The Common UNIX Print System (CUPS) has become quite popular. All major Linux distributions now ship it as their default printing system. To many, it is still a mystical tool. Mostly, it just works. People tend to regard it as a “black box” that they do not want to look into as long as it works. But once there is a little problem, they have trouble finding out where to start debugging it. Refer to Classical Printing, which contains much information that is also relevant to CUPS.

CUPS sports quite a few unique and powerful features. While its basic functions may be grasped quite easily, they are also new. Because it is different from other, more traditional printing systems, it is best not to try to apply any prior knowledge about printing to this new system. Rather, try to understand CUPS from the beginning. This documentation will lead you to a complete understanding of CUPS. Let's start with the most basic things first.

CUPS is more than just a print spooling system. It is a complete printer management system that complies with the new Internet Printing Protocol (IPP). IPP is an industry and Internet Engineering Task Force (IETF) standard for network printing. Many of its functions can be managed remotely (or locally) via a Web browser (giving you platform-independent access to the CUPS print server). Additionally, it has the traditional command line and several more modern GUI interfaces (GUI interfaces developed by third parties, like KDE's overwhelming KDEPrint).

CUPS allows creation of raw printers (i.e., no print file format translation) as well as smart printers (i.e., CUPS does file format conversion as required for the printer). In many ways, this gives CUPS capabilities similar to the MS Windows print monitoring system. Of course, if you are a CUPS advocate, you would argue that CUPS is better! In any case, let us now explore how to configure CUPS for interfacing with MS Windows print clients via Samba.

Samba has a special relationship to CUPS. Samba can be compiled with CUPS library support. Most recent installations have this support enabled. By default, CUPS linking is compiled into smbd and other Samba binaries. Of course, you can use CUPS even if Samba is not linked against libcups.so but there are some differences in required or supported configuration.

When Samba is compiled and linked with libcups, printcap = cups uses the CUPS API to list printers, submit jobs, query queues, and so on. Otherwise it maps to the System V commands with an additional -oraw option for printing. On a Linux system, you can use the ldd utility to find out if smbd has been linked with the libcups library (ldd may not be present on other OS platforms, or its function may be embodied by a different command):

root# ldd `which smbd`
libssl.so.0.9.6 => /usr/lib/libssl.so.0.9.6 (0x4002d000)
libcrypto.so.0.9.6 => /usr/lib/libcrypto.so.0.9.6 (0x4005a000)
libcups.so.2 => /usr/lib/libcups.so.2 (0x40123000)
[....]

The line libcups.so.2 => /usr/lib/libcups.so.2 (0x40123000) shows there is CUPS support compiled into this version of Samba. If this is the case, and printing = cups is set, then any otherwise manually set print command in smb.conf is ignored. This is an important point to remember!

To summarize, the Simplest Printing-Related smb.conf file shows the simplest printing-related setup for smb.conf to enable basic CUPS support:

This is all you need for basic printing setup for CUPS. It will print all graphic, text, PDF, and PostScript files submitted from Windows clients. However, most of your Windows users would not know how to send these kinds of files to print without opening a GUI application. Windows clients tend to have local printer drivers installed, and the GUI application's print buttons start a printer driver. Your users also rarely send files from the command line. Unlike UNIX clients, they rarely submit graphic, text, or PDF formatted files directly to the spooler. They nearly exclusively print from GUI applications with a “printer driver” hooked between the application's native format and the print data stream. If the backend printer is not a PostScript device, the print data stream is “binary,” sensible only for the target printer. Read on to learn what problem this may cause and how to avoid it.

The Overriding Global CUPS Settings for One Printer example is a slightly more complex printing-related setup for smb.conf. It enables general CUPS printing support for all printers, but defines one printer share, which is set up differently.

This special share is only for testing purposes. It does not write the print job to a file. It just logs the job parameters known to Samba into the /tmp/smbprn.log file and deletes the job-file. Moreover, the printer admin of this share is “kurt” (not the “@ntadmins” group), guest access is not allowed, the share isn't published to the Network Neighborhood (so you need to know it is there), and it allows access from only three hosts. To prevent CUPS from kicking in and taking over the print jobs for that share, we need to set printing = sysv and printcap = lpstat.

Many small office or home networks, as well as badly organized larger environments, allow each client a direct access to available network printers. This is generally a bad idea. It often blocks one client's access to the printer when another client's job is printing. It might freeze the first client's application while it is waiting to get rid of the job. Also, there are frequent complaints about various jobs being printed with their pages mixed with each other. A better concept is the use of a print server: it routes all jobs through one central system, which responds immediately, takes jobs from multiple concurrent clients, and transfers them to the printer(s) in the correct order.

Most traditionally configured UNIX print servers acting on behalf of Samba's Windows clients represented a really simple setup. Their only task was to manage the “raw” spooling of all jobs handed to them by Samba. This approach meant that the Windows clients were expected to prepare the print job file that is ready to be sent to the printing device. In this case, a native (vendor-supplied) Windows printer driver needs to be installed on each and every client for the target device.

It is possible to configure CUPS, Samba, and your Windows clients in the same traditional and simple way. When CUPS printers are configured for raw print-through mode operation, it is the responsibility of the Samba client to fully render the print job (file). The file must be sent in a format that is suitable for direct delivery to the printer. Clients need to run the vendor-provided drivers to do this. In this case, CUPS will not do any print file format conversion work.

The easiest printing configuration possible is raw print-through. This is achieved by installation of the printer as if it were physically attached to the Windows client. You then redirect output to a raw network print queue. This procedure may be followed to achieve this:

The printer drivers on the Windows clients may be installed in two functionally different ways:

The second method is recommended for use over the first.

If you use the first option (drivers are installed on the client side), there is one setting to take care of: CUPS needs to be told that it should allow “raw” printing of deliberate (binary) file formats. The CUPS files that need to be correctly set for raw mode printers to work are:

Both contain entries (at the end of the respective files) that must be uncommented to allow RAW mode operation. In /etc/cups/mime.types, make sure this line is present:

application/octet-stream

In /etc/cups/mime.convs, have this line:

application/octet-stream   application/vnd.cups-raw   0   - 

If these two files are not set up correctly for raw Windows client printing, you may encounter the dreaded Unable to convert file 0 in your CUPS error_log file.

Background.  That CUPS is a more security-aware printing system than traditional ones does not by default allow a user to send deliberate (possibly binary) data to printing devices. This could be easily abused to launch a “Denial of Service” attack on your printer(s), causing at least the loss of a lot of paper and ink. “Unknown” data are tagged by CUPS as MIME type: application/octet-stream and not allowed to go to the printer. By default, you can only send other (known) MIME types “raw.” Sending data “raw” means that CUPS does not try to convert them and passes them to the printer untouched.

This is all you need to know to get the CUPS/Samba combo printing “raw” files prepared by Windows clients, which have vendor drivers locally installed. If you are not interested in background information about more advanced CUPS/Samba printing, simply skip the remaining sections of this chapter.

This section describes three familiar methods, plus one new one, by which printer drivers may be uploaded.

If you want to use the MS-RPC-type printing, you must upload the drivers onto the Samba server first ([print$] share). For a discussion on how to deposit printer drivers on the Samba host (so the Windows clients can download and use them via “Point'n'Print”), please refer to the Classical Printing chapter of this book. There you will find a description or reference to three methods of preparing the client drivers on the Samba server:

These three methods apply to CUPS all the same. The cupsaddsmb utility is a new and more convenient way to load the Windows drivers into Samba and is provided if you use CUPS.

cupsaddsmb is discussed in much detail later in this chapter. But we first explore the CUPS filtering system and compare the Windows and UNIX printing architectures.

Network printing is one of the most complicated and error-prone day-to-day tasks any user or administrator may encounter. This is true for all OS platforms, and there are reasons it is so.

You can't expect to throw just any file format at a printer and have it get printed. A file format conversion must take place. The problem is that there is no common standard for print file formats across all manufacturers and printer types. While PostScript (trademark held by Adobe) and, to an extent, PCL (trademark held by Hewlett-Packard) have developed into semi-official “standards” by being the most widely used page description languages (PDLs), there are still many manufacturers who “roll their own” (their reasons may be unacceptable license fees for using printer-embedded PostScript interpreters, and so on).

In Windows OS, the format conversion job is done by the printer drivers. On MS Windows OS platforms all application programmers have at their disposal a built-in API, the graphical device interface (GDI), as part and parcel of the OS itself to base themselves on. This GDI core is used as one common unified ground for all Windows programs to draw pictures, fonts, and documents on screen as well as on paper (print). Therefore, printer driver developers can standardize on a well-defined GDI output for their own driver input. Achieving WYSIWYG (What You See Is What You Get) is relatively easy, because the on-screen graphic primitives, as well as the on-paper drawn objects, come from one common source. This source, the GDI, often produces a file format called Enhanced MetaFile (EMF). The EMF is processed by the printer driver and converted to the printer-specific file format.

