Chapter 4. Wayland Protocol and Model of Operation

Table of Contents

Basic Principles
Code Generation
Wire Format
Interfaces
Versioning
Connect Time
Security and Authentication
Creating Objects
Compositor
Surfaces
Input
Output
Data sharing between clients

Basic Principles

The Wayland protocol is an asynchronous object oriented protocol. All requests are method invocations on some object. The requests include an object ID that uniquely identifies an object on the server. Each object implements an interface and the requests include an opcode that identifies which method in the interface to invoke.

The protocol is message-based. A message sent by a client to the server is called request. A message from the server to a client is called event. A message has a number of arguments, each of which has a certain type (see the section called “Wire Format” for a list of argument types).

Additionally, the protocol can specify enums which associate names to specific numeric enumeration values. These are primarily just descriptive in nature: at the wire format level enums are just integers. But they also serve a secondary purpose to enhance type safety or otherwise add context for use in language bindings or other such code. This latter usage is only supported so long as code written before these attributes were introduced still works after; in other words, adding an enum should not break API, otherwise it puts backwards compatibility at risk.

enums can be defined as just a set of integers, or as bitfields. This is specified via the bitfield boolean attribute in the enum definition. If this attribute is true, the enum is intended to be accessed primarily using bitwise operations, for example when arbitrarily many choices of the enum can be ORed together; if it is false, or the attribute is omitted, then the enum arguments are a just a sequence of numerical values.

The enum attribute can be used on either uint or int arguments, however if the enum is defined as a bitfield, it can only be used on uint args.

The server sends back events to the client, each event is emitted from an object. Events can be error conditions. The event includes the object ID and the event opcode, from which the client can determine the type of event. Events are generated both in response to requests (in which case the request and the event constitutes a round trip) or spontaneously when the server state changes.

  • State is broadcast on connect, events are sent out when state changes. Clients must listen for these changes and cache the state. There is no need (or mechanism) to query server state.

  • The server will broadcast the presence of a number of global objects, which in turn will broadcast their current state.

Code Generation

The interfaces, requests and events are defined in protocol/wayland.xml. This xml is used to generate the function prototypes that can be used by clients and compositors.

The protocol entry points are generated as inline functions which just wrap the wl_proxy_* functions. The inline functions aren't part of the library ABI and language bindings should generate their own stubs for the protocol entry points from the xml.

Wire Format

The protocol is sent over a UNIX domain stream socket, where the endpoint usually is named wayland-0 (although it can be changed via WAYLAND_DISPLAY in the environment). Beginning in Wayland 1.15, implementations can optionally support server socket endpoints located at arbitrary locations in the filesystem by setting WAYLAND_DISPLAY to the absolute path at which the server endpoint listens.

Every message is structured as 32-bit words; values are represented in the host's byte-order. The message header has 2 words in it:

  • The first word is the sender's object ID (32-bit).

  • The second has 2 parts of 16-bit. The upper 16-bits are the message size in bytes, starting at the header (i.e. it has a minimum value of 8).The lower is the request/event opcode.

The payload describes the request/event arguments. Every argument is always aligned to 32-bits. Where padding is required, the value of padding bytes is undefined. There is no prefix that describes the type, but it is inferred implicitly from the xml specification.

The representation of argument types are as follows:

int, uint

The value is the 32-bit value of the signed/unsigned int.

fixed

Signed 24.8 decimal numbers. It is a signed decimal type which offers a sign bit, 23 bits of integer precision and 8 bits of decimal precision. This is exposed as an opaque struct with conversion helpers to and from double and int on the C API side.

string

Starts with an unsigned 32-bit length (including null terminator), followed by the string contents, including terminating null byte, then padding to a 32-bit boundary. A null value is represented with a length of 0.

object

32-bit object ID. A null value is represented with an ID of 0.

new_id

The 32-bit object ID. Generally, the interface used for the new object is inferred from the xml, but in the case where it's not specified, a new_id is preceded by a string specifying the interface name, and a uint specifying the version.

array

Starts with 32-bit array size in bytes, followed by the array contents verbatim, and finally padding to a 32-bit boundary.

fd

The file descriptor is not stored in the message buffer, but in the ancillary data of the UNIX domain socket message (msg_control).

The protocol does not specify the exact position of the ancillary data in the stream, except that the order of file descriptors is the same as the order of messages and fd arguments within messages on the wire.

In particular, it means that any byte of the stream, even the message header, may carry the ancillary data with file descriptors.

Clients and compositors should queue incoming data until they have whole messages to process, as file descriptors may arrive earlier or later than the corresponding data bytes.

