NAME

Xdmx - Distributed Multi-head X server

SYNOPSIS

Xdmx [:display] [option ...]

DESCRIPTION

Xdmx is a proxy X server that uses one or more other X servers as its display devices. It provides multi-head X functionality for displays that might be located on different machines. Xdmx functions as a front-end X server that acts as a proxy to a set of back-end X servers. All of the visible rendering is passed to the back-end X servers. Clients connect to the Xdmx front-end, and everything appears as it would in a regular multi-head configuration. If Xinerama is enabled (e.g., with +xinerama on the command line), the clients see a single large screen.

Xdmx communicates to the back-end X servers using the standard X11 protocol, and standard and/or commonly available X server extensions.

OPTIONS

In addition to the normal X server options described in the Xserver(1) manual page, Xdmx accepts the following command line switches:

-display display-name

This specifies the name(s) of the back-end X server display(s) to connect to. This option may be specified multiple times to connect to more than one back-end display. The first is used as screen 0, the second as screen 1, etc. If this option is omitted, the $DISPLAY environment variable is used as the single back-end X server display.

 

-xinput input-source

This specifies the source to use for XInput extension devices. The choices are the same as for -input , described below, except that core devices on backend servers cannot be treated as XInput extension devices. (Although extension devices on backend and console servers are supported as extension devices under Xdmx).

 

-input input-source

This specifies the source to use for the core input devices. The choices are:

-shadowfb

This option turns on (legacy) support for the shadow frame buffer.

 

-noshadowfb

This option turns off (legacy) support for the shadow frame buffer. Note that this option has been deprecated and will be removed in the next release.

 

-nomulticursor

This option turns off support for displaying multiple cursors on overlapped back-end displays. This option is available for testing and benchmarking purposes.

 

-fontpath

This option sets the Xdmx server's default font path. This option can be specified multiple times to accommodate multiple font paths. See the FONT PATHS section below for very important information regarding setting the default font path.

 

-configfile filename

Specify the configuration file that should be read. Note that if the -display command-line option is used, then the configuration file will be ignored.

 

-config name

Specify a configuration to use. The name will be the name following the virtual keyword in the configuration file.

 

-stat interval screens

This option enables the display of performance statistics. The interval is in seconds. The screens is a count of the number of back-end screens for which data is printed each interval. Specifying 0 for screens will display data for all screens.

 

For each screen, the following information is printed: the screen number, an absolute count of the number of XSync() calls made (SyncCount), the rate of these calls during the previous interval (Sync/s), the average round-trip time (in microseconds) of the last 10 XSync() calls (avSync), the maximum round-trip time (in microseconds) of the last 10 XSync calls (mxSync), the average number of XSync() requests that were pending but not yet processed for each of the last 10 processed XSync() calls, the maximum number of XSync() requests that were pending but not yet processed for each of the last 10 processed XSync() calls, and a histogram showing the distribution of the times of all of the XSync() calls that were made during the previous interval.

 

(The length of the moving average and the number and value of histogram bins are configurable at compile time in the dmxstat.h header file.)

 

-syncbatch interval

This option sets the interval in milliseconds for XSync() batching. An interval less than or equal to 0 will disable XSync() batching. The default interval is 100 ms.

 

-nooffscreenopt

This option disables the offscreen optimization. Since the lazy window creation optimization requires the offscreen optimization to be enabled, this option will also disable the lazy window creation optimization.

 

-nowindowopt

This option disables the lazy window creation optimization.

 

-nosubdivprims

This option disables the primitive subdivision optimization.

 

-noxkb

Disable use of the XKB extension for communication with the back end displays. (Combine with -kb to disable all use of XKB.)

 

-depth int

This option sets the root window's default depth. When choosing a default visual from those available on the back-end X server, the first visual with that matches the depth specified is used.

 

This option can be combined with the -cc option, which specifies the default color visual class, to force the use of a specific depth and color class for the root window.

 

-norender

This option disables the RENDER extension.

 

-noglxproxy

This option disables GLX proxy -- the build-in GLX extension implementation that is DMX aware.