The example in Windows Printing to a Local Printer illustrates local Windows printing.

Figure 21.1. Windows Printing to a Local Printer.

In UNIX and Linux, there is no comparable layer built into the OS kernel(s) or the X (screen display) server. Every application is responsible for itself to create its print output. Fortunately, most use PostScript and that at least gives some common ground. Unfortunately, there are many different levels of quality for this PostScript. And worse, there is a huge difference (and no common root) in the way the same document is displayed on screen and how it is presented on paper. WYSIWYG is more difficult to achieve. This goes back to the time, decades ago, when the predecessors of X.org, designing the UNIX foundations and protocols for graphical user interfaces, refused to take responsibility for “paper output”, as some had demanded at the time, and restricted itself to “on-screen only.” (For some years now, the “Xprint” project has been under development, attempting to build printing support into the X framework, including a PostScript and a PCL driver, but it is not yet ready for prime time.) You can see this unfavorable inheritance up to the present day by looking into the various “font” directories on your system; there are separate ones for fonts used for X display and fonts to be used on paper.

Background.  The PostScript programming language is an “invention” by Adobe, but its specifications have been published extensively. Its strength lies in its powerful abilities to describe graphical objects (fonts, shapes, patterns, lines, curves, and dots), their attributes (color, linewidth), and the way to manipulate (scale, distort, rotate, shift) them. Because of its open specification, anybody with the skill can start writing his or her own implementation of a PostScript interpreter and use it to display PostScript files on screen or on paper. Most graphical output devices are based on the concept of “raster images” or “pixels” (one notable exception is pen plotters). Of course, you can look at a PostScript file in its textual form and you will be reading its PostScript code, the language instructions that need to be interpreted by a rasterizer. Rasterizers produce pixel images, which may be displayed on screen by a viewer program or on paper by a printer.

So UNIX is lacking a common ground for printing on paper and displaying on screen. Despite this unfavorable legacy for UNIX, basic printing is fairly easy if you have PostScript printers at your disposal. The reason is that these devices have a built-in PostScript language “interpreter,” also called a raster image processor (RIP), (which makes them more expensive than other types of printers; throw PostScript toward them, and they will spit out your printed pages. The RIP does all the hard work of converting the PostScript drawing commands into a bitmap picture as you see it on paper, in a resolution as done by your printer. This is no different than PostScript printing a file from a Windows origin.

Figure 21.2. Printing to a PostScript Printer.

However, there are other types of printers out there. These do not know how to print PostScript. They use their own PDL, often proprietary. To print to them is much more demanding. Since your UNIX applications mostly produce PostScript, and since these devices do not understand PostScript, you need to convert the print files to a format suitable for your printer on the host before you can send it away.

Here is where Ghostscript kicks in. Ghostscript is the traditional (and quite powerful) PostScript interpreter used on UNIX platforms. It is a RIP in software, capable of doing a lot of file format conversions for a very broad spectrum of hardware devices as well as software file formats. Ghostscript technology and drivers are what enable PostScript printing to non-PostScript hardware. This is shown in Ghostscript as a RIP for Non-PostScript Printers.

Figure 21.3. Ghostscript as a RIP for Non-PostScript Printers.

While PostScript in essence is a PDL to represent the page layout in a device-independent way, real-world print jobs are always ending up being output on hardware with device-specific features. To take care of all the differences in hardware and to allow for innovations, Adobe has specified a syntax and file format for PostScript Printer Description (PPD) files. Every PostScript printer ships with one of these files.

PPDs contain all the information about general and special features of the given printer model: Which different resolutions can it handle? Does it have a duplexing unit? How many paper trays are there? What media types and sizes does it take? For each item, it also names the special command string to be sent to the printer (mostly inside the PostScript file) in order to enable it.

Information from these PPDs is meant to be taken into account by the printer drivers. Therefore, installed as part of the Windows PostScript driver for a given printer is the printer's PPD. Where it makes sense, the PPD features are presented in the drivers' UI dialogs to display to the user a choice of print options. In the end, the user selections are somehow written (in the form of special PostScript, PJL, JCL, or vendor-dependent commands) into the PostScript file created by the driver.

CUPS can handle all spec-compliant PPDs as supplied by the manufacturers for their PostScript models. Even if a vendor does not mention our favorite OS in his or her manuals and brochures, you can safely trust this: If you get the Windows NT version of the PPD, you can use it unchanged in CUPS and thus access the full power of your printer just like a Windows NT user could!

CUPS also uses specially crafted PPDs to handle non-PostScript printers. These PPDs are usually not available from the vendors (and no, you can't just take the PPD of a PostScript printer with the same model name and hope it works for the non-PostScript version too). To understand how these PPDs work for non-PS printers, we first need to dive deeply into the CUPS filtering and file format conversion architecture. Stay tuned.

CUPS reads the file /etc/cups/mime.types (and all other files carrying a *.types suffix in the same directory) upon startup. These files contain the MIME type recognition rules that are applied when CUPS runs its autotyping routines. The rule syntax is explained in the man page for mime.types and in the comments section of the mime.types file itself. A simple rule reads like this:

application/pdf         pdf string(0,%PDF)

This means if a filename has a .pdf suffix or if the magic string %PDF is right at the beginning of the file itself (offset 0 from the start), then it is a PDF file (application/pdf). Another rule is this:

application/postscript  ai eps ps string(0,%!) string(0,<04>%!)

If the filename has one of the suffixes .ai, .eps, .ps, or if the file itself starts with one of the strings %! or <04>%!, it is a generic PostScript file (application/postscript).

CUPS can handle ASCII text, HP-GL, PDF, PostScript, DVI, and many image formats (GIF, PNG, TIFF, JPEG, Photo-CD, SUN-Raster, PNM, PBM, SGI-RGB, and more) and their associated MIME types with its filters.

CUPS reads the file /etc/cups/mime.convs (and all other files named with a *.convs suffix in the same directory) upon startup. These files contain lines naming an input MIME type, an output MIME type, a format conversion filter that can produce the output from the input type, and virtual costs associated with this conversion. One example line reads like this:

application/pdf         application/postscript   33   pdftops

This means that the pdftops filter will take application/pdf as input and produce application/postscript as output; the virtual cost of this operation is 33 CUPS-$. The next filter is more expensive, costing 66 CUPS-$:

application/vnd.hp-HPGL application/postscript   66   hpgltops

This is the hpgltops, which processes HP-GL plotter files to PostScript.

application/octet-stream

Here are two more examples:

application/x-shell     application/postscript   33    texttops
text/plain              application/postscript   33    texttops

The last two examples name the texttops filter to work on text/plain as well as on application/x-shell. (Hint: This differentiation is needed for the syntax highlighting feature of texttops).

There are many more combinations named in mime.convs. However, you are not limited to use the ones predefined there. You can plug in any filter you like to the CUPS framework. It must meet, or must be made to meet, some minimal requirements. If you find (or write) a cool conversion filter of some kind, make sure it complies with what CUPS needs and put in the right lines in mime.types and mime.convs; then it will work seamlessly inside CUPS.

As previously stated, PostScript is the central file format to any UNIX-based printing system. From PostScript, CUPS generates raster data to feed non-PostScript printers.

But what happens if you send one of the supported non-PS formats to print? Then CUPS runs “prefilters” on these input formats to generate PostScript first. There are prefilters to create PostScript from ASCII text, PDF, DVI, or HP-GL. The outcome of these filters is always of MIME type application/postscript (meaning that any device-specific print options are not yet embedded into the PostScript by CUPS and that the next filter to be called is pstops). Another prefilter is running on all supported image formats, the imagetops filter. Its outcome is always of MIME type application/vnd.cups-postscript (not application/postscript), meaning it has the print options already embedded into the file. This is shown in Prefiltering in CUPS to Form PostScript.

Figure 21.4. Prefiltering in CUPS to Form PostScript.

pstops is a filter that is used to convert application/postscript to application/vnd.cups-postscript. As stated earlier, this filter inserts all device-specific print options (commands to the printer to ask for the duplexing of output, or stapling and punching it, and so on) into the PostScript file. An example is illustrated in Adding Device-Specific Print Options.

Figure 21.5. Adding Device-Specific Print Options.

This is not all. Other tasks performed by it are:

pstoraster is at the core of the CUPS filtering system. It is responsible for the first stage of the rasterization process. Its input is of MIME type application/vnd.cups-postscript; its output is application/vnd.cups-raster. This output format is not yet meant to be printable. Its aim is to serve as a general-purpose input format for more specialized raster drivers that are able to generate device-specific printer data. This is shown in the PostScript to Intermediate Raster Format diagram.

Figure 21.6. PostScript to Intermediate Raster Format.

