astrid-runtime/astrid · error

WinFsp service control endpoint is already present

Error message

WinFsp service control endpoint is already present

What it means

Before starting the private service, validate_service_launch checks that no stale control endpoint (Unix-domain-style socket) already exists at control_path. If local_transport::endpoint_is_present reports the path is occupied, startup is aborted to avoid binding conflicts or hijacking a pre-existing socket.

Solutions

1. Stop the existing daemon that owns the socket, or delete the stale process-control.sock file, then relaunch.
2. Check for a running duplicate instance (ps / process manager) and use it instead of spawning a new one.
3. Ensure shutdown paths unlink the control endpoint so a clean restart leaves no file behind.

Example fix

// before
spawn_daemon(&lease, &launch)?; // fails: socket still present
// after
if local_transport::endpoint_is_present(&launch.control_path)? {
    std::fs::remove_file(&launch.control_path)?; // stale socket from crashed daemon
}
spawn_daemon(&lease, &launch)?;

if local_transport::endpoint_is_present(&launch.control_path)? {
    std::fs::remove_file(&launch.control_path)?; // only after confirming no live owner
}

fn endpoint_free(path: &Path) -> bool {
    !path.exists()
}

match service_main(&launch).await {
    Err(e) if e.to_string().contains("already present") => {
        cleanup_stale_socket(&launch.control_path)?;
        service_main(&launch).await
    },
    other => other,
}

When it happens

Trigger: service_main -> validate_service_launch when a previous daemon crashed without cleaning up its socket, or when two instances are launched with the same lease resource path.

Common situations: Killed daemon leaving an orphaned process-control.sock; double-starting the mount service; running a second mount for the same resource concurrently; container restarts with a persistent volume that kept the socket file.

Related errors

• WinFsp daemon exited immediately after readiness
• build WinFsp callback filesystem
• candidate generation for
• Daemon exited prematurely
• Daemon is still starting after

AI-assisted analysis of astrid-runtime/astrid@affd8760f4 (2026-09-09). Data as JSON: /api/errors/0cab9eff17504c70. Report an issue: GitHub.

Appendix: source

Thrown at crates/astrid-storage-provider-winfsp/src/win.rs:333

            .control_path
            .components()
            .any(|component| matches!(component, std::path::Component::ParentDir))
        || launch.control_path != lease.resource_path.join("process-control.sock")
    {
        bail!("WinFsp service control path is malformed");
    }
    let control_parent = launch
        .control_path
        .parent()
        .context("WinFsp service control path has no parent")?;
    platform_fs::validate_private_directory(control_parent)
        .context("validate private WinFsp control parent")?;
    platform_fs::verify_no_redirects(&launch.control_path)
        .context("reject redirected WinFsp control path")?;
    if local_transport::endpoint_is_present(&launch.control_path)
        .context("inspect WinFsp service control endpoint")?
    {
        bail!("WinFsp service control endpoint is already present");
    }
    Ok(())
}

async fn probe_callback(launch: &StorageProviderServiceLaunchV1) -> Result<()> {
    let mut stream = local_transport::connect(&launch.lease.callback_path)
        .await
        .context("connect WinFsp lease callback")?;
    let request = StorageFilesystemRequestV2 {
        protocol_version: STORAGE_FILESYSTEM_PROTOCOL_V2,
        request_id: format!("winfsp-service-{}", launch.lease.mount_id),
        lease_token: launch.lease.lease_token.clone(),
        operation: StorageFilesystemOperationV2::Stat {
            path: String::new(),
        },
    };
    let bytes = serde_json::to_vec(&request).context("encode WinFsp callback probe")?;
    let length = u32::try_from(bytes.len()).context("WinFsp callback probe is too large")?;

View on GitHub (pinned to affd8760f4)