 

-noglxswapgroup

This option disables the swap group and swap barrier extensions in GLX proxy.

 

-glxsyncswap

This option enables synchronization after a swap buffers call by waiting until all X protocol has been processed. When a client issues a glXSwapBuffers request, Xdmx relays that request to each back-end X server, and those requests are buffered along with all other protocol requests. However, in systems that have large network buffers, this buffering can lead to the set of back-end X servers handling the swap buffers request asynchronously. With this option, an XSync() request is issued to each back-end X server after sending the swap buffers request. The XSync() requests will flush all buffered protocol (including the swap buffers requests) and wait until the back-end X servers have processed those requests before continuing. This option does not wait until all GL commands have been processed so there might be previously issued commands that are still being processed in the GL pipe when the XSync() request returns. See the -glxfinishswap option below if Xdmx should wait until the GL commands have been processed.

 

-glxfinishswap

This option enables synchronization after a swap buffers call by waiting until all GL commands have been completed. It is similar to the -glxsyncswap option above; however, instead of issuing an XSync(), it issues a glFinish() request to each back-end X server after sending the swap buffers requests. The glFinish() request will flush all buffered protocol requests, process both X and GL requests, and wait until all previously called GL commands are complete before returning.

 

-ignorebadfontpaths

This option ignores font paths that are not available on all back-end servers by removing the bad font path(s) from the default font path list. If no valid font paths are left after removing the bad paths, an error to that effect is printed in the log.

 

-addremovescreens

This option enables the dynamic addition and removal of screens, which is disabled by default. Note that GLXProxy and Render do not yet support dynamic addition and removal of screens, and must be disabled via the -noglxproxy and -norender command line options described above.

 

-param

This option specifies parameters on the command line. Currently, only parameters dealing with XKEYBOARD configuration are supported. These parameters apply only to the core keyboard. Parameter values are installation-dependent. Please see /usr/X11R6/lib/X11/xkb or a similar directory for complete information.

CONFIGURATION FILE GRAMMAR

The following words and tokens are reserved:

Comments start with a # mark and extend to the end of the line. They may appear anywhere. If a configuration file is read into xdmxconfig, the comments in that file will be preserved, but will not be editable.

The grammar is as follows:

The name following virtual is used as an identifier for the configuration, and may be passed to Xdmx using the -config command line option. The name of a display should be standard X display name, although no checking is performed (e.g., "machine:0").

For names, double quotes are optional unless the name is reserved or contains spaces.

The first dimension following wall is the dimension for tiling (e.g., 2x4 or 4x4). The second dimension following wall is the dimension of each display in the wall (e.g., 1280x1024).

The first geometry following display is the geometry of the screen window on the backend server. The second geometry, which is always preceeded by a slash, is the geometry of the root window. By default, the root window has the same geometry as the screen window.

The option line can be used to specify any command-line options (e.g., -input). (It cannot be used to specify the name of the front-end display.) The option line is processed once at server startup, just line command line options. This behavior may be unexpected.

CONFIGURATION FILE EXAMPLES

Two displays being used for a desktop may be specified in any of the following formats: A 4x4 wall of 16 total displays could be specified as follows (if no tiling dimension is specified, an approximate square is used):

FONT PATHS

The font path used by the Xdmx front-end server will be propagated to each back-end server,which requires that each back-end server have access to the exact same font paths as the front-end server. This can be most easily handled by either using a font server (e.g., xfs) or by remotely mounting the font paths on each back-end server, and then setting the Xdmx server's default font path with the -I "-fontpath" command line option described above.

For example, if you specify a font path with the following command line:

Then, /usr/fonts/75dpi/ and /usr/fonts/Type1/ must be valid font paths on the Xdmx server and all back-end server, which is d0 in this example.

Font servers can also be specified with the -fontpath option. For example, let's assume that a properly configured font server is running on host d0. Then, the following command line

will initialize the front-end Xdmx server and each of the back-end servers to use the font server on d0.