CUPS raster is a generic raster format with powerful features. It is able to include per-page information, color profiles, and more, to be used by the downstream raster drivers. Its MIME type is registered with IANA and its specification is, of course, completely open. It is designed to make it quite easy and inexpensive for manufacturers to develop Linux and UNIX raster drivers for their printer models should they choose to do so. CUPS always takes care of the first stage of rasterization so these vendors do not need to care about Ghostscript complications (in fact, there are currently more than one vendor financing the development of CUPS raster drivers). This is illustrated in the CUPS-Raster Production Using Ghostscript illustration.

Figure 21.7. CUPS-Raster Production Using Ghostscript.

CUPS versions before version 1.1.15 shipped a binary (or source code) standalone filter, named pstoraster. pstoraster, which was derived from GNU Ghostscript 5.50 and could be installed instead of and in addition to any GNU or AFPL Ghostscript package without conflicting.

Since version 1.1.15, this feature has changed. The functions for this filter have been integrated back into Ghostscript (now based on GNU Ghostscript version 7.05). The pstoraster filter is now a simple shell script calling gs with the -sDEVICE=cups parameter. If your Ghostscript fails when this command is executed: gs -h |grep cups, you might not be able to print, update your Ghostscript.

In the section about prefilters, we mentioned the prefilter that generates PostScript from image formats. The imagetoraster filter is used to convert directly from image to raster, without the intermediate PostScript stage. It is used more often than the previously mentioned prefilters. We summarize in a flowchart the image file filtering in the Image Format to CUPS-Raster Format Conversion illustration.

Figure 21.8. Image Format to CUPS-Raster Format Conversion.

CUPS ships with quite a variety of raster drivers for processing CUPS raster. On my system, I find in /usr/lib/cups/filter/ the following: rastertoalps, rastertobj, rastertoepson, rastertoescp, rastertopcl, rastertoturboprint, rastertoapdk, rastertodymo, rastertoescp, rastertohp, and rastertoprinter. Don't worry if you have fewer drivers than this; some of these are installed by commercial add-ons to CUPS (like rastertoturboprint), and others (like rastertoprinter) by third-party driver development projects (such as Gimp-Print) wanting to cooperate as closely as possible with CUPS. See the Raster to Printer-Specific Formats illustration.

Figure 21.9. Raster to Printer-Specific Formats.

The last part of any CUPS filtering chain is a backend. Backends are special programs that send the print-ready file to the final device. There is a separate backend program for any transfer protocol for sending print jobs over the network, and one for every local interface. Every CUPS print queue needs to have a CUPS “device-URI” associated with it. The device URI is the way to encode the backend used to send the job to its destination. Network device-URIs use two slashes in their syntax, local device URIs only one, as you can see from the following list. Keep in mind that local interface names may vary greatly from my examples, if your OS is not Linux:

It is easy to write your own backends as shell or Perl scripts if you need any modification or extension to the CUPS print system. One reason could be that you want to create “special” printers that send the print jobs as email (through a “mailto:/” backend), convert them to PDF (through a “pdfgen:/” backend) or dump them to “/dev/null”. (In fact, I have the systemwide default printer set up to be connected to a devnull:/ backend: there are just too many people sending jobs without specifying a printer, and scripts and programs that do not name a printer. The systemwide default deletes the job and sends a polite email back to the $USER asking him or her to always specify the correct printer name.)

Not all of the mentioned backends may be present on your system or usable (depending on your hardware configuration). One test for all available CUPS backends is provided by the lpinfo utility. Used with the -v parameter, it lists all available backends:

	$ lpinfo -v
	

cupsomatic filters may be the most widely used on CUPS installations. You must be clear that these were not developed by the CUPS people. They are a third-party add-on to CUPS. They utilize the traditional Ghostscript devices to render jobs for CUPS. When troubleshooting, you should know about the difference. Here the whole rendering process is done in one stage, inside Ghostscript, using an appropriate device for the target printer. cupsomatic uses PPDs that are generated from the Foomatic Printer & Driver Database at Linuxprinting.org.

You can recognize these PPDs from the line calling the cupsomatic filter:

*cupsFilter: "application/vnd.cups-postscript  0  cupsomatic"

You may find this line among the first 40 or so lines of the PPD file. If you have such a PPD installed, the printer shows up in the CUPS Web interface with a foomatic namepart for the driver description. cupsomatic is a Perl script that runs Ghostscript with all the complicated command-line options autoconstructed from the selected PPD and command line options give to the print job.

However, cupsomatic is now deprecated. Its PPDs (especially the first generation of them, still in heavy use out there) are not meeting the Adobe specifications. You might also suffer difficulties when you try to download them with “Point'n'Print” to Windows clients. A better and more powerful successor is now in a stable beta-version: it is called foomatic-rip. To use foomatic-rip as a filter with CUPS, you need the new type of PPDs, which have a similar but different line:

*cupsFilter: "application/vnd.cups-postscript  0  foomatic-rip"

The PPD-generating engine at Linuxprinting.org has been revamped. The new PPDs comply with the Adobe spec. They also provide a new way to specify different quality levels (hi-res photo, normal color, grayscale, and draft) with a single click, whereas before you could have required five or more different selections (media type, resolution, inktype, and dithering algorithm). There is support for custom-size media built in. There is support to switch print options from page to page in the middle of a job. And the best thing is that the new foomatic-rip works seamlessly with all legacy spoolers too (like LPRng, BSD-LPD, PDQ, PPR, and so on), providing for them access to use PPDs for their printing.

If you want to see an overview of all the filters and how they relate to each other, the complete picture of the puzzle is at the end of this chapter.

CUPS autoconstructs all possible filtering chain paths for any given MIME type and every printer installed. But how does it decide in favor of or against a specific alternative? (There may be cases where there is a choice of two or more possible filtering chains for the same target printer.) Simple. You may have noticed the figures in the third column of the mime.convs file. They represent virtual costs assigned to this filter. Every possible filtering chain will sum up to a total “filter cost.” CUPS decides for the most “inexpensive” route.

You can tell CUPS to print (nearly) any file “raw”. “Raw” means it will not be filtered. CUPS will send the file to the printer “as is” without bothering if the printer is able to digest it. Users need to take care themselves that they send sensible data formats only. Raw printing can happen on any queue if the “-o raw” option is specified on the command line. You can also set up raw-only queues by simply not associating any PPD with it. This command:

$ lpadmin -P rawprinter -v socket://11.12.13.14:9100 -E

sets up a queue named “rawprinter”, connected via the “socket” protocol (a.k.a. “HP JetDirect”) to the device at IP address 11.12.1.3.14, using port 9100. (If you had added a PPD with -P /path/to/PPD to this command line, you would have installed a “normal” print queue.)

CUPS will automatically treat each job sent to a queue as a “raw” one if it can't find a PPD associated with the queue. However, CUPS will only send known MIME types (as defined in its own mime.types file) and refuse others.

Any MIME type with no rule in the /etc/cups/mime.types file is regarded as unknown or application/octet-stream and will not be sent. Because CUPS refuses to print unknown MIME types by default, you will probably have experienced that print jobs originating from Windows clients were not printed. You may have found an error message in your CUPS logs like:

Unable to convert file 0 to printable format for job

To enable the printing of application/octet-stream files, edit these two files:

Both contain entries (at the end of the respective files) that must be uncommented to allow raw mode operation for application/octet-stream. In /etc/cups/mime.types make sure this line is present:

application/octet-stream

This line (with no specific autotyping rule set) makes all files not otherwise auto-typed a member of application/octet-stream. In /etc/cups/mime.convs, have this line:

application/octet-stream   application/vnd.cups-raw   0   -

This line tells CUPS to use the Null Filter (denoted as “-”, doing nothing at all) on application/octet-stream, and tag the result as application/vnd.cups-raw. This last one is always a green light to the CUPS scheduler to now hand the file over to the backend connecting to the printer and sending it over.

Background.  That CUPS is a more security-aware printing system than traditional ones does not by default allow one to send deliberate (possibly binary) data to printing devices. (This could be easily abused to launch a Denial of Service attack on your printer(s), causing at least the loss of a lot of paper and ink.) “Unknown” data are regarded by CUPS as MIME type application/octet-stream. While you can send data “raw”, the MIME type for these must be one that is known to CUPS and allowed by it. The file /etc/cups/mime.types defines the “rules” of how CUPS recognizes MIME types. The file /etc/cups/mime.convs decides which file conversion filter(s) may be applied to which MIME types.

Originally PPDs were meant to be used for PostScript printers only. Here, they help to send device-specific commands and settings to the RIP, which processes the job file. CUPS has extended this scope for PPDs to cover non-PostScript printers too. This was not difficult, because it is a standardized file format. In a way it was logical too: CUPS handles PostScript and uses a PostScript RIP (Ghostscript) to process the job files. The only difference is that a PostScript printer has the RIP built-in, for other types of printers the Ghostscript RIP runs on the host computer.

