HP OpenVMS Systems Documentation

Content starts here

VMS DECwindows Device Driver Manual


Previous Contents Index

1.4 Driver Architecture

DECwindows workstation device driver software divides into multiple driver modules, as follows:

  • Class input driver
  • Port input driver
  • Output driver
  • Common driver

The driver modularity provides flexibility and ease of coding during driver development and ease of driver upgrades for the varied workstation types and devices. The various driver modules communicate by means of vector tables and shared data in the unit control block (UCB).

The architecture of the software modules or basic subsystems for workstation families with busless CPUs is shown in Figure 1-2. Input drivers process data from the keyboard or mouse and pass it to the server. The input drivers also process output data, such as keyboard LED information, from the server to the keyboard. An output driver processes graphics and windowing requests in the form of output data that passes from the server to the monitor.

Workstation families with Q22 bus-based CPUs and VCB01/VCB02 video controllers use the GxADRIVER modules. For these, the driver architecture differs slightly in that the port input driver software is part of the output driver module, as shown in Figure 1-3. However, the class/port input characteristic of a DECwindows device interface remains the same.

Figure 1-2 DECwindows Driver Architecture for Busless CPUs


Figure 1-3 DECwindows Driver Architecture for Q22-Bus CPUs


1.4.1 Driver/Server Interface

As shown in Figure 1-2 and Figure 1-3, the DECwindows driver presents two interfaces to the server: a queue interface and a $QIO interface.

The queue interface is an input event queue in memory that is shared between the server and drivers. This queue is interlocked for correct operation on multiprocessor workstations. The input queue receives all input events as they are generated by the drivers. A driver time-stamps each input event as it inserts the event in the queue. Asynchronous input events supported by the drivers include the following:

  • Key presses and releases
  • Mouse movement
  • Mouse button presses and releases

The use of a single queue ensures that all input is correctly time-ordered when the server reads it.

The $QIO interface is the second driver/server interface, which is also common to VMS device drivers. Generally only the DECwindows server should make $QIO calls or access the queue. Applications should use the DECwindows library (Xlib) routines to perform these functions.

1.4.2 Common Driver Function

The common driver (INDRIVER) is the interface between an input or an output driver and the DECwindows server. It monitors the input queue and supports the server protocol of the interface. The common driver handles device-independent processing and functions that are common to all workstations. Common driver $QIO service routines support $QIO calls from the server. Getting device information or parsing $QIO parameters for all the device drivers are examples of the common driver function.

1.4.3 Class/Port Input Driver Function

The input drivers handle input data transfers from the keyboard, mouse, tablet, and other input devices to the common driver and on to the server. They are based on a modular class/port driver interface designed for VMS that allows for new input devices or serial line hardware to be easily added. Class/port input drivers are bound together by means of the input device's UCB data structure and form a class/port interface. Note that parts of the VMS terminal class/port interface model are not used by DECwindows. The VMS terminal class/port interface is described in the OpenVMS VAX Device Support Manual.

The port drivers provided by VMS receive interrupts from and transmit data to the hardware ports. Port drivers service input serial lines only. These serial lines may be part of the graphics hardware (VCB02 controller) or a standard serial line controller (VAXstation 2000 DZ controller). Data received by the port driver is passed to the appropriate class driver for interpretation.

The class drivers provided by VMS include the keyboard driver (IKDRIVER) and the mouse driver (IMDRIVER). A class driver interprets a byte stream from an input device and then formats the data into an event packet for the input queue. Because each input device, such as a keyboard or tablet, must interpret a different byte stream protocol, there is one class driver per input device type on a system.

1.4.4 Output Driver Function

The output driver processes graphics and windowing requests from the server to the screen. Output drivers manage the output functions of the video controller. For example, the output driver performs all device-dependent processing, such as receiving device interrupts, manipulating the color map, drawing, and managing the current state of the graphics hardware. Note that the output modules for Q22-bus CPUs (GxADRIVERs) also contain port input driver software, as shown in Figure 1-3.

Some DECwindows video devices use an output queue. For instance, drivers for color devices support an output queue, while monochrome drivers do not. This queue is the interface for drawing operations from the server (or applications). Like the input queue, the output queue is in nonpaged pool shared by the driver and the server. Drawing packets are inserted into the queue by the server. The driver removes the packets from the queue and executes them in the queued order.


Chapter 2
Common Driver/Server Interface

The common driver/server link defines two interface types; the common input queue interface and the common $QIO interface. The main DECwindows device driver interface to the server is a buffer containing a queue of event packets formatted for the X11 standard protocol. This chapter describes the buffer/input queue and the protocol of the driver/server interface. Also described is the $QIO interface. The service mechanism that supports the $QIO calls is described in Chapter 5. Data structures referenced in this chapter are described in detail in Appendix A.