Interfaces

The protocol includes several interfaces which are used for interacting with the server. Each interface provides requests, events, and errors (which are really just special events) as described above. Specific compositor implementations may have their own interfaces provided as extensions, but there are several which are always expected to be present.

Core interfaces:

wl_display
core global object
wl_registry
global registry object
wl_callback
callback object
wl_compositor
the compositor singleton
wl_shm_pool
a shared memory pool
wl_shm
shared memory support
wl_buffer
content for a wl_surface
wl_data_offer
offer to transfer data
wl_data_source
offer to transfer data
wl_data_device
data transfer device
wl_data_device_manager
data transfer interface
wl_shell
create desktop-style surfaces
wl_shell_surface
desktop-style metadata interface
wl_surface
an onscreen surface
wl_seat
group of input devices
wl_pointer
pointer input device
wl_keyboard
keyboard input device
wl_touch
touchscreen input device
wl_output
compositor output region
wl_region
region interface
wl_subcompositor
sub-surface compositing
wl_subsurface
sub-surface interface to a wl_surface

Versioning

Every interface is versioned and every protocol object implements a particular version of its interface. For global objects, the maximum version supported by the server is advertised with the global and the actual version of the created protocol object is determined by the version argument passed to wl_registry.bind(). For objects that are not globals, their version is inferred from the object that created them.

In order to keep things sane, this has a few implications for interface versions:

  • The object creation hierarchy must be a tree. Otherwise, inferring object versions from the parent object becomes a much more difficult to properly track.

  • When the version of an interface increases, so does the version of its parent (recursively until you get to a global interface)

  • A global interface's version number acts like a counter for all of its child interfaces. Whenever a child interface gets modified, the global parent's interface version number also increases (see above). The child interface then takes on the same version number as the new version of its parent global interface.

To illustrate the above, consider the wl_compositor interface. It has two children, wl_surface and wl_region. As of wayland version 1.2, wl_surface and wl_compositor are both at version 3. If something is added to the wl_region interface, both wl_region and wl_compositor will get bumpped to version 4. If, afterwards, wl_surface is changed, both wl_compositor and wl_surface will be at version 5. In this way the global interface version is used as a sort of "counter" for all of its child interfaces. This makes it very simple to know the version of the child given the version of its parent. The child is at the highest possible interface version that is less than or equal to its parent's version.

It is worth noting a particular exception to the above versioning scheme. The wl_display (and, by extension, wl_registry) interface cannot change because it is the core protocol object and its version is never advertised nor is there a mechanism to request a different version.

Connect Time

There is no fixed connection setup information, the server emits multiple events at connect time, to indicate the presence and properties of global objects: outputs, compositor, input devices.

Security and Authentication

  • mostly about access to underlying buffers, need new drm auth mechanism (the grant-to ioctl idea), need to check the cmd stream?

  • getting the server socket depends on the compositor type, could be a system wide name, through fd passing on the session dbus. or the client is forked by the compositor and the fd is already opened.

Creating Objects

Each object has a unique ID. The IDs are allocated by the entity creating the object (either client or server). IDs allocated by the client are in the range [1, 0xfeffffff] while IDs allocated by the server are in the range [0xff000000, 0xffffffff]. The 0 ID is reserved to represent a null or non-existent object. For efficiency purposes, the IDs are densely packed in the sense that the ID N will not be used until N-1 has been used. Any ID allocation algorithm that does not maintain this property is incompatible with the implementation in libwayland.

Compositor

The compositor is a global object, advertised at connect time.

See the section called “wl_compositor - the compositor singleton” for the protocol description.

Surfaces

A surface manages a rectangular grid of pixels that clients create for displaying their content to the screen. Clients don't know the global position of their surfaces, and cannot access other clients' surfaces.

Once the client has finished writing pixels, it 'commits' the buffer; this permits the compositor to access the buffer and read the pixels. When the compositor is finished, it releases the buffer back to the client.

See the section called “wl_surface - an onscreen surface” for the protocol description.

Input

A seat represents a group of input devices including mice, keyboards and touchscreens. It has a keyboard and pointer focus. Seats are global objects. Pointer events are delivered in surface-local coordinates.

The compositor maintains an implicit grab when a button is pressed, to ensure that the corresponding button release event gets delivered to the same surface. But there is no way for clients to take an explicit grab. Instead, surfaces can be mapped as 'popup', which combines transient window semantics with a pointer grab.

To avoid race conditions, input events that are likely to trigger further requests (such as button presses, key events, pointer motions) carry serial numbers, and requests such as wl_surface.set_popup require that the serial number of the triggering event is specified. The server maintains a monotonically increasing counter for these serial numbers.