PPDs for a non-PostScript printer have a few lines that are unique to CUPS. The most important one looks similar to this:

*cupsFilter: application/vnd.cups-raster  66   rastertoprinter

It is the last piece in the CUPS filtering puzzle. This line tells the CUPS daemon to use as a last filter rastertoprinter. This filter should be served as input an application/vnd.cups-raster MIME type file. Therefore, CUPS should autoconstruct a filtering chain, which delivers as its last output the specified MIME type. This is then taken as input to the specified rastertoprinter filter. After the last filter has done its work (rastertoprinter is a Gimp-Print filter), the file should go to the backend, which sends it to the output device.

CUPS by default ships only a few generic PPDs, but they are good for several hundred printer models. You may not be able to control different paper trays, or you may get larger margins than your specific model supports. See Table 21.1??? for summary information.

Native CUPS rasterization works in two steps:

Often this produces better quality (and has several more advantages) than other methods. This is shown in the cupsomatic/foomatic Processing Versus Native CUPS illustration.

Figure 21.10. cupsomatic/foomatic Processing Versus Native CUPS.

One other method is the cupsomatic/foomatic-rip way. Note that cupsomatic is not made by the CUPS developers. It is an independent contribution to printing development, made by people from Linuxprinting.org.^[6] cupsomatic is no longer developed, maintained, or supported. It now been replaced by foomatic-rip. foomatic-rip is a complete rewrite of the old cupsomatic idea, but very much improved and generalized to other (non-CUPS) spoolers. An upgrade to foomatic-rip is strongly advised, especially if you are upgrading to a recent version of CUPS, too.

Like the old cupsomatic method, the foomatic-rip (new) method from Linuxprinting.org uses the traditional Ghostscript print file processing, doing everything in a single step. It therefore relies on all the other devices built into Ghostscript. The quality is as good (or bad) as Ghostscript rendering is in other spoolers. The advantage is that this method supports many printer models not supported (yet) by the more modern CUPS method.

Of course, you can use both methods side by side on one system (and even for one printer, if you set up different queues) and find out which works best for you.

cupsomatic kidnaps the print file after the application/vnd.cups-postscript stage and deviates it through the CUPS-external, systemwide Ghostscript installation. Therefore, the print file bypasses the pstoraster filter (and also bypasses the CUPS raster drivers rastertosomething). After Ghostscript finished its rasterization, cupsomatic hands the rendered file directly to the CUPS backend. cupsomatic/foomatic Processing Versus Native CUPS, illustrates the difference between native CUPS rendering and the Foomatic/cupsomatic method.

Here are a few examples of commonly occurring filtering chains to illustrate the workings of CUPS.

Assume you want to print a PDF file to an HP JetDirect-connected PostScript printer, but you want to print pages 3-5, 7, and 11-13 only, and you want to print them “two-up” and “duplex”:

The resulting filter chain, therefore, is as shown in the PDF to socket chain illustration.

Figure 21.11. PDF to Socket Chain.

Assume you want to print the same filter to an USB-connected Epson Stylus Photo Printer installed with the CUPS stphoto2.ppd. The first few filtering stages are nearly the same:

The resulting filter chain therefore is as shown in the PDF to USB Chain illustration.

Figure 21.12. PDF to USB Chain.

On the Internet you can now find many thousands of CUPS-PPD files (with their companion filters), in many national languages supporting more than 1,000 non-PostScript models.

CUPS also supports the use of “interface scripts” as known from System V AT&T printing systems. These are often used for PCL printers, from applications that generate PCL print jobs. Interface scripts are specific to printer models. They have a role similar to PPDs for PostScript printers. Interface scripts may inject the Escape sequences as required into the print data stream if the user, for example, selects a certain paper tray, or changes paper orientation, or uses A3 paper. Interface scripts are practically unknown in the Linux realm. On HP-UX platforms they are more often used. You can use any working interface script on CUPS too. Just install the printer with the -i option:

root# lpadmin -p pclprinter -v socket://11.12.13.14:9100 \
          -i /path/to/interface-script

Interface scripts might be the “unknown animal” to many. However, with CUPS they provide the easiest way to plug in your own custom-written filtering script or program into one specific print queue (some information about the traditional use of interface scripts is found at http://playground.sun.com/printing/documentation/interface.html).

Windows clients printing to an NT-based print server have two options. They may:

Both print paths are shown in the flowcharts in Print Driver Execution on the Client, and Print Driver Execution on the Server.

In the first case, the print server must spool the file as raw, meaning it shouldn't touch the job file and try to convert it in any way. This is what a traditional UNIX-based print server can do too, and at a better performance and more reliably than an NT print server. This is what most Samba administrators probably are familiar with. One advantage of this setup is that this “spooling-only” print server may be used even if no driver(s) for UNIX is available. It is sufficient to have the Windows client drivers available and installed on the clients. This is illustrated in the Print Driver Execution on the Client diagram.

Figure 21.13. Print Driver Execution on the Client.

The other path executes the printer driver on the server. The client transfers print files in EMF format to the server. The server uses the PostScript, PCL, ESC/P, or other driver to convert the EMF file into the printer-specific language. It is not possible for UNIX to do the same. Currently, there is no program or method to convert a Windows client's GDI output on a UNIX server into something a printer could understand. This is illustrated in the Print Driver Execution on the Server diagram.

Figure 21.14. Print Driver Execution on the Server.

However, something similar is possible with CUPS, so read on.

Here is a simple recipe showing how you can take advantage of CUPS's powerful features for the benefit of your Windows network printing clients:

This requires the clients to use a PostScript driver (even if the printer is a non-PostScript model. It also requires that you have a driver on the CUPS server.

First, to enable CUPS-based printing through Samba, the following options should be set in your smb.conf file [global] section:

When these parameters are specified, all manually set print directives (like print command or lppause command) in smb.conf (as well as in Samba itself) will be ignored. Instead, Samba will directly interface with CUPS through its application program interface (API), as long as Samba has been compiled with CUPS library (libcups) support. If Samba has not been compiled with CUPS support, and if no other print commands are set up, then printing will use the System V AT&T command set, with the -oraw option automatically passing through (if you want your own defined print commands to work with a Samba server that has CUPS support compiled in, simply use classicalprinting = sysv). This is illustrated in the Printing via CUPS/Samba Server diagram.

Figure 21.15. Printing via CUPS/Samba Server.

Samba must use its own spool directory (it is set by a line similar to path = /var/spool/samba, in the [printers] or [printername] section of smb.conf). Samba receives the job in its own spool space and passes it into the spool directory of CUPS (the CUPS spool directory is set by the RequestRoot directive in a line that defaults to RequestRoot /var/spool/cups). CUPS checks the access rights of its spool directory and resets it to healthy values with every restart. We have seen quite a few people who used a common spooling space for Samba and CUPS, and struggled for weeks with this “problem.”

A Windows user authenticates only to Samba (by whatever means is configured). If Samba runs on the same host as CUPS, you only need to allow “localhost” to print. If it runs on different machines, you need to make sure the Samba host gets access to printing on CUPS.

CUPS does not limit itself to “real” PostScript printers in its use of PPDs. The CUPS developers have extended the scope of the PPD concept to also describe available device and driver options for non-PostScript printers through CUPS-PPDs.

This is logical, because CUPS includes a fully featured PostScript interpreter (RIP). This RIP is based on Ghostscript. It can process all received PostScript (and additionally many other file formats) from clients. All CUPS-PPDs geared to non-PostScript printers contain an additional line, starting with the keyword *cupsFilter. This line tells the CUPS print system which printer-specific filter to use for the interpretation of the supplied PostScript. Thus CUPS lets all its printers appear as PostScript devices to its clients, because it can act as a PostScript RIP for those printers, processing the received PostScript code into a proper raster print format.

CUPS-PPDs can also be used on Windows clients, on top of a “core” PostScript driver (now recommended is the CUPS PostScript Driver for Windows NT/200x/XP; you can also use the Adobe one, with limitations). This feature enables CUPS to do a few tricks no other spooler can do:

Using CUPS PPDs on Windows clients enables them to control all print job settings just as a UNIX client can do.

Windows NT printer drivers, which run in “kernel mode”, introduce a high risk for the stability of the system if the driver is not really stable and well-tested. And there are a lot of bad drivers out there! Especially notorious is the example of the PCL printer driver that had an additional sound module running to notify users via soundcard of their finished jobs. Do I need to say that this one was also reliably causing “blue screens of death” on a regular basis?

PostScript drivers are generally well-tested. They are not known to cause any problems, even though they also run in kernel mode. This might be because until now there have been only two different PostScript drivers: the one from Adobe and the one from Microsoft. Both are well-tested and are as stable as you can imagine on Windows. The CUPS driver is derived from the Microsoft one.