2.1 Driver/Server Common Buffer

The common driver manages an input buffer that the driver shares with the server. Figure 2-1 illustrates the input buffer structure with its queues. The shared input buffer is a block allocated in nonpaged pool. It contains a control block or header and two queues: an input event queue and a free queue. The queues are self-relative interlocked queues that provide an efficient communication path for frequent driver/server operations. Using the input buffer control block (INB), the common driver monitors each queue containing input event packets (INPs). Each packet stored in the queue contains a forward and a backward pointer (FLINK and BLINK) to the next and previous packet in the self-relative queue.

Figure 2-1 Input Buffer General Structure


2.1.1 Input Queue and Motion History Buffer

The server may create a pointer motion history buffer (MHB) to improve system response to pointer movement. The input queue always maintains the most recent motion events along with other input device events for the server. However, if the motion compression feature is enabled, the server may not receive all of the motion events generated in the input queue. If the server requires motion events that were not delivered because of motion compression, the server can access the motion history buffer for the older motion events.

Like the input buffer, the MHB is allocated in nonpaged pool. The server issues an Initialize Motion History $QIO call (described in this chapter) specifying the desired size in pages. The motion history buffer (shown in Figure 2-2) contains a control block or header in the first 16 bytes, followed by a ring of 8-byte motion history packets (MHPs) throughout the remaining allocated space. Like the input buffer, the motion history buffer contains active event packets and free packets. When a server/driver searches the queue, the oldest motion history event packet and free packet are located in the ring with put and get pointers in the MHB header. Each motion history packet contains the x and y movement with an event timestamp. Refer to Appendix A for detailed field information.

Once an MHB is created, a pointer motion event is first stored in a motion history packet in the MHB and then copied into an input packet in the input buffer shared with the server. However, the server can disable the MHB by setting the INB$V_MHB_BUSY bit, which forces the buffering of all events by way of the input buffer only.

When motion event compression is enabled (by the Set Motion Compression $QIO), the motion event decoder removes the oldest motion event packet from the input queue as it inserts the newest event. Thus, the removed events (the oldest of a large burst of pointer motion, those not yet retrieved by the server, or both) are lost, yielding motion compression. The number of lost motion events or motion compression hits is stored in counter DWI$L_PTR_MOTION_COMP_HIT. If necessary, the server program can recover lost events by accessing the motion history buffer instead of the input queue.

Figure 2-2 Motion History Buffer General Structure


2.1.2 Input Queue Event Packet

The input packet structure (INP) defines the packet format used in the input queue interface between the device driver and the DECwindows server. The basic DECwindows format of the input packet, shown in Figure 2-3, is compatible with the X event in the X Window System protocol.

Depending on the driver, some fields in the input packet of certain events may vary. The packet illustrated in Figure 2-3 is a typical keyboard- or mouse-generated input event for key/button transitions and mouse motion. The first 12 bytes (3 longwords) are common to all event types. The event information is always 32 bytes long, excluding the forward/backward pointers (FLINK/BLINK). The FLINK and BLINK pointers link (in proper order) all the event packets of the input queue. Refer to Appendix A for detailed field information.

Figure 2-3 Queue Event Packet Format


2.1.3 Queue Processing of Input

There is one input queue and one free queue for each keyboard/mouse pair for input. As each event occurs in the device, the class driver gets a free packet from the free queue and inserts it into the input queue. The class driver links all active keyboard and pointer event packets in the input queue using the forward link (FLINK) and backward link (BLINK) INP fields (see Figure 2-4).

Because there is a single input queue shared by the input devices, packet-link pointers ensure that all events are correctly time-ordered when they are read by the server. The size of the input queue varies inversely with the size of the free queue; as more packets move to the input queue, the number of free packets diminishes.

The common driver checks the queue at timed intervals to see if there is input. During the hardware vertical retrace interval (VSYNC), the driver checks the INPUT_QUEUE_FLINK pointer in the input buffer control block (INB). If the queue is not empty or the queue is not being accessed by the server, the driver wakes the server to signal the presence of input.

When the server responds, the server processes the event data, removes the event packet from the input queue, and inserts the packet on the free queue. The server processes each event packet in the queue until the queue is empty.

Figure 2-4 Input Queue and Free Queue


2.2 $QIO Common Interface

The INDRIVER module contains function decision table (FDT) routines that make up a $QIO common interface. The $QIO common interface provides for initialization and information requests from the server or server extension to a device. The $QIO interface is used for infrequent operations that do not require synchronization with input or output requests or when notification upon completion of a request is needed.

2.3 $QIO Calls to DECwindows Drivers

This section presents the output $QIO calls used in a server that are supported by services within the DECwindows common driver. The $QIO system service format is presented first.

$QIO calls must be issued to a physical device, as they cannot be directed to a pseudodevice (such as IKA0 for the keyboard decoder). Initially, using the $ASSIGN system service, the appropriate device name is assigned to an I/O channel. The channel number entry is required for the chan parameter in the $QIO system service call. Physical device names on a GPX workstation are GAA0 for output to the screen, GAA1 for the keyboard, and GAA2 for the mouse (see Table 1-1). Note that the DECwindows environment provides logical names (DECW$SERVER_SCREENS, DECW$KEYBOARD, and DECW$POINTER) located in DECW$SERVERn_TABLES to point to the physical device.


