmd(&settings).await,
        }
    }
}

#[tokio::main]
async fn main() -> Result<()> {
    let settings = Settings::new().wrap_err("could not load client settings")?;
    let db_path = PathBuf::from(settings.db_path.as_str());
    let record_store_path = PathBuf::from(settings.record_store_path.as_str());

    let db = ClientSqlite::new(db_path, settings.local_timeout).await?;
    let sqlite_store = SqliteStore::new(record_store_path, settings.local_timeout).await?;

    Cmd::parse().run(settings, sqlite_store, db).await
}

const DAEMON_VERSION: &str = env!("CARGO_PKG_VERSION");
const DAEMON_PROTOCOL_VERSION: u32 = 1;
const STARTUP_POLL: Duration = Duration::from_millis(40);
const LOCK_POLL: Duration = Duration::from_millis(20);
const LEGACY_DAEMON_RESTART_MESSAGE: &str = "legacy daemon detected; restart daemon manually";

struct PidfileGuard {
    file: File,
}

impl PidfileGuard {
    fn acquire(path: &Path) -> Result<Self> {
        let mut file = open_lock_file(path)?;

        if !file.try_lock_exclusive()? {
            bail!(
                "daemon already running (pidfile lock busy at {})",
                path.display()
            );
        }

        file.set_len(0)
            .wrap_err_with(|| format!("could not truncate daemon pidfile {}", path.display()))?;
        writeln!(file, "{}", std::process::id())
            .and_then(|()| writeln!(file, "{DAEMON_VERSION}"))
            .wrap_err_with(|| format!("could not write daemon pidfile {}", path.display()))?;

        Ok(Self { file })
    }
}

impl Drop for PidfileGuard {
    fn drop(&mut self) {
        drop(self.file.unlock());
    }
}

enum Probe {
    Ready(HistoryClient),
    NeedsRestart(String),
    Unreachable(eyre::Report),
}

fn daemon_matches_expected(version: &str, protocol: u32) -> bool {
    version == DAEMON_VERSION && protocol == DAEMON_PROTOCOL_VERSION
}

fn daemon_mismatch_message(version: &str, protocol: u32) -> String {
    if protocol == DAEMON_PROTOCOL_VERSION {
        format!("daemon is out of date: expected {DAEMON_VERSION}, got {version}")
    } else {
        format!("daemon protocol mismatch: expected {DAEMON_PROTOCOL_VERSION}, got {protocol}")
    }
}

fn is_legacy_daemon_error(err: &eyre::Report) -> bool {
    matches!(classify_error(err), DaemonClientErrorKind::Unimplemented)
}

fn open_lock_file(path: &Path) -> Result<File> {
    if let Some(parent) = path.parent() {
        fs::create_dir_all(parent)
            .wrap_err_with(|| format!("could not create lock directory {}", parent.display()))?;
    }

    OpenOptions::new()
        .read(true)
        .write(true)
        .create(true)
        .truncate(false)
        .open(path)
        .wrap_err_with(|| format!("could not open lock file {}", path.display()))
}

async fn wait_for_lock(path: &Path, timeout: Duration) -> Result<File> {
    let file = open_lock_file(path)?;
    let start = Instant::now();

    loop {
        match file.try_lock_exclusive() {
            Ok(true) => return Ok(file),
            Ok(false) => {
                if start.elapsed() >= timeout {
                    bail!("timed out waiting for lock at {}", path.display());
                }

                sleep(LOCK_POLL).await;
            }
            Err(err) => {
                return Err(eyre!("could not lock {}: {err}", path.display()));
            }
        }
    }
}

async fn wait_for_pidfile_available(path: &Path, timeout: Duration) -> Result<()> {
    let file = wait_for_lock(path, timeout).await?;
    file.unlock()
        .wrap_err_with(|| format!("failed to unlock {}", path.display()))?;
    Ok(())
}

async fn connect_client(settings: &Settings) -> Result<HistoryClient> {
    HistoryClient::new(
        #[cfg(unix)]
        settings.daemon.socket_path.clone(),
    )
    .await
}

async fn probe(settings: &Settings) -> Probe {
    let mut client = match connect_client(settings).await {
        Ok(client) => client,
        Err(err) => return Probe::Unreachable(err),
    };

    match client.status().await {
        Ok(status) => {
            if daemon_matches_expected(&status.version, status.protocol) {
                Probe::Ready(client)
            } else {
                Probe::NeedsRestart(daemon_mismatch_message(&status.version, status.protocol))
            }
        }
        Err(err) => Probe::Unreachable(err),
    }
}

async fn request_shutdown(settings: &Settings) {
    if let Ok(mut client) = connect_client(settings).await {
        drop(client.shutdown().await);
    }
}

fn startup_timeout(settings: &Settings) -> Duration {
    Duration::from_secs_f64(settings.local_timeout.max(0.5) + 2.0)
}

pub(crate) async fn start_history(settings: &Settings, history: History) -> Result<String> {
    match async {
        connect_client(settings)
            .await?
            .start_history(history.clone())
            .await
    }
    .await
    {
        Ok(resp) => {
            if daemon_matches_expected(&resp.version, resp.protocol) {
                return Ok(resp.id);
            }

