Phoenix-RTOS implements device drivers as device servers running on the user level. The communication with drivers is done via message passing using a well-defined interface for exchanging data between programs and drivers.
To control the device, two mechanisms are necessary and should be provided by the operating system kernel: access to device hardware registers and method of handling interrupts triggered by it. The architecture of these Phoenix-RTOS mechanisms has been presented in the chapter’s subsections.
Generic driver architecture
Each typical driver in Phoenix-RTOS consists of the main function which typically parses command line parameters, creates a driver communication port, registers it in the operating system, namespace, initializes a device, and starts processing the incoming messages.
The message is fetched from the queue using the
msgRecv syscall. This syscall blocks the thread until a message is
available, so the device server should implement a thread pool to process the incoming request. The client is blocked
until the server responds to the message using the
msgRespond syscall. There are two approaches to message
processing - concurrent processing and iterative processing.
Messages which can be processed instantly may be serviced by one of the main loop threads.
If delays are expected (e.g. waiting for I/O) message should be inserted into the server's internal queue and processed by the internal thread.
The source code of Phoenix-RTOS device drivers could be obtained using the following command
git clone http://git.phoenix-rtos.com/phoenix-rtos-devices
The documentation of a particular driver is provided within its source code repository.