In an attempt to work around problems, site administrators have resorted to restricting the allowed drivers installed on their WTS to one generic PCL and one PostScript driver. This, however, restricts the the number of printer options available for clients to use. Often they can't get out more than simplex prints from one standard paper tray, while their devices could do much better if driven by a different driver!

Using a PostScript driver, enabled with a CUPS-PPD, seems to be a very elegant way to overcome all these shortcomings. There are, depending on the version of Windows OS you use, up to three different PostScript drivers now available: Adobe, Microsoft, and CUPS PostScript drivers. None of them is known to cause major stability problems on WTS (even if used with many different PPDs). The clients will be able to (again) choose paper trays, duplex printing, and other settings. However, there is a certain price for this too: a CUPS server acting as a PostScript RIP for its clients requires more CPU and RAM than when just acting as a “raw spooling” device. Plus, this setup is not yet widely tested, although the first feedbacks look very promising.

More recent printer drivers on W200x and XP no longer run in kernel mode (unlike Windows NT). However, both operating systems can still use the NT drivers, running in kernel mode (you can roughly tell which is which as the drivers in subdirectory “2” of “W32X86” are “old” ones). As was said before, the Adobe as well as the Microsoft PostScript drivers are not known to cause any stability problems. The CUPS driver is derived from the Microsoft one. There is a simple reason for this: the MS DDK (Device Development Kit) for Windows NT (which used to be available at no cost to licensees of Visual Studio) includes the source code of the Microsoft driver, and licensees of Visual Studio are allowed to use and modify it for their own driver development efforts. This is what the CUPS people have done. The license does not allow them to publish the whole of the source code. However, they have released the “diff” under the GPL, and if you are the owner of an “MS DDK for Windows NT,” you can check the driver yourself.

The cupsaddsmb utility (shipped with all current CUPS versions) is an alternative method to transfer printer drivers into the Samba [print$] share. Remember, this share is where clients expect drivers deposited and set up for download and installation. It makes the sharing of any (or all) installed CUPS printers quite easy. cupsaddsmb can use the Adobe PostScript driver as well as the newly developed CUPS PostScript driver for Windows NT/200x/XP. cupsaddsmb does not work with arbitrary vendor printer drivers, but only with the exact driver files that are named in its man page.

The CUPS printer driver is available from the CUPS download site. Its package name is cups-samba-[version].tar.gz. It is preferred over the Adobe drivers because it has a number of advantages:

However, currently only Windows NT, 2000, and XP are supported by the CUPS drivers. You will also need to get the respective part of the Adobe driver if you need to support Windows 95, 98, and Me clients.

Prior to running cupsaddsmb, you need the settings in smb.conf as shown in the smb.conf for cupsaddsmb Usage.

CUPS users may get the exact same package from http://www.cups.org/software.html. It is a separate package from the CUPS-based software files, tagged as CUPS 1.1.x Windows NT/200x/XP Printer Driver for Samba (tar.gz, 192k). The filename to download is cups-samba-1.1.x.tar.gz. Upon untar and unzipping, it will reveal these files:

root# tar xvzf cups-samba-1.1.19.tar.gz
cups-samba.install
cups-samba.license
cups-samba.readme
cups-samba.remove
cups-samba.ss

These have been packaged with the ESP meta-packager software EPM. The *.install and *.remove files are simple shell scripts, which untar the *.ss (the *.ss is nothing else but a tar archive, which can be untarred by “tar” too). Then it puts the content into /usr/share/cups/drivers/. This content includes three files:

root# tar tv cups-samba.ss
cupsdrvr.dll
cupsui.dll
cups.hlp  

The cups-samba.install shell scripts are easy to handle:

root# ./cups-samba.install
[....]
Installing software...
Updating file permissions...
Running post-install commands...
Installation is complete.       

The script should automatically put the driver files into the /usr/share/cups/drivers/ directory:

root# cp /usr/share/drivers/cups.hlp /usr/share/cups/drivers/

This new CUPS PostScript driver is currently binary only, but free of charge. No complete source code is provided (yet). The reason is that it has been developed with the help of the Microsoft DDK and compiled with Microsoft Visual Studio 6. Driver developers are not allowed to distribute the whole of the source code as free software. However, CUPS developers released the “diff” in source code under the GPL, so anybody with a license for Visual Studio and a DDK will be able to compile for himself or herself.

The CUPS drivers do not support the older Windows 95/98/Me, but only the Windows NT/2000/XP client.

Windows NT, 2000, and XP are supported by:

Adobe drivers are available for the older Windows 95/98/Me as well as for Windows NT/2000/XP clients. The set of files is different from the different platforms.

Windows 95, 98, and ME are supported by:

Windows NT, 2000, and XP are supported by:

Acquiring the Adobe driver files seems to be unexpectedly difficult for many users. They are not available on the Adobe Web site as single files, and the self-extracting and/or self-installing Windows-.exe is not easy to locate either. You probably need to use the included native installer and run the installation process on one client once. This will install the drivers (and one generic PostScript printer) locally on the client. When they are installed, share the generic PostScript printer. After this, the client's [print$] share holds the Adobe files, which you can get with smbclient from the CUPS host.

Users of the ESP Print Pro software are able to install the ESP print drivers package as an alternative to the Adobe PostScript drivers. To do so, retrieve the driver files from the normal download area of the ESP Print Pro software at Easy Software web site. You need to locate the link labeled “SAMBA” among the Download Printer Drivers for ESP Print Pro 4.x area and download the package. Once installed, you can prepare any driver by simply highlighting the printer in the Printer Manager GUI and selecting Export Driver... from the menu. Of course, you need to have prepared Samba beforehand to handle the driver files; that is, set up the [print$] share, and so on. The ESP Print Pro package includes the CUPS driver files as well as a (licensed) set of Adobe drivers for the Windows 95/98/Me client family.

Once you have run the install script (and possibly manually moved the cups.hlp file to /usr/share/cups/drivers/), the driver is ready to be put into Samba's [print$] share (which often maps to /etc/samba/drivers/ and contains a subdirectory tree with WIN40 and W32X86 branches). You do this by running cupsaddsmb (see also man cupsaddsmb for CUPS since release 1.1.16).

Once the driver files are in the [print$] share and are initialized, they are ready to be downloaded and installed by the Windows NT/200x/XP clients.

Are you interested in a comparison between the CUPS and the Adobe PostScript drivers? For our purposes, these are the most important items that weigh in favor of CUPS:

The cupsaddsmb command copies the needed files into your [print$] share. Additionally, the PPD associated with this printer is copied from /etc/cups/ppd/ to [print$]. There the files wait for convenient Windows client installations via Point'n'Print. Before we can run the command successfully, we need to be sure that we can authenticate toward Samba. If you have a small network, you are probably using user-level security (security = user).

Here is an example of a successfully run cupsaddsmb command:

root# cupsaddsmb -U root infotec_IS2027
Password for root required to access localhost via Samba: ['secret']

To share all printers and drivers, use the -a parameter instead of a printer name. Since cupsaddsmb “exports” the printer drivers to Samba, it should be obvious that it only works for queues with a CUPS driver associated.

Probably you want to see what's going on. Use the -v parameter to get a more verbose output. The output below was edited for better readability: all “\” at the end of a line indicate that I inserted an artificial line break plus some indentation here:

root# cupsaddsmb -U root -v infotec_2105
Password for root required to access localhost via GANDALF:
Running command: smbclient //localhost/print\$ -N -U'root%secret' \
    -c 'mkdir W32X86; \
    put /var/spool/cups/tmp/3e98bf2d333b5 W32X86/infotec_2105.ppd; \
	put /usr/share/cups/drivers/cupsdrvr.dll W32X86/cupsdrvr.dll; \
    put /usr/share/cups/drivers/cupsui.dll W32X86/cupsui.dll; \
    put /usr/share/cups/drivers/cups.hlp W32X86/cups.hlp'
added interface ip=10.160.51.60 bcast=10.160.51.255 nmask=255.255.252.0
Domain=[CUPS-PRINT] OS=[UNIX] Server=[Samba 2.2.7a]
NT_STATUS_OBJECT_NAME_COLLISION making remote directory \W32X86
putting file /var/spool/cups/tmp/3e98bf2d333b5 as \W32X86/infotec_2105.ppd
putting file /usr/share/cups/drivers/cupsdrvr.dll as \W32X86/cupsdrvr.dll
putting file /usr/share/cups/drivers/cupsui.dll as \W32X86/cupsui.dll
putting file /usr/share/cups/drivers/cups.hlp as \W32X86/cups.hlp
  
Running command: rpcclient localhost -N -U'root%secret' 
   -c 'adddriver "Windows NT x86"   \
   "infotec_2105:cupsdrvr.dll:infotec_2105.ppd:cupsui.dll:cups.hlp:NULL: \
    RAW:NULL"'
cmd = adddriver "Windows NT x86" \
   "infotec_2105:cupsdrvr.dll:infotec_2105.ppd:cupsui.dll:cups.hlp:NULL: \
	RAW:NULL"
Printer Driver infotec_2105 successfully installed.
  