Some fonts might not be supported by either the front-end or the back-end servers. For example, let's assume the front-end Xdmx server includes support Type1 fonts, but one of the back-end servers does not. Let's also assume that the default font path for Xdmx includes Type1 fonts in its font path. Then, when Xdmx initializes the default font path to load the default font, the font path that includes Type1 fonts (along with the other default font paths that are used by the Xdmx server) is sent to the back-end server that cannot handle Type1 fonts. That back-end server then rejects the font path and sends an error back to the Xdmx server. Xdmx then prints an error message and exits because it failed to set the default font path and was unable load the default font.

To fix this error, the offending font path must be removed from the default font path by using a different -fontpath command line option.

The -fontpath option can also be added to the configuration file as described above.

COMMAND-LINE EXAMPLES

The back-end machines are d0 and d1, core input is from the pointer and keyboard attached to d0, clients will refer to :1 when opening windows:

As above, except with core input from d1:

As above, except with core input from a console window on the local display:

As above, except with core input from the local keyboard and mouse:

Note that local input can be used under Linux while another X session is running on :0 (assuming the user can access the Linux console tty and mouse devices): a new (blank) VC will be used for keyboard input on the local machine and the Ctrl-Alt-F* sequence will be available to change to another VC (possibly back to another X session running on the local machine). Using Ctrl-Alt-Backspace on the blank VC will terminate the Xdmx session and return to the original VC.

This example uses the configuration file shown in the previous section:

With this configuration file line: the command line can be shortened to:

USING THE USB DEVICE DRIVERS

The USB device drivers use the devices called /dev/input/event0, /dev/input/event1, etc. under Linux. These devices are driven using the evdev Linux kernel module, which is part of the hid suite. Please note that if you load the mousedev or kbddev Linux kernel modules, then USB devices will appear as core Linux input devices and you will not be able to select between using the device only as an Xdmx core device or an Xdmx XInput extension device. Further, you may be unable to unload the mousedev Linux kernel module if XFree86 is configured to use /dev/input/mice as an input device (this is quite helpful for laptop users and is set up by default under some Linux distributions, but should be changed if USB devices are to be used with Xdmx).

The USB device drivers search through the Linux devices for the first mouse, keyboard, or non-mouse-non-keyboard Linux device and use that device.

KEYBOARD INITIALIZATION

If Xdmx was invoked with -xkb or was not compiled to use the XKEYBOARD extension, then a keyboard on a backend or console will be initialized using the map that the host X server provides.

If the XKEYBOARD extension is used for both Xdmx and the host X server for the keyboard (i.e., the backend or console X server), then the type of the keyboard will be obtained from the host X server and the keyboard under Xdmx will be initialized with that information. Otherwise, the default type of keyboard will be initialized. In both cases, the map from the host X server will not be used. This means that different initial behavior may be noted with and without XKEYBOARD. Consistent and expected results will be obtained by running XKEYBOARD on all servers and by avoiding the use of xmodmap on the backend or console X servers prior to starting Xdmx.

If -xkbmap is specified on the Xdmx command line, then that map will currently be used for all keyboards.

MULTIPLE CORE KEYBOARDS

X was not designed to support multiple core keyboards. However, Xdmx provides some support for multiple core keyboards. Best results will be obtained if all of the keyboards are of the same type and are using the same keyboard map. Because the X server passes raw key code information to the X client, key symbols for keyboards with different key maps would be different if the key code for each keyboard was sent without translation to the client. Therefore, Xdmx will attempt to translate the key code from a core keyboard to the key code for the key with the same key symbol of the first core keyboard that was loaded. If the key symbol appears in both maps, the results will be expected. Otherwise, the second core keyboard will return a NoSymbol key symbol for some keys that would have been translated if it was the first core keyboard.

SEE ALSO

DMX(3X), X(__miscmansuffix__), Xserver(1), xdmxconfig(1), vdltodmx(1), xfs(1), xkbcomp(1)

AUTHORS

Kevin E. Martin <kem@redhat.com>, David H. Dawes <dawes@xfree86.org>, and Rickard E. (Rik) Faith <faith@redhat.com>.

Portions of Xdmx are based on code from The XFree86 Project (http://www.xfree86.org) and X.Org (http://www.x.org).