Struct Connection

pub struct Connection { /* private fields */ }

A D-Bus connection.

A connection to a D-Bus bus, or a direct peer.

Once created, the connection is authenticated and negotiated and messages can be sent or received, such as method calls or signals.

For higher-level message handling (typed functions, introspection, documentation reasons etc), it is recommended to wrap the low-level D-Bus messages into Rust functions with the proxy and interface macros instead of doing it directly on a Connection.

Typically, a connection is made to the session bus with Connection::session, or to the system bus with Connection::system. Then the connection is used with crate::Proxy instances or the on-demand ObjectServer instance that can be accessed through Connection::object_server.

Connection implements Clone and cloning it is a very cheap operation, as the underlying data is not cloned. This makes it very convenient to share the connection between different parts of your code. Connection also implements std::marker::Sync and std::marker::Send so you can send and share a connection instance across threads as well.

Connection keeps internal queues of incoming message. The default capacity of each of these is 64. The capacity of the main (unfiltered) queue is configurable through the set_max_queued method. When the queue is full, no more messages can be received until room is created for more. This is why it’s important to ensure that all crate::MessageStream and crate::blocking::MessageIterator instances are continuously polled and iterated on, respectively.

For sending messages you can use the Connection::send method.

To gracefully close a connection while waiting for any outstanding method calls to complete, use Connection::graceful_shutdown. To immediately close a connection in a way that will disrupt any outstanding method calls, use Connection::close. If you do not need the shutdown to be immediate and do not care about waiting for outstanding method calls, you can also simply drop the Connection instance, which will act similarly to spawning graceful_shutdown in the background.

Examples

Get the session bus ID

use zbus::Connection;

let connection = Connection::session().await?;

let reply_body = connection
    .call_method(
        Some("org.freedesktop.DBus"),
        "/org/freedesktop/DBus",
        Some("org.freedesktop.DBus"),
        "GetId",
        &(),
    )
    .await?
    .body();

let id: &str = reply_body.deserialize()?;
println!("Unique ID of the bus: {}", id);

Monitoring all messages

Let’s eavesdrop on the session bus 😈 using the Monitor interface:

use futures_util::stream::TryStreamExt;
use zbus::{Connection, MessageStream};

let connection = Connection::session().await?;

connection
    .call_method(
        Some("org.freedesktop.DBus"),
        "/org/freedesktop/DBus",
        Some("org.freedesktop.DBus.Monitoring"),
        "BecomeMonitor",
        &(&[] as &[&str], 0u32),
    )
    .await?;

let mut stream = MessageStream::from(connection);
while let Some(msg) = stream.try_next().await? {
    println!("Got message: {}", msg);
}

This should print something like:

Got message: Signal NameAcquired from org.freedesktop.DBus
Got message: Signal NameLost from org.freedesktop.DBus
Got message: Method call GetConnectionUnixProcessID from :1.1324
Got message: Error org.freedesktop.DBus.Error.NameHasNoOwner:
             Could not get PID of name ':1.1332': no such name from org.freedesktop.DBus
Got message: Method call AddMatch from :1.918
Got message: Method return from org.freedesktop.DBus

Implementations

impl Connection

pub async fn send(&self, msg: &Message) -> Result<()>

Send msg to the peer.

pub async fn call_method<'d, 'p, 'i, 'm, D, P, I, M, B>( &self, destination: Option<D>, path: P, interface: Option<I>, method_name: M, body: &B, ) -> Result<Message>

where D: TryInto<BusName<'d>>, P: TryInto<ObjectPath<'p>>, I: TryInto<InterfaceName<'i>>, M: TryInto<MemberName<'m>>, D::Error: Into<Error>, P::Error: Into<Error>, I::Error: Into<Error>, M::Error: Into<Error>, B: Serialize + DynamicType,

Send a method call.

Create a method-call message, send it over the connection, then wait for the reply.

On successful reply, an Ok(Message) is returned. On error, an Err is returned. D-Bus error replies are returned as Error::MethodError.

pub async fn emit_signal<'d, 'p, 'i, 'm, D, P, I, M, B>( &self, destination: Option<D>, path: P, interface: I, signal_name: M, body: &B, ) -> Result<()>

where D: TryInto<BusName<'d>>, P: TryInto<ObjectPath<'p>>, I: TryInto<InterfaceName<'i>>, M: TryInto<MemberName<'m>>, D::Error: Into<Error>, P::Error: Into<Error>, I::Error: Into<Error>, M::Error: Into<Error>, B: Serialize + DynamicType,

Emit a signal.

Create a signal message, and send it over the connection.

pub async fn reply<B>(&self, call: &Header<'_>, body: &B) -> Result<()>

where B: Serialize + DynamicType,

Reply to a message.