Input events also carry timestamps with millisecond granularity. Their base is undefined, so they can't be compared against system time (as obtained with clock_gettime or gettimeofday). They can be compared with each other though, and for instance be used to identify sequences of button presses as double or triple clicks.

See the section called “wl_seat - group of input devices” for the protocol description.

Talk about:

  • keyboard map, change events

  • xkb on Wayland

  • multi pointer Wayland

A surface can change the pointer image when the surface is the pointer focus of the input device. Wayland doesn't automatically change the pointer image when a pointer enters a surface, but expects the application to set the cursor it wants in response to the pointer focus and motion events. The rationale is that a client has to manage changing pointer images for UI elements within the surface in response to motion events anyway, so we'll make that the only mechanism for setting or changing the pointer image. If the server receives a request to set the pointer image after the surface loses pointer focus, the request is ignored. To the client this will look like it successfully set the pointer image.

Setting the pointer image to NULL causes the cursor to be hidden.

The compositor will revert the pointer image back to a default image when no surface has the pointer focus for that device.

What if the pointer moves from one window which has set a special pointer image to a surface that doesn't set an image in response to the motion event? The new surface will be stuck with the special pointer image. We can't just revert the pointer image on leaving a surface, since if we immediately enter a surface that sets a different image, the image will flicker. If a client does not set a pointer image when the pointer enters a surface, the pointer stays with the image set by the last surface that changed it, possibly even hidden. Such a client is likely just broken.

Output

An output is a global object, advertised at connect time or as it comes and goes.

See the section called “wl_output - compositor output region” for the protocol description.

  • laid out in a big (compositor) coordinate system

  • basically xrandr over Wayland

  • geometry needs position in compositor coordinate system

  • events to advertise available modes, requests to move and change modes

Data sharing between clients

The Wayland protocol provides clients a mechanism for sharing data that allows the implementation of copy-paste and drag-and-drop. The client providing the data creates a wl_data_source object and the clients obtaining the data will see it as wl_data_offer object. This interface allows the clients to agree on a mutually supported mime type and transfer the data via a file descriptor that is passed through the protocol.

The next section explains the negotiation between data source and data offer objects. the section called “Data devices” explains how these objects are created and passed to different clients using the wl_data_device interface that implements copy-paste and drag-and-drop support.

See the section called “wl_data_offer - offer to transfer data”, the section called “wl_data_source - offer to transfer data”, the section called “wl_data_device - data transfer device” and the section called “wl_data_device_manager - data transfer interface” for protocol descriptions.

MIME is defined in RFC's 2045-2049. A registry of MIME types is maintained by the Internet Assigned Numbers Authority (IANA).

Data negotiation

A client providing data to other clients will create a wl_data_source object and advertise the mime types for the formats it supports for that data through the wl_data_source.offer request. On the receiving end, the data offer object will generate one wl_data_offer.offer event for each supported mime type.

The actual data transfer happens when the receiving client sends a wl_data_offer.receive request. This request takes a mime type and a file descriptor as arguments. This request will generate a wl_data_source.send event on the sending client with the same arguments, and the latter client is expected to write its data to the given file descriptor using the chosen mime type.

Data devices

Data devices glue data sources and offers together. A data device is associated with a wl_seat and is obtained by the clients using the wl_data_device_manager factory object, which is also responsible for creating data sources.

Clients are informed of new data offers through the wl_data_device.data_offer event. After this event is generated the data offer will advertise the available mime types. New data offers are introduced prior to their use for copy-paste or drag-and-drop.

Selection

Each data device has a selection data source. Clients create a data source object using the device manager and may set it as the current selection for a given data device. Whenever the current selection changes, the client with keyboard focus receives a wl_data_device.selection event. This event is also generated on a client immediately before it receives keyboard focus.

The data offer is introduced with wl_data_device.data_offer event before the selection event.

Drag and Drop

A drag-and-drop operation is started using the wl_data_device.start_drag request. This requests causes a pointer grab that will generate enter, motion and leave events on the data device. A data source is supplied as argument to start_drag, and data offers associated with it are supplied to clients surfaces under the pointer in the wl_data_device.enter event. The data offer is introduced to the client prior to the enter event with the wl_data_device.data_offer event.

Clients are expected to provide feedback to the data sending client by calling the wl_data_offer.accept request with a mime type it accepts. If none of the advertised mime types is supported by the receiving client, it should supply NULL to the accept request. The accept request causes the sending client to receive a wl_data_source.target event with the chosen mime type.

When the drag ends, the receiving client receives a wl_data_device.drop event at which it is expected to transfer the data using the wl_data_offer.receive request.