Running command: smbclient //localhost/print\$ -N -U'root%secret' \
-c 'mkdir WIN40; \
    put /var/spool/cups/tmp/3e98bf2d333b5 WIN40/infotec_2105.PPD; \
	put /usr/share/cups/drivers/ADFONTS.MFM WIN40/ADFONTS.MFM;   \
    put /usr/share/cups/drivers/ADOBEPS4.DRV WIN40/ADOBEPS4.DRV; \
    put /usr/share/cups/drivers/ADOBEPS4.HLP WIN40/ADOBEPS4.HLP; \
    put /usr/share/cups/drivers/DEFPRTR2.PPD WIN40/DEFPRTR2.PPD; \
	put /usr/share/cups/drivers/ICONLIB.DLL WIN40/ICONLIB.DLL; \
	put /usr/share/cups/drivers/PSMON.DLL WIN40/PSMON.DLL;'
  added interface ip=10.160.51.60 bcast=10.160.51.255 nmask=255.255.252.0
  Domain=[CUPS-PRINT] OS=[UNIX] Server=[Samba 2.2.7a]
  NT_STATUS_OBJECT_NAME_COLLISION making remote directory \WIN40
  putting file /var/spool/cups/tmp/3e98bf2d333b5 as \WIN40/infotec_2105.PPD
  putting file /usr/share/cups/drivers/ADFONTS.MFM as \WIN40/ADFONTS.MFM
  putting file /usr/share/cups/drivers/ADOBEPS4.DRV as \WIN40/ADOBEPS4.DRV
  putting file /usr/share/cups/drivers/ADOBEPS4.HLP as \WIN40/ADOBEPS4.HLP
  putting file /usr/share/cups/drivers/DEFPRTR2.PPD as \WIN40/DEFPRTR2.PPD
  putting file /usr/share/cups/drivers/ICONLIB.DLL as \WIN40/ICONLIB.DLL
  putting file /usr/share/cups/drivers/PSMON.DLL as \WIN40/PSMON.DLL
  
  Running command: rpcclient localhost -N -U'root%secret' \
   -c 'adddriver "Windows 4.0"      \
   "infotec_2105:ADOBEPS4.DRV:infotec_2105.PPD:NULL:ADOBEPS4.HLP: \
   PSMON.DLL:RAW:ADOBEPS4.DRV,infotec_2105.PPD,ADOBEPS4.HLP,PSMON.DLL, \
    ADFONTS.MFM,DEFPRTR2.PPD,ICONLIB.DLL"'
	cmd = adddriver "Windows 4.0" "infotec_2105:ADOBEPS4.DRV:\
	infotec_2105.PPD:NULL:ADOBEPS4.HLP:PSMON.DLL:RAW:ADOBEPS4.DRV,\
	infotec_2105.PPD,ADOBEPS4.HLP,PSMON.DLL,ADFONTS.MFM,DEFPRTR2.PPD,\
	ICONLIB.DLL"
  Printer Driver infotec_2105 successfully installed.
  
  Running command: rpcclient localhost -N -U'root%secret'  \
   -c 'setdriver infotec_2105 infotec_2105'
  cmd = setdriver infotec_2105 infotec_2105
  Successfully set infotec_2105 to driver infotec_2105.

If you look closely, you'll discover your root password was transferred unencrypted over the wire, so beware! Also, if you look further, you may discover error messages like NT_STATUS_OBJECT_NAME_COLLISION in the output. This will occur when the directories WIN40 and W32X86 already existed in the [print$] driver download share (from a previous driver installation). These are harmless warning messages.

What has happened? What did cupsaddsmb do? There are five stages of the procedure:

root# cupsaddsmb -H sambaserver -h cupsserver -v printer

You must always check if the utility completed successfully in all fields. You need at minimum these three messages among the output:

These messages are probably not easily recognized in the general output. If you run cupsaddsmb with the -a parameter (which tries to prepare all active CUPS printer drivers for download), you might miss if individual printer drivers had problems installing properly. A redirection of the output will help you analyze the results in retrospective.

If you get:

SetPrinter call failed!
result was WERR_ACCESS_DENIED

it means that you might have set use client driver = yes for this printer. Setting it to “no” will solve the problem. Refer to the smb.conf man page for explanation of the use client driver.

Can't get the standard cupsaddsmb command to run on a Samba PDC? Are you asked for the password credential again and again, and the command just will not take off at all? Try one of these variations:

root# cupsaddsmb -U MIDEARTH\\root -v printername
root# cupsaddsmb -H SAURON -U MIDEARTH\\root -v printername
root# cupsaddsmb -H SAURON -U MIDEARTH\\root -h cups-server -v printername

(Note the two backslashes: the first one is required to “escape” the second one).

The cupsaddsmb Flowchart shows a chart about the procedures, command flows, and data flows of the cupaddsmb command. Note again: cupsaddsmb is not intended to, and does not work with, raw print queues!

Figure 21.16. cupsaddsmb Flowchart.

After cupsaddsmb is completed, your driver is prepared for the clients to use. Here are the steps you must perform to download and install it via Point'n'Print. From a Windows client, browse to the CUPS/Samba server:

After a few seconds, there should be a new printer in your client's local Printers folder. On Windows XP it will follow a naming convention of PrinterName on SambaServer. (In my current case it is infotec_2105 on kde-bitshop). If you want to test it and send your first job from an application like Winword, the new printer appears in a \\SambaServer\PrinterName entry in the drop-down list of available printers.

cupsaddsmb will only reliably work with CUPS version 1.1.15 or higher and with Samba version 2.2.4, or later. If it does not work, or if the automatic printer driver download to the clients does not succeed, you can still manually install the CUPS printer PPD on top of the Adobe PostScript driver on clients. Then point the client's printer queue to the Samba printer share for a UNC type of connection:

C:\> net use lpt1: \\sambaserver\printershare /user:ntadmin

should you desire to use the CUPS networked PostScript RIP functions. (Note that user “ntadmin” needs to be a valid Samba user with the required privileges to access the printershare.) This sets up the printer connection in the traditional LanMan way (not using MS-RPC).

Printing works, but there are still problems. Most jobs print well, some do not print at all. Some jobs have problems with fonts, which do not look very good. Some jobs print fast and some are dead-slow. Many of these problems can be greatly reduced or even completely eliminated if you follow a few guidelines. Remember, if your print device is not PostScript-enabled, you are treating your Ghostscript installation on your CUPS host with the output your client driver settings produce. Treat it well:

First let's check the rpcclient man page. Here are two relevant passages:

adddriver <arch> <config> Execute an AddPrinterDriver() RPC to install the printer driver information on the server. The driver files should already exist in the directory returned by getdriverdir. Possible values for arch are the same as those for the getdriverdir command. The config parameter is defined as follows:

Long Printer Name:\
Driver File Name:\
Data File Name:\
Config File Name:\
Help File Name:\
Language Monitor Name:\
Default Data Type:\
Comma Separated list of Files

Any empty fields should be entered as the string “NULL”.

Samba does not need to support the concept of print monitors, since these only apply to local printers whose drivers can use a bidirectional link for communication. This field should be “NULL”. On a remote NT print server, the print monitor for a driver must already be installed before adding the driver or else the RPC will fail.

setdriver <printername> <drivername> Execute a SetPrinter() command to update the printer driver associated with an installed printer. The printer driver must already be correctly installed on the print server.

See also the enumprinters and enumdrivers commands to obtain a list of installed printers and drivers.

The exact format isn't made too clear by the man page, since you have to deal with some parameters containing spaces. Here is a better description for it. We have line-broken the command and indicated the breaks with “\”. Usually you would type the command in one line without the line breaks:

adddriver "Architecture" \
   "LongPrinterName:DriverFile:DataFile:ConfigFile:HelpFile:\
   LanguageMonitorFile:DataType:ListOfFiles,Comma-separated"

What the man pages denote as a simple <config> keyword in reality consists of eight colon-separated fields. The last field may take multiple (in some very insane cases, even 20 different additional) files. This might sound confusing at first. What the man pages call the “LongPrinterName” in reality should be called the “Driver Name”. You can name it anything you want, as long as you use this name later in the rpcclient ... setdriver command. For practical reasons, many name the driver the same as the printer.

It isn't simple at all. I hear you asking: “How do I know which files are Driver File”, “Data File”, “Config File”, “Help File” and “Language Monitor File in each case?” For an answer, you may want to have a look at how a Windows NT box with a shared printer presents the files to us. Remember that this whole procedure has to be developed by the Samba Team by listening to the traffic caused by Windows computers on the wire. We may as well turn to a Windows box now and access it from a UNIX workstation. We will query it with rpcclient to see what it tells us and try to understand the man page more clearly.