Given an existing message (likely a method call), send a reply back to the caller with the given body.

pub async fn reply_error<'e, E, B>( &self, call: &Header<'_>, error_name: E, body: &B, ) -> Result<()>

where B: Serialize + DynamicType, E: TryInto<ErrorName<'e>>, E::Error: Into<Error>,

Reply an error to a message.

Given an existing message (likely a method call), send an error reply back to the caller with the given error_name and body.

pub async fn reply_dbus_error( &self, call: &Header<'_>, err: impl DBusError, ) -> Result<()>

Reply an error to a message.

Given an existing message (likely a method call), send an error reply back to the caller using one of the standard interface reply types.

pub async fn request_name<'w, W>(&self, well_known_name: W) -> Result<()>

where W: TryInto<WellKnownName<'w>>, W::Error: Into<Error>,

Register a well-known name for this connection.

When connecting to a bus, the name is requested from the bus. In case of p2p connection, the name (if requested) is used for self-identification.

You can request multiple names for the same connection. Use Connection::release_name for deregistering names registered through this method.

Note that exclusive ownership without queueing is requested (using RequestNameFlags::ReplaceExisting and RequestNameFlags::DoNotQueue flags) since that is the most typical case. If that is not what you want, you should use Connection::request_name_with_flags instead (but make sure then that name is requested after you’ve set up your service implementation with the ObjectServer).

Caveats

The associated ObjectServer will only handle method calls destined for the unique name of this connection or any of the registered well-known names. If no well-known name is registered, the method calls destined to all well-known names will be handled.

Since names registered through any other means than Connection or Builder API are not known to the connection, method calls destined to those names will only be handled by the associated ObjectServer if none of the names are registered through Connection* API. Simply put, either register all the names through Connection* API or none of them.

Errors

Fails with zbus::Error::NameTaken if the name is already owned by another peer.

pub async fn request_name_with_flags<'w, W>( &self, well_known_name: W, flags: BitFlags<RequestNameFlags>, ) -> Result<RequestNameReply>

where W: TryInto<WellKnownName<'w>>, W::Error: Into<Error>,

Register a well-known name for this connection.

This is the same as Connection::request_name but allows to specify the flags to use when requesting the name.

If the RequestNameFlags::DoNotQueue flag is not specified and request ends up in the queue, you can use fdo::NameAcquiredStream to be notified when the name is acquired. A queued name request can be cancelled using Connection::release_name.

If the RequestNameFlags::AllowReplacement flag is specified, the requested name can be lost if another peer requests the same name. You can use fdo::NameLostStream to be notified when the name is lost

Example

use zbus::{Connection, fdo::{DBusProxy, RequestNameFlags, RequestNameReply}};
use enumflags2::BitFlags;
use futures_util::stream::StreamExt;

let name = "org.freedesktop.zbus.QueuedNameTest";
let conn1 = Connection::session().await?;
// This should just work right away.
conn1.request_name(name).await?;

let conn2 = Connection::session().await?;
// A second request from the another connection will fail with `DoNotQueue` flag, which is
// implicit with `request_name` method.
assert!(conn2.request_name(name).await.is_err());

// Now let's try w/o `DoNotQueue` and we should be queued.
let reply = conn2
    .request_name_with_flags(name, RequestNameFlags::AllowReplacement.into())
    .await?;
assert_eq!(reply, RequestNameReply::InQueue);
// Another request should just give us the same response.
let reply = conn2
    // The flags on subsequent requests will however be ignored.
    .request_name_with_flags(name, BitFlags::empty())
    .await?;
assert_eq!(reply, RequestNameReply::InQueue);
let mut acquired_stream = DBusProxy::new(&conn2)
    .await?
    .receive_name_acquired()
    .await?;
assert!(conn1.release_name(name).await?);
// This would have waited forever if `conn1` hadn't just release the name.
let acquired = acquired_stream.next().await.unwrap();
assert_eq!(acquired.args().unwrap().name, name);

// conn2 made the mistake of being too nice and allowed name replacemnt, so conn1 should be
// able to take it back.
let mut lost_stream = DBusProxy::new(&conn2)
    .await?
    .receive_name_lost()
    .await?;
conn1.request_name(name).await?;
let lost = lost_stream.next().await.unwrap();
assert_eq!(lost.args().unwrap().name, name);

Caveats

• Same as that of Connection::request_name.
• If you wish to track changes to name ownership after this call, make sure that the fdo::NameAcquired and/or fdo::NameLostStream instance(s) are created before calling this method. Otherwise, you may loose the signal if it’s emitted after this call but just before the stream instance get created.

pub async fn release_name<'w, W>(&self, well_known_name: W) -> Result<bool>

where W: TryInto<WellKnownName<'w>>, W::Error: Into<Error>,

Deregister a previously registered well-known name for this service on the bus.