            Err(eyre!(
                "{}. Restart the daemon manually",
                daemon_mismatch_message(&resp.version, resp.protocol)
            ))
        }
        Err(err) => Err(err),
    }
}

pub(crate) async fn end_history(
    settings: &Settings,
    id: String,
    duration: u64,
    exit: i64,
) -> Result<()> {
    match async {
        connect_client(settings)
            .await?
            .end_history(id.clone(), duration, exit)
            .await
    }
    .await
    {
        Ok(resp) => {
            if daemon_matches_expected(&resp.version, resp.protocol) {
                return Ok(());
            }

            Err(eyre!(
                "{}. Restart the daemon manually",
                daemon_mismatch_message(&resp.version, resp.protocol)
            ))
        }
        Err(err) => Err(err),
    }
}

/// Emit a daemon event.
pub(crate) async fn emit_event(settings: &Settings, event: DaemonEvent) {
    // Try to connect and send
    match ControlClient::from_settings(settings).await {
        Ok(mut client) => {
            if let Err(e) = client.send_event(event).await {
                tracing::debug!(?e, "failed to send event to daemon");
            }
        }
        Err(e) => {
            tracing::debug!(?e, "daemon not available, skipping event emission");
        }
    }
}

pub(crate) async fn tail_client(settings: &Settings) -> Result<HistoryClient> {
    match probe(settings).await {
        Probe::Ready(client) => Ok(client),
        Probe::NeedsRestart(reason) => {
            bail!("{reason}. Restart the daemon manually");
        }
        Probe::Unreachable(err) if is_legacy_daemon_error(&err) => {
            Err(err.wrap_err(LEGACY_DAEMON_RESTART_MESSAGE))
        }
        Probe::Unreachable(err) => Err(err),
    }
}

async fn status_cmd(settings: &Settings) -> Result<()> {
    match probe(settings).await {
        Probe::Ready(mut client) => {
            let status = client.status().await?;
            println!("Daemon running");
            println!("  PID:      {}", status.pid);
            println!("  Version:  {}", status.version);
            println!("  Protocol: {}", status.protocol);
            println!("  Healthy:  {}", status.healthy);
            #[cfg(unix)]
            println!("  Socket:   {}", settings.daemon.socket_path);
        }
        Probe::NeedsRestart(reason) => {
            println!("Daemon running (needs restart)");
            println!("  Reason: {reason}");
        }
        Probe::Unreachable(_) => {
            println!("Daemon is not running");
        }
    }

    Ok(())
}

async fn stop_cmd(settings: &Settings) -> Result<()> {
    let Ok(mut client) = connect_client(settings).await else {
        println!("Daemon is not running");
        return Ok(());
    };

    match client.shutdown().await {
        Ok(true) => {
            println!("Shutdown requested");

            let pidfile_path = PathBuf::from(&settings.daemon.pidfile_path);
            let timeout = Duration::from_secs(5);
            match wait_for_pidfile_available(&pidfile_path, timeout).await {
                Ok(()) => println!("Daemon stopped"),
                Err(_) => println!("Daemon may still be shutting down"),
            }

            Ok(())
        }
        Ok(false) => bail!("Daemon rejected shutdown request"),
        Err(err) => Err(err.wrap_err("Failed to send shutdown request")),
    }
}

async fn run(settings: Settings, store: SqliteStore, history_db: ClientSqlite) -> Result<()> {
    let pidfile_path = PathBuf::from(&settings.daemon.pidfile_path);
    let _pidfile_guard = PidfileGuard::acquire(&pidfile_path)?;

    turtle_daemon::boot(settings, store, history_db).await?;

    Ok(())
}

#[cfg(test)]
mod tests {
    use super::{
        DAEMON_PROTOCOL_VERSION, DAEMON_VERSION, daemon_matches_expected, daemon_mismatch_message,
    };

    #[test]
    fn test_version_matches() {
        assert!(daemon_matches_expected(
            DAEMON_VERSION,
            DAEMON_PROTOCOL_VERSION
        ));
    }

    #[test]
    fn test_version_mismatch() {
        assert!(!daemon_matches_expected("0.0.0", DAEMON_PROTOCOL_VERSION));
        assert!(!daemon_matches_expected(DAEMON_VERSION, 999));
        assert!(!daemon_matches_expected("0.0.0", 999));
    }

    #[test]
    fn test_mismatch_message_version() {
        let msg = daemon_mismatch_message("0.0.0", DAEMON_PROTOCOL_VERSION);
        assert!(msg.contains("out of date"), "got: {msg}");
        assert!(msg.contains("0.0.0"));
        assert!(msg.contains(DAEMON_VERSION));
    }

    #[test]
    fn test_mismatch_message_protocol() {
        let msg = daemon_mismatch_message(DAEMON_VERSION, 999);
        assert!(msg.contains("protocol mismatch"), "got: {msg}");
    }
}