We could run rpcclient with a getdriver or a getprinter subcommand (in level 3 verbosity) against it. Just sit down at a UNIX or Linux workstation with the Samba utilities installed, then type the following command:

root# rpcclient -U'user%secret' NT-SERVER -c 'getdriver printername 3'

From the result it should become clear which is which. Here is an example from my installation:

root# rpcclient -U'Danka%xxxx' W200xSERVER \
    -c'getdriver "DANKA InfoStream Virtual Printer" 3'
    cmd = getdriver "DANKA InfoStream Virtual Printer" 3

 [Windows NT x86]
 Printer Driver Info 3:
         Version: [2]
         Driver Name: [DANKA InfoStream]
         Architecture: [Windows NT x86]
         Driver Path: [C:\WINNT\System32\spool\DRIVERS\W32X86\2\PSCRIPT.DLL]
         Datafile: [C:\WINNT\System32\spool\DRIVERS\W32X86\2\INFOSTRM.PPD]
         Configfile: [C:\WINNT\System32\spool\DRIVERS\W32X86\2\PSCRPTUI.DLL]
         Helpfile: [C:\WINNT\System32\spool\DRIVERS\W32X86\2\PSCRIPT.HLP]
 
         Dependentfiles: []
         Dependentfiles: []
         Dependentfiles: []
         Dependentfiles: []
         Dependentfiles: []
         Dependentfiles: []
         Dependentfiles: []
 
         Monitorname: []
         Defaultdatatype: []

Some printer drivers list additional files under the label Dependentfiles, and these would go into the last field ListOfFiles,Comma-separated. For the CUPS PostScript drivers, we do not need any (nor would we for the Adobe PostScript driver); therefore, the field will get a “NULL” entry.

From the man page (and from the quoted output of cupsaddsmb above) it becomes clear that you need to have certain conditions in order to make the manual uploading and initializing of the driver files succeed. The two rpcclient subcommands (adddriver and setdriver) need to encounter the following preconditions to complete successfully:

We are going to install a printer driver now by manually executing all required commands. Because this may seem a rather complicated process at first, we go through the procedure step by step, explaining every single action item as it comes up.

The setdriver command will fail if in Samba's mind the queue is not already there. A successful installation displys the promising message that the:

Printer Driver ABC successfully installed.

following the adddriver parts of the procedure. But you may also see a disappointing message like this one: result was NT_STATUS_UNSUCCESSFUL

It is not good enough that you can see the queue in CUPS, using the lpstat -p ir85wm command. A bug in most recent versions of Samba prevents the proper update of the queue list. The recognition of newly installed CUPS printers fails unless you restart Samba or send a HUP to all smbd processes. To verify if this is the reason why Samba does not execute the setdriver command successfully, check if Samba “sees” the printer:

root# rpcclient transmeta -N -U'root%xxxx' -c 'enumprinters 0'|grep ir85wm
        printername:[ir85wm]

An alternate command could be this:

root# rpcclient transmeta -N -U'root%secret' -c 'getprinter ir85wm' 
        cmd = getprinter ir85wm
        flags:[0x800000]
        name:[\\transmeta\ir85wm]
        description:[\\transmeta\ir85wm,ir85wm,DPD]
        comment:[CUPS PostScript-Treiber for Windows NT/200x/XP]

By the way, you can use these commands, plus a few more, of course, to install drivers on remote Windows NT print servers too!

A Windows NT (print) server keeps track of all information needed to serve its duty toward its clients by storing entries in the Windows registry. Client queries are answered by reading from the registry, Administrator or user configuration settings that are saved by writing into the registry. Samba and UNIX obviously do not have such a Registry. Samba instead keeps track of all client-related information in a series of *.tdb files. (TDB stands for trivial data base). These are often located in /var/lib/samba/ or /var/lock/samba/. The printing-related files are ntprinters.tdb, printing.tdb,ntforms.tdb, and ntdrivers.tdb.

*.tdb files are not human readable. They are written in a binary format. “Why not ASCII?”, you may ask. “After all, ASCII configuration files are a good and proven tradition on UNIX.” The reason for this design decision by the Samba Team is mainly performance. Samba needs to be fast; it runs a separate smbd process for each client connection, in some environments many thousands of them. Some of these smbds might need to write-access the same *.tdb file at the same time. The file format of Samba's *.tdb files allows for this provision. Many smbd processes may write to the same *.tdb file at the same time. This wouldn't be possible with pure ASCII files.

It is very important that all *.tdb files remain consistent over all write and read accesses. However, it may happen that these files do get corrupted. (A kill -9 `pidof smbd' while a write access is in progress could do the damage, as could a power interruption, etc.). In cases of trouble, a deletion of the old printing-related *.tdb files may be the only option. After that, you need to re-create all print-related setups unless you have made a backup of the *.tdb files in time.

Samba ships with a little utility that helps the root user of your system to backup your *.tdb files. If you run it with no argument, it prints a usage message:

root# tdbbackup
 Usage: tdbbackup [options] <fname...>
 
 Version:3.0a
   -h            this help message
   -s suffix     set the backup suffix
   -v            verify mode (restore if corrupt)

Here is how I backed up my printing.tdb file:

root# ls
.              browse.dat     locking.tdb     ntdrivers.tdb printing.tdb
..             share_info.tdb connections.tdb messages.tdb  ntforms.tdb
printing.tdbkp unexpected.tdb brlock.tdb      gmon.out      namelist.debug  
ntprinters.tdb sessionid.tdb
 
root# tdbbackup -s .bak printing.tdb
 printing.tdb : 135 records
 
root# ls -l printing.tdb*
 -rw-------    1 root     root        40960 May  2 03:44 printing.tdb
 -rw-------    1 root     root        40960 May  2 03:44 printing.tdb.bak

Nowadays, most Linux distributions rely on the utilities from the Linuxprinting.org to create their printing-related software (which, by the way, works on all UNIXes and on Mac OS X and Darwin, too). The utilities from this sire have a very end-user-friendly interface that allows for an easy update of drivers and PPDs for all supported models, all spoolers, all operating systems, and all package formats (because there is none). Its history goes back a few years.

Recently, Foomatic has achieved the astonishing milestone of 1,000 listed printer models. Linuxprinting.org keeps all the important facts about printer drivers, supported models, and which options are available for the various driver/printer combinations in its Foomatic database. Currently there are 245 drivers in the database. Many drivers support various models, and many models may be driven by different drivers its your choice!

Here are the steps to install a foomatic-rip-driven LaserJet 4 Plus-compatible printer in CUPS (note that recent distributions of SuSE, UnitedLinux and Mandrake may ship with a complete package of Foomatic-PPDs plus the foomatic-rip utility. Going directly to Linuxprinting.org ensures that you get the latest driver/PPD files).

Once you print to a print queue set up with the Foomatic PPD, CUPS will insert the appropriate commands and comments into the resulting PostScript job file. foomatic-rip is able to read and act upon these and uses some specially encoded Foomatic comments embedded in the job file. These in turn are used to construct (transparently for you, the user) the complicated Ghostscript command line telling the printer driver exactly how the resulting raster data should look and which printer commands to embed into the data stream. You need:

This is an example command of how root would set a print quota in CUPS, assuming an existing printer named “quotaprinter”:

root# lpadmin -p quotaprinter -o job-quota-period=604800 \
	-o job-k-limit=1024 -o job-page-limit=100

This would limit every single user to print no more than 100 pages or 1024 KB of data (whichever comes first) within the last 604,800 seconds ( = 1 week).

For CUPS to count correctly, the printfile needs to pass the CUPS pstops filter; otherwise it uses a dummy count of “one”. Some print files do not pass it (e.g., image files), but then those are mostly one-page jobs anyway. This also means that proprietary drivers for the target printer running on the client computers and CUPS/Samba, which then spool these files as “raw” (i.e., leaving them untouched, not filtering them), will be counted as one-pagers too!

You need to send PostScript from the clients (i.e., run a PostScript driver there) to have the chance to get accounting done. If the printer is a non-PostScript model, you need to let CUPS do the job to convert the file to a print-ready format for the target printer. This is currently working for about a thousand different printer models. Linuxprinting.org has a driver list.

Before CUPS 1.1.16, your only option was to use the Adobe PostScript driver on the Windows clients. The output of this driver was not always passed through the pstops filter on the CUPS/Samba side, and therefore was not counted correctly (the reason is that it often, depending on the PPD being used, wrote a PJL-header in front of the real PostScript, which caused CUPS to skip pstops and go directly to the pstoraster stage).