Use this method to deregister a well-known name, registered through Connection::request_name.

Unless an error is encountered, returns Ok(true) if name was previously registered with the bus through self and it has now been successfully deregistered, Ok(false) if name was not previously registered or already deregistered.

pub fn is_bus(&self) -> bool

Check if self is a connection to a message bus.

This will return false for p2p connections. When the p2p feature is disabled, this will always return true.

pub fn unique_name(&self) -> Option<&OwnedUniqueName>

The unique name of the connection, if set/applicable.

The unique name is assigned by the message bus or set manually using Connection::set_unique_name.

pub fn set_unique_name<U>(&self, unique_name: U) -> Result<()>

where U: TryInto<OwnedUniqueName>, U::Error: Into<Error>,

Set the unique name of the connection (if not already set).

This is mainly provided for bus implementations. All other users should not need to use this method. Hence why this method is only available when the bus-impl feature is enabled.

Panics

This method panics if the unique name is already set. It will always panic if the connection is to a message bus as it’s the bus that assigns peers their unique names.

pub fn max_queued(&self) -> usize

The capacity of the main (unfiltered) queue.

pub fn set_max_queued(&mut self, max: usize)

Set the capacity of the main (unfiltered) queue.

pub fn server_guid(&self) -> &OwnedGuid

The server’s GUID.

pub fn executor(&self) -> &Executor<'static>

The underlying executor.

When a connection is built with internal_executor set to false, zbus will not spawn a thread to run the executor. You’re responsible to continuously tick the executor. Failure to do so will result in hangs.

Examples

Here is how one would typically run the zbus executor through tokio’s scheduler:

use zbus::connection::Builder;
use tokio::task::spawn;

#[tokio::main]
async fn main() {
    let conn = Builder::session()
        .unwrap()
        .internal_executor(false)
        .build()
        .await
        .unwrap();
    {
       let conn = conn.clone();
       spawn(async move {
           loop {
               conn.executor().tick().await;
           }
       });
    }

    // All your other async code goes here.
}

Note: zbus 2.1 added support for tight integration with tokio. This means, if you use zbus with tokio, you do not need to worry about this at all. All you need to do is enable tokio feature. You should also disable the (default) async-io feature in your Cargo.toml to avoid unused dependencies. Also note that prior to zbus 3.0, disabling async-io was required to enable tight tokio integration.

pub fn object_server(&self) -> &ObjectServer

Get a reference to the associated ObjectServer.

The ObjectServer is created on-demand.

Note: Once the ObjectServer is created, it will be replying to all method calls received on self. If you want to manually reply to method calls, do not use this method (or any of the ObjectServer related API).

pub async fn session() -> Result<Self>

Create a Connection to the session/user message bus.

pub async fn system() -> Result<Self>

Create a Connection to the system-wide message bus.

pub fn monitor_activity(&self) -> EventListener

Return a listener, notified on various connection activity.

This function is meant for the caller to implement idle or timeout on inactivity.

pub async fn peer_credentials(&self) -> Result<ConnectionCredentials>

Return the peer credentials.

The fields are populated on the best effort basis. Some or all fields may not even make sense for certain sockets or on certain platforms and hence will be set to None.

Caveats

Currently unix_group_ids and linux_security_label fields are not populated.

pub async fn close(self) -> Result<()>

Close the connection.

After this call, all reading and writing operations will fail.

pub async fn graceful_shutdown(self)

Gracefully close the connection, waiting for all other references to be dropped.

This will not disrupt any incoming or outgoing method calls, and will await their completion.

Caveats

• This will not prevent new incoming messages from keeping the connection alive (and indefinitely delaying this method’s completion).

• The shutdown will not complete until the underlying connection is fully dropped, so beware of deadlocks if you are holding any other clones of this Connection.

Example

let conn = Builder::session()?
    .name("foo.bar.baz")?
    .serve_at("/foo/bar/baz", MyInterface)?
    .build()
    .await?;

some_exit_condition.await;

conn.graceful_shutdown().await;

Trait Implementations

impl Clone for Connection

fn clone(&self) -> Connection

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Connection

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl From<&Connection> for MessageStream

fn from(conn: &Connection) -> Self

Converts to this type from the input type.

impl From<&MessageStream> for Connection

fn from(stream: &MessageStream) -> Connection

Converts to this type from the input type.

impl From<Connection> for Connection

fn from(conn: Connection) -> Self

Converts to this type from the input type.

impl From<Connection> for Connection

fn from(conn: Connection) -> Self

Converts to this type from the input type.

impl From<Connection> for MessageStream

fn from(conn: Connection) -> Self

Converts to this type from the input type.

impl From<MessageStream> for Connection

fn from(stream: MessageStream) -> Connection

Converts to this type from the input type.