$QIO System Service

The Queue I/O Request system service queues an I/O request to a channel associated with a device. The SYS$QIO format described next applies to all the $QIO calls presented in this chapter. For more information on SYS$QIO refer to OpenVMS System Services Reference Manual.

FORMAT

SYS$QIO [efn],chan,func,[iosb],[astadr],[astprm]

,p1,p2,p3[,p4][,p5][,p6]

arguments efn is the event flag number of the I/O operation. The efn argument is a longword containing the number of the event flag.

chan is the I/O channel assigned ($ASSIGN) to the device name to which the request is directed. The chan argument is a longword containing the number of the I/O channel; however, $QIO uses only the low-order word.

func is the device-specific function code specifying the operation to be performed. The func argument is a longword containing the function code.

iosb is the I/O status block to receive the final completion status of the I/O operation. The iosb argument is the address of the quadword I/O status block.

astadr is the asynchronous system trap (AST) service routine to be executed when the I/O completes. The astadr argument is the address of a longword that is the entry mask to the AST routine.

astprm is the AST parameter to be passed to the AST service routine. The astprm argument is a longword containing the AST parameter.

p1 is the function modifier specifying the service being called within the basic function code (IO$K_DECW_xxx).

p2 to p6 are the function-specific parameters being passed.

2.4 Sense-Mode Calls

The FDT sense-mode routines within the common driver service the $QIO sense mode function calls from a server. The following sense-mode calls are supported by the DECwindows common driver:
  • Get Device Information
  • Sense Cursor Bounds
  • Sense Keyboard Information
  • Sense Keyboard LED
  • Sense Motion Compression
  • Sense Operator Window Key
  • Sense Pointer Acceleration
  • Sense Pseudomouse Key
  • Sense Screen Saver Timeout

This section defines the specific argument data required for each $QIO call within the sense-mode functions serviced by the common driver. Each of these calls requires the IO$_SENSEMODE function code.


Get Device Information

The Get Device Information $QIO sense-mode function returns the address and size of the device information block (DVI). The target device is the output display. The function returns a pointer to the (read-only) DVI block. The DVI contains static device information such as the size of the memory frame buffer, the resolution of the screen, the number of bits per pixel in the frame buffer, the number of cursor planes, and the width and height of the cursor bitmap. See the DVI block in Figure A-1 and Table A-1 for more detailed field information. Table 2-1 provides the required argument information for the Get Device Information $QIO call.

Table 2-1 Argument Data for Get Device Information$QIO Call
$QIO Argument Required Data
func IO$_SENSEMODE function code
p1 IO$K_DECW_DEVICE_INFO function modifier
p2 Address that stores the DVI block address
p3 Address that stores the DVI block length
p4, p5, p6 Set to 0

Sense Cursor Bounds

The Sense Cursor Bounds $QIO sense-mode function returns the cursor (pointer) x and y screen boundaries. In a multiscreen environment, the call is made from the server/screen that the cursor is confined to.

If the confine-cursor flag (FLAG$V_CURSOR_BOUNDED) is set to 1, the cursor boundaries are confined to the values pointed to by longword parameters p3 to p6. If the confine-cursor flag is set to 0, the cursor is free to travel anywhere, including other screens in multiscreen environments.

Much of the cursor information is stored in the UCB common output extension (DECW).

Table 2-2 provides the required argument information for the Sense Cursor Bounds $QIO call.

Table 2-2 Argument Data for Sense Cursor Bounds$QIO Call
$QIO Argument Required Data
func IO$_SENSEMODE function code
p1 IO$K_DECW_CURSOR_BOUNDS function modifier
p2 Address of the confine cursor flag bit (FLAG$V_CURSOR_BOUNDED of UCB$L_DECW_CTRL)
p3 Address of the x-axis left boundary (UCB$W_DECW_X1_BOUNDRY)
p4 Address of the y-axis top boundary (UCB$W_DECW_Y1_BOUNDRY)
p5 Address of the x-axis right boundary (UCB$W_DECW_X2_BOUNDRY)
p6 Address of the y-axis bottom boundary (UCB$W_DECW_Y2_BOUNDRY)

Sense Keyboard Information

The Sense Keyboard Information $QIO function returns the current functional characteristics or information concerning the keyboard device. Table 2-3 provides the argument information required for the Sense Keyboard Information $QIO call.

Figure A-4 and Table A-4 show and define the data structure that passes the keyboard information requested for the $QIO call.

Table 2-3 Argument Data for Sense Keyboard Information$QIO Call
$QIO Argument Required Data
func IO$_SENSEMODE function code
p1 IO$K_DECW_KB_INFO function modifier
p2 Address of the keyboard information (characteristics) block (KIB)
p3 Address of the longword that stores the length of the keyboard information block
p4, p5, p6 Set to 0


Previous Next Contents Index