From CUPS 1.1.16 and later releases, you can use the CUPS PostScript driver for Windows NT/200x/XP clients (which is tagged in the download area of http://www.cups.org/ as the cups-samba-1.1.16.tar.gz package). It does not work for Windows 9x/Me clients, but it guarantees:

You can read more about the setup of this combination in the man page for cupsaddsmb (which is only present with CUPS installed, and only current from CUPS 1.1.16).

These are the items CUPS logs in the page_log for every page of a job:

Here is an extract of my CUPS server's page_log file to illustrate the format and included items:

tec_IS2027 kurt 401 [22/Apr/2003:10:28:43 +0100] 1 3 #marketing 10.160.50.13
tec_IS2027 kurt 401 [22/Apr/2003:10:28:43 +0100] 2 3 #marketing 10.160.50.13
tec_IS2027 kurt 401 [22/Apr/2003:10:28:43 +0100] 3 3 #marketing 10.160.50.13
tec_IS2027 kurt 401 [22/Apr/2003:10:28:43 +0100] 4 3 #marketing 10.160.50.13
Dig9110 boss 402 [22/Apr/2003:10:33:22 +0100] 1 440 finance-dep 10.160.51.33

This was job ID 401, printed on tec_IS2027 by user kurt, a 64-page job printed in three copies, billed to #marketing, and sent from IP address 10.160.50.13. The next job had ID 402, was sent by user boss from IP address 10.160.51.33, printed from one page 440 copies, and is set to be billed to finance-dep.

What flaws or shortcomings are there with this quota system?

This is the best system currently available, and there are huge improvements under development for CUPS 1.2:

Other accounting tools that can be used includes: PrintAnalyzer, pyKota, printbill, LogReport. For more information regarding these tools you can try a Google search.

Some important parameter settings in the CUPS configuration file cupsd.conf are:

(There are also additional settings for MaxJobsPerUser and MaxJobsPerPrinter.)

For everything to work as it should, you need to have three things:

If you want to do things manually, replace the printing = cups by printing = bsd. Then your manually set commands may work (I haven't tested this), and a print command = lp -d %P %s; rm %s may do what you need.

For Windows 9x/Me, clients require the printer names to be eight characters (or “8 plus 3 chars suffix”) max; otherwise, the driver files will not get transferred when you want to download them from Samba.

Have you set security = user? Have you used smbpasswd to give root a Samba account? You can do two things: open another terminal and execute smbpasswd -a root to create the account and continue entering the password into the first terminal. Or, break out of the loop by pressing Enter twice (without trying to type a password).

If the error is “Tree connect failed: NT_STATUS_BAD_NETWORK_NAME”, you may have forgotten to create the /etc/samba/drivers directory.

If cupsaddsmb, or rpcclient addriver emit the error message WERR_BAD_PASSWORD, refer to the previous common error.

The use of “cupsaddsmb” gives “No PPD file for printer...” message while PPD file is present. What might the problem be?

Have you enabled printer sharing on CUPS? This means, do you have a <Location /printers>....</Location> section in CUPS server's cupsd.conf that does not deny access to the host you run “cupsaddsmb” from? It could be an issue if you use cupsaddsmb remotely, or if you use it with a -h parameter: cupsaddsmb -H sambaserver -h cupsserver -v printername.

Is your TempDir directive in cupsd.conf set to a valid value, and is it writable?

Use smbstatus to check which user you are from Samba's point of view. Do you have the privileges to write into the [print$] share?

Once you are connected as the wrong user (for example, as nobody, which often occurs if you have map to guest = bad user), Windows Explorer will not accept an attempt to connect again as a different user. There will not be any bytes transferred on the wire to Samba, but still you'll see a stupid error message that makes you think Samba has denied access. Use smbstatus to check for active connections. Kill the PIDs. You still can't re-connect, and you get the dreaded You can't connect with a second account from the same machine message as soon as you try. And you do not see a single byte arriving at Samba (see logs; use “ethereal”) indicating a renewed connection attempt. Shut all Explorer Windows. This makes Windows forget what it has cached in its memory as established connections. Then reconnect as the right user. The best method is to use a DOS terminal window and first do net use z: \\GANDALF\print$ /user:root. Check with smbstatus that you are connected under a different account. Now open the Printers folder (on the Samba server in the Network Neighborhood), right-click on the printer in question, and select Connect.....

You see per smbstatus that you are connected as user nobody, but you want to be root or printer admin. This is probably due to map to guest = bad user, which silently connected you under the guest account when you gave (maybe by accident) an incorrect username. Remove map to guest if you want to prevent this.

This information came from a mailing list posting regarding problems experienced when upgrading from Adobe drivers to CUPS drivers on Microsoft Windows NT/200x/XP clients.

First delete all old Adobe-using printers. Then delete all old Adobe drivers. (On Windows 200x/XP, right-click in the background of Printers folder, select Server Properties..., select tab Drivers, and delete here).

Do you use the “naked” root user name? Try to do it this way: cupsaddsmb -U DOMAINNAME\\root -v printername> (note the two backslashes: the first one is required to “escape” the second one).

Deleting a printer on the client will not delete the driver too (to verify, right-click on the white background of the Printers folder, select Server Properties and click on the Drivers tab). These same old drivers will be re-used when you try to install a printer with the same name. If you want to update to a new driver, delete the old ones first. Deletion is only possible if no other printer uses the same driver.

Local security policies may not allow the installation of unsigned drivers “local security policies” may not allow the installation of printer drivers at all.

Windows XP handles SMB printers on a “per-user” basis. This means every user needs to install the printer himself or herself. To have a printer available for everybody, you might want to use the built-in IPP client capabilities of Win XP. Add a printer with the print path of http://cupsserver:631/printers/printername. We're still looking into this one. Maybe a logon script could automatically install printers for all users.

For print change, notify functions on NT++ clients. These need to run the Server service first (renamed to File & Print Sharing for MS Networks in XP).

Win XP-SP1 introduced a Point and Print Restriction Policy (this restriction does not apply to “Administrator” or “Power User” groups of users). In Group Policy Object Editor, go to User Configuration -> Administrative Templates -> Control Panel -> Printers. The policy is automatically set to Enabled and the Users can only Point and Print to machines in their Forest . You probably need to change it to Disabled or Users can only Point and Print to these servers to make driver downloads from Samba possible.

How are you doing it? I bet the wrong way (it is not easy to find out, though). There are three different ways to bring you to a dialog that seems to set everything. All three dialogs look the same, yet only one of them does what you intend. You need to be Administrator or Print Administrator to do this for all users. Here is how I do in on XP:

Do you see any difference? I don't either. However, only the last one, which you arrived at with steps “C.1. to C.6.”, will save any settings permanently and be the defaults for new users. If you want all clients to get the same defaults, you need to conduct these steps as Administrator (printer admin in smb.conf) before a client downloads the driver (the clients can later set their own per-user defaults by following the procedures A or B).

Don't use Optimize for Speed, but use Optimize for Portability instead (Adobe PS Driver). Don't use Page Independence: No. Always settle with Page Independence: Yes (Microsoft PS Driver and CUPS PS Driver for Windows NT/200x/XP). If there are problems with fonts, use Download as Softfont into printer (Adobe PS Driver). For TrueType Download Options choose Outline. Use PostScript Level 2 if you are having trouble with a non-PS printer and if there is a choice.

Symptom: The last command of cupsaddsmb does not complete successfully. If the cmd = setdriver printername printername result was NT_STATUS_UNSUCCESSFUL, then possibly the printer was not yet recognized by Samba. Did it show up in Network Neighborhood? Did it show up in rpcclient hostname -c `enumprinters'? Restart smbd (or send a kill -HUP to all processes listed by smbstatus, and try again.

Have you ever by accident set the CUPS spool directory to the same location (RequestRoot /var/spool/samba/ in cupsd.conf or the other way round: /var/spool/cups/ is set as path> in the [printers] section)? These must be different. Set RequestRoot /var/spool/cups/ in cupsd.conf and path = /var/spool/samba in the [printers] section of smb.conf. Otherwise, cupsd will sanitize permissions to its spool directory with each restart and printing will not work reliably.

In this case a print queue called “lp” intermittently swallows jobs and spits out completely different ones from what was sent.

It is a bad idea to name any printer “lp”. This is the traditional UNIX name for the default printer. CUPS may be set up to do an automatic creation of Implicit Classes. This means, to group all printers with the same name to a pool of devices and load-balance the jobs across them in a round-robin fashion. Chances are high that someone else has a printer named “lp” too. You may receive that person's jobs and send your own to his or her device unwittingly. To have tight control over the printer names, set BrowseShortNames No. It will present any printer as printername@cupshost, which gives you better control over what may happen in a large networked environment.

Use smbclient to connect to any Windows box with a shared PostScript printer: smbclient //windowsbox/print\$ -U guest. You can navigate to the W32X86/2 subdir to mget ADOBE* and other files or to WIN40/0 to do the same. Another option is to download the *.exe packaged files from the Adobe Web site.