Driver runner

The driver runner is the runner responsible for launching components that run in the driver host environment.

Using the driver runner

To use the driver runner, the component's manifest must include a program block similar to the following:

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
    }
}

A driver component's program block requires the following fields at a minimum:

runner – This field must be set to the string driver.
binary – The path to the driver‘s binary output in the component’s package.
bind – The path to the compiled bind program in the component's package.

Optional fields

In additional to the required fields, the driver runner accepts a set of optional fields, which are used to specify metadata or configure the runtime environment of the driver component.

Colocation

If the colocate field is set to the string true, the driver will be put in the same driver host as its parent driver if possible. However this is advisory. The driver manager may still put the driver in a separate driver host, for instance, if the parent device has MUST_ISOLATE set. In DFv1, a driver is always colocated if the parent device is a composite – isolation may still be enforced by setting MUST_ISOLATE on the primary fragment of the composite.

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}colocate: "true"{{ '</strong>' }}
    }
}

If the colocate field is not specified, its value defaults to the string false.

colocate is mutually exclusive to the host_restart_on_crash field. Only one of them can be true for a driver.

Default dispatcher options

The default_dispatcher_opts field provides the options which are used when creating the driver's default dispatcher, for example:

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}default_dispatcher_opts: [ "allow_sync_calls" ]{{ '</strong>' }}
    }
}

The options in this field correspond to the flags defined in this types.h file. Today, the supported options are:

allow_sync_calls: This option indicates that the dispatcher may not share Zircon threads with other drivers. This setting allows the driver to make synchronous Banjo or FIDL calls on the dispatcher without deadlocking.

Default dispatcher scheduler role

The default_dispatcher_scheduler_role field provides the options that are used when creating the driver's default dispatcher, for example:

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}default_dispatcher_scheduler_role: "fuchsia.graphics.display.driver"{{ '</strong>' }}
    }
}

Make sure that the scheduler roles that you specify match what a component would send through the fuchsia.scheduler/RoleManager.SetRole FIDL API.

Allowed scheduler roles

The allowed_scheduler_roles field dictates what is allowed to be passed in as a scheduler_role when creating new dispatchers, for example:

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}allowed_scheduler_roles: "fuchsia.graphics.display.driver"{{ '</strong>' }}
    }
}

This would allow the driver to create a new dispatcher at runtime and specify the "fuchsia.graphics.display.driver" scheduler_role.

Fallback

If the fallback field is set to the string true, this fallback driver will only attempt to bind once all the base driver packages are indexed. Furthermore, if this driver matches to a node and a non-fallback driver matches to the same node, the non-fallback driver will bind to the node instead.

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}fallback: "true"{{ '</strong>' }}
    }
}

If the fallback field is not specified, its value defaults to the string false.

Next vDSO

If the use_next_vdso field is set to the string true, the driver will be put in a driver host with the next vdso dynamic linked in. The driver must also have colocate set to true or this field is ignored.

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        colocate: "true"
        {{ '<strong>' }}use_next_vdso: "true"{{ '</strong>' }}
    }
}

If the use_next_vdso field is not specified, its value defaults to the string false.

Device categories

The device_categories field provides metadata indicating the device categories that the driver controls, for example:

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}device_categories: [
            { category: "board", subcategory: "i2c" },
            { category: "sensor", subcategory: "temperature" },
        ]{{ '</strong>' }}
    }
}

This metadata is used to determine the tests that the driver will undergo during its certification process. See the full list of device categories and subcategories in the FHCP schema.

Host restart on crash {:#host-restart-on-crash}

The host_restart_on_crash field tells the driver framework that it should restart the driver host for the node that the driver binds to should the driver go down unexpectedly.

This includes if:

The driver host crashes.
The driver closes its client end to the fuchsia.driver.framework/Node protocol while running.

Because this affects the driver host, it can only be set by the root driver of the host. The root driver is the driver for which the host was created. This is the case if and only if the colocate field is set to false.

Therefore host_restart_on_crash and colocate are mutually exclusive. Only one of them can be true for a driver.

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}host_restart_on_crash: "true"{{ '</strong>' }}
    }
}

If the host_restart_on_crash field is not specified, its value defaults to the string false.

When host_restart_on_crash is false, the node is removed from the driver framework's node topology if the driver goes down unexpectedly.

Service Connect Validation {:#service-connect-validation}

The service_connect_validation field is used by the driver sdk's DriverBase to enable availability valdations to run on service capability connections.

It does this by looking through the offers available to the bound node of the drive, and ensuring all incoming()->Connect() requests are trying to connect to a valid offer.

In single parent cases, this just ensures the service is available to the node, as all requests should be going to a "default" instance, when no instance is specified by the user.

In composite cases, this ensures that the instance name requested, and the "default" instance name case, have a corresponding offer from that parent.

If these validations fail, the Connect() method will return ZX_ERR_NOT_FOUND immediately, instead of making a connection that will fail only when a two-way method is called on it.

{
    program: {
        runner: "driver",
        binary: "driver/example.so",
        bind: "meta/bind/example.bindbc",
        {{ '<strong>' }}service_connect_validation: "true"{{ '</strong>' }}
    }
}

If this field is not set, the validations are